-- DISCONTINUED PRODUCT --
LogiCORE™ IP
1-Gigabit Ethernet
MAC v8.5
User Guide
UG144 April 24, 2009
R
Download from Www.Somanuals.com. All Manuals Search And Download.
-- DISCONTINUED PRODUCT --
Revision History
The following table shows the revision history for this document.
Date
Version
1.0
Revision
09/30/04
04/28/05
01/18/06
07/13/06
09/21/06
02/15/07
08/08/07
Initial Xilinx release.
2.0
Updated to 1-Gigabit Ethernet MAC version 6.0, Xilinx tools v7.1i SP1.
Updated to 1-Gigabit Ethernet MAC version 7.0, Xilinx tools v8.1i.
Updated to 1-Gigabit Ethernet MAC version 8.0, Xilinx tools v8.2i.
Updated to 1-Gigabit Ethernet MAC version 8.1, added support for Spartan®-3A platform.
Updated to 1-Gigabit Ethernet MAC version 8.2, Xilinx tools v9.1i.
3.0
4.0
4.1
4.2
5.0
Advanced core version to 8.3, updated various tool versions and trademarks for the
IP1 I Jade Minor release.
03/24/08
04/24/09
6.0
7.0
Updated core version to 8.4; Xilinx tools 10.1.
Updated core version to 8.5; Xilinx tools 11.1.
The product was discontinued as of August 31, 2009.
1-Gigabit Ethernet MAC v8.5 User Guide
UG144 April 24, 2009
Download from Www.Somanuals.com. All Manuals Search And Download.
-- DISCONTINUED PRODUCT --
1-Gigabit Ethernet MAC v8.5 User Guide
UG144 April 24, 2009
Download from Www.Somanuals.com. All Manuals Search And Download.
-- DISCONTINUED PRODUCT --
Table of Contents
Schedule of Figures. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
Schedule of Tables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
Guide Contents . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
Conventions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
Typographical. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
Online Document . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
List of Acronyms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
About the Core. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
Recommended Design Experience. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
Additional Core Resources . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
Related Xilinx Ethernet Products and Services . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
Specifications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
Technical Support. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
Feedback. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
GEMAC Core . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
Document . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
System Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
Core Components . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
Core Interfaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
GMAC Core with Optional Management Interface . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
Client Side Interface. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
Physical Side Interface. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
Graphical User Interface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
Component Name . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
Management Interface. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
Address Filter . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
Number of Address Table Entries . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
Physical Interface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
Parameter Values in the XCO File . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
Output Generation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
1-Gigabit Ethernet MAC v8.5 User Guide
UG144 April 24, 2009
Download from Www.Somanuals.com. All Manuals Search And Download.
-- DISCONTINUED PRODUCT --
R
General Design Guidelines . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
Design Steps . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
Know the Degree of Difficulty . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
Keep it Registered . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
Recognize Timing Critical Signals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
Use Supported Design Flows . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
Make Only Allowed Modifications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
Receiving Inbound Frames . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
Normal Frame Reception . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
rx_good_frame, rx_bad_frame timing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
Frame Reception with Errors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
Client-Supplied FCS Passing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
VLAN Tagged Frames. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
Maximum Permitted Frame Length . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
Length/Type Field Error Checks. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
Address Filter . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
Receiver Statistics Vector . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
Transmitting Outbound Frames . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
Normal Frame Transmission . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
Padding . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
Client-Supplied FCS Passing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
Client Underrun . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
VLAN Tagged Frames. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
Maximum Permitted Frame Length . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
Inter-Frame Gap Adjustment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
Transmitter Statistics Vector. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
Overview of Flow Control. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
Flow Control Requirement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
Flow Control Basics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
Pause Control Frames . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
Flow Control Operation of the GEMAC . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
Transmitting a PAUSE Control Frame . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
Receiving a Pause Control Frame . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
Flow Control Implementation Example. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
Implementing External GMII. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
GMII Transmitter Logic . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
GMII Receiver Logic . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
Implementing External RGMII . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
RGMII Transmitter Logic . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
RGMII Receiver Logic . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70
RGMII Inband Status Decoding Logic . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75
Using the MDIO interface. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76
1-Gigabit Ethernet MAC v8.5 User Guide
UG144 April 24, 2009
Download from Www.Somanuals.com. All Manuals Search And Download.
-- DISCONTINUED PRODUCT --
R
Connecting the MDIO to an Internally Integrated PHY . . . . . . . . . . . . . . . . . . . . . . . . 76
Connecting the MDIO to an External PHY . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76
Using the Optional Management Interface. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77
Host Clock Frequency . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77
Configuration Registers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78
MDIO Interface. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86
Access without the Management Interface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90
Required Constraints. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
Device, Package, and Speedgrade Selection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
I/O Location Constraints . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
Placement Constraints. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
Timing Constraints . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
Constraints when Implementing an External GMII . . . . . . . . . . . . . . . . . . . . . . . . . . . . 96
Constraints when Implementing an External RGMII. . . . . . . . . . . . . . . . . . . . . . . . . . 101
Clocking the Core . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109
With Internal GMII . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109
With External GMII . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109
With RGMII. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 110
Multiple Cores . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 110
With External GMII . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 110
With RGMII. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111
Reset Conditions. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112
Ethernet 1000Base-X PCS/PMA or SGMII Core . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113
Integration to Provide 1000BASE-X PCS with TBI . . . . . . . . . . . . . . . . . . . . . . . . . . . . 114
Integration to Provide SGMII Functionality . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 119
Ethernet Statistics Core. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 119
1-Gigabit Ethernet MAC v8.5 User Guide
UG144 April 24, 2009
Download from Www.Somanuals.com. All Manuals Search And Download.
-- DISCONTINUED PRODUCT --
R
Pre-implementation Simulation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 123
Using the Simulation Model. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 123
Synthesis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 123
XST—VHDL . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 123
XST—Verilog . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 124
Implementation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 124
Generating the Xilinx Netlist . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 124
Mapping the Design . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 124
Placing-and-Routing the Design . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125
Static Timing Analysis. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125
Generating a Bitstream . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125
Post-Implementation Simulation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125
Generating a Simulation Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125
Using the Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 126
Other Implementation Information. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 126
Interfaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 128
Transmit FIFO. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 128
Receive FIFO . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 129
Overview of LocalLink Interface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 130
Data Flow . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 130
Functional Operation. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 131
Clock Requirements. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 131
Receive FIFO . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 131
Transmit FIFO. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 131
Expanding Maximum Frame Size . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 132
User Interface Data Width Conversion. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 132
Verification by Simulation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 133
Hardware Verification . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 133
DCM Phase-Shifting . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 135
Finding the Ideal Phase-Shift. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 135
Transmit Path Latency . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 137
Receive Path Latency . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 137
1-Gigabit Ethernet MAC v8.5 User Guide
UG144 April 24, 2009
Download from Www.Somanuals.com. All Manuals Search And Download.
-- DISCONTINUED PRODUCT --
Schedule of Figures
Figure 2-1: Block Diagram . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
and with Optional Address Filter. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
Optional Address Filter . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
Figure 3-1: 1-Gigabit Ethernet MAC Main Screen. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
Figure 4-1: 1-Gigabit Ethernet MAC Core Example Design . . . . . . . . . . . . . . . . . . . . . . . . 36
Figure 5-1: Normal Frame Reception . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
Figure 5-2: Frame Reception with Error . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
Figure 5-3: Frame Reception with In-Band FCS Field. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
Figure 5-4: Reception of a VLAN Tagged Frame . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
Figure 5-5: Receiver Statistics Vector Timing. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
Figure 5-6: Normal Frame Transmission . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
Figure 5-7: Frame Transmission with Client-supplied FCS. . . . . . . . . . . . . . . . . . . . . . . . . 48
Figure 5-8: Frame Transmission with Underrun . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
Figure 5-9: Transmission of a VLAN Tagged Frame. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
Figure 5-10: Inter-Frame Gap Adjustment. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
Figure 5-11: Transmitter Statistic Vector Timing. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
Figure 6-1: Requirement for Flow Control . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
Figure 6-2: MAC Control Frame Format . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
Figure 6-3: Pause Request Timing. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
1-Gigabit Ethernet MAC v8.5 User Guide
9
UG144 April 24, 2009
Download from Www.Somanuals.com. All Manuals Search And Download.
-- DISCONTINUED PRODUCT --
R
Figure 7-1: External GMII Transmitter Logic . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
Spartan-3A Devices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
Figure 7-3: External GMII Receiver Logic for Virtex-5 Devices . . . . . . . . . . . . . . . . . . . . . 65
Figure 7-4: External RGMII Transmitter Logic. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
Figure 7-5: External RGMII Transmitter Logic in Virtex-4 Devices. . . . . . . . . . . . . . . . . . 68
Figure 7-6: External RGMII Transmitter Logic in Virtex-5 Devices. . . . . . . . . . . . . . . . . . 69
Figure 7-7: External RGMII Receiver Logic . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
Figure 7-8: External RGMII Receiver Logic for Virtex-4 Devices. . . . . . . . . . . . . . . . . . . . 73
Figure 7-9: External RGMII Receiver Logic for Virtex-5 Devices. . . . . . . . . . . . . . . . . . . . 74
Figure 7-10: RGMII Inband Status Decoding Logic . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75
Figure 7-11: Creating an External MDIO Interface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76
Figure 8-1: Configuration Register Write Timing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84
Figure 8-2: Configuration Register Read Timing. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84
Figure 8-3: Address Table Write Timing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85
Figure 8-4: Address Table Read Timing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86
Figure 8-5: Typical MDIO-managed System . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87
Figure 8-6: MDIO Write Transaction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87
Figure 8-7: MDIO Read Transaction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88
Figure 8-8: MDIO Access through Management Interface . . . . . . . . . . . . . . . . . . . . . . . . . 89
Figure 9-1: Input GMII Timing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97
Figure 9-2: Timing Report Setup/Hold Illustration. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101
Figure 9-3: Input RGMII Timing. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 102
Figure 9-4: Timing Report Setup/Hold Illustration. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107
Figure 10-1: Clock Management Logic with External GMII . . . . . . . . . . . . . . . . . . . . . . . 109
Figure 10-2: Clock Management with External RGMII . . . . . . . . . . . . . . . . . . . . . . . . . . . 110
Figure 10-5: Reset Circuit for a Single Clock/reset Domain. . . . . . . . . . . . . . . . . . . . . . . . 112
10
1-Gigabit Ethernet MAC v8.5 User Guide
UG144 April 24, 2009
Download from Www.Somanuals.com. All Manuals Search And Download.
-- DISCONTINUED PRODUCT --
R
using a RocketIO Transceiver . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 115
using a RocketIO transceiver. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 117
using the RocketIO transceiver. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 118
Figure A-2: Frame Transfer across LocalLink Interface . . . . . . . . . . . . . . . . . . . . . . . . . . . 130
Figure A-3: Frame Transfer with Flow Control . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 130
1-Gigabit Ethernet MAC v8.5 User Guide
11
UG144 April 24, 2009
Download from Www.Somanuals.com. All Manuals Search And Download.
-- DISCONTINUED PRODUCT --
R
12
1-Gigabit Ethernet MAC v8.5 User Guide
UG144 April 24, 2009
Download from Www.Somanuals.com. All Manuals Search And Download.
-- DISCONTINUED PRODUCT --
Schedule of Tables
Table 2-1: Transmitter Client Interface Signal Pins. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
Table 2-2: Receive Client Interface Signal Pins . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
Table 2-3: Flow Control Interface Signal Pinout . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
Table 2-4: Optional Management Interface Signal Pinout. . . . . . . . . . . . . . . . . . . . . . . . . . 28
Table 2-5: Optional MAC Unicast Address Signal Pinout . . . . . . . . . . . . . . . . . . . . . . . . . . 28
Table 2-6: Optional Configuration Vector Signal Pinout . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
Table 2-7: Reset Signal. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
Table 2-8: GMII Interface Signal Pinout . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
Table 2-9: MDIO Interface Signal Pinout . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
Table 3-1: XCO File Values and Default Values. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
Table 4-1: Degree of Difficulty for Various Implementations . . . . . . . . . . . . . . . . . . . . . . 38
Table 5-1: Abbreviations Used in Timing Diagrams. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
Table 5-2: Bit Definition for the Receiver Statistics Vector . . . . . . . . . . . . . . . . . . . . . . . . . 45
Table 5-4: Bit Definition for the Transmitter Statistics Vector . . . . . . . . . . . . . . . . . . . . . . 51
Table 8-1: Management Interface Transaction Types . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77
Table 8-2: Configuration Registers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78
Table 8-3: Receiver Configuration Word 0. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79
Table 8-4: Receiver Configuration Word 1. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79
Table 8-5: Transmitter Configuration Word . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
Table 8-6: Flow Control Configuration Word . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81
Table 8-7: Management Configuration Word . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82
1-Gigabit Ethernet MAC v8.5 User Guide
13
UG144 April 24, 2009
Download from Www.Somanuals.com. All Manuals Search And Download.
-- DISCONTINUED PRODUCT --
R
Table 8-8: Unicast Address Word 0 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82
Table 8-9: Unicast Address Word 1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82
Table 8-10: Address Table Configuration Word 0 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83
Table 8-11: Address Table Configuration Word 1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83
Table 8-12: Address Filter Mode . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83
Table 8-13: Configuration Vector Bit Definition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90
Table 9-1: Input GMII Timing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97
Table 9-2: Input RGMII Timing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 102
Table 11-1: Management Interface Transaction Types . . . . . . . . . . . . . . . . . . . . . . . . . . . . 121
Table A-1: Transmit FIFO Client Interface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 128
Table A-2: Transmit FIFO LocalLink Interface. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 128
Table A-3: Receive FIFO Client Interface. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 129
Table A-4: Receive FIFO LocalLink Interface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 129
14
1-Gigabit Ethernet MAC v8.5 User Guide
UG144 April 24, 2009
Download from Www.Somanuals.com. All Manuals Search And Download.
-- DISCONTINUED PRODUCT --
R
Preface
About This Guide
The LogiCORE™ IP 1-Gigabit Ethernet MAC User Guide provides information about
generating the core, customizing and simulating the core utilizing the provided example
design, and running the design files through implementation using the Xilinx tools.
Guide Contents
This guide contains the following chapters:
•
•
Preface, “About this Guide” introduces the organization and purpose of the guide
and the conventions used in this document.
Chapter 1, “Introduction” describes the core and related information, including
recommended design experience, additional resources, technical support, and
submitting feedback to Xilinx.
•
•
•
Chapter 2, “Core Architecture” provides an overview of the core and discusses the
Physical/Client signal interfaces.
Chapter 3, “Generating the Core” describes the graphical user interface options used
to generate the core.
describe design parameters, including how to initialize the core, generate and
consume core packets, and how to operate the Management Interface.
•
•
Chapter 9, “Constraining the Core” describes the constraints associated with the core.
Chapter 10, “Clocking and Resetting” discusses special design considerations
associated with clock management logic, including the Gigabit Media Independent
Interface (GMII) and Reduced Gigabit Media Independent Interface (RGMII) options.
•
•
Chapter 11, “Interfacing to Other Cores” describes how to interface the 1-Gigabit
Ethernet MAC core to the Ethernet 1000BASE-X PCS/PMA or SGMII core and the
Ethernet Statistics core.
Chapter 12, “Implementing Your Design” provides instructions for how to set up
synthesis, simulation, and implementation environments and how to generate a
bitstream through the design flow.
•
•
•
•
Appendix A, “Using the Client-Side FIFO” describes the FIFO provided in the
example design that accompanies the GEMAC core.
core was verified and certified for compliance.
Appendix C, “Calculating DCM Phase-Shifting” provides information about how to
calculate the system timing requirements when using DCMs with the core.
Appendix D, “Core Latency” describes the latency of the core.
1-Gigabit Ethernet MAC v8.5 User Guide
15
UG144 April 24, 2009
Download from Www.Somanuals.com. All Manuals Search And Download.
-- DISCONTINUED PRODUCT --
R
Preface: About This Guide
Conventions
This document uses the following conventions. An example illustrates each convention.
Typographical
The following typographical conventions are used in this document:
Convention
Meaning or Use
Example
Messages, prompts, and
program files that the system speed grade: - 100
Courier font
displays
Literal commands you enter in
ngdbuild design_name
a syntactical statement
Courier bold
Variables in a syntax
See the Development System
statement for which you must Reference Guide for more
supply values
information.
Italic font
References to other manuals
See the User Guide for details.
If a wire is drawn so that it
overlaps the pin of a symbol,
the two nets are not connected.
Emphasis in text
Items that are not supported
or reserved
Dark Shading
This feature is not supported
An optional entry or
parameter. However, in bus
specifications, such as
ngdbuild [option_name]
design_name
Square brackets [ ]
Braces { }
bus[7:0], they are required.
A list of items from which you
must choose one or more
lowpwr ={on|off}
lowpwr ={on|off}
Separates items in a list of
choices
Vertical bar
|
IOB #1: Name = QOUT’
IOB #2: Name = CLKIN’
.
.
.
Vertical ellipsis
.
.
.
Repetitive material that has
been omitted
allowblock block_name
loc1 loc2 ... locn;
Horizontal ellipsis . . . Omitted repetitive material
The prefix ‘0x’ or the suffix ‘h’
A read of address
0x00112975 returned
45524943h.
indicate hexadecimal notation
Notations
An ‘_n’ means the signal is
active low
usr_teof_nis active low.
16
1-Gigabit Ethernet MAC v8.5 User Guide
UG144 April 24, 2009
Download from Www.Somanuals.com. All Manuals Search And Download.
-- DISCONTINUED PRODUCT --
R
Conventions
Online Document
The following linking conventions are used in this document:
Convention
Meaning or Use
Example
See the section “Additional
Resources” for details.
Cross-reference link to a
location in the current
document
Blue text
See “Title Formats” in
Chapter 1 for details.
Go to www.xilinx.com for the
latest speed files.
Blue, underlined text
Hyperlink to a website (URL)
List of Acronyms
The following table describes acronyms used in this manual.
Acronym
CLB
Spelled Out
Configurable Logic Block
Digital Clock Manager
Double Data Rate
DCM
DDR
FCS
Frame Check Sequence
Field Programmable Gate Array.
Gigabit Interface Converter
Gigabit per second
FPGA
GBIC
Gbps
GEMAC
GMII
HDL
IO
Gigabit Ethernet Media Access Controller
Gigabit Media Independent Interface
Hardware Description Language
Input/Output
IOB
Input/Output Block
IP
Intellectual Property
ISE®
LSW
MAC
MDIO
MHz
MMD
ms
Integrated Software Environment
Least Significant Word
Media Access Controller
Management Data Input/Output
Mega Hertz
MDIO Managed Device
milliseconds
MSW
Most Significant Word
1-Gigabit Ethernet MAC v8.5 User Guide
17
UG144 April 24, 2009
Download from Www.Somanuals.com. All Manuals Search And Download.
-- DISCONTINUED PRODUCT --
R
Preface: About This Guide
Acronym
NCD
NGC
NGD
ns
Spelled Out
Native Circuit Description
Native Generic Circuit
Native Generic Database
nanoseconds
PCB
Printed Circuit Board
Physical Coding Sublayer
physical-side interface
Physical Medium Attachment
Physical Medium Dependent
PCS
PHY
PMA
PMD
RGMII
SGMII
VHDL
Reduced Gigabit Media Independent Interface
Serial Gigabit Media Independent Interface
VHSIC Hardware Description Language
(VHSIC an acronym for Very High-Speed
Integrated Circuits).
VCS
Verilog Compiled Simulator
VLAN
Virtual LAN (Local Area Network)
18
1-Gigabit Ethernet MAC v8.5 User Guide
UG144 April 24, 2009
Download from Www.Somanuals.com. All Manuals Search And Download.
-- DISCONTINUED PRODUCT --
R
Chapter 1
Introduction
The 1-Gigabit Ethernet MAC (GEMAC) core is a fully verified solution that supports
Verilog-HDL and VHDL. In addition, the example design provided with the core is
provided in both Verilog and VHDL.
This chapter introduces the GEMAC core and provides other related information,
including recommended design experience, additional resources, technical support, and
ways to submit feedback to Xilinx.
About the Core
The GEMAC core is a Xilinx CORE Generator™ IP core, included in the latest IP Update on
the Xilinx IP Center. For detailed information about the core, see the GEMAC product
page. For information about licensing options, see Chapter 2, “Licensing the Core,” in the
1-Gigabit Ethernet MAC Getting Started Guide.
Recommended Design Experience
Although the GEMAC core is a fully verified solution, the challenge associated with
implementing a complete design varies, depending on the configuration and functionality
of the application. For best results, previous experience building high performance,
pipelined FPGA designs using Xilinx implementation software and user constraint files
(UCFs) is recommended.
Contact your local Xilinx representative for a closer review and estimation for your specific
requirements.
Additional Core Resources
For detailed information and updates about the GEMAC core, see the following
•
•
1-Gigabit Ethernet MAC Data Sheet
1-Gigabit Ethernet MAC Getting Started Guide
After generating the core, the 1-Gigabit Ethernet MAC Release Notes are available from the
document directory.
Related Xilinx Ethernet Products and Services
1-Gigabit Ethernet MAC v8.5 User Guide
19
UG144 April 24, 2009
Download from Www.Somanuals.com. All Manuals Search And Download.
-- DISCONTINUED PRODUCT --
R
Chapter 1: Introduction
Specifications
•
IEEE 802.3 2005
Reduced Gigabit Media Independent Interface (RGMII) version 2.0
•
Technical Support
with expertise using the GEMAC core.
Xilinx will provide technical support for use of this product as described in the 1-Gigabit
Ethernet MAC User Guide and the 1-Gigabit Ethernet MAC Getting Started Guide. Xilinx
cannot guarantee timing, functionality, or support of this product for designs that do not
follow these guidelines.
Feedback
Xilinx welcomes comments and suggestions about the GEMAC core and the
documentation supplied with the core.
GEMAC Core
For comments or suggestions about the GEMAC core, please submit a WebCase from
support.xilinx.com/. Be sure to include the following information:
•
•
•
Product name
Core version number
Explanation of your comments
Document
For comments or suggestions about this document, please submit a WebCase from
support.xilinx.com/. Be sure to include the following information:
•
•
•
•
Document title
Document number
Page number(s) to which your comments refer
Explanation of your comments
20
1-Gigabit Ethernet MAC v8.5 User Guide
UG144 April 24, 2009
Download from Www.Somanuals.com. All Manuals Search And Download.
-- DISCONTINUED PRODUCT --
R
Chapter 2
Core Architecture
This chapter describes the GEMAC core architecture, including the major functional blocks
and all interfaces.
System Overview
Figure 2-1 illustrates a block diagram of the GEMAC core with all the major functional
blocks and interfaces. Descriptions of the functional blocks and interfaces are provided in
the sections that follow.
Gigabit Ethernet MAC Core
Client
Transmitter
Interface
Transmit Engine
Flow Control
To Physical
Sublayers
Client
Receive Engine
Receiver
Interface
Optional
Address
Filter
Optional Management
Client
Management
Interface
Configuration
MDIO
Figure 2-1: Block Diagram
1-Gigabit Ethernet MAC v8.5 User Guide
21
UG144 April 24, 2009
Download from Www.Somanuals.com. All Manuals Search And Download.
-- DISCONTINUED PRODUCT --
R
Chapter 2: Core Architecture
Core Components
Transmit Engine
The Transmit Engine accepts Ethernet frame data from the Client Transmitter Interface,
adds the preamble field to the start of the frame, adds padding bytes (if required) to ensure
that the frame meets the minimum frame length requirements, and adds the frame check
sequence (when configured to do so). The transmitter also ensures that the inter-frame
spacing between successive frames is at least the minimum specified. The frame is then
converted into a format that is compatible with the GMII and sent to the GMII Block.
Receive Engine
The Receive Engine accepts Ethernet frame data from the GMII Block, removes the
preamble field at the start of the frame, removes padding bytes and Frame Check Sequence
(if required, and when configured to do so). The receiver also performs error detection on
the received frame using information such as the frame check sequence field, received
GMII error codes, and legal frame size boundaries.
Flow Control
The Flow Control block is designed to clause 31 of the IEEE 802.3-2005 standard. The MAC
may be configured to send pause frames and to act upon their reception. These two
behaviors can be configured independently.
Address Filter
The Address Filter checks the address of incoming frames into the receiver. If the Address
Filter is enabled, the device will not pass frames that do not contain one of a set of known
addresses to the client.
Management Interface
The optional processor-independent Management Interface has standard address, data,
and control signals. It may be used as is, or you can apply a logical shim to interface to
common bus architectures. See Chapter 8, “Configuration and Status.”
This interface is used to access the following blocks.
•
Configuration Register After power up or reset, the client may reconfigure the core
parameters from their defaults. Configuration changes can be written at any time.
•
MDIO Interface The Management Interface is also used to access the MDIO interface
of the GEMAC core; this interface is typically connected to the MDIO port of a
physical layer device (PHY) to access its configuration and status registers. The MDIO
format is defined in IEEE802.3 clause 22.
GMII Block
This implements GMII style signaling for the physical interface of the core and is typically
attached to a physical layer device (PHY), either off-chip or internally integrated.
22
1-Gigabit Ethernet MAC v8.5 User Guide
UG144 April 24, 2009
Download from Www.Somanuals.com. All Manuals Search And Download.
-- DISCONTINUED PRODUCT --
R
Core Interfaces
Core Interfaces
GMAC Core with Optional Management Interface
Figure 2-2 shows the pinout for the GEMAC core using the optional Management
Interface. The interface is unchanged, regardless of whether the optional Address Filter is
included.
Client Side Interface
gtx_clk domain
Physical Side Interface (GMII)
gtx_clk
tx_data[7:0]
tx_data_valid
gmii_txd[7:0]
gmii_tx_en
gmii_tx_er
tx_ifg_delay[7:0]
tx_ack
tx_underrun
tx_statistics_vector[31:0]
tx_statistics_valid
pause_req
pause_val[15:0]
gmii_rx_clk domain
gmii_rx_clk
rx_data[7:0]
rx_data_valid
gmii_rxd[7:0]
gmii_rx_dv
gmii_rx_er
rx_good_frame
rx_bad_frame
rx_statistics_vector[27:0]
rx_statistics_valid
host_clk domain
host_clk
host_opcode[1:0]
host_addr[9:0]
host_wr_data[31:0]
host_rd_data[31:0]
host_miim_sel
host_req
mdc
mdio_in
mdio_out
mdio_tri
host_miim_rdy
reset
Figure 2-2: Component Pinout for MAC with Optional Management Interface
1-Gigabit Ethernet MAC v8.5 User Guide
23
UG144 April 24, 2009
Download from Www.Somanuals.com. All Manuals Search And Download.
-- DISCONTINUED PRODUCT --
R
Chapter 2: Core Architecture
GMAC Core Without Management Interface and With Address Filter
Figure 2-3 shows the pinout for the GEMAC core when the optional Management Interface
is omitted and the optional Address Filter is included in the core.
The configuration_vector[67:0]input provides the method for configuration of
the core, and mac_unicast_address[47:0]input provides the method of setting the
unicast address used by the Address Filter.
Client Side Interface
Physical Side Interface (GMII)
gtx_clk domain
gtx_clk
gmii_txd[7:0]
gmii_tx_en
gmii_tx_er
tx_data[7:0]
tx_data_valid
tx_ifg_delay[7:0]
tx_ack
tx_underrun
tx_statistics_vector[31:0]
tx_statistics_valid
pause_req
pause_val[15:0]
gmii_rx_clk domain
gmii_rx_clk
rx_data[7:0]
rx_data_valid
rx_good_frame
gmii_rxd[7:0]
gmii_rx_dv
gmii_rx_er
rx_bad_frame
rx_statistics_vector[27:0]
rx_statistics_valid
mac_unicast_address[47:0]
configuration_vector[67:0]
reset
Figure 2-3: Component Pinout for MAC without Optional Management Interface
and with Optional Address Filter
24
1-Gigabit Ethernet MAC v8.5 User Guide
UG144 April 24, 2009
Download from Www.Somanuals.com. All Manuals Search And Download.
-- DISCONTINUED PRODUCT --
R
Core Interfaces
GEMAC Core Without Management Interface and Without Address Filter
Figure 2-4 shows the pinout for the GEMAC core when the optional Management Interface
is omitted and the optional Address Filter is omitted.
The configuration_vector[67:0]input provides the method for configuration of
the core.
Client Side Interface
Physical Side Interface (GMII)
gtx_clk domain
gtx_clk
gmii_txd[7:0]
gmii_tx_en
gmii_tx_er
tx_data[7:0]
tx_data_valid
tx_ifg_delay[7:0]
tx_ack
tx_underrun
tx_statistics_vector[31:0]
tx_statistics_valid
pause_req
pause_val[15:0]
gmii_rx_clk domain
gmii_rx_clk
rx_data[7:0]
rx_data_valid
rx_good_frame
gmii_rxd[7:0]
gmii_rx_dv
gmii_rx_er
rx_bad_frame
rx_statistics_vector[27:0]
rx_statistics_valid
configuration_vector[67:0]
reset
Figure 2-4: Component Pinout for MAC without Optional Management Interface or Optional Address Filter
1-Gigabit Ethernet MAC v8.5 User Guide
25
UG144 April 24, 2009
Download from Www.Somanuals.com. All Manuals Search And Download.
-- DISCONTINUED PRODUCT --
R
Chapter 2: Core Architecture
All ports of the core are internal connections in FPGA fabric. An HDL example design is
delivered with the core that will add IBUFs, OBUFs, and IOB flip-flops to the external
signals of the Gigabit Media Independent Interface (GMII) or Reduced Gigabit Media
Independent Interface (RGMII).
All clock management logic is placed in this example design, which allows for more
flexibility in implementation (for example, in designs using multiple cores). This example
design is provided in both VHDL and Verilog. For more information about example
designs, see the 1-Gigabit Ethernet MAC Getting Started Guide.
Client Side Interface
Transmitter Interface
Table 2-1 describes the client-side transmitter signals of the GEMAC core. These signals are
The Transmitter Interface is designed to be connected to internal device logic only.
Attempting to add external ports to this interface will result in a breakdown of the
handshaking protocol used by this interface.
Table 2-1: Transmitter Client Interface Signal Pins
Clock
Domain
Signal
Direction
Description
gtx_clk
Input
n/a
Clock signal provided to the core at
125 MHz. Tolerance must be
within IEEE 802.3-2005
specification. This clock signal is
used by all of the transmitter logic.
tx_data[7:0]
Input
gtx_clk
Frame data to be transmitted is
supplied on this port.
tx_data_valid
Input
Input
gtx_clk
gtx_clk
Control signal for tx_data port.
tx_ifg_delay[7:0]
Control signal for configurable
Inter Frame Gap adjustment.
tx_ack
Output
gtx_clk
Handshaking signalassertedwhen
the current data on tx_data has
been accepted.
tx_underrun
Input
Output
Output
gtx_clk
gtx_clk
gtx_clk
Asserted by client to force GEMAC
core to corrupt the current frame.
tx_statistics_vector[31:0]
tx_statistics_valid
Provides statistical information
about the last frame transmitted.
Asserted at end of frame
transmission, indicating that the
tx_statistics_vector is valid.
26
1-Gigabit Ethernet MAC v8.5 User Guide
UG144 April 24, 2009
Download from Www.Somanuals.com. All Manuals Search And Download.
-- DISCONTINUED PRODUCT --
R
Core Interfaces
Receiver Interface
Table 2-2 describes the client-side receiver signals of the GEMAC core. These signals are
Table 2-2: Receive Client Interface Signal Pins
Signal
rx_data[7:0]
Direction Clock Domain
Description
Output
Output
Output
gmii_rx_clk
gmii_rx_clk
gmii_rx_clk
Frame data received is
supplied on this port.
rx_data_valid
Control signal for the rx_data
port.
rx_good_frame
Asserted at end of frame
reception to indicate that the
frame should be processed by
the MAC client.
rx_bad_frame
Output
gmii_rx_clk
Asserted at end of frame
reception to indicate that the
frame should be discarded by
the MAC client.
rx_statistics_vector[27:0]
rx_statistics_valid
Output
Output
gmii_rx_clk
gmii_rx_clk
Provides statistical
information about the last
frame received.
Asserted at end of frame
reception, indicating that the
rx_statistics_vector is valid.
Flow Control Interface
Table 2-3 describes the signals used by the client to request a flow control action from the
Table 2-3: Flow Control Interface Signal Pinout
Signal
pause_req
Direction Clock Domain
Description
Input
gtx_clk
Pause request. sends a pause
frame down the link.
pause_val[15:0]
Input
gtx_clk
Pause value; inserted into the
parameter field of the
transmitted pause frame.
1-Gigabit Ethernet MAC v8.5 User Guide
27
UG144 April 24, 2009
Download from Www.Somanuals.com. All Manuals Search And Download.
-- DISCONTINUED PRODUCT --
R
Chapter 2: Core Architecture
Management Interface (Optional)
Table 2-4 describes the optional signals used by the client to access the management
Table 2-4: Optional Management Interface Signal Pinout
Clock
Domain
Signal
Direction
Description
host_clk
Input
n/a
Clock for the Management
Interface; must be 10 MHz or
above.
host_opcode[1:0]
Input
host_clk
Defines operation to be performed
over MDIO interface. Bit 1 is also
used as a read/write control
signal for configuration register
access.
host_addr[9:0]
Input
Input
host_clk
host_clk
host_clk
host_clk
Address of register to be accessed.
Data to write to register .
host_wr_data[31:0]
host_rd_data[31:0]
host_miim_sel
Output
Input
Data read from register.
When asserted, the MDIO
interface is accessed. When not
asserted, the configuration
registers are accessed.
host_req
Input
host_clk
host_clk
Used to signal a transaction on the
MDIO interface.
host_miim_rdy
Output
When high, the MDIO interface
has completed any pending
transaction and is ready for a new
transaction.
MAC Unicast Address (Optional)
Table 2-5 describes the alternative method of access to the unicast address registers when
the optional Management Interface is not present.
Table 2-5: Optional MAC Unicast Address Signal Pinout
Signal
Direction
Description
mac_unicast_address[47:0]
Input
Used to assess the MAC unicast
address registers when the
Management Interface is not used
Note: All bits are registered on input but may be treated as asynchronous inputs.
28
1-Gigabit Ethernet MAC v8.5 User Guide
UG144 April 24, 2009
Download from Www.Somanuals.com. All Manuals Search And Download.
-- DISCONTINUED PRODUCT --
R
Core Interfaces
Configuration Vector (Optional)
Table 2-6 describes the alternative to the optional Management Interface signals. The
Configuration Vector uses direct inputs to the core to replace the functionality of the MAC
Table 2-6: Optional Configuration Vector Signal Pinout
Signal
Direction
Description
configuration_vector[67:0]
Input
Used to replace the functionality of
the MAC Configuration Registers
when the Management Interface is
not used
Note: All bits are registered on input but may be treated as asynchronous inputs.
Asynchronous Reset
Table 2-7 describes the asynchronous reset signal for the entire core.
Table 2-7: Reset Signal
Signal
reset
Direction
Clock Domain
Description
Input
n/a
Asynchronous reset for entire core
Physical Side Interface
GMII
Table 2-8: GMII Interface Signal Pinout
Signal
gmii_txd[7:0]
gmii_tx_en
gmii_tx_er
gmii_rx_clk
Direction
Output
Output
Output
Input
Clock Domain
gtx_clk
Description
Transmit data from MAC
gtx_clk
Transmit control signal from MAC
Transmit control signal from MAC
gtx_clk
n/a
Receive clock from external PHY (125
MHz)
gmii_rxd[7:0]
gmii_rx_dv
gmii_rx_er
Input
Input
Input
gmii_rx_clk
gmii_rx_clk
gmii_rx_clk
Received data to MAC
Received control signal to MAC
Received control signal to MAC
1-Gigabit Ethernet MAC v8.5 User Guide
29
UG144 April 24, 2009
Download from Www.Somanuals.com. All Manuals Search And Download.
-- DISCONTINUED PRODUCT --
R
Chapter 2: Core Architecture
MDIO Interface
76.
Table 2-9: MDIO Interface Signal Pinout
Clock
Domain
Signal
mdc
mdio_in
Direction
Description
Output
host_clk
Management Clock: programmable
frequency derived from host_clk.
1
Input
host_clk
Input data signal for communication with
PHY configuration and status. Tie high if
unused.
mdio_out
Output
Output
host_clk
host_clk
Output data signal for communication
with PHY configuration and status.
mdio_tri
Tristate control for MDIO signals; 0 signals
that the value on mdio_out should be
asserted onto the MDIO bus.
1. mdio_in, mdio_outand mdio_trican be connected to a Tri-state buffer to create a bi-directional mdio
signal suitable for connection to an external PHY.
30
1-Gigabit Ethernet MAC v8.5 User Guide
UG144 April 24, 2009
Download from Www.Somanuals.com. All Manuals Search And Download.
-- DISCONTINUED PRODUCT --
R
Chapter 3
Generating the Core
The GEMAC core is generated through the Xilinx CORE Generator™ using a graphical
user interface (GUI). This chapter describes the GUI options used to generate and
customize the core.
Graphical User Interface
Figure 3-1 shows the main GEMAC core user GUI screen.
Figure 3-1: 1-Gigabit Ethernet MAC Main Screen
For general help starting and using CORE Generator on your system, see the
documentation supplied with the Xilinx ISE® software, including the CORE Generator
1-Gigabit Ethernet MAC v8.5 User Guide
31
UG144 April 24, 2009
Download from Www.Somanuals.com. All Manuals Search And Download.
-- DISCONTINUED PRODUCT --
R
Chapter 3: Generating the Core
Component Name
The component name is used as the base name of the output files generated for the core.
Names must begin with a letter and must be composed from the following characters:
a through z, 0 through 9 and “_”.
Management Interface
Select this option to include the optional Management Interface (see “Using the Optional
Management Interface,” on page 77). If this option is not selected, the core is generated
with a replacement configuration vector (see “Access without the Management Interface,”
on page 90). The default is to use the Management Interface.
Address Filter
Select this option to include the optional Address Filter. This prevents the reception of
is to use the Address Filter.
Number of Address Table Entries
The Address Filter can be instantiated with an address table that holds up to 4 additional
valid addresses. You may select an integer between 0 and 4 to define the number of
addresses that are present in the table.
This option is only available when the Management Interface and Address Filter have been
selected. The default is to use 4 address table entries.
Physical Interface
Depending on the target Xilinx FPGA architecture, it may be possible to select from two
different physical interface choices for the core:
•
•
The choice of physical interface determines the content of the example design delivered
with the core. The external GMII or RGMII is added in the HDL top-level design file. There
is no change in the core netlist for this option. The default is the GMII physical interface.
Parameter Values in the XCO File
XCO file parameter names and their values are identical to the names and values shown in
the GUI, except that underscore characters (_) are used instead of spaces. The text in an
XCO file is not case-sensitive.
Table 3-1 defines the XCO file parameters and values and summarizes the GUI defaults.
The following is an example of the CSET parameters in an XCO file.
CSET component_name = abc123
CSET physical_interface = gmii
CSET management_interface = true
CSET address_filter = true
CSET no_of_address_table_entries = 4
32
1-Gigabit Ethernet MAC v8.5 User Guide
UG144 April 24, 2009
Download from Www.Somanuals.com. All Manuals Search And Download.
-- DISCONTINUED PRODUCT --
R
Output Generation
Table 3-1: XCO File Values and Default Values
Parameter
XCO File Values
Default GUI Setting
component_name
ASCII text starting with a letter and
based upon the following character
set: a..z, 0..9 and _
gig_eth_mac_v8_5
physical_interface
One of the following keywords: gmii, gmii
rgmii
management_interface One of the following keywords: true,
false
true
true
4
address_filter
One of the following keywords: true,
false
no_of_address_table_
entries
Integer in the range 0 - 4
Output Generation
The output files generated from the CORE Generator tool are placed in the CORE
Generator project directory. The list of output files includes the following items.
•
•
•
•
•
Netlist file for the core
Supporting CORE Generator files
Release notes and documentation
Subdirectories containing an HDL example design
Scripts to run the core through the back-end tools and to simulate the core using either
Mentor Graphics ModelSim, Cadence IUS or Synopsys VCS simulators.
See the 1-Gigabit Ethernet MAC Getting Started Guide for more information about the CORE
Generator output files and for details on the HDL example design.
1-Gigabit Ethernet MAC v8.5 User Guide
33
UG144 April 24, 2009
Download from Www.Somanuals.com. All Manuals Search And Download.
-- DISCONTINUED PRODUCT --
R
Chapter 3: Generating the Core
34
1-Gigabit Ethernet MAC v8.5 User Guide
UG144 April 24, 2009
Download from Www.Somanuals.com. All Manuals Search And Download.
-- DISCONTINUED PRODUCT --
R
Chapter 4
Designing with the Core
This chapter provides general guidelines for creating designs using the GEMAC core. To
work with the example design included with the GEMAC core, see the 1-Gigabit Ethernet
MAC Getting Started Guide.
General Design Guidelines
This section describes the steps required to turn a GEMAC core into a fully functioning
design integrated with user-application logic. Not all implementations require all the
design steps described in this chapter. The following sections discuss the design steps
required for various implementations. For best results, carefully follow the logic design
guidelines.
Design Steps
Generate the core from the Xilinx CORE Generator™. See Chapter 3, “Generating the
Using the Example Design as a Starting Point
The GEMAC core is delivered through the CORE Generator with an HDL example design
built around the core, allowing the functionality of the core to be demonstrated using
block diagram of the example design. For more information about the example design, see
the 1-Gigabit Ethernet MAC Getting Started Guide.
The example design illustrates how to:
•
•
•
Instantiate the core from HDL.
Source and use the client-side interface ports of the core from application logic.
Connect the physical-side interface of the core (GMII or RGMII) to device IOBs
•
Derive the clock management logic, as described in Chapter 10, “Clocking and
1-Gigabit Ethernet MAC v8.5 User Guide
35
UG144 April 24, 2009
Download from Www.Somanuals.com. All Manuals Search And Download.
-- DISCONTINUED PRODUCT --
R
Chapter 4: Designing with the Core
Statistics Vectors
Interface
<component_name>_example_design
<component_name>_locallink
<component_name>_block
Clock/
Reset
Circuitry
1-Gigabit Ethernet
MAC Core
Physical
Interface
10 Mbps, 100 Mbps
1 Gbps Ethernet FIFO
Client
Interface
Tx Client
FIFO
GMII/ RGMII
Interface
Logic,
IOBs and
Clock
Management
Rx Client
FIFO
Address
Swap
Module
Management
Interface
Figure 4-1: 1-Gigabit Ethernet MAC Core Example Design
Using the example design as a starting point, you can do the following:
•
Edit the HDL top level of the example design file to:
♦
♦
♦
♦
Change the clocking scheme.
Add/remove IOBs as required.
Replace the client loopback logic with your specific application logic.
Adapt the 10 Mbps, 100 Mbps, 1 Gbps Ethernet FIFO to suit your specific
•
•
Synthesize the entire design.
The Xilinx Synthesis Tool (XST) script and Project file in the /implementdirectory
may be adapted to include any HDL files you may want to add.
Run the implement script in the /implementdirectory to create a top-level netlist for
the design. The script may also run the Xilinx tools map, par, and bitgen, creating a
bitstream that can be downloaded to a Xilinx device. SimPrim-based simulation
models for the entire design are also produced by the implement scripts.
•
•
Simulate the entire design using the demonstration test bench provided as a template
in the /simulationdirectory.
Download the bitstream to a target device.
36
1-Gigabit Ethernet MAC v8.5 User Guide
UG144 April 24, 2009
Download from Www.Somanuals.com. All Manuals Search And Download.
-- DISCONTINUED PRODUCT --
R
General Design Guidelines
Implementing the 1-Gigabit Ethernet MAC in Your Application
The example design can be studied as an example of how to do the following:
•
•
•
Instantiate the core from HDL.
Source and use the client-side interface ports of the core from application logic.
Connect the physical-side interface of the core (GMII or RGMII) to device IOBs to
create an external interface.
•
Derive the clock management logic.
After working with the example design, you can write your own HDL application, using
single or multiple instances of the GEMAC core. Client-side interfaces and operation of the
core are detailed later in this chapter. For more information, see:
•
•
Clock Management Logic in Chapter 10, “Clocking and Resetting.”
Using the GEMAC core in conjunction with the Ethernet 1000BASE-X PCS/PMA or
SGMII core in Chapter 11, “Interfacing to Other Cores.”
•
•
Using the GEMAC core in conjunction with the Ethernet Statistics core in Chapter 11,
10 Mbps, 100 Mbps, 1 Gbps Ethernet FIFO in Appendix A, “Using the Client-Side
You can synthesize the entire design using any synthesis tool. The GEMAC core is pre-
synthesized and is delivered as an NGC netlist (which appears as a black box to synthesis
tools).
Run the Xilinx tools map, par, and bitgento create a bitstream that can be downloaded to
a Xilinx device. Care must be taken to constrain the design correctly, and the UCF
produced by the CORE Generator should be used as the basis for the your own UCF. See
You can simulate the entire design and download the bitstream to the target device.
Know the Degree of Difficulty
A 1-Gigabit Ethernet MAC implementation is challenging to implement in any technology,
and all applications require careful attention to system performance requirements.
Pipelining, logic mapping, placement constraints, and logic duplication are all methods
that help boost system performance.
Virtex® device families. These designs relate to meeting the core required system clock
frequency of 125 MHz.
1-Gigabit Ethernet MAC v8.5 User Guide
37
UG144 April 24, 2009
Download from Www.Somanuals.com. All Manuals Search And Download.
-- DISCONTINUED PRODUCT --
R
Chapter 4: Designing with the Core
and Spartan-3A device setup and hold requirements for external GMII.
Table 4-1: Degree of Difficulty for Various Implementations
Device Family
Spartan-3A/3AN/3A DSP platforms
Spartan-3E platform
Spartan-3 device
Difficulty
Difficult
Difficult
Difficult
Easy
Virtex-4 device
Virtex-5 device
Easy
Keep it Registered
To simplify timing and increase system performance in an FPGA design, keep all inputs
and outputs registered between the user application and the core. This means that all
inputs and outputs from the user application should come from, or connect to, a flip-flop.
While registering signals may not be possible for all paths, it simplifies timing analysis and
makes it easier for the Xilinx tools to place-and-route the design.
Recognize Timing Critical Signals
The UCF provided with the example design identifies the critical signals and timing
constraints that should be applied. See Chapter 9, “Constraining the Core”
Use Supported Design Flows
The core is pre-synthesized and delivered as an NGC netlist. The example implementation
scripts provided use XST 11.1 as the synthesis tool for the HDL example design. Other
synthesis tools may be used for the user application logic. The core is always unknown to
the synthesis tool and should appear as a black box. For post synthesis, only ISE® 11.1 tools
are supported.
Make Only Allowed Modifications
The GEMAC core should not be modified by you. Any modifications may have adverse
effects on system timing and protocol compliance. Supported user configurations of the
GEMAC core can only be made by selecting the options in the CORE Generator when the
core is generated. See Chapter 3, “Generating the Core.”
38
1-Gigabit Ethernet MAC v8.5 User Guide
UG144 April 24, 2009
Download from Www.Somanuals.com. All Manuals Search And Download.
-- DISCONTINUED PRODUCT --
R
Chapter 5
Using the Client Side Data Path
This chapter provides general guidelines for creating designs using the GEMAC core,
including a detailed description of each client-side data flow interface of the core.
Definitions of the abbreviations used throughout the remainder of this chapter are defined
Table 5-1: Abbreviations Used in Timing Diagrams
Abbreviation
Definition
Destination address; 6 bytes
DA
SA
Source address; 6 bytes
L/T
FCS
Length/type field; 2 bytes
Frame check sequence; 4 bytes
Receiving Inbound Frames
Received Ethernet frames are presented to the client logic on the Receiver subset of the
Normal Frame Reception
Figure 5-1 illustrates the timing of a normal inbound frame transfer. The client must be
prepared to accept data at any time; there is no buffering within the MAC to allow for
latency in the receive client. Once frame reception begins, data is transferred on
consecutive clock cycles to the receive client until the frame is complete. The MAC asserts
the rx_good_framesignal to indicate that the frame was successfully received and that
the frame should be analyzed by the client.
1-Gigabit Ethernet MAC v8.5 User Guide
39
UG144 April 24, 2009
Download from Www.Somanuals.com. All Manuals Search And Download.
-- DISCONTINUED PRODUCT --
Chapter 5: Using the Client Side Data Path
R
gmii_rx_clk
rx_data[7:0]
DA
SA
L/T
DATA
rx_data_valid
rx_good_frame
rx_bad_frame
Figure 5-1: Normal Frame Reception
Frame parameters (destination address, source address, length/type and optionally FCS)
are supplied on the data bus according to the timing diagram.
If the length/type field in the frame has a length interpretation, indicating that the
inbound frame has been padded to meet the Ethernet minimum frame size, the padding is
not passed to the client in the data payload. The exception to this is where FCS passing is
enabled. See “Client-Supplied FCS Passing.”
When Client-Supplied FCS passing is disabled, rx_data_validis equal to zero between
frames for the duration of the padding field (if present), the FCS field, carrier extension (if
present), the interframe gap following the frame, and the preamble field of the next frame.
When Client-Supplied FCS passing is enabled, rx_data_valid is equal to zero between
frames for the duration of carrier extension (if present), the interframe gap, and the
preamble field of the following frame.
rx_good_frame, rx_bad_frame timing
shortly after the last valid data on rx_data, this is not always the case. The
rx_good_frameor rx_bad_framesignals are asserted only after all frame checks are
completed. This is after the FCS field has been received (and after reception of carrier
extension, if present).
Therefore, either rx_good_frameor rx_bad_frameis asserted following frame
reception at the beginning of the interframe gap.
40
1-Gigabit Ethernet MAC v8.5 User Guide
UG144 April 24, 2009
Download from Www.Somanuals.com. All Manuals Search And Download.
-- DISCONTINUED PRODUCT --
R
Receiving Inbound Frames
Frame Reception with Errors
Figure 5-2 illustrates an unsuccessful frame reception (for example, a fragment frame or a
frame with an incorrect FCS). In this case, the rx_bad_framesignal is asserted to the
client at the end of the frame. It is then the responsibility of the client to drop the data
already transferred for this frame.
The following conditions cause the assertion of rx_bad_frame:
•
•
•
•
•
FCS errors occur.
Packets are shorter than 64 bytes (undersize or fragment frames).
Jumbo frames are received when jumbo frames are not enabled.
VLAN frames of length 1519-1522 are received when VLAN frames are not enabled.
A value of 0x0000 to 0x002D is in the type/length field. In this situation, the frame
should be padded to minimum length. If it is not padded to exactly minimum frame
length, the frame is marked as bad (when length/type checking is enabled).
•
•
A value of 0x002E to 0x0600 is in the type/length field, but the real length of the
received frame does not match this value (when length/type checking is enabled).
Any control frame that is received is not exactly the minimum frame length (unless
control frame length checks are disabled).
• gmii_rx_er is asserted at any point during frame reception.
•
•
An error code is received in the 1-Gigabit frame extension field.
A valid pause frame, addressed to the MAC, is received when flow control is enabled.
gmii_rx_clk
rx_data[7:0]
DA
SA
L/T
DATA
rx_data_valid
rx_good_frame
rx_bad_frame
Figure 5-2: Frame Reception with Error
1-Gigabit Ethernet MAC v8.5 User Guide
41
UG144 April 24, 2009
Download from Www.Somanuals.com. All Manuals Search And Download.
-- DISCONTINUED PRODUCT --
Chapter 5: Using the Client Side Data Path
R
Client-Supplied FCS Passing
If the GEMAC core is configured to pass the FCS field to the client (see “Configuration
inserted into the frame to meet Ethernet minimum frame length specifications will be left
intact and passed to the client.
Even though the FCS is passed up to the client, it is also verified by the GEMAC core, and
rx_bad_frameasserted if the FCS check fails.
gmii_rx_clk
rx_data[7:0]
DA
SA
L/T
DATA
FCS
rx_data_valid
rx_good_frame
rx_bad_frame
Figure 5-3: Frame Reception with In-Band FCS Field
VLAN Tagged Frames
Figure 5-4 illustrates the reception of a VLAN tagged frame (if enabled). The VLAN frame
is passed to the client so that the frame may be identified as VLAN tagged. This is followed
by the Tag Control Information bytes, V1 and V2. More information on the interpretation
of these bytes may be found in IEEE 802.3-2005 standard.
gmii_rx_clk
rx_data[7:0]
8100V1V2
DA
SA
VLAN
tag
L/T
DATA
rx_data_valid
rx_good_frame
rx_bad_frame
Figure 5-4: Reception of a VLAN Tagged Frame
42
1-Gigabit Ethernet MAC v8.5 User Guide
UG144 April 24, 2009
Download from Www.Somanuals.com. All Manuals Search And Download.
-- DISCONTINUED PRODUCT --
R
Receiving Inbound Frames
Maximum Permitted Frame Length
The maximum legal length of a frame specified in IEEE 802.3-2005 is 1518 bytes for non-
VLAN tagged frames. VLAN tagged frames may be extended to 1522 bytes. When jumbo
frame handling is disabled and the core receives a frame which exceeds the maximum
legal length, rx_bad_frameis asserted. When jumbo frame handling is enabled, frames
which are longer than the legal maximum are received in the same way as shorter frames.
For more information about enabling and disabling jumbo frame handling, see
Length/Type Field Error Checks
Enabled
Default operation is with the length/type error checking enabled (see “Receiver
Configuration,” on page 79). In this mode, the following checks are made on all frames
received. If either of these checks fail, the frame is marked as bad.
•
A value in the length/type field that is greater than or equal to decimal 46, but less
than decimal 1536 (a length interpretation), is checked against the actual data length
received.
•
A value in the length/type field that is less than decimal 46 is checked to see that the
data field is padded to exactly 46 bytes (so that the resultant frame is a minimum
frame size of 64 bytes total in length).
Furthermore, if padding is indicated (the length/type field is less than decimal 46) and
client-supplied FCS passing is disabled, the length value in the length/type field will be
used to deassert rx_data_validafter the indicated number of data bytes so that the
padding bytes are removed from the frame. See “Client-Supplied FCS Passing.”
Disabled
When the length/type error checking is disabled and the length/type field has a length
interpretation, the MAC does not check the length value against the actual data length
marked as good.
However, if the length/type field is less than decimal 46 then the MAC will mark a frame
as bad if it is not the minimum frame size of 64 bytes.
If padding is indicated and client-supplied FCS passing is disabled, then a length value in
the length/type field will not be used to deassert rx_data_valid. Instead,
rx_data_validis deasserted before the start of the FCS field, and any padding is not
removed from the frame.
1-Gigabit Ethernet MAC v8.5 User Guide
43
UG144 April 24, 2009
Download from Www.Somanuals.com. All Manuals Search And Download.
-- DISCONTINUED PRODUCT --
Chapter 5: Using the Client Side Data Path
R
Address Filter
If the optional Address Filter is included in the core, the MAC is able to reject frames that
do not contain a known address in their destination address field. If a frame is rejected, the
rx_data_validsignal is not asserted for the duration of the frame. In addition, neither
rx_good_frameor rx_bad_frameare asserted at the end of the frame. The statistics
vectors are still output with a valid pulse at the end of the rejected frame.
If the Address Filter is not in promiscuous mode, it will reject frames in which the
destination address does not meet any of the following criteria:
•
•
•
It is equal to the broadcast address defined in the IEEE 802.3-2005 specification.
It is equal to the pause multicast address defined in the IEEE 802.3-2005 specification.
The destination address field contains the pause frame MAC source address specified
in the Receiver Configuration Word 0 and Word 1.
•
It is equal to the MAC unicast address. When the optional Management Interface is
Table 8-9, page 82). If the Management Interface is not present the unicast address is
input on the mac_unicast_addressinput.
•
It matches any of the addresses stored in the MAC address table. The address table is
only present when the MAC contains the optional Management Interface and the core
was built with one or more address table entries.
Receiver Statistics Vector
The statistics for the frame received are contained within the rx_statistics_vector.
The vector is driven synchronously by the receiver clock, gmii_rx_clk, following frame
All bit fields, with the exception of byte valid, are valid only when the
significant on every gmii_rx_clkcycle.
Caution! The statistic vectors in this release have been made compatible with the Tri-Mode
Ethernet MAC core. They are not backwards compatible with previous versions of the 1-Gigabit
gmii_rx_clk
rx_statistics_valid
rx_statistics_vector[27:0]
Figure 5-5: Receiver Statistics Vector Timing
44
1-Gigabit Ethernet MAC v8.5 User Guide
UG144 April 24, 2009
Download from Www.Somanuals.com. All Manuals Search And Download.
-- DISCONTINUED PRODUCT --
R
Receiving Inbound Frames
Table 5-2: Bit Definition for the Receiver Statistics Vector
rx_statistics_vector
Name
Description
bit(s)
27
Address Match
If the optional Address Filter is included in the
core, this bit is asserted if the address of the
incoming frame matches one of the stored or
pre-set addresses in the Address Filter. If the
Address Filter is omitted from the core, or is
configured in promiscuous mode, this line is
held high.
26
25
Reserved
Always at logic 0.
Length/Type
Out of Range
If the length/type field contained a length
value that did not match the number of MAC
client data bytes received and the length/type
field checks are enabled, then this bit is
asserted.
This bit is also asserted if the length/type field
is less than 46 and the frame is not padded to
exactly 64 bytes. This is independent of
whether or not the length/type field checks
are enabled.
24
23
Bad Opcode
Asserted if the previous frame was error-free
and contained the special control frame
identifier in the length/type field, but
contained an opcode that is unsupported by
the MAC (any opcode other than Pause).
Flow Control
Frame
Asserted if the previous frame met all the
following conditions:
• error-free
• contained the special control frame
identifier in the length/type field
• contained a destination address that
matched either the MAC Control Multicast
Address or the configured source address
of the MAC
• contained the supported Pause opcode
• was acted upon by the MAC
22
21
Byte Valid
Asserted if a MAC frame byte (DA to FCS
inclusive) is in the process of being received.
This is valid on every clock cycle.
Do not use this as an enable signal to indicate
that data is present on rx_data.
VLAN frame
Asserted if the previous frame contained a
VLAN identifier in the length/type field when
receiver VLAN operation is enabled.
1-Gigabit Ethernet MAC v8.5 User Guide
45
UG144 April 24, 2009
Download from Www.Somanuals.com. All Manuals Search And Download.
-- DISCONTINUED PRODUCT --
Chapter 5: Using the Client Side Data Path
R
Table 5-2: Bit Definition for the Receiver Statistics Vector
rx_statistics_vector
Name
Description
bit(s)
20
Out of Bounds
Asserted if the previous frame exceeded the
specified IEEE802.3-2005 maximum legal
length (see “Maximum Permitted Frame
Length”). This is only valid if jumbo frames
are disabled.
19
18:5
4
Control Frame
Frame Length
Asserted if the previous frame contained the
special control frame identifier in the
length/type field.
The length of the previous frame in number of
bytes. The count will stick at 16,383 for any
jumbo frames larger than this value.
Multicast Frame Asserted if the previous frame contained a
multicast address in the destination address
field.
3
Broadcast Frame Asserted if the previous frame contained the
broadcast address in the destination address
field.
2
FCS Error
Asserted if the previous frame received had an
incorrect FCS value or the MAC detected error
codes during frame reception.
1
0
Bad Frame
Asserted if the previous frame received
contained errors.
Good Frame
Asserted if the previous frame received was
error-free.
Table 5-3 provides conversion information against previous versions of the GEMAC.
Table 5-3: Rx Statistics conversion to previous core GEMAC core versions
Version 8.5
rx_statistics_vector
bit(s)
Version 8.4 (and earlier)
rx_statistics_vector
bit(s)
Notes
27
26
26
Bit 27 is equivalent to bit 26
of all previous core
versions.
N/A
25:0
Bit 26 (reserved) has been
inserted into version 8.5 for
statistic vector
compatibility with the Tri-
Mode Ethernet MAC
LogiCORE™.
25:0
No differences
46
1-Gigabit Ethernet MAC v8.5 User Guide
UG144 April 24, 2009
Download from Www.Somanuals.com. All Manuals Search And Download.
-- DISCONTINUED PRODUCT --
R
Transmitting Outbound Frames
Transmitting Outbound Frames
Ethernet frames to be transmitted are presented to the client logic on the Transmitter subset
Normal Frame Transmission
Figure 5-6 illustrates the timing of a normal outbound frame transfer. When the client
wishes to transmit a frame, it places the first column of data onto the tx_dataport and
asserts a ‘1’ onto tx_data_valid.
When the GEMAC core has read this first byte of data, and in accordance with flow control
requests and interpacket gap requirements, it will assert the tx_acksignal; on the next
and subsequent rising clock edges, the client must provide the remainder of the data for
the frame.
The end of frame is signalled to the GEMAC core by taking tx_data_validlow.
For maximum flexibility in switching and routing applications, the Ethernet frame
parameters (destination address, source address, length/type and optionally FCS) are
encoded within the same data stream that the frame payload is transferred upon, rather
than on separate ports. This is illustrated in the timing diagrams. Definitions of the
gtx_clk
tx_data[7:0]
DA
SA
L/T
DATA
tx_data_valid
tx_ack
tx_underrun
Figure 5-6: Normal Frame Transmission
Padding
When fewer than 46 bytes of data are supplied by the client to the GEMAC core, the
transmitter module will add padding up to the minimum frame length. The exception to
this is when the GEMAC core is configured for client-passed FCS; in this case, the client
must also supply the padding to maintain the minimum frame length. See “Client-
Supplied FCS Passing” for more information.
1-Gigabit Ethernet MAC v8.5 User Guide
47
UG144 April 24, 2009
Download from Www.Somanuals.com. All Manuals Search And Download.
-- DISCONTINUED PRODUCT --
Chapter 5: Using the Client Side Data Path
R
Client-Supplied FCS Passing
the FCS field passed in by the client. In this case, it is the responsibility of the client to
ensure that the frame meets the Ethernet minimum frame length requirements as the
GEMAC core will not perform any padding of the payload. See “Configuration Registers,”
gtx_clk
tx_data[7:0]
DA
SA
L/T
DATA
FCS
tx_data_valid
tx_ack
tx_underrun
Figure 5-7: Frame Transmission with Client-supplied FCS
Client Underrun
Figure 5-8 illustrates the timing of an aborted transfer. An example of this situation is a
FIFO connected to the client interface that empties before a frame transfer is complete.
When the client asserts tx_underrunduring a frame transmission, the GEMAC core
inserts an error code to corrupt the current frame, and then falls back to idle transmission.
It is the responsibility of the client to re-queue the aborted frame for transmission.
When an underrun occurs, tx_data_validmay be asserted on the clock cycle after the
tx_underrunassertion to request a new transmission.
gtx_clk
tx_data[7:0]
DA
SA
L/T
DATA
tx_data_valid
tx_ack
tx_underrun
Figure 5-8: Frame Transmission with Underrun
48
1-Gigabit Ethernet MAC v8.5 User Guide
UG144 April 24, 2009
Download from Www.Somanuals.com. All Manuals Search And Download.
-- DISCONTINUED PRODUCT --
R
Transmitting Outbound Frames
VLAN Tagged Frames
Figure 5-9 illustrates transmission of a VLAN tagged frame (if enabled). The handshaking
signals across the interface do not change; however, the VLAN type tag 81-00 must be
supplied by the client to signify that the frame is VLAN tagged. The client also supplies the
two bytes of Tag Control Information, V1 and V2, at the appropriate times in the data
stream. More information on the contents of these two bytes can be found in IEEE 802.3-
2005. For more information about enabling and disabling jumbo frame handling, see
gtx_clk
tx_data[7:0]
8100V1V2
VLAN
tag
L/T
DA
SA
DATA
tx_data_valid
tx_ack
tx_underrun
Figure 5-9: Transmission of a VLAN Tagged Frame
Maximum Permitted Frame Length
The maximum legal length of a frame specified in IEEE 802.3-2005 is 1518 bytes for non-
VLAN tagged frames. VLAN tagged frames may be extended to 1522 bytes. When jumbo
frame handling is disabled and the client attempts to transmit a frame that exceeds the
maximum legal length, the GEMAC core will insert an error code to corrupt the current
frame, and the frame will be truncated to the maximum legal length. When jumbo frame
handling is enabled, frames longer than the legal maximum are transmitted error-free. For
more information on enabling and disabling Jumbo frame handling, see “Configuration
Inter-Frame Gap Adjustment
A configuration bit in the transmitter control register (see “Configuration Registers,” on
page 78) allows you to control the length of the inter-frame gap transmitted by the MAC on
the physical interface. If this function is selected, the MAC exerts back pressure on the
client interface to delay the transmission of the next frame until the requested number of
idle cycles has elapsed. The number of idle cycles is controlled by the value on the
tx_ifg_delay port seen at the start of frame transmission on the client interface.
Figure 5-10 shows the MAC operating in this mode.
Reducing the interframe gap to below the IEEE 802.3-2005 minimum of 12 idles is
supported, but the MAC will transmit an absolute minimum of 4 idles. If the Ethernet
Statistics core is used with the MAC, then accuracy cannot be guaranteed if the interframe
gap adjustment is set to less than 12 idles. However, the tx_statistic_vectorand
rx_statistic_vectorvalues will always remain correct.
1-Gigabit Ethernet MAC v8.5 User Guide
49
UG144 April 24, 2009
Download from Www.Somanuals.com. All Manuals Search And Download.
-- DISCONTINUED PRODUCT --
R
Chapter 5: Using the Client Side Data Path
gtx_clk
tx_data[7:0]
DA
SA
DA
tx_data_valid
tx_ack
tx_ifg_delay
0x0D
IFG ADJUST VALUE
Next IFG ADJUST VALUE
13 Idles inserted between the
end of frame and the preamble
field of the following frame
Figure 5-10: Inter-Frame Gap Adjustment
Transmitter Statistics Vector
The statistics for the transmitted frame are contained within the tx_statistic_vector.
The vector is driven synchronously by the transmitter clock, gtx_clk, following frame
All bit fields, with the exception of byte valid, are valid only when
gtx_clkcycle.
Caution! The statistic vectors in this release have been made compatible with the Tri-Mode
Ethernet MAC core. They are not backwards compatible with previous versions of the 1-Gigabit
gtx_clk
tx_statistic_valid
tx_statistic_vector[31:0]
Figure 5-11: Transmitter Statistic Vector Timing
50
1-Gigabit Ethernet MAC v8.5 User Guide
UG144 April 24, 2009
Download from Www.Somanuals.com. All Manuals Search And Download.
-- DISCONTINUED PRODUCT --
R
Transmitting Outbound Frames
Table 5-4: Bit Definition for the Transmitter Statistics Vector
tx_statistics_vector
bit(s)
Name
Description
31
30
Pause Frame
Asserted if the previous frame was a pause
frame that the MAC itself initiated in
response to a pause_req assertion.
Byte Valid
Asserted if a MAC frame byte (DA to FCS
inclusive) is in the process of being
transmitted. This is valid on every clock
cycle.
Do not use this as an enable signal to
indicate that data is present on gmii_txd.
29:20
19
Reserved
Always at logic 0.
VLAN Frame
Asserted if the previous frame contained a
VLAN identifier in the length/type field
when transmitter VLAN operation is
enabled.
18:5
4
Frame Length
Control Frame
The length of the previous frame in number
of bytes. The count will stick at 16,383 for
any jumbo frames larger than this value.
Asserted if the previous frame had the
special MAC Control Type code 88-08 in the
length/type field.
3
2
Underrun Frame Asserted if the previous frame contained an
underrun error.
Multicast Frame
Asserted if the previous frame contained a
multicast address in the destination address
field.
1
0
Broadcast Frame Asserted if the previous frame contained a
broadcast address in the destination address
field.
Successful Frame Asserted if the previous frame was
transmitted without error.
1-Gigabit Ethernet MAC v8.5 User Guide
51
UG144 April 24, 2009
Download from Www.Somanuals.com. All Manuals Search And Download.
-- DISCONTINUED PRODUCT --
Chapter 5: Using the Client Side Data Path
R
Table 5-5 provides conversion information against previous versions of the GEMAC.
Table 5-5: Tx Statistics conversion to previous core GEMAC core versions
Version 8.5
tx_statistics_vector
bit(s)
Version 8.4 (and earlier)
tx_statistics_vector
bit(s)
Notes
31
30
21
Bit 31 is equivalent to bit 21
of all previous core
versions.
20
Bit 30 is equivalent to bit 20
of all previous core
versions.
29:20
N/A
Reserved: have been
inserted into version 8.5 for
statistic vector
compatibility with the Tri-
Mode Ethernet MAC
LogiCORE.
19:0
19:0
No differences
52
1-Gigabit Ethernet MAC v8.5 User Guide
UG144 April 24, 2009
Download from Www.Somanuals.com. All Manuals Search And Download.
-- DISCONTINUED PRODUCT --
R
Chapter 6
Using Flow Control
This chapter describes the operation of the flow-control logic of the GEMAC core. The flow
control block is designed to clause 31 of the IEEE 802.3-2005 standard. The MAC may be
configured to transmit pause requests and to act on their reception; these modes of
operation can be independently enabled or disabled. See “Flow Control Configuration,”
Overview of Flow Control
Flow Control Requirement
Figure 6-1 illustrates the requirements for Flow Control.
User System
Client Logic
User MAC
Link Partner MAC
125MHz -100ppm
Tx
Rx
Rx
Tx
FIFO
125MHz +100ppm
Figure 6-1: Requirement for Flow Control
1-Gigabit Ethernet MAC v8.5 User Guide
53
UG144 April 24, 2009
Download from Www.Somanuals.com. All Manuals Search And Download.
-- DISCONTINUED PRODUCT --
R
Chapter 6: Using Flow Control
The user MAC on the left side has a reference clock slightly slower than the nominal
125 MHz. The link partner MAC on the right side has a reference clock slightly faster than
the nominal 125 MHz. As a result, the user MAC receives data at a faster line rate than that
at which it can transmit. The MAC on the left is shown performing a loopback
implementation which results in the FIFO filling up over time. Without Flow Control, this
FIFO will eventually fill and overflow, resulting in the corruption or loss of Ethernet
frames. Enabling Flow Control in the MAC provides a mechanism to solve this data rate
matching problem.
Flow Control Basics
A MAC may transmit a pause control frame to request that its link partner cease
transmission for a defined period of time. For example, the user MAC on the left side of
Figure 6-1 may initiate a pause request when its client FIFO (illustrated) reaches a nearly
full state.
A MAC should respond to received pause control frames by ceasing transmission of
frames for the period of time defined in the received pause control frame. For example, the
frame transmitted by the user MAC. In a well designed system, the link partner MAC
would cease transmission before the client FIFO experienced an overflow condition. This
provides time for the FIFO to be emptied to a safe level before normal operation resumes,
thus safeguarding the system against FIFO overflow conditions and frame loss.
54
1-Gigabit Ethernet MAC v8.5 User Guide
UG144 April 24, 2009
Download from Www.Somanuals.com. All Manuals Search And Download.
-- DISCONTINUED PRODUCT --
R
Overview of Flow Control
Pause Control Frames
Control frames are a unique type of Ethernet frame, defined in clause 31 of the IEEE 802.3-
2005 standard. Control frames are differentiated from other frame types by a defined value
frame format.
6 OCTETS
6 OCTETS
2 OCTETS
2 OCTETS
DESTINATION
ADDRESS
SOURCE
ADDRESS
LENGTH/TYPE
FRAME
TRANSMITTED
TOP-TO-BOTTOM
MAC CONTROL
OPCODE
MAC CONTROL
PARAMETERS
44 OCTETS
4 OCTETS
RESERVED
(transmitted as zeroes)
FRAME CHECK
SEQUENCE
Figure 6-2: MAC Control Frame Format
A pause control frame is a unique type of control frame, identified by a defined value
placed in the MAC Control opcode field.
Note: MAC Control opcodes other than for pause (flow control) frames have recently been defined
for Ethernet Passive Optical Networks.
The MAC control parameter field of the pause control frame contains a 16-bit field
containing a binary value directly related to the pause duration. This defines the number of
pause_quantum(512 bit times of the particular implementation). For 1-Gigabit Ethernet,
a single pause_quantumcorresponds to 512 ns.
1-Gigabit Ethernet MAC v8.5 User Guide
55
UG144 April 24, 2009
Download from Www.Somanuals.com. All Manuals Search And Download.
-- DISCONTINUED PRODUCT --
R
Chapter 6: Using Flow Control
Flow Control Operation of the GEMAC
Transmitting a PAUSE Control Frame
Core-initiated Pause Request
If the GEMAC core is configured to support transmit flow control, the client can initiate a
gtx_clk
pause_req
pause_val[15:0]
Figure 6-3: Pause Request Timing
This action causes the core to construct and transmit a pause control frame on the link with
•
•
•
The destination address used is an IEEE802.3 globally assigned multicast address (to
which any flow control-capable MAC will respond).
The source address used is the configurable pause frame MAC address (see“Receiver
The value sampled from the pause_val[15:0]port at the time of the pause_req
assertion will be encoded into the MAC control parameter field to select the duration
of the pause (in units of pause_quantum).
If the transmitter is inactive at the time of the pause request, this pause control frame is
transmitted immediately. If the transmitter is currently busy, the current frame being
transmitted is allowed to complete, followed by the pause control frame (in preference to
any pending client-supplied frame).
A pause control frame initiated by this method is transmitted even if the transmitter has
ceased in response to receiving an inbound pause request.
Note: Only a single pause control frame request is stored by the transmitter. if pause_reqis
asserted numerous times in a short time period (before the control pause frame transmission has
begun), only a single pause control frame is transmitted. The most recent value sampled will be the
pause_val[15:0]value used.
Client Initiated Pause Request
For maximum flexibility, flow control logic can be disabled in the core and alternatively
implemented in the client logic connected to the core (see “Flow Control Configuration,”
on page 81). Any type of control frame can be transmitted through the core through the
client interface using the same transmission procedure as a standard Ethernet frame (see
56
1-Gigabit Ethernet MAC v8.5 User Guide
UG144 April 24, 2009
Download from Www.Somanuals.com. All Manuals Search And Download.
-- DISCONTINUED PRODUCT --
Flow Control Operation of the GEMAC
R
Receiving a Pause Control Frame
Core Initiated Response to a Pause Request
pass all standard receiver frame checks (for example, FCS field checking). In addition, the
control frame received must be exactly 64-bytes in length (from destination address
through to the FCS field inclusive: this is minimum legal Ethernet MAC frame size and the
defined size for control frames).
Any control frame received that does not conform to these checks contains an error and is
passed to the receiver client with the rx_bad_framesignal asserted.
Pause Frame Reception Disabled
When pause control reception is disabled, an error free control frame is received through
Pause Frame Reception Enabled
When pause control reception is enabled, and an error-free frame is received by the
decoding functions are performed:
•
The destination address field is matched against the IEEE 802.3 globally assigned
multicast address or the configurable pause frame MAC address (see “Configuration
•
•
The length/type field is matched against the MAC control type code.
The opcode field contents are matched against the Pause opcode.
If any of the previously described checks are false, the frame is ignored by the Flow Control
logic and passed up to the client logic for interpretation by marking it with
rx_good_frameasserted. It is then the responsibility of the MAC client logic to decode,
act on (if required), and drop this control frame.
If all the previously described checks are true, the 16-bit binary value in the MAC Control
Parameters field of the control frame is then used to inhibit transmitter operation for the
required number of pause_quantum. This inhibit is implemented by delaying the assertion
of tx_ackat the transmitter client interface until the requested pause duration has
expired. The received pause frame is then passed on to the client with rx_bad_frame
asserted to indicate to the client that the pause frame can be dropped.
Note: Any frame in which the length/type field contains the MAC control type should be dropped by
the receiver client logic. All control frames are indicated by rx_statistic_vectorbit 19 (see
Client Initiated Response to a Pause Request
For maximum flexibility, flow control logic can be disabled in the core and alternatively
implemented in the client logic connected to the core (see “Flow Control Configuration,”
on page 81). Any type of error-free control frame is then passed through the core with
rx_good_frameasserted. The frame is passed to the client for interpretation. It is then the
responsibility of the client to drop this control frame and to act on it by ceasing
transmission through the core, if applicable.
1-Gigabit Ethernet MAC v8.5 User Guide
57
UG144 April 24, 2009
Download from Www.Somanuals.com. All Manuals Search And Download.
-- DISCONTINUED PRODUCT --
R
Chapter 6: Using Flow Control
Flow Control Implementation Example
This section provides a basic overview of a Flow Control implementation, using Figure 6-1
as a sample. To summarize the example, the user MAC on the left hand side of the figure
cannot match the full line rate of the link partner MAC on the right hand side due to clock
tolerances. Over time, the FIFO illustrated will fill and overflow. The goal is to implement
a flow control method which will (over a long time period) reduce the average line rate of
the link partner MAC to that of the user MAC.
Method
1. Choose a FIFO nearly full to occupancy threshold (7/8 occupancy is used in this
description—but the choice of threshold is implementation specific). When the
occupancy of the FIFO exceeds this occupancy, initiate a single pause control frame,
from the user MAC, with 0xFFFF used as the pause_quantum duration (0xFFFF is
placed on pause_val[15:0]). This is the maximum pause duration. This causes the
link partner MAC to cease transmission, and the FIFO of the user system will start to
empty.
2. Choose a second FIFO occupancy threshold (3/4 is used in this description—but the
choice of threshold is implementation specific). When the occupancy of the FIFO falls
below this occupancy, initiate a second pause control frame from the user MAC, with
0x0000 used as the pause_quantum duration (0x0000 is placed on pause_val[15:0]).
This indicates a zero pause duration, and upon receiving this pause control frame, the
link partner MAC immediately resumes transmission (it does not wait for the original
requested pause duration to expire). This pause control frame can therefore be
considered a “pause cancel” command.
58
1-Gigabit Ethernet MAC v8.5 User Guide
UG144 April 24, 2009
Download from Www.Somanuals.com. All Manuals Search And Download.
-- DISCONTINUED PRODUCT --
Flow Control Implementation Example
R
Operation
Figure 6-4 illustrates the FIFO occupancy over a period of time.
Full
A
C
7/8
3/4
5/8
1/2
B
time
Figure 6-4: Flow Control Implementation Triggered from FIFO Occupancy
The following describes the sequence of flow control operation.
1. The average FIFO occupancy of the user system gradually increases over time due to
the clock tolerances. At point A, the occupancy has reached the threshold of 7/8
occupancy. This triggers the maximum duration pause control frame request.
2. On receiving the pause control frame, the link partner MAC ceases transmission.
3. After the link partner MAC ceases transmission, the occupancy of the FIFO in the user
system rapidly empties. The occupancy falls to the second threshold of 3/4 occupancy
at point B. This triggers the zero duration pause control frame request (the pause
cancel command).
4. On receiving this second pause control frame, the link partner MAC resumes
transmission.
5. Normal operation resumes and the FIFO occupancy again gradually increases over
time. At point C, this Flow Control cycle repeats.
1-Gigabit Ethernet MAC v8.5 User Guide
59
UG144 April 24, 2009
Download from Www.Somanuals.com. All Manuals Search And Download.
-- DISCONTINUED PRODUCT --
R
Chapter 6: Using Flow Control
60
1-Gigabit Ethernet MAC v8.5 User Guide
UG144 April 24, 2009
Download from Www.Somanuals.com. All Manuals Search And Download.
-- DISCONTINUED PRODUCT --
R
Chapter 7
Using the Physical Side Interface
This chapter provides general guidelines for creating designs using the Physical Side
Interface of the GEMAC core. The physical side interface implements GMII-style signaling
and is typically attached to a physical layer device (PHY), either off-chip or internally
internal interface in conjunction with the Ethernet 1000BASE-X PCS/PMA or SGMII core,
The remainder of this chapter describes how to use the core with an external GMII or
Implementing External GMII
The HDL example design that is delivered with the core will implement an external GMII
when GMII is selected from the CORE Generator™ GUI (see Chapter 3, “Generating the
Getting Started Guide.
GMII Transmitter Logic
Figure 7-1 illustrates how to use the physical transmitter interface of the core to create an
external GMII in a Spartan®-3 device. The signal names and logic shown in this figure
exactly match those delivered with the example design when the GMII is selected. If other
families are chosen, equivalent primitives and logic specific to that family is used in the
example design.
Figure 7-1 shows that the output transmitter signals are registered in device IOBs before
driving them to the device pads. The logic required to forward the transmitter clock is also
shown. This logic uses an IOB output Double-Data-Rate (DDR) register so that the clock
signal produced incurs exactly the same delay as the data and control signals. This clock
signal, gmii_tx_clk, is inverted with respect to gtx_clkso that the rising edge of
gmii_tx_clkwill occur in the centre of the data valid window, therefore maximizing
setup and hold times across the interface.
1-Gigabit Ethernet MAC v8.5 User Guide
61
UG144 April 24, 2009
Download from Www.Somanuals.com. All Manuals Search And Download.
-- DISCONTINUED PRODUCT --
Chapter 7: Using the Physical Side Interface
R
IOB LOGIC
FDDRRSE
IOB LOGIC
'0'
BUFG
D
Q
Q
IBUFG
gtx_clk
IPAD
OBUF
gtx_clk_bufg
gmii_tx_clk
OPAD
'1'
D
1-Gigabit Ethernet MAC LogiCORE
OBUF
OBUF
OBUF
gmii_txd[0]
OPAD
gmii_txd_reg[0]
gmii_tx_en_reg
gmii_tx_er_reg
gmii_txd_int[0]
gtx_clk
D
D
D
Q
Q
Q
gmii_txd[0]
gmii_tx_en
OPAD
gmii_tx_en_int
gmii_tx_en
gmii_tx_er
OPAD
gmii_tx_er_int
gmii_tx_er
Figure 7-1: External GMII Transmitter Logic
62
1-Gigabit Ethernet MAC v8.5 User Guide
UG144 April 24, 2009
Download from Www.Somanuals.com. All Manuals Search And Download.
-- DISCONTINUED PRODUCT --
R
Implementing External GMII
GMII Receiver Logic
Spartan-3, Spartan-3E, Spartan-3A and Virtex-4 Devices
the input setup and hold requirements for GMII. This is performed by the example designs
may optionally be used in other families.
Phase-shifting may then be applied to the DCM to fine-tune the setup and hold times at the
GMII IOB input flip-flops. Fixed phase-shift is applied to the DCM with the example UCF
for the example design. See Appendix C, “Calculating DCM Phase-Shifting.”
IOB LOGIC
BUFG
IBUFG
DCM
gmii_rx_clk
IPAD
gmii_rx_clk0
CLKIN
CLK0
1-Gigabit Ethernet MAC LogiCORE
FB
gmii_rx_clk
gmii_rx_clk_bufg
IOB LOGIC
IBUF
IBUF
IBUF
gmii_rxd[0]
IPAD
gmii_rxd_reg[0]
gmii_rx_dv_reg
gmii_rx_er_reg
gmii_rxd[0]
gmii_rx_dv
gmii_rx_er
Q
Q
Q
D
D
D
gmii_rx_dv
IPAD
gmii_rx_er
IPAD
Figure 7-2: External GMII Receiver Logic for Spartan-3, Spartan-3E, and Spartan-3A Devices
1-Gigabit Ethernet MAC v8.5 User Guide
63
UG144 April 24, 2009
Download from Www.Somanuals.com. All Manuals Search And Download.
-- DISCONTINUED PRODUCT --
Chapter 7: Using the Physical Side Interface
R
DCM Reset circuitry
example design next to the DCM. Since this logic must be reliable whatever the
reset/locked status of the DCM, the module requires a reliable reference clock. In the
example design for GMII, the global transmitter clock is therefore used (gtx_clk_bufg
family, for the receiver DCM under the following conditions:
•
The lockedsignal from the DCM is constantly monitored. Following a high to low
transition on this signal, indicating that the DCM has lost lock, a reset will be issued.
•
A timeout counter is enabled when the DCM is in the loss of lock state. If, following
the timeout period, the DCM has not obtained lock, another DCM reset will be issued.
This timeout counter will time a > 1ms interval. This timeout functionality is required
for DCMs connected to Ethernet PHYs since the PHYs may source discontinuous
clocks under certain network conditions (for example, when no ethernet cable is
connected).
Spartan-3 Devices
For Spartan-3 families, the reset pulse is transferred into the DCM input clock
and routed to the reset input of the DCM.
Virtex-4 Devices
For Virtex®-4 families, the generated reset from within the DCM reset module is further
complicated. Here the DCM reset pulse duration must be asserted for a minimum of 200
ms (see the Virtex-4 Datasheet). Consequently, a 200 ms duration timer is also included in
the DCM reset module to extend the reset pulse. This reset signal is output from the DCM
reset module and should be routed to the DCM reset input port.
Caution! In the example designs for Virtex-4 families, the 200ms reset pulse from the DCM
reset module is NOT connected directly to the DCM: this signal must be connected when
implementing the core in real hardware.
For Virtex-4 FPGA example designs, the reason why the 200ms input is not routed directly
to the DCM is so that simulations can occur in a timely manner.
The 200ms reset pulse takes the signal name reset_200msin the example design HDL
files. It is routed from the DCM reset module instantiation up through all levels of
hierarchy to the top level. It is then left unconnected in the example design instantiation
from within the demonstration test bench. Furthermore, an accompanying signal,
reset_200ms_in, is routed down from the top level of the example design hierarchy to
the DCM instantiation, where it is connected to the DCM reset. From the demonstration
test bench, this reset_200ms_insignal is driven for a short duration to enable fast
simulation start up. However, to re-iterate, when implementing the design in real
hardware, the DCM reset signal must be connected correctly:
•
This can be achieved by connecting the reset_200mssignal to the
reset_200ms_insignal at any level of example design HDL hierarchy.
64
1-Gigabit Ethernet MAC v8.5 User Guide
UG144 April 24, 2009
Download from Www.Somanuals.com. All Manuals Search And Download.
-- DISCONTINUED PRODUCT --
R
Implementing External GMII
Virtex-5 Devices
An IODELAY component may be used on the clock, data and control paths, as illustrated
and control signals to meet the setup and hold requirements of GMII and to allow for any
bus skew across the data and control inputs.The IODELAY components are used in fixed
delay mode, where the attribute IDELAY_VALUE determines the tap delay value. An
IDELAYCTRL primitive must be instantiated for this mode of operation. See the Virtex-5
FPGA User Guide for more information on the use of IDELAYCTRL and IODELAY
components.
IOB LOGIC
BUFG
IBUFG
gmii_rx_clk
IPAD
IODELAY
1-Gigabit Ethernet MAC LogiCORE
gmii_rx_clk
IBUF
IBUF
IBUF
gmii_rxd[0]
IPAD
gmii_rxd_reg[0]
gmii_rx_dv_reg
gmii_rx_er_reg
IODELAY
IODELAY
IODELAY
gmii_rxd[0]
gmii_rx_dv
gmii_rx_er
Q
Q
Q
D
D
D
gmii_rx_dv
IPAD
gmii_rx_er
IPAD
Figure 7-3: External GMII Receiver Logic for Virtex-5 Devices
1-Gigabit Ethernet MAC v8.5 User Guide
65
UG144 April 24, 2009
Download from Www.Somanuals.com. All Manuals Search And Download.
-- DISCONTINUED PRODUCT --
Chapter 7: Using the Physical Side Interface
R
Implementing External RGMII
The HDL example design delivered with the core implements an external RGMII when
For more information about using the example design, see the 1-Gigabit Ethernet MAC
Getting Started Guide.
RGMII Transmitter Logic
Spartan-3, Spartan-3E, Spartan-3A and Spartan-3A DSP Devices
Figure 7-4 illustrates how to use the physical transmitter interface of the core to create an
external RGMII in a Spartan-3 device. The signal names and logic precisely match those
delivered with the example design when the RGMII is selected. If other families are used,
equivalent primitives and logic specific to that family is used in the example design.
IOB LOGIC
FDDRRSE
BUFG
IOB LOGIC
rgmii_tx_clk_bufg
gtx_clk_bufg
DCM
CLK90
CLK0
'1'
D
D
Q
Q
IBUFG
OBUF
gtx_clk
IPAD
rgmii_txc
OPAD
CLKIN
FB
'0'
IOB LOGIC
FDDRRSE
1-Gigabit Ethernet MAC Core
gtx_clk
gmii_txd_int[0]
D
D
Q
Q
gmii_txd[0]
D
Q
OBUF
rgmii_txd[0]
OPAD
gmii_txd_int[4]
gmii_txd[4]
D
Q
D
Q
FDDRRSE
gmii_tx_en_int
gmii_tx_er_int
D
Q
Q
gmii_tx_en
gmii_tx_er
D
Q
OBUF
rgmii_tx_ctl
OPAD
D
Q
D
D
Q
Figure 7-4: External RGMII Transmitter Logic
66
1-Gigabit Ethernet MAC v8.5 User Guide
UG144 April 24, 2009
Download from Www.Somanuals.com. All Manuals Search And Download.
-- DISCONTINUED PRODUCT --
R
Implementing External RGMII
the FPGA fabric, including the encoded rgmii_tx_ctl_intsignal, derived from the
logical xorof gmii_tx_en_intand gmii_tx_er_int. The signals to be transmitted on
the RGMII falling clock edge are then registered on the falling edge of this clock. This
ensures that the data is presented to the Double Data Rate registers at the correct time.
Finally, the transmitter signals are registered by an IOB output Double-Data-Rate (DDR)
register before being driven to output pads.
The logic required to forward the transmitter clock is also shown. This uses an IOB output
DDR register so the clock signal produced incurs on exactly the same delay as the data and
control signals. The rgmii_tx_clk clock signal is phase-shifted by 90 degrees in the
DCM with respect to gtx_clk_bufg. This means that the rising edge of rgmii_txc
occurs in the center of the data valid window—which maximizes setup and hold times
across the interface, as specified in the Reduced Gigabit Media Independent Interface (RGMII)
Version 2.0 specification.
Virtex-4 Devices
Figure 7-5 shows using the physical transmitter interface of the core to create an external
RGMII in a Virtex-4 device. The signal names and logic shown exactly match those
delivered with the example design when the RGMII is selected
Figure 7-5 also shows that the output transmitter signals are registered in the IOBs in
ODDR components. These components convert the input signals into one double-data-rate
signal. These signals are then output through OBUFs before being driven to output pads.
1-Gigabit Ethernet MAC v8.5 User Guide
67
UG144 April 24, 2009
Download from Www.Somanuals.com. All Manuals Search And Download.
-- DISCONTINUED PRODUCT --
Chapter 7: Using the Physical Side Interface
R
IOB LOGIC
ODDR
ODDR
ODDR
BUFGMUX
rgmii_tx_clk_bufg
IOB LOGIC
DCM
CLK90
CLK0
OBUF
OBUF
OBUF
'1'
'0'
D1
D2
IBUFG
rgmii_txc
OPAD
gtx_clk
IPAD
Q
Q
Q
CLKIN
FB
C
gtx_clk_bufg
IOB LOGIC
1-Gigabit Ethernet MAC Core
gmii_txd_int[0]
gmii_txd_int[4]
gmii_txd[0]
D1
D2
rgmii_txd[0]
OPAD
gmii_txd[4]
gtx_clk
C
IOB LOGIC
gmii_tx_en_int
gmii_tx_er_int
D1
D2
gmii_tx_en
gmii_tx_er
rgmii_tx_ctl
OPAD
C
Figure 7-5: External RGMII Transmitter Logic in Virtex-4 Devices
The logic required to forward the transmitter clock is also shown: this uses an ODDR
register so that the clock signal produced incurs exactly on the same delay as the data and
control signals. The rgmii_tx_clk clock signal is phase-shifted by 90 degrees in the
DCM with respect to gtx_clk_bufg. This means that the rising edge of rgmii_txc
occurs in the center of the data valid window—which maximizes setup and hold times
across the interface, as specified in the RGMII v2.0 specification.
The use of the BUFGMUX shown, with one input connected to the DCM CLK90 output is
included so that a reliable 125MHz clock source is always provided on global routing
(when the DCM is held in reset, the DCM input clock is instead selected). This is required
68
1-Gigabit Ethernet MAC v8.5 User Guide
UG144 April 24, 2009
Download from Www.Somanuals.com. All Manuals Search And Download.
-- DISCONTINUED PRODUCT --
R
Implementing External RGMII
Virtex-5 Devices
devices. However, an alternative solution has been adopted for the example design
core to create an external RGMII in a Virtex-5 device. The signal names and logic shown
exactly match those delivered with the example design when the RGMII is selected.
Figure 7-6 also shows that the output transmitter signals are registered in the IOBs in
ODDR components. These components convert the input signals into one double-data-rate
signal. The ODDR outputs are passed through IODELAYs—and these can be used to
adjust the relationship between the individual signals. These signals are then output
through OBUFs before being driven to output pads.
IOB LOGIC
ODDR
OBUF
D1
D2
IOB LOGIC
‘1’
‘0’
rgmii_txc
OPAD
IODELAY
Q
IBUFG
gtx_clk
IPAD
gtx_clk_bufg
C
IOB LOGIC
ODDR
1-Gigabit Ethernet MAC Core
gmii_txd_int[0]
gmii_txd[0]
D1
D2
OBUF
rgmii_txd[0]
OPAD
gmii_txd_int[4]
gmii_txd[4]
gtx_clk
Q
IODELAY
C
IOB LOGIC
ODDR
gmii_tx_en_int
gmii_tx_er_int
OBUF
gmii_tx_en
gmii_tx_er
D1
D2
rgmii_tx_ctl
OPAD
Q
IODELAY
C
Figure 7-6: External RGMII Transmitter Logic in Virtex-5 Devices
1-Gigabit Ethernet MAC v8.5 User Guide
69
UG144 April 24, 2009
Download from Www.Somanuals.com. All Manuals Search And Download.
-- DISCONTINUED PRODUCT --
Chapter 7: Using the Physical Side Interface
R
The logic required to forward the transmitter clock is also shown. It has matching logic to
the data and control signals to provide a known relationship between the signals. An
IODELAY component is used to phase-shift the rgmii_txc clock signal by 90 degrees
with respect to gtx_clk_bufg. This allows the rising edge of rgmii_txc to occur in the
center of the data valid window—which maximizes setup and hold times across the
interface, as specified in the RGMII v2.0 specification. The IODELAY component is used in
fixed delay mode, where the attribute ODELAY_VALUEdetermines the tap delay value. An
IDELAYCTRL primitive must be instantiated for this mode of operation. See the Virtex-5
User Guide for more information on the use of IDELAYCTRL and IODELAY components.
RGMII Receiver Logic
Spartan-3, Spartan-3E, Spartan-3A and Spartan-3A DSP Devices
Figure 7-7 shows using the physical receiver interface of the core to create an external
RGMII in a Spartan-3 device. The signal names and logic exactly match those delivered
with the example design when the RGMII is selected. If other families are used, equivalent
primitives and logic specific to that family is used in the example design.
Figure 7-7 also shows that the input receiver signals are registered in device IOBs on rising
and falling edges of gmii_rx_clk_bufg. The signals are then registered inside the FPGA
fabric, before a final register stage to synchronize signals to the rising edge clock. To
achieve the required setup and hold times across the interface, the DCM uses a phase-shift
to adjust the clock relative to the data. See Appendix C, “Calculating DCM Phase-
DCM Reset circuitry
example design next to the DCM. Since this logic must be reliable whatever the
reset/locked status of the DCM, the module requires a reliable reference clock. In the
example design for RGMII, a transmitter clock source is therefore used for this receiver
DCM.
This reset circuitry will generate an appropriate reset pulse for the receiver DCM of
Figure 7-7 under the following conditions:
•
The lockedsignal from the DCM is constantly monitored. Following a high to low
transition on this signal, indicating that the DCM has lost lock, a reset will be issued.
•
A timeout counter is enabled when the DCM is in the loss of lock state. If, following
the timeout period, the DCM has not obtained lock, another DCM reset will be issued.
This timeout counter will time a > 1ms interval. This timeout functionality is required
for DCMs connected to Ethernet PHYs since the PHYs may source discontinuous
clocks under certain network conditions (for example, when no ethernet cable is
connected).
For Spartan-3 families, the reset pulse is transferred into the DCM input clock (rgmii_rxc
the reset input of the DCM.
70
1-Gigabit Ethernet MAC v8.5 User Guide
UG144 April 24, 2009
Download from Www.Somanuals.com. All Manuals Search And Download.
-- DISCONTINUED PRODUCT --
R
Implementing External RGMII
IOB LOGIC
BUFG
IBUFG
DCM
CLKIN
rgmii_rxc
IPAD
CLK0
1-Gigabit Ethernet MAC Core
FB
gmii_rx_clk_bufg
gmii_rx_clk
IOB LOGIC
IBUF
rgmii_rxd_ddr[0]
rgmii_rxd_reg[0]
rgmii_rxd[0]
IPAD
gmii_rxd_reg[0]
gmii_rxd[0]
Q
Q
Q
D
D
D
Q
D
rgmii_rxd_ddr[4]
rgmii_rxd_reg[4]
Q
D
Q
D
gmii_rxd_reg[4]
gmii_rxd[4]
IBUF
rgmii_rx_dv_reg rgmii_rx_dv_ddr
rgmii_rx_ctl
IPAD
gmii_rx_dv_reg
gmii_rx_er_reg
gmii_rx_dv
gmii_rx_er
Q
Q
D
D
Q
D
rgmii_rx_ctl_ddr
rgmii_rx_ctl_reg
Q
Q
D
D
Q
D
Figure 7-7: External RGMII Receiver Logic
1-Gigabit Ethernet MAC v8.5 User Guide
UG144 April 24, 2009
71
Download from Www.Somanuals.com. All Manuals Search And Download.
-- DISCONTINUED PRODUCT --
Chapter 7: Using the Physical Side Interface
R
Virtex-4 Devices
Figure 7-8 shows using the physical receiver interface of the core to create an external
RGMII in a Virtex-4 device. The signal names and logic exactly match those delivered with
the example design when RGMII is selected.
Figure 7-8 also shows that the input receiver signals are registered in the IOBs in IDDR
components. These components convert the input double data rate signals into GMII
specification signals. The gmii_rx_er_intsignal is derived in the FPGA fabric from
the outputs of the control IDDR component.
To achieve the required setup and hold times across the interface, the DCM uses a phase-
shift to adjust the clock relative to the data. See Appendix C, “Calculating DCM Phase-
DCM Reset circuitry
example design next to the DCM. Since this logic must be reliable whatever the
reset/locked status of the DCM, the module requires a reliable reference clock. In the
example design for RGMII, a transmitter clock source is therefore used for this receiver
DCM. This is obtained from the BUFGMUX output which is connected to the DCM CLK90
clock is always present on this global clock route, even when the RGMII transmitter DCM
This reset circuitry will generate an appropriate reset pulse for the receiver DCM of
Figure 7-8 under the following conditions:
•
The lockedsignal from the DCM is constantly monitored. Following a high to low
transition on this signal, indicating that the DCM has lost lock, a reset will be issued.
•
A timeout counter is enabled when the DCM is in the loss of lock state. If, following
the timeout period, the DCM has not obtained lock, another DCM reset will be issued.
This timeout counter will time a > 1ms interval. This timeout functionality is required
for DCMs connected to Ethernet PHYs since the PHYs may source discontinuous
clocks under certain network conditions (for example, when no ethernet cable is
connected).
For Virtex-4 families, the generated reset from within the DCM reset module must be
asserted for a minimum of 200 ms (see the Virtex-4 Datasheet). Consequently, a 200 ms
duration timer is also included in the DCM reset module to extend the reset pulse. This
reset signal is output from the DCM reset module and should be routed to the DCM reset
input port.
Caution! In the example designs for Virtex-4 families, the 200ms reset pulse from the DCM
reset module is NOT connected directly to the DCM: this signal must be connected when
implementing the core in real hardware. The reason why the 200ms input is not routed directly to
the DCM is so that simulations can occur in a timely manner.
The 200ms reset pulse takes the signal name reset_200msin the example design HDL
files. It is routed from the DCM reset module instantiation up through all levels of
hierarchy to the top level. It is then left unconnected in the example design instantiation
from within the demonstration test bench. Furthermore, an accompanying signal,
reset_200ms_in, is routed down from the top level of the example design hierarchy to
the DCM instantiation, where it is connected to the DCM reset. From the demonstration
test bench, this reset_200ms_insignal is driven for a short duration to enable fast
simulation start up. However, to re-iterate, when implementing the design in real
hardware, the DCM reset signal must be connected correctly:
72
1-Gigabit Ethernet MAC v8.5 User Guide
UG144 April 24, 2009
Download from Www.Somanuals.com. All Manuals Search And Download.
-- DISCONTINUED PRODUCT --
R
Implementing External RGMII
•
This can be achieved by connecting the reset_200mssignal to the
reset_200ms_insignal at any level of example design HDL hierarchy.
IOB LOGIC
BUFG
IBUFG
DCM
rgmii_rxc
IPAD
CLKIN
CLK0
FB
1-Gigabit Ethernet MAC Core
gmii_rx_clk_bufg
gmii_rx_clk
IOB LOGIC
IDDR
gmii_rxd_int[0]
gmii_rxd_int[4]
IBUF
Q1
Q2
gmii_rxd[0]
gmii_rxd[4]
rgmii_rxd[0]
IPAD
D
C
IOB LOGIC
IDDR
gmii_rx_dv_int
gmii_rx_er_int
IBUF
gmii_rx_dv
gmii_rx_er
Q1
Q2
rgmii_rx_ctl
IPAD
D
C
Figure 7-8: External RGMII Receiver Logic for Virtex-4 Devices
1-Gigabit Ethernet MAC v8.5 User Guide
73
UG144 April 24, 2009
Download from Www.Somanuals.com. All Manuals Search And Download.
-- DISCONTINUED PRODUCT --
Chapter 7: Using the Physical Side Interface
R
Virtex-5 Devices
Figure 7-9 shows using the physical receiver interface of the core to create an external
RGMII in a Virtex-5 device. The signal names and logic exactly match those delivered with
the example design when RGMII is selected.
Figure 7-9 also shows that the input receiver signals are registered in the IOBs in IDDR
components. These components convert the input double data rate signals into GMII
specification signals. The gmii_rx_er_intsignal is derived in the FPGA fabric from the
outputs of the control IDDR component.
IOB LOGIC
BUFG
IBUFG
IODELAY
rgmii_rxc
IPAD
1-Gigabit Ethernet MAC Core
gmii_rx_clk_bufg
gmii_rx_clk
IOB LOGIC
IDDR
gmii_rxd_int[0]
gmii_rxd_int[4]
IBUF
Q1
Q2
gmii_rxd[0]
gmii_rxd[4]
IODELAY
rgmii_rxd[0]
D
IPAD
C
IOB LOGIC
IDDR
gmii_rx_dv_int
gmii_rx_er_int
IBUF
gmii_rx_dv
gmii_rx_er
Q1
Q2
IODELAY
rgmii_rx_ctl
IPAD
D
C
Figure 7-9: External RGMII Receiver Logic for Virtex-5 Devices
The IODELAY components are used to phase-shift the input RGMII clock, data and control
signals to meet the setup and hold margins. The IODELAY components are used in fixed
delay mode, where the attribute IDELAY_VALUEdetermines the tap delay value. An
IDELAYCTRL primitive must be instantiated for this mode of operation. See the Virtex-5
User Guide for more information on the use of IDELAYCTRL and IODELAY components.
74
1-Gigabit Ethernet MAC v8.5 User Guide
UG144 April 24, 2009
Download from Www.Somanuals.com. All Manuals Search And Download.
-- DISCONTINUED PRODUCT --
R
Implementing External RGMII
RGMII Inband Status Decoding Logic
the decoding of RGMII inband status information. This information is received through
the RGMII interface between frames. The signal names and logic shown exactly match
those delivered with the example design when the RGMII is selected.
OBUF
inband_link_status
1-Gigabit Ethernet MAC Core
D
CE
Q
Q
Q
Q
OPAD
OBUF
OBUF
OBUF
inband_clock_speed[0]
OPAD
D
CE
inband_clock_speed[1]
OPAD
D
CE
inband_duplex_status
OPAD
D
CE
gmii_rx_clk_bufg
gmii_rxd_reg[0]
gmii_rx_clk
gmii_rxd[0]
RGMII RECEIVER LOGIC
gmii_rxd_reg[1]
gmii_rxd_reg[2]
gmii_rxd[1]
gmii_rxd[2]
gmii_rxd_reg[3]
gmii_rxd[3]
gmii_rx_dv_reg
gmii_rx_er_reg
gmii_rx_dv
gmii_rx_er
Figure 7-10: RGMII Inband Status Decoding Logic
1-Gigabit Ethernet MAC v8.5 User Guide
75
UG144 April 24, 2009
Download from Www.Somanuals.com. All Manuals Search And Download.
-- DISCONTINUED PRODUCT --
Chapter 7: Using the Physical Side Interface
R
Using the MDIO interface
The MDIO interface is accessed through the optional management interface and is
typically connected to the MDIO port of a physical-layer device to access its configuration
IEEE 802.3, clause 22.
Connecting the MDIO to an Internally Integrated PHY
The MDIO ports of the GEMAC core can be connected to the MDIO ports of an internally
integrated physical-layer device, such as the MDIO port of the Xilinx Ethernet 1000BASE-
information.
Connecting the MDIO to an External PHY
The MDIO ports of the GEMAC core can be connected to the MDIO of an external
physical-layer device. In this situation, mdio_in, mdio_out, and mdio_trimust be
connected to a tri-state buffer to create a bidirectional wire, mdio. This tri-state buffer can
be either external to the FPGA, or internally integrated by using an IOB IOBUF component
IOB LOGIC
OBUF
1-Gigabit Ethernet MAC Core
mdc
OPAD
MDC
I
O
IOB LOGIC
IOBUF
mdio_tri
mdio_out
mdio_in
T
IOPAD
MDIO
I
IO
O
Figure 7-11: Creating an External MDIO Interface
76
1-Gigabit Ethernet MAC v8.5 User Guide
UG144 April 24, 2009
Download from Www.Somanuals.com. All Manuals Search And Download.
-- DISCONTINUED PRODUCT --
R
Chapter 8
Configuration and Status
This chapter provides general guidelines for configuring and monitoring the GEMAC
core, including a detailed description of the client-side management interface and registers
present in the core. It also describes the alternative to the optional management interface
which is the Configuration Vector.
Using the Optional Management Interface
The Management Interface is a processor-independent interface with standard address,
data, and control signals. It may be used as is, or a wrapper (not supplied) may be used to
interface to common bus architectures. For port definition, see “Management Interface
This interface is used for:
•
•
Configuring of the GEMAC core via the configuration registers.
Access through the MDIO interface to the management registers located in the PHY
connected to the GEMAC core.
The Management Interface can be accessed in different ways, depending on the type of
transaction. A truth table showing which access method is required for each transaction
Table 8-1: Management Interface Transaction Types
Transaction
Configuration
host_miim_sel
host_addr[9]
0
1
1
MIIM access
X
Host Clock Frequency
The Management Interface clock, host_clk, is used to derive the MDIO clock, mdc, and
for this reason is subject to the following frequency restriction:
≥ 10 MHz
1-Gigabit Ethernet MAC v8.5 User Guide
77
UG144 April 24, 2009
Download from Www.Somanuals.com. All Manuals Search And Download.
-- DISCONTINUED PRODUCT --
R
Chapter 8: Configuration and Status
Configuration Registers
After a power-up or system reset, the client may reconfigure the core parameters using
their defaults. Configuration changes can be written at any time. Both the receiver and
transmitter logic responds only to configuration changes during inter-frame gaps. The
exceptions to this are the configurable resets that take effect immediately.
Configuration of the GEMAC core is performed through a register bank that is accessed
through the management interface. The configuration registers available in the core are
address in the ranges shown performs a 32-bit read or write from the same configuration
word.
Table 8-2: Configuration Registers
Address
Description
Receiver Configuration (Word 0)
0x200-0x23F
0x240-0x27F
0x280-0x2BF
0x2C0-0x2FF
0x300-0x33F
0x340-0x37F
0x380-0x383
0x384-0x387
0x388-0x38B
Receiver Configuration (Word 1)
Transmitter Configuration
Flow Control Configuration
Reserved
Management Configuration
Unicast Address (Word 0) (if Address Filter is present)
Unicast Address (Word 1) (if Address Filter is present)
Address Table Configuration (Word 0) (if Address Filter is
present)
0x38C-0x38F
0x390-0x393
Address Table Configuration (Word 1) (if Address Filter is
present)
Address Filter Mode (if Address Filter is present)
78
1-Gigabit Ethernet MAC v8.5 User Guide
UG144 April 24, 2009
Download from Www.Somanuals.com. All Manuals Search And Download.
-- DISCONTINUED PRODUCT --
Using the Optional Management Interface
R
Receiver Configuration
Table 8-3: Receiver Configuration Word 0
Default
Value
Bit
Description
31-0
All 0s
Pause frame MAC Source Address[31:0]. This address is
used by the MAC to match against the destination address
of any incoming flow control frames. It is also used by the
flow control block as the source address (SA) for any
outbound flow control frames. See Chapter 6, “Using Flow
The address is ordered so the first byte
transmitted/received is the lowest positioned byte in the
register; for example, a MAC address of AA-BB-CC-DD-
EE-FF would be stored in Address[47:0] as
0xFFEEDDCCBBAA.
Table 8-4: Receiver Configuration Word 1
Default
Value
Bit
Description
15-0
23-16
24
All 0s
n/a
0
Pause frame MAC Source Address[47:32]
Reserved
Control Frame Length Check Disable When this bit is set
to ‘1,’ the core will not mark control frames as ‘bad’ if they
are greater than the minimum frame length.
25
0
Length/Type Error Check Disable When this bit is set to
‘1,’ the core will not perform the length/type field error
checks as described in “Length/Type Field Error Checks,”
on page 43. When this bit is set to ‘0,’ the length/type field
checks will be performed; this is normal operation.
26
27
n/a
0
Reserved
VLAN Enable When this bit is set to ‘1,’ VLAN tagged
frames will be accepted by the receiver.
28
1
Receiver Enable. If set to ‘1,’ the receiver block will be
operational. If set to ‘0,’ the block will ignore activity on the
physical interface RX port.
1-Gigabit Ethernet MAC v8.5 User Guide
79
UG144 April 24, 2009
Download from Www.Somanuals.com. All Manuals Search And Download.
-- DISCONTINUED PRODUCT --
R
Chapter 8: Configuration and Status
Table 8-4: Receiver Configuration Word 1 (Continued)
Default
Value
Bit
Description
29
30
31
0
In-band FCS Enable When this bit is ‘1,’ the MAC receiver
will pass the FCS field up to the client. When at ‘0,’ the
client will not be passed the FCS. In both cases, the FCS
will be verified on the frame.
0
0
Jumbo Frame Enable When this bit is set to ‘1,’ the MAC
receiver will accept frames over the specified IEEE 802.3-
2005 maximum legal length. When this bit is ‘0,’ the MAC
will only accept frames up to the specified maximum.
Reset When this bit is set to ‘1,’ the receiver will be reset.
The bit will then automatically revert to ‘0.’ This reset also
sets all of the receiver configuration registers to their
default values.
Transmitter Configuration
Table 8-5: Transmitter Configuration Word
Default
Value
Bit
Description
24-0
25
n/a
Reserved
0
Interframe Gap Adjust Enable If ‘1,’ the transmitter will
read the value on the port tx_ifg_delay at the start of frame
transmission and adjust the interframe gap following the
frame accordingly.
26
27
n/a
0
Reserved
VLAN Enable When this bit is set to ‘1,’ the transmitter
will allow the transmission of VLAN tagged frames.
28
1
Transmit Enable When this bit is ‘1,’ the transmitter is
operational. When it is ‘0,’ the transmitter is disabled.
80
1-Gigabit Ethernet MAC v8.5 User Guide
UG144 April 24, 2009
Download from Www.Somanuals.com. All Manuals Search And Download.
-- DISCONTINUED PRODUCT --
Using the Optional Management Interface
R
Table 8-5: Transmitter Configuration Word (Continued)
Default
Value
Bit
Description
29
0
0
0
In-band FCS Enable When this bit is ‘1,’ the MAC
transmitter will expect the FCS field to be passed in by the
client. When this bit is ‘0,’ the MAC transmitter will
append padding as required, compute the FCS and
append it to the frame.
30
31
Jumbo Frame Enable When this bit is set to ‘1,’ the MAC
transmitter will send frames that are greater than the
specified IEEE 802.3-2005 maximum legal length. When
this bit is ‘0,’ the MAC will only send frames up to the
specified maximum.
Reset When set to ‘1,’ the transmitter will be reset. The bit
will then automatically revert to ‘0.’ This reset will also set
all of the transmitter configuration registers to their default
values.
Flow Control Configuration
Table 8-6: Flow Control Configuration Word
Default
Bit
Description
Value
n/a
1
28-0
29
Reserved
Flow Control Enable (RX) When this bit is ‘1,’ received
flow control frames will inhibit the transmitter operation.
When this bit is ‘0,’ received flow control frames will
always be passed up to the client.
30
31
1
Flow Control Enable (TX) When this bit is ‘1,’ asserting the
PAUSE_REQ signal will send a flow control frame out
from the transmitter. When this bit is ‘0,’ asserting the
PAUSE_REQ signal has no effect.
n/a
Reserved
1-Gigabit Ethernet MAC v8.5 User Guide
81
UG144 April 24, 2009
Download from Www.Somanuals.com. All Manuals Search And Download.
-- DISCONTINUED PRODUCT --
R
Chapter 8: Configuration and Status
MDIO Configuration
Table 8-7: Management Configuration Word
Default
Value
Bits
Description
4-0
5
All 0s
Clock Divide[4:0] This value enters a logical equation
which enables the mdc frequency to be set as a divided
down ratio of the host_clk frequency.
0
MDIO Enable When this bit is ‘1,’ the MDIO interface can
be used to access attached PHY devices. When this bit is
‘0,’ the MDIO interface is disabled and the MDIO signals
remain inactive.
31-6
n/a
Reserved
Address Filter Configuration
configuration when the GEMAC core implemented with an Address Filter. The register
Table 8-8: Unicast Address Word 0
Default
Value
Bits
Description
31-0
All 0s
Address filter unicast address[31:0].
The address is ordered so the first byte received is the
lowest positioned byte in the register; for example, a MAC
address of AA-BB-CC-DD-EE-FF would be stored in
Address[47:0] as 0xFFEEDDCCBBAA.
Table 8-9: Unicast Address Word 1
Default
Value
Bits
Description
15-0
All 0s
N/A
Address filter unicast address[47:32].
Reserved
31-16
82
1-Gigabit Ethernet MAC v8.5 User Guide
UG144 April 24, 2009
Download from Www.Somanuals.com. All Manuals Search And Download.
-- DISCONTINUED PRODUCT --
Using the Optional Management Interface
R
The Address Filter can be programmed to respond to four separate additional addresses
contents of the address table are set.
Table 8-10: Address Table Configuration Word 0
Default
Value
Bits
31–0
Description
All 0s
MAC Address[31:0].
The address is ordered so the first byte received is the
lowest positioned byte in the register; for example, a MAC
address of AA-BB-CC-DD-EE-FF would be stored in
Address[47:0] as 0xFFEEDDCCBBAA.
Table 8-11: Address Table Configuration Word 1
Default
Value
Bits
Description
15–0
All 0s
All 0s
MAC Address[47:32].
17–16
The location in the address table that the MAC address is
to be written to or read from. There are up to 4 entries in
the table (Location 0 to 3).
22–18
23
N/A
0
Reserved
Read not write This bit is set to ‘1’ to read from the address
table. If it is set to ‘1,’ the contents of the table entry that is
being accessed by the bits 17-16 will be output on the
hostrddata bus in consecutive cycles (least significant
word first). If it is set to ‘0,’ the data on bits 15-0 is written
into the table at the address specified by bits 17-16.
31–24
N/A
Reserved
Table 8-12: Address Filter Mode
Default
Bits
Description
Value
N/A
0
30–0
31
Reserved
Promiscuous Mode If this bit is set to ‘1,’ the Address Filter
operates in promiscuous mode. All frames are passed to
the receiver client, regardless of the destination address.
Writing and Reading to and from the Configuration Registers
Writing to the configuration registers through the management interface is shown in
host_miim_sel= ‘0’), the upper bit of host_opcodefunctions as an active low write
enable signal. The lower host_opcodebit is a don’t care bit.
1-Gigabit Ethernet MAC v8.5 User Guide
83
UG144 April 24, 2009
Download from Www.Somanuals.com. All Manuals Search And Download.
-- DISCONTINUED PRODUCT --
R
Chapter 8: Configuration and Status
host_clk
host_miim_sel
host_opcode[1]
host_addr[8:0]
host_addr[9]
host_wr_data[31:0]
Figure 8-1: Configuration Register Write Timing
Reading from the configuration register words is similar, but the upper host_opcodebit
host_rd_datathe host_clkedge after the register address is asserted onto
host_addr.
host_clk
host_miim_sel
host_opcode[1]
host_addr[8:0]
host_addr[9]
host_rd_data[31:0]
Figure 8-2: Configuration Register Read Timing
84
1-Gigabit Ethernet MAC v8.5 User Guide
UG144 April 24, 2009
Download from Www.Somanuals.com. All Manuals Search And Download.
-- DISCONTINUED PRODUCT --
Using the Optional Management Interface
R
Accessing the Address Table
To write to a specific entry in the address table, you must first write the least significant 32-
bits of the address into the Address Table Configuration (Word 0) register. You then write
the most significant 16 bits together with the location in the table (bits 17–16) to the
Address Table Configuration (Word1) register with bit 23 (read not write) set to ‘0.’ This is
writes to be on adjacent cycles.
hostclk
hostmiimsel
hostopcode[1]
0x188
0x18C
hostaddr[8:0]
hostaddr[9]
ADDR[31:0]
hostwrdata[31:0]
BIT31
BIT0
BITS15..0 = ADDR[47:32]
BITS17..16 = LOCATION
BIT23 = 0
Figure 8-3: Address Table Write Timing
1) with the location set to the desired table entry and bit 23 set to ‘1’ to read from the
address table. On the next cycle the least significant word appears on the hostrddata
bus. One cycle afterwards, the most significant 16-bits are output on the lower 16 bits of the
bus.
1-Gigabit Ethernet MAC v8.5 User Guide
85
UG144 April 24, 2009
Download from Www.Somanuals.com. All Manuals Search And Download.
-- DISCONTINUED PRODUCT --
R
Chapter 8: Configuration and Status
hostclk
hostmiimsel
hostopcode[1]
0x18C
hostaddr[8:0]
hostaddr[9]
hostwrdata[23]
hostwrdata[17:16]
hostrddata[31:0]
LOCATION
31 : 0
47 : 32
Figure 8-4: Address Table Read Timing
MDIO Interface
Introduction to MDIO
The MDIO interface for 1 Gbps operation (and slower speeds) is defined in IEEE 802.3
clause 22. This is a two wire interface consisting of a clock, mdc, and a shared serial data
line, mdio.
An MDIO bus in a system consists of a single Station Management (STA) master
management entity and a number of MDIO Managed Device (MMD) slave entities.
GEMAC core implements a STA and can be connected to MMDs (PHY devices) to access
their management registers.
86
1-Gigabit Ethernet MAC v8.5 User Guide
UG144 April 24, 2009
Download from Www.Somanuals.com. All Manuals Search And Download.
-- DISCONTINUED PRODUCT --
Using the Optional Management Interface
R
MAC 1
MAC 2
mdio
mdc
STA
MMD
MMD
MMD
MMD
MMD
MMD
Figure 8-5: Typical MDIO-managed System
There are two different transaction types of MDIO for write and read. They are described
in this section.
Abbreviations Used
The following abbreviations apply for the remainder of this chapter.
•
•
•
•
•
•
PRE Preamble
ST Start of frame
OP Operation code
PHYAD PHY address
REGAD Register address
TA Turnaround
Write Transaction
Figure 8-6 shows a write transaction across the MDIO, as defined by OP=’01.’ The
addressed MMD (PHYAD) device takes the 16-bit word in the data field and writes it to the
register at REGAD.
STA drives MDIO
mdc
mdio
Z
Z
1
1
1
0
1
0
1 P4 P3 P2 P1 P0R4R3R2R1R0 1 0 D15 D13 D11 D9
D7
D5
D3
D1
Z
Z
D14 D12 D10 D8
D6
D4
D2
D0
IDLE 32 bits
PRE
ST
OP
PHYAD
REGAD
TA
16-bit WRITE DATA
IDLE
Figure 8-6: MDIO Write Transaction
1-Gigabit Ethernet MAC v8.5 User Guide
87
UG144 April 24, 2009
Download from Www.Somanuals.com. All Manuals Search And Download.
-- DISCONTINUED PRODUCT --
R
Chapter 8: Configuration and Status
Read Transaction
Figure 8-7 shows a Read transaction; this is defined by OP=”10”. The addressed MMD
(PHYAD) device returns the 16-bit word from the register at REGAD.
STA drives MDIO
MMD drives MDIO
mdc
mdio
Z
Z
1
1
1
0
1
1
0 P4 P3 P2 P1 P0R4R3R2R1R0 Z 0 D15 D13 D11 D9
D7
D5
D3
D1
Z
Z
D14 D12 D10 D8
D6
D4
D2
D0
IDLE 32 bits
PRE
ST
OP
PRTAD
REGAD
TA
16-bit READ DATA
IDLE
Figure 8-7: MDIO Read Transaction
For details of the register map of MMD (PHY layer devices) and a detailed description of
the operation of the MDIO Interface itself, see IEEE 802.3-2005.
Accessing MDIO With GEMAC
More information about MDIO with GEMAC can be found in the following sections of this
guide:
•
•
•
For the GEMAC port definition of the MDIO, see “MDIO Interface” in Chapter 2
The management interface is also used to access the MDIO interface of the GEMAC core.
The MDIO interface supplies a clock to the connected PHY, mdc. This clock is derived from
the host_clksignal using the value in the Clock Divide[4:0]configuration register.
The frequency of mdcis given by the following equation:
fHOST_CLK
fMDC = -------------------------------------------------------------------
(1 + Clock Divide[4:0]) × 2
The frequency of mdcgiven by this equation should not exceed 2.5 MHz to comply with
the IEEE 802.3-2005 specification for this interface. To prevent mdcfrom being out of
specification, the Clock Divide[4:0]value powers up at 00000. While this value is in
the register, it is impossible to enable the MDIO interface.
For details of the register map of PHY layer devices and a detailed description of the
operation of the MDIO interface itself, see IEEE 802.3-2005.
88
1-Gigabit Ethernet MAC v8.5 User Guide
UG144 April 24, 2009
Download from Www.Somanuals.com. All Manuals Search And Download.
-- DISCONTINUED PRODUCT --
Using the Optional Management Interface
R
Figure 8-8 shows access to the MDIO interface through the Management Interface.
host_clk
host_miim_sel
host_req
host_opcode1:0]
host_addr9:0]
host_wr_data15:0]
*
host_rdy
host_rd_data[15:0]
*
* If a read transaction is initiated, the host_rd_data bus is valid
at the point indicated. If a write transaction is initiated, the
host_wr_data bus must be valid at the indicated point.
Simultaneous read and write is not permitted.
Figure 8-8: MDIO Access through Management Interface
For MDIO transactions, the following applies:
• host_miim_selis ‘1’
• host_opcode[1:0]maps to the OP (opcode) field of the MDIO frame
• host_addrmaps to the two address fields of the MDIO frame; PHYAD is
host_addr[9:5], and REGAD is host_addr[4:0]
• host_wr_data[15:0]maps into the data field of the MDIO frame when
performing a write operation
•
The data field of the MDIO frame maps into host_rd_data[15:0]when
performing a read operation
The GEMAC core signals the host that it is ready for an MDIO transaction by asserting
host_miim_rdy. A read or write transaction on the MDIO is initiated by a pulse on the
host_reqsignal. This pulse is ignored if the MDIO interface already has a transaction in
progress. The GEMAC core then deasserts the host_miim_rdysignal while the
transaction across the MDIO is in progress. When the transaction across the MDIO
interface has been completed, the host_miim_rdysignal is asserted by the GEMAC core;
if the transaction is a read, the data is available on the host_rd_data[15:0]bus at this
time.
1-Gigabit Ethernet MAC v8.5 User Guide
89
UG144 April 24, 2009
Download from Www.Somanuals.com. All Manuals Search And Download.
-- DISCONTINUED PRODUCT --
R
Chapter 8: Configuration and Status
Access without the Management Interface
If the optional management interface is omitted from the core, all of the relevant
as signals. These signals are bundled into the configuration_vector[67:0]signal as
These signals may permanently set by connecting to logic 0 or 1, or may be changed by the
user application at any time; however, with the exception of the reset and the flow control
configuration signals, any changes do not take effect until the current frame has completed
transmission or reception.
registered into before use by the core. It is not necessary to drive the signal from this clock
domain.
Table 8-13: Configuration Vector Bit Definition
Configuration
Bit(s) Register cross
reference
Clock
Description
47:0
“Receiver
gmii_rx_clk Pause frame MAC Source Address[47:0]
This address is used by the GEMAC core to
match against the destination address of any
incoming flow control frames, and as the
source address for any outbound flow control
frames.
and “Receiver
The address is ordered such that the first byte
transmitted or received is the least significant
byte in the register; for example, a MAC
address of AA-BB-CC-DD-EE-FF will be
stored in bite [47:0] as 0xFFEEDDCCBBAA.
48
49
n/a
n/a
This input is unused.
“Receiver
gmii_rx_clk Receiver VLAN Enable When this bit is set to
'1,’ VLAN tagged frames are accepted by the
receiver.
50
51
“Receiver
gmii_rx_clk Receiver Enable If set to '1,’ the receiver
block is operational. If set to '0,’ the block
ignores activity on the physical interface RX
port.
“Receiver
gmii_rx_clk Receiver In-band FCS Enable When this bit
is ‘1,’ the MAC receiver will pass the FCS
field up to the client. When it is ‘0,’ the MAC
receiver will not pass the FCS field. In both
cases, the FCS field will be verified on the
frame.
52
“Receiver
gmii_rx_clk Receiver Jumbo Frame Enable When this bit
is ‘0,’ the receiver will not pass frames longer
than the maximum legal frame size specified
in IEEE 802.3-2005. At ‘1,’ the receiver will not
have an upper limit on frame size.
90
1-Gigabit Ethernet MAC v8.5 User Guide
UG144 April 24, 2009
Download from Www.Somanuals.com. All Manuals Search And Download.
-- DISCONTINUED PRODUCT --
Access without the Management Interface
R
Table 8-13: Configuration Vector Bit Definition (Continued)
Configuration
Bit(s) Register cross
reference
Clock
n/a
Description
53
“Receiver
Receiver Reset. When this bit is ‘1,’ the
receiver is held in reset. This signal is an input
to the reset circuit for the receiver block.
54
“Transmitter
gtx_clk
Transmitter Interframe Gap Adjust Enable
If ‘1,’ then the transmitter will read the value
of the tx_ifg_delay port and set the interframe
gap accordingly. If ‘0,’ the transmitter will
always insert at least the legal minimum
interframe gap.
55
56
n/a
n/a
This input is unused.
“Transmitter
gtx_clk
Transmitter VLAN Enable When this bit is
set to ‘1,’ the transmitter allows the
transmission of VLAN tagged frames.
57
58
“Transmitter
gtx_clk
gtx_clk
Transmitter Enable When this bit is ‘1,’ the
transmitter will be operational. When it is ‘0,’
the transmitter is disabled.
“Transmitter
Transmitter In-Band FCS Enable When this
bit is ‘1,’ the MAC transmitter will expect the
FCS field to be pass in by the client. When it
is ‘0,’ the MAC transmitter will append
padding as required, compute the FCS and
append it to the frame.
59
“Transmitter
gtx_clk
Transmitter Jumbo Frame Enable When this
bit is ‘1,’ the MAC transmitter will allow
frames larger than the maximum legal frame
length specified in IEEE 802.3-2005 to be sent.
When set to ‘0,’ the MAC transmitter will
only allow frames up to the legal maximum
to be sent.
60
61
“Transmitter
n/a
Transmitter Reset. When this bit is ‘1,’ the
MAC transmitter is held in reset. This signal
is an input to the reset circuit for the
transmitter block.
“Flow Control
gtx_clk
Transmit Flow Control Enable. When this
bit is ‘1,’ asserting the pause_req signal
causes the GEMAC core to send a flow
control frame out from the transmitter. When
this bit is ‘0,’ asserting the pause_req signal
will have no effect.
1-Gigabit Ethernet MAC v8.5 User Guide
91
UG144 April 24, 2009
Download from Www.Somanuals.com. All Manuals Search And Download.
-- DISCONTINUED PRODUCT --
R
Chapter 8: Configuration and Status
Table 8-13: Configuration Vector Bit Definition (Continued)
Configuration
Bit(s) Register cross
reference
Clock
gtx_clk
Description
62
63
“Flow Control
Receive Flow Control Enable. When this bit
is ‘1,’ received flow control frames will inhibit
the transmitter operation. When at ‘0,’
received flow frames are passed up to the
client.
“Receiver
gtx_clk
Length/Type Error Check Disable. When
this bit is set to ‘1,’ the core will not perform
the length/type field error checks as
described in “Length/Type Field Error
Checks,” on page 43. When this bit is set to
‘0,’ the length/type field checks will be
performed: this is normal operation.
64
Mode” bit 31
gmii_rx_clk Address Filter Enable. When this bit is ‘0,’
the Address Filter is enabled. If it is set to ‘1,’
the Address Filter will operate in
promiscuous mode.
66:65 n/a
67 “Receiver
n/a
This input is unused.
gtx_clk
Control Frame Length Check Disable When
this bit is set to ‘1,’ the core will not mark
control frames as ‘bad’ if they are greater than
the minimum frame length.
92
1-Gigabit Ethernet MAC v8.5 User Guide
UG144 April 24, 2009
Download from Www.Somanuals.com. All Manuals Search And Download.
-- DISCONTINUED PRODUCT --
R
Chapter 9
Constraining the Core
This chapter defines the GEMAC core constraint requirements. An example UCF that
implements the constraints defined in this chapter is provided with the HDL example
design for the core.
See the 1-Gigabit Ethernet MAC Getting Started Guide for more information about the CORE
Generator™ output files and detailed information about the HDL example design.
Required Constraints
Device, Package, and Speedgrade Selection
The GEMAC can be implemented in Virtex®-4, Virtex-5, Spartan®-3, Spartan-3E, and
Spartan-3A devices with the following attributes:
•
•
•
Large enough to accommodate the core
Contains a sufficient number of IOBs
Operates at the following speed grades:
♦
♦
♦
–4 for Spartan-3, Spartan-3E, and Spartan-3A devices
–10 for Virtex-4 FPGA
–1 for Virtex-5 FPGA
I/O Location Constraints
No specific I/O location constraints are required.
Placement Constraints
No specific placement constraints are required.
Timing Constraints
The core can have up to three separate clock domains: gtx_clkfor the transmitter logic,
gmii_rx_clkfor the receiver logic, and host_clkfor the optional management logic.
These clock nets and the signals within the core that cross these clock domains must be
constrained appropriately in a UCF.
The constraints defined in this section are implemented in the UCF for the example design
delivered with the core. Sections from this UCF are copied into the following descriptions
to provide examples. These examples should be studied in conjunction with the HDL
source code for the example design.
1-Gigabit Ethernet MAC v8.5 User Guide
93
UG144 April 24, 2009
Download from Www.Somanuals.com. All Manuals Search And Download.
-- DISCONTINUED PRODUCT --
R
Chapter 9: Constraining the Core
PERIOD Constraints for Clock Nets
gtx_clk
The clock provided to gtx_clkmust be constrained for a clock frequency of 125 MHz.
The following UCF syntax shows the necessary constraints being applied to the
gtx_clk_bufgsignal (which is routed to the gtx_clk port of the core):
# Set the Transmitter clock period constraints: please do not relax
NET "gtx_clk_bufg"
TIMEGRP "tx_clock"
TIMESPEC "TS_tx_clk"
TNM_NET = "clk_tx";
= "clk_tx";
= PERIOD "tx_clock" 8000 ps HIGH 50 %;
gmii_rx_clk (for GMII Example Designs)
The clock provided to gmii_rx_clkmust be constrained for a clock frequency of 125 MHz.
The following UCF syntax shows the necessary constraints being applied to the clock signal
(which is routed to the gmii_rx_clk port of the core)
# Set the Receiver clock period constraints: please do not relax
NET "gmii_rx_clk" TNM_NET = "gmii_rx_clk";
TIMESPEC "TS_gmii_rx_clk" = PERIOD "gmii_rx_clk" 8000 ps HIGH 50 %;
TIMEGRP "rx_clock"
= "gmii_rx_clk";
rgmii_rxc (for RGMII Example Designs)
The receiver clock provided by the RGMII must be constrained for a clock frequency of 125
MHz.
The following UCF syntax shows the necessary constraints being applied to the clock signal
(which is routed to the gmii_rx_clk port of the core):
# Set the Receiver clock period constraints: please do not relax
NET "gmii_rx_clk_bufg"
TIMEGRP "rx_clock"
TNM_NET = "clk_rx";
= "clk_rx";
TIMESPEC "TS_rx_clk"
= PERIOD "rx_clock" 8000 ps HIGH 50 %;
host_clk
The clock provided to host_clkmust be constrained to the desired frequency within the
The following UCF syntax shows a 100 MHz period constraint being applied to the
host_clksignal (which is routed to the host_clk port of the core):
# Set the Management Clock period constraints: relax as required
NET "host_clk" TNM_NET
TIMEGRP "host"
TIMESPEC "TS_host_clk"
= "host_clk";
= "host_clk" EXCEPT "mdio_logic";
= PERIOD "host" 10000 ps HIGH 50 %;
MDIO Logic
MDIO logic is clocked with host_clk, but is sent a clock enable pulse at the mdc
frequency. To prevent this logic being over constrained by the host_clk period, the
relevant flip-flops for the MDIO logic can be grouped together and removed from the
host_clk period constraint. This is shown in the previous UCF syntax for host_clk.
94
1-Gigabit Ethernet MAC v8.5 User Guide
UG144 April 24, 2009
Download from Www.Somanuals.com. All Manuals Search And Download.
-- DISCONTINUED PRODUCT --
R
Required Constraints
The UCF syntax which follows targets the MDIO logic flip-flops and groups them together.
Reduced clock period constraints are then applied.
############################################################
# MDIO Constraints: please do not edit
#
############################################################
# Place the MDIO logic in it's own timing groups
INST "*gmac_core/BU2/U0/MANIFGEN?MANAGEN/PHY/ENABLE_REG"
INST "*gmac_core/BU2/U0/MANIFGEN?MANAGEN/PHY/READY_INT"
TNM = "mdc_rising";
TNM = "mdc_rising";
INST "*gmac_core/BU2/U0/MANIFGEN?MANAGEN/PHY/STATE_COUNT*" TNM = FFS "mdc_rising";
INST "*gmac_core/BU2/U0/MANIFGEN?MANAGEN/PHY/MDIO_TRISTATE" TNM = "mdc_falling";
INST "*gmac_core/BU2/U0/MANIFGEN?MANAGEN/PHY/MDIO_OUT"
TIMEGRP "mdio_logic" = "mdc_rising" "mdc_falling";
TNM = "mdc_falling";
TIMESPEC "TS_mdio1" = PERIOD "mdio_logic" 400 ns;
TIMESPEC "TS_mdio2" = FROM "mdc_rising" TO "mdc_falling" 200 ns;
Timespecs for Critical Logic within the Core
Signals must cross clock domains at certain points in the core, as described in the following
sections.
Flow Control
Pause requests are received and decoding in the gmii_rx_clkdomain and must be
transferred into the gtx_clkdomain to pause the transmitter. Therefore, whenever
gmii_rx_clk and gtx_clk are derived from different clock sources, the following
constraints must always be applied:
# Flow Control logic reclocking
INST "*gmac_core/BU2/U0/FLOW/RX_PAUSE/GOOD_FRAME_TO_TX"
INST "*gmac_core/BU2/U0/FLOW/RX_PAUSE/PAUSE_REQ_TO_TX"
TNM="flow_rx_to_tx";
TNM="flow_rx_to_tx";
INST "*gmac_core/BU2/U0/FLOW/RX_PAUSE_/AUSE_VALUE_TO_TX*" TNM="flow_rx_to_tx";
TIMESPEC "TS_flow_rx_to_tx" = FROM "flow_rx_to_tx" TO "tx_clock" 8000 ps;
Configuration
When the optional Management Interface is used with the core, configuration information
is written synchronously to host_clk. Receiver configuration data must be transferred
onto the gmii_rx_clk clock domain for use with the receiver and into the gtx_clk
clock domain for use with the transmitter. The following UCF syntax targets this logic and
a timing ignore attribute (TIG) is applied. It does not matter when configuration changes
take place—the current configurations are sampled between frames by both the receiver
and transmitter.
# Configuration Register reclocking
INST "*gmac_core/BU2/U0/MANIFGEN?MANAGEN/CONF/RX0_OUT*"
INST "*gmac_core/BU2/U0/MANIFGEN?MANAGEN/CONF/RX1_OUT*"
INST "*gmac_core/BU2/U0/MANIFGEN?MANAGEN/CONF/FC_OUT_29"
TIMESPEC "TS_config_to_rx" = FROM "config_to_rx" TO "rx_clock" TIG;
TNM="config_to_rx";
TNM="config_to_rx";
TNM="config_to_rx";
INST "*gmac_core/BU2/U0/MANIFGEN?MANAGEN/CONF/TX_OUT*"
INST "*gmac_core/BU2/U0/MANIFGEN?MANAGEN/CONF/FC_OUT_30"
TIMESPEC "TS_config_to_tx" = FROM "config_to_tx" TO "tx_clock" TIG;
TNM="config_to_tx";
TNM="config_to_tx";
1-Gigabit Ethernet MAC v8.5 User Guide
95
UG144 April 24, 2009
Download from Www.Somanuals.com. All Manuals Search And Download.
-- DISCONTINUED PRODUCT --
R
Chapter 9: Constraining the Core
Timespecs for Reset Logic within the Core
Internally, the core is divided into clock/reset domains that group elements with common
clock and reset signals. The reset circuitry for one of these domains is illustrated in
Figure 10-5. This circuit provides controllable skews on the reset nets within the design.
The following UCF syntax identifies the relevant reset logic and groups them together.
Timing constraints are then applied to constrain the skews on the reset nets:
INST "*gmac_core/BU2/U0/SYNC_RX_RESET_I/RESET_OUT" TNM = "reset_dist_grp";
INST "*gmac_core/BU2/U0/SYNC_TX_RESET_I/RESET_OUT" TNM = "reset_dist_grp";
INST "*gmac_core/BU2/U0/G_SYNC_MGMT_RESET?SYNC_MGMT_RESET_HOST_I/RESET_OUT" TNM =
"reset_dist_grp";
TIMESPEC "TS_reset_dist" = FROM "reset_dist_grp" 6100 ps;
Note: The third line in the preceding UCF syntax is only required when the optional Management
Interface is used.
Constraints when Implementing an External GMII
The constraints defined in this section are implemented in the UCF for the example design
delivered with the core. Sections from this UCF are copied into the following descriptions
to provide examples. These examples should be studied in conjunction with the HDL
source code for the example design and with the description “Implementing External
GMII IOB Constraints
The following constraints target the flip-flops that are inferred in the top-level HDL file for
the example design; constraints are set to ensure that these are placed in IOBs.
# GMII Transmitter Constraints: place flip-flops in IOB
INST "*gmii_interface/gmii_txd_reg*"
IOB = true;
INST "*gmii_interface/gmii_tx_en_reg" IOB = true;
INST "*gmii_interface/gmii_tx_er_reg" IOB = true;
# GMII Receiver Constraints: place flip-flops in IOB
INST "*gmii_interface/gmii_rxd_reg*"
IOB = true;
INST "*gmii_interface/gmii_rx_dv_reg" IOB = true;
INST "*gmii_interface/gmii_rx_er_reg" IOB = true;
The GMII is a 3.3 volt signal-level interface which is not the default SelectIO™ standard.
Therefore, use the following constraints with the device IO Banking rules.
INST "gmii_txd<?>"
INST "gmii_tx_en"
INST "gmii_tx_er"
IOSTANDARD = LVTTL;
IOSTANDARD = LVTTL;
IOSTANDARD = LVTTL;
INST "gmii_rxd<?>"
INST "gmii_rx_dv"
INST "gmii_rx_er"
IOSTANDARD = LVTTL;
IOSTANDARD = LVTTL;
IOSTANDARD = LVTTL;
INST "gmii_tx_clk"
INST "gmii_rx_clk"
IOSTANDARD = LVTTL;
IOSTANDARD = LVTTL;
In addition, the example design provides pad locking on the GMII for several families.
This is a provided as a guideline only; there are no specific I/O location constraints for this
core.
96
1-Gigabit Ethernet MAC v8.5 User Guide
UG144 April 24, 2009
Download from Www.Somanuals.com. All Manuals Search And Download.
-- DISCONTINUED PRODUCT --
R
Required Constraints
GMII Input Setup/Hold Timing
signals. This is the worst-case data valid window presented to the FPGA device pins.
GMII_RX_CLK
GMII_RXD[7:0],
GMII_RX_DV,
GMII_RX_ER t
SETUP
tHOLD
Figure 9-1: Input GMII Timing
Observe that there is a 2 ns data valid window which is presented across the GMII input
bus. This must be correctly sampled by the FPGA devices.
Table 9-1: Input GMII Timing
Symbol
Min
2.00
0.00
Max
Units
ns
t
t
-
-
SETUP
HOLD
ns
The following constraints are provided in the UCF for GMII Example Designs. These
constraints invoke the tools to analyze the setup/hold requirements (though, if failing, the
tools are NOT capable of fixing them: meeting these constraints is a manual process - see
the following family specific sections for details):
INST "gmii_rxd<?>" TNM = IN_GMII;
INST "gmii_rx_er" TNM = IN_GMII;
INST "gmii_rx_dv" TNM = IN_GMII;
TIMEGRP "IN_GMII" OFFSET = IN 2.1 ns VALID 2.2 ns BEFORE "gmii_rx_clk";
The constraints defined in the preceding lines have a built-in 10% tolerance: this allows all
combinations of the example designs to pass timing. However, these should be tightened
in a real customer design as per the following syntax. Manual adjustments to IDELAY
values or DCM phase shift requirements may be required to meet these constraints (see the
following family specific sections for details).
TIMEGRP "IN_GMII" OFFSET = IN 2 ns VALID 2 ns BEFORE "gmii_rx_clk";
Spartan-3, Spartan-3E, Spartan-3A and Virtex-4 Devices
The GMII design uses a DCM on the receiver clock domain for Spartan-3, Spartan-3E,
Spartan-3A, and Virtex-4 devices. Phase-shifting is then applied to the DCM to align the
resultant clock so that it will correctly sample the 2 ns GMII data valid window at the input
flip-flops.
1-Gigabit Ethernet MAC v8.5 User Guide
97
UG144 April 24, 2009
Download from Www.Somanuals.com. All Manuals Search And Download.
-- DISCONTINUED PRODUCT --
R
Chapter 9: Constraining the Core
The fixed phase-shift is applied to the DCM using the following UCF syntax:
INST *gmii_interface/gmii_rxc_dcm CLKOUT_PHASE_SHIFT = FIXED;
INST *gmii_interface/gmii_rxc_dcm PHASE_SHIFT = 0;
The value of PHASE_SHIFT is preconfigured in the example designs to meet the setup and
hold constraints for the example GMII pinout in the particular device. The setup/hold
timing which is achieved after place-and-route is reported in the data sheet section of the
TRCE report (created by the implement script).
For customers fixing their own pinout, the setup and hold figures reported in the TRCE
report can be used to initially setup the approximate DCM phase-shift. Appendix C,
“Calculating DCM Phase-Shifting” describes a more accurate method for fixing the phase-
shift by using hardware measurement of a unique PCB design.
Virtex-5 Devices
The GMII design uses IODELAY components on the receiver clock, data and control
signals for Virtex-5 devices. A fixed tap delay can be applied to either delay the data and
control signals or delay the clock so that the data/control are correctly sampled by the
gmii_rx_clkclock at the IOB flip-flop, meeting GMII setup and hold timing.
The choice of delaying data/control or clock is dependant upon a number of factors, not
least being the required shift. There are trade-offs to be made with either choice. Delaying
the clock is clock-period specific as we move the clock to line up each edge with data from
the following edge. Delaying the data/control introduces more jitter which degrades the
overall setup/hold window. The interface timing report in the two cases is also quite
different. See “Understanding Timing Reports for GMII Setup/Hold Timing.”
The following constraint shows an example of setting the delay value for one of these
IODELAY components. Data/Control bits can be adjusted individually, if desired, to
compensate for any PCB routing skew.
INST *gmii_interface/delay_gmii_rx_dv IDELAY_VALUE = 33;
The value of IDELAY_VALUEis preconfigured in the example designs to meet the setup
and hold constraints for the example GMII pinout in the particular device. The setup/hold
timing which is achieved after place-and-route is reported in the datasheet section of the
TRCE report (created by the implement script).
When IDELAY or IODELAY primitives are instantiated with a fixed delay attribute, an
IDELAYCTRL component must be also instantiated to continuously calibrate the
individual input delay elements. The IDELAYCTRL module requires a reference clock,
which is assumed to be an input to the example design delivered by CORE Generator. The
most efficient way to use the IDELAYCTRL module is to lock the placement of the instance
to the clock region of the device where the IDELAY/IODELAY components are placed. An
example LOC constraint for the IDELAYCTRL module is shown commented out in the
UCF. See the Virtex-5 User Guide and code comments for more information.
The following constraints are provided in the example design to link the instance of the
IDELAYCTRL to the IODELAY components used on the GMII. These constraints aid the
Xilinx tools in automatic IDELAYCTRL placement:
# Group IODELAY and IDELAYCTRL components to aid placement
INST "*gemac_block/gmii_interface/delay_gmii_rx_clk" HIODELAY_GROUP ="HIODELAY_GMII_GRP1";
INST "*gemac_block/gmii_interface/delay_gmii_rx_dv" HIODELAY_GROUP = "HIODELAY_GMII_GRP1";
INST "*gemac_block/gmii_interface/delay_gmii_rx_er" HIODELAY_GROUP = "HIODELAY_GMII_GRP1";
INST "*gemac_block/gmii_interface/*delay_gmii_rxd" HIODELAY_GROUP = "HIODELAY_GMII_GRP1";
INST "*gemac_block/dlyctrl" HIODELAY_GROUP = "HIODELAY_GMII_GRP1";
98
1-Gigabit Ethernet MAC v8.5 User Guide
UG144 April 24, 2009
Download from Www.Somanuals.com. All Manuals Search And Download.
-- DISCONTINUED PRODUCT --
R
Required Constraints
Understanding Timing Reports for GMII Setup/Hold Timing
Non-Virtex-5 devices
Setup and Hold results for the GMII input bus can be found in the data sheet section of the
Timing Report. The results are self-explanatory and it is easy to see how they relate to
Figure 9-1. Following is an example for the GMII report from a Virtex-4 device. The
implementation requires 1.531 ns of setup—this is less than the 2 ns required so there is
some slack. The implementation requires -0.125 ns of hold— this is less than the 0 ns
required so there is some slack.
Data Sheet report:
-----------------
All values displayed in nanoseconds (ns)
Setup/Hold to clock gmii_rx_clk
------------+------------+------------+------------------+--------+
| Setup to | Hold to |
| clk (edge) | clk (edge) |Internal Clock(s) | Phase |
------------+------------+------------+------------------+--------+
| Clock |
Source
gmii_rx_dv |
gmii_rx_er |
gmii_rxd<0> |
gmii_rxd<1> |
gmii_rxd<2> |
gmii_rxd<3> |
gmii_rxd<4> |
gmii_rxd<5> |
gmii_rxd<6> |
gmii_rxd<7> |
1.531(R)| -0.141(R)|gmii_rx_clk_bufg | 0.000|
1.531(R)| -0.141(R)|gmii_rx_clk_bufg | 0.000|
1.531(R)| -0.141(R)|gmii_rx_clk_bufg | 0.000|
1.525(R)| -0.135(R)|gmii_rx_clk_bufg | 0.000|
1.531(R)| -0.141(R)|gmii_rx_clk_bufg | 0.000|
1.525(R)| -0.135(R)|gmii_rx_clk_bufg | 0.000|
1.515(R)| -0.125(R)|gmii_rx_clk_bufg | 0.000|
1.515(R)| -0.125(R)|gmii_rx_clk_bufg | 0.000|
1.520(R)| -0.130(R)|gmii_rx_clk_bufg | 0.000|
1.520(R)| -0.130(R)|gmii_rx_clk_bufg | 0.000|
------------+------------+------------+------------------+--------+
Virtex-5 devices with Delayed Data/Control
Setup and Hold results for the GMII input bus can be found in the data sheet section of the
Timing Report. The results are self-explanatory and it is easy to see how they relate to
Figure 9-1. Here follows an example for the GMII report from a Virtex-5 device. The
implementation requires 1.962 ns of setup—this is less than the 2 ns required, so there is
slack. The implementation requires –0.008 ns of hold—this is less than the 0 ns required, so
there is slack.
1-Gigabit Ethernet MAC v8.5 User Guide
99
UG144 April 24, 2009
Download from Www.Somanuals.com. All Manuals Search And Download.
-- DISCONTINUED PRODUCT --
R
Chapter 9: Constraining the Core
Data Sheet report:
-----------------
All values displayed in nanoseconds (ns)
Setup/Hold to clock gmii_rx_clk
------------+------------+------------+------------------+--------+
| Setup to | Hold to | | Clock |
| clk (edge) | clk (edge) |Internal Clock(s) | Phase |
------------+------------+------------+------------------+--------+
Source
gmii_rx_dv |
gmii_rx_er |
gmii_rxd<0>|
gmii_rxd<1>|
gmii_rxd<2>|
gmii_rxd<3>|
gmii_rxd<4>|
gmii_rxd<5>|
gmii_rxd<6>|
gmii_rxd<7>|
1.955(R)| -0.017(R)| gmii_rx_clk_bufg | 0.000|
1.962(R)| -0.031(R)| gmii_rx_clk_bufg | 0.000|
1.949(R)| -0.013(R)| gmii_rx_clk_bufg | 0.000|
1.944(R)| -0.009(R)| gmii_rx_clk_bufg | 0.000|
1.947(R)| -0.012(R)| gmii_rx_clk_bufg | 0.000|
1.942(R)| -0.008(R)| gmii_rx_clk_bufg | 0.000|
1.950(R)| -0.015(R)| gmii_rx_clk_bufg | 0.000|
1.962(R)| -0.026(R)| gmii_rx_clk_bufg | 0.000|
1.957(R)| -0.022(R)| gmii_rx_clk_bufg | 0.000|
1.952(R)| -0.020(R)| gmii_rx_clk_bufg | 0.000|
------------+------------+------------+------------------+--------+
Virtex-5 Devices with Delayed Clock
Setup and Hold results for the GMII input bus can be found in the data sheet section of the
Timing Report. However, depending on how the setup/hold requirements have been met,
example for the GMII report from a Virtex-5 device where the clock has been delayed to
meet the setup/hold requirements.
Data Sheet report:
-----------------
All values displayed in nanoseconds (ns)
Setup/Hold to clock gmii_rx_clk
------------+------------+------------+------------------+--------+
| Setup to | Hold to |
| clk (edge) | clk (edge) |Internal Clock(s) | Phase |
------------+------------+------------+------------------+--------+
| Clock |
Source
gmii_rx_dv | -6.198(R)|
gmii_rx_er | -6.225(R)|
gmii_rxd<0> | -6.149(R)|
gmii_rxd<1> | -6.152(R)|
gmii_rxd<2> | -6.206(R)|
gmii_rxd<3> | -6.207(R)|
gmii_rxd<4> | -6.134(R)|
gmii_rxd<5> | -6.134(R)|
gmii_rxd<6> | -6.170(R)|
gmii_rxd<7> | -6.170(R)|
7.526(R)| gmii_rx_clk_bufg | 0.000|
7.554(R)| gmii_rx_clk_bufg | 0.000|
7.484(R)| gmii_rx_clk_bufg | 0.000|
7.486(R)| gmii_rx_clk_bufg | 0.000|
7.532(R)| gmii_rx_clk_bufg | 0.000|
7.533(R)| gmii_rx_clk_bufg | 0.000|
7.476(R)| gmii_rx_clk_bufg | 0.000|
7.476(R)| gmii_rx_clk_bufg | 0.000|
7.506(R)| gmii_rx_clk_bufg | 0.000|
7.506(R)| gmii_rx_clk_bufg | 0.000|
------------+------------+------------+------------------+--------+
figure of 1.866 ns relative to the following rising edge of the clock (since the IDELAY has
acted to delay the clock by an entire period when measured from the input flip-flop). This
is less than the 2 ns required, so there is slack.
100
1-Gigabit Ethernet MAC v8.5 User Guide
UG144 April 24, 2009
Download from Www.Somanuals.com. All Manuals Search And Download.
-- DISCONTINUED PRODUCT --
R
Required Constraints
figure of -0.446 ns relative to the following rising edge of the clock (since the IDELAY has
acted to delay the clock by an entire period when measured from the input flip-flop). This
is less than the 0 ns required so there is slack.
GMII_RX_CLK
GMII_RXD[7:0],
GMII_RX_DV
GMII_RX_ER
8 ns
-6.134 ns
= 8 - 6.134
= 1.866 ns
tSETUP
t
HOLD = 7.554 - 8
= -0.446 ns
7.554 ns
8 ns
Figure 9-2: Timing Report Setup/Hold Illustration
Constraints when Implementing an External RGMII
The constraints defined in this section are implemented in the UCF for the example design
delivered with the core. Sections from this UCF are copied into the following descriptions
to provide examples. These examples should be studied in conjunction with the HDL
source code for the example design and with the description “Implementing External
RGMII IOB Constraints
The following constraints target the flip-flops that are inferred in the top-level HDL file for
the example design. Constraints are set to ensure that these are placed in IOBs. The DDR
register constraints are not present for a Virtex-4 device or Virtex-5 device where DDR
components are instantiated rather than inferred.
# RGMII Receiver Constraints: place DDR registers in IOB
INST "*rgmii_interface/rgmii_rxd_ddr*" IOB = true;
INST "*rgmii_interface/rgmii_rx_dv_ddr" IOB = true;
INST "*rgmii_interface/rgmii_rx_ctl_ddr" IOB = true;
# Inband Status Registers: place registers in IOB
INST "*rgmii_interface/link_status" IOB = true;
INST "*rgmii_interface/clock_speed*" IOB = true;
INST "*rgmii_interface/duplex_status" IOB = true;
1-Gigabit Ethernet MAC v8.5 User Guide
101
UG144 April 24, 2009
Download from Www.Somanuals.com. All Manuals Search And Download.
-- DISCONTINUED PRODUCT --
R
Chapter 9: Constraining the Core
The RGMII v2.0 is a 1.5 volt signal-level interface. The 1.5 volt HSTL Class I SelectIO
standard is used for RGMII interface pins. Use the following constraints with the device IO
Banking rules. The IO slew rate is set to fast to ensure that the interface can meet setup and
hold times.
INST "rgmii_txd<?>"
INST "rgmii_tx_ctl"
INST "rgmii_rxd<?>"
INST "rgmii_rx_ctl"
IOSTANDARD = HSTL_I;
IOSTANDARD = HSTL_I;
IOSTANDARD = HSTL_I;
IOSTANDARD = HSTL_I;
INST "rgmii_txc"
INST "rgmii_rxc"
IOSTANDARD = HSTL_I;
IOSTANDARD = HSTL_I;
INST "rgmii_txd<?>"
INST "rgmii_tx_ctl"
INST "rgmii_txc"
SLEW = FAST;
SLEW = FAST;
SLEW = FAST;
In addition, the example design provides pad locking on the RGMII for several families.
This is provided as a guideline only; there are no specific I/O location constraints for this
core.
RGMII Input Setup/Hold Timing
signals. This is the worst-case data valid window presented to the FPGA device pins.
RGMII_RXC
RGMII_RXD[3:0],
RGMII_RX_CTL
tSETUP
t
SETUPt
HOLD
tHOLD
Figure 9-3: Input RGMII Timing
The 2 ns data valid window, which is presented across the RGMII input bus, must be
correctly sampled on both clock edges by the FPGA devices.
Table 9-2: Input RGMII Timing
Symbol
Min
Typical
Units
t
t
1.0
1.0
2.0
2.0
ns
ns
SETUP
HOLD
For RGMII, the lower data bits, rgmii_rxd[3:0], should be sampled internally on the
rising edge of rgmii_rxc, and the upper data bits, rgmii_rxd[7:4], should be
sampled internally on the falling edge of rgmii_rxc.
The following constraints are provided in the UCF for RGMII Example Designs. These
constraints invoke the tools to analyze the setup/hold requirements (though, if failing, the
tools are NOT capable of fixing them: meeting these constraints is a manual process - see
the following family specific sections for details):
102
1-Gigabit Ethernet MAC v8.5 User Guide
UG144 April 24, 2009
Download from Www.Somanuals.com. All Manuals Search And Download.
-- DISCONTINUED PRODUCT --
R
Required Constraints
INST "rgmii_rxd<?>" TNM = IN_RGMII;
INST "rgmii_rx_ctl" TNM = IN_RGMII;
TIMEGRP "DDR_RISING" = FFS;
TIMEGRP "DDR_FALLING" = FALLING FFS;
TIMEGRP "IN_RGMII" OFFSET = IN 1.1 ns VALID 2.2 ns BEFORE "rgmii_rxc" TIMEGRP "DDR_RISING";
TIMEGRP "IN_RGMII" OFFSET = IN -2.9 ns VALID 2.2 ns BEFORE "rgmii_rxc" TIMEGRP "DDR_FALLING"
The constraints defined in the preceding lines have a built in 10% tolerance: this allows all
combinations of the example designs to pass timing. However, these should be tightened
in a real customer design as per the following syntax. Manual adjustments to IDELAY
values or DCM phase shift requirements may be required to meet these constraints (see the
following family specific sections for details).
TIMEGRP "IN_RGMII" OFFSET = IN 1 ns VALID 2 ns BEFORE "rgmii_rxc" TIMEGRP "DDR_RISING";
TIMEGRP "IN_RGMII" OFFSET = IN -3 ns VALID 2 ns BEFORE "rgmii_rxc" TIMEGRP "DDR_FALLING"
Spartan-3, Spartan-3E, Spartan-3A, and Virtex-4 Devices
The RGMII design uses a DCM on the receiver clock domain for all devices except Virtex-
5. Phase-shifting is then applied to the DCM to align the resultant clock so that it will
correctly sample the 2 ns RGMII data valid window at the input flip-flops.
The fixed phase-shift is applied to the DCM using the following UCF syntax.
INST *rgmii_interface/rgmii_rxc_dcm CLKOUT_PHASE_SHIFT = FIXED;
INST *rgmii_interface/rgmii_rxc_dcm PHASE_SHIFT = 40;
The value of PHASE_SHIFT is preconfigured in the example designs to meet the setup and
hold constraints for the example RGMII pinout in the particular device. The setup/hold
timing which is achieved after place-and-route is reported in the data sheet section of the
TRCE report (created by the implement script).
For customers fixing their own pinout, the setup and hold figures reported in the TRCE
report can be used to initially setup the approximate DCM phase-shift. Appendix C,
“Calculating DCM Phase-Shifting” describes a more accurate method for fixing the phase-
shift by using hardware measurement of a unique PCB design.
Virtex-5 Devices
The RGMII design uses IODELAY components on both the receiver and transmitter clock
domains for Virtex-5 devices. A fixed tap delay is applied to the rgmii_txcoutput clock
to move the rising edge of this clock to the centre of the output data window. For the
receiver clock, data and control signals, a fixed tap delay can be applied to either delay the
data and control signals or delay the clock so that the data/control are correctly sampled
by the rgmii_rxc clock at the IOB IDDR registers, meeting RGMII setup and hold timing.
The choice of delaying data/control or clock is dependant upon a number of factors, not
least being the required shift. There are trade-offs to be made with either choice. Delaying
the clock is clock-period specific as we move the clock to line up each edge with data from
the following edge. Delaying the data/control introduces more jitter which degrades the
overall setup/hold window. The interface timing report in the two cases is also quite
different. See “Understanding Timing Reports for RGMII Setup/Hold timing.”
The following constraint shows an example of setting the delay value for two of these
IODELAY components. Data/Control bits can be adjusted individually to compensate for
any PCB routing skew, if desired.
1-Gigabit Ethernet MAC v8.5 User Guide
103
UG144 April 24, 2009
Download from Www.Somanuals.com. All Manuals Search And Download.
-- DISCONTINUED PRODUCT --
R
Chapter 9: Constraining the Core
INST *rgmii_interface/delay_rgmii_tx_clk IDELAY_TYPE = “FIXED”;
INST *rgmii_interface/delay_rgmii_tx_clk ODELAY_VALUE = 25;
INST *rgmii_interface/delay_rgmii_tx_clk DELAY_SRC = “O”;
INST *rgmii_interface/delay_rgmii_rx_ctl IDELAY_TYPE = “FIXED”;
INST *rgmii_interface/delay_rgmii_rx_ctl IDELAY_VALUE = 20;
INST *rgmii_interface/delay_rgmii_rx_ctl DELAY_SRC = “I”;
The value of IDELAY_VALUEis preconfigured in the example designs to meet the setup
and hold constraints for the example RGMII pinout in the particular device. The
setup/hold timing which is achieved after place-and-route is reported in the data sheet
section of the TRCE report (created by the implement script).
When IDELAY or IODELAY primitives are instantiated with a fixed delay attribute, an
IDELAYCTRL component must be also instantiated to continuously calibrate the
individual input delay elements. The IDELAYCTRL module requires a reference clock,
which is assumed to be an input to the example design delivered by CORE Generator. The
most efficient way to use the IDELAYCTRL module is to lock the placement of the instance
to the clock region of the device where the IDELAY/IODELAY components are placed. An
example LOC constraint for the IDELAYCTRL module is shown commented-out in the
UCF. See the Virtex-5 User Guide and code comments for more information.
The following constraints are provided in the example design to link the instance of the
IDELAYCTRL to the IODELAY components used on the RGMII. These constraints aid the
Xilinx tools in automatic IDELAYCTRL placement:
# Group IODELAY and IDELAYCTRL components to aid placement
INST "*gemac_block/rgmii_interface/delay_rgmii_rx_clk"HIODELAY_GROUP =
"HIODELAY_RGMII_GRP1";
INST "*gemac_block/rgmii_interface/*delay_rgmii_rxd" HIODELAY_GROUP =
"HIODELAY_RGMII_GRP1";
INST "*gemac_block/rgmii_interface/delay_rgmii_rx_ctl" HIODELAY_GROUP =
"HIODELAY_RGMII_GRP1";
INST "*gemac_block/dlyctrl" IHODELAY_GROUP = "HIODELAY_RGMII_GRP1";
RGMII DDR Constraints
The following constraints are present for RGMII designs in all devices with the exception
of Virtex-4 and Virtex-5 devices. Due to the use of IDDR and ODDR primitives in the
Virtex-4 design, these extra clocking constraints are not required.
The RGMII design requires further clock crossing constraints to ensure timing is met when
crossing from rising to falling clock edges and vice versa. A stringent time constraint
ensures that timing is met with the worst-case timing allowed in the RGMII specification.
INST "*rgmii_interface/rgmii_rxd_reg_4"
INST "*rgmii_interface/rgmii_rxd_reg_5"
INST "*rgmii_interface/rgmii_rxd_reg_6"
INST "*rgmii_interface/rgmii_rxd_reg_7"
INST "*rgmii_interface/rgmii_rx_ctl_reg"
INST "*rgmii_interface/gmii_rxd_reg_4"
INST "*rgmii_interface/gmii_rxd_reg_5"
INST "*rgmii_interface/gmii_rxd_reg_6"
INST "*rgmii_interface/gmii_rxd_reg_7"
INST "*rgmii_interface/gmii_rx_er_reg"
TNM="rgmii_falling";
TNM="rgmii_falling";
TNM="rgmii_falling";
TNM="rgmii_falling";
TNM="rgmii_falling";
TNM="rgmii_rising";
TNM="rgmii_rising";
TNM="rgmii_rising";
TNM="rgmii_rising";
TNM="rgmii_rising";
TIMESPEC "TS_rgmii_falling_to_rising" = FROM "rgmii_falling" TO
"rgmii_rising" 3200 ps;
104
1-Gigabit Ethernet MAC v8.5 User Guide
UG144 April 24, 2009
Download from Www.Somanuals.com. All Manuals Search And Download.
-- DISCONTINUED PRODUCT --
R
Required Constraints
Understanding Timing Reports for RGMII Setup/Hold timing
Non-Virtex-5 Devices
Setup and Hold results for the RGMII input bus can be found in the data sheet section of
the Timing Report. The results are self-explanatory and it is easy to see how they relate to
Following is an example for the RGMII report from a Virtex-4 device. Each Input lists two
sets of values—one corresponding to the –ve edge of the clock and one to the +ve edge. The
first set listed corresponds to –ve edge which occurs at time 4 ns. The implementation
requires 0.648 ns of setup to the –ve edge and 0.661 ns to the +ve edge. This is less than the
1 ns required, so there is slack. The implementation requires 0.300 ns of hold to the –ve
edge and 0.316 ns to the +ve edge. This is less than the 1 ns required, so there is slack.
Data Sheet report:
-----------------
All values displayed in nanoseconds (ns)
Setup/Hold to clock rgmii_rxc
------------+------------+------------+---------------------+--------+
| Setup to | Hold to |
| clk (edge) | clk (edge) |Internal Clock(s)
| Clock |
| Phase |
Source
------------+------------+------------+---------------------+--------+
rgmii_rx_ctl| -3.352(R)| 4.300(R)|not_rgmii_rx_clk_bufg| 4.938|
0.661(R)| 0.284(R)|rgmii_rx_clk_bufg | 0.938|
rgmii_rxd<0>| -3.384(R)| 4.332(R)|not_rgmii_rx_clk_bufg| 4.938|
0.629(R)| 0.316(R)|rgmii_rx_clk_bufg | 0.938|
rgmii_rxd<1>| -3.348(R)| 4.296(R)|not_rgmii_rx_clk_bufg| 4.938|
0.665(R)| 0.280(R)|rgmii_rx_clk_bufg | 0.938|
rgmii_rxd<2>| -3.360(R)| 4.308(R)|not_rgmii_rx_clk_bufg| 4.938|
0.653(R)| 0.292(R)|rgmii_rx_clk_bufg | 0.938|
rgmii_rxd<3>| -3.428(R)| 4.382(R)|not_rgmii_rx_clk_bufg| 4.938|
0.585(R)| 0.366(R)|rgmii_rx_clk_bufg | 0.938|
|
|
|
|
|
------------+------------+------------+---------------------+--------+
Virtex-5 devices with delayed Data/Control
Setup and Hold results for the RGMII input bus can be found in the data sheet section of
the Timing Report. The results are self-explanatory and it is easy to see how they relate to
Following is an example for the RGMII report from a Virtex-5 device. Each Input lists two
sets of values—one corresponding to the –ve edge of the clock and one to the +ve edge. The
first set listed corresponds to +ve edge which occurs at time 0 ns. The implementation
requires 0.818 ns of setup to the +ve edge and 0.794 ns to the –ve edge. This is less than the
1 ns required, so there is slack. The implementation requires 0.946 ns of hold to the –ve
edge and 0.972 ns to the +ve edge. This is less than the 1 ns required, so there is slack.
1-Gigabit Ethernet MAC v8.5 User Guide
105
UG144 April 24, 2009
Download from Www.Somanuals.com. All Manuals Search And Download.
-- DISCONTINUED PRODUCT --
R
Chapter 9: Constraining the Core
Data Sheet report:
-----------------
All values displayed in nanoseconds (ns)
Setup/Hold to clock rgmii_rxc
------------+------------+------------+------------------+--------+
| Setup to | Hold to | | Clock |
| clk (edge) | clk (edge) |Internal Clock(s) | Phase |
------------+------------+------------+------------------+--------+
Source
rgmii_rx_ctl|
| -3.214(F)|
rgmii_rxd<0>|
| -3.213(F)|
rgmii_rxd<1>|
| -3.223(F)|
rgmii_rxd<2>|
0.810(R)|
0.933(R)| rgmii_rx_clk_bufg| 0.000|
4.959(F)| rgmii_rx_clk_bufg| 4.000|
0.940(R)| rgmii_rx_clk_bufg| 0.000|
4.966(F)| rgmii_rx_clk_bufg| 4.000|
0.946(R)| rgmii_rx_clk_bufg| 0.000|
4.972(F)| rgmii_rx_clk_bufg| 4.000|
0.929(R)| rgmii_rx_clk_bufg| 0.000|
4.955(F)| rgmii_rx_clk_bufg| 4.000|
0.936(R)| rgmii_rx_clk_bufg| 0.000|
4.962(F)| rgmii_rx_clk_bufg| 4.000|
0.811(R)|
0.801(R)|
0.818(R)|
| -3.206(F)|
rgmii_rxd<3>|
0.809(R)|
| -3.215(F)|
------------+------------+------------+------------------+--------+
Virtex-5 Devices with Delayed Clock
Setup and Hold results for the RGMII input bus can be found in the data sheet section of
the Timing Report. However, depending on how the setup/hold requirements have been
an example for the RGMII report from a Virtex-5 device where the clock has been delayed
to meet the setup/hold requirements.
Data Sheet report:
-----------------
All values displayed in nanoseconds (ns)
Setup/Hold to clock rgmii_rxc
------------+------------+------------+------------------+--------+
| Setup to | Hold to |
| clk (edge) | clk (edge) |Internal Clock(s) | Phase |
------------+------------+------------+------------------+--------+
rgmii_rx_ctl| -7.178(R)| 8.880(R)| rgmii_rx_clk_bufg| 0.000|
| -11.178(F)| 12.880(F)| rgmii_rx_clk_bufg| 4.000|
| Clock |
Source
rgmii_rxd<0>| -7.192(R)|
8.893(R)| rgmii_rx_clk_bufg| 0.000|
| -11.192(F)| 12.893(F)| rgmii_rx_clk_bufg| 4.000|
rgmii_rxd<1>| -7.182(R)|
8.884(R)| rgmii_rx_clk_bufg| 0.000|
| -11.182(F)| 12.884(F)| rgmii_rx_clk_bufg| 4.000|
rgmii_rxd<2>| -7.180(R)|
8.882(R)| rgmii_rx_clk_bufg| 0.000|
| -11.180(F)| 12.882(F)| rgmii_rx_clk_bufg| 4.000|
rgmii_rxd<3>| -7.179(R)|
8.881(R)| rgmii_rx_clk_bufg| 0.000|
| -11.179(F)| 12.881(F)| rgmii_rx_clk_bufg| 4.000|
------------+------------+------------+------------------+--------+
Each Input lists two sets of values—one corresponding to the +ve edge of the clock and one
to the –ve edge. The first set listed corresponds to +ve edge, which occurs at time 8 ns as we
have delayed the clock to use the following +ve edge.
this represents 0.821 ns relative to the following rising edge of the clock (since the IDELAY
has acted to delay the clock by an entire period when measured from the input flip-flop).
106
1-Gigabit Ethernet MAC v8.5 User Guide
UG144 April 24, 2009
Download from Www.Somanuals.com. All Manuals Search And Download.
-- DISCONTINUED PRODUCT --
R
Required Constraints
This is less than the 1 ns required, so there is slack. Equally for the –ve edge, we have
–11.179 ns of setup—this edge is at time 12 ns and therefore this equates to a setup of
0.821 ns.
this represents 0.893 ns relative to the following rising edge of the clock (since the IDELAY
has acted to delay the clock by an entire period when measured from the input flip-flop).
This is less than the 1 ns required, so there is slack. Equally for the –ve edge, we have
12.893 ns of hold —this edge is at time 12 ns and therefore equates to a hold time of
0.893 ns.
RGMII_RXC
RGMII_RXD[3:0],
RGMII_RX_CTL
8 ns
-6.134 ns
= 8 - 7.179
= 0.821 ns
tSETUP
t
HOLD = 8.893 - 8
= 0.893 ns
8.893 ns
8 ns
12 ns
-11.179 ns
12 ns
12.893 ns
t
HOLD = 12.893 - 12
= 0.893 ns
= 12 - 11.179
= 0.821 ns
tSETUP
Figure 9-4: Timing Report Setup/Hold Illustration
1-Gigabit Ethernet MAC v8.5 User Guide
107
UG144 April 24, 2009
Download from Www.Somanuals.com. All Manuals Search And Download.
-- DISCONTINUED PRODUCT --
R
Chapter 9: Constraining the Core
108
1-Gigabit Ethernet MAC v8.5 User Guide
UG144 April 24, 2009
Download from Www.Somanuals.com. All Manuals Search And Download.
-- DISCONTINUED PRODUCT --
R
Chapter 10
Clocking and Resetting
This chapter describes clock management considerations that are associated with
implementing the GEMAC core. It describes the clock management logic for all
implementations of the core and how clock management logic can be shared across
multiple instantiations of the core. The reset circuitry within the core is also described.
Clocking the Core
With Internal GMII
When the GMII-style interface of the core is used as an internal interface (for example, with
an internally connected PHY core), it is likely that gtx_clk and gmii_rx_clk will be
example.
With External GMII
Figure 10-1 illustrates the clock management used with an external GMII interface. All
clocks illustrated have a frequency of 125 MHz. The clock gtx_clkmust be provided to
the GEMAC core. This is a high-quality clock that satisfies the IEEE 802.3-2005
requirements. It is expected that this clock will be derived from an external oscillator and
When an external GMII, gmii_rx_clkwill usually be derived from a different clock
source to gtx_clk. In this case, gmii_rx_clkis received through an IBUFG. This clock is
then usually routed onto a global clock network by connecting it to a BUFG. The resulting
global clock is used by all MAC receiver logic. Some families contain a DCM on the
1-Gigabit Ethernet MAC
BUFG
IBUFG
IBUFG
BUFG
gtx_clk
gtx_clk
gmii_rx_clk
gmii_rx_clk
Figure 10-1: Clock Management Logic with External GMII
1-Gigabit Ethernet MAC v8.5 User Guide
109
UG144 April 24, 2009
Download from Www.Somanuals.com. All Manuals Search And Download.
-- DISCONTINUED PRODUCT --
R
Chapter 10: Clocking and Resetting
With RGMII
Standard Clocking Scheme
Figure 10-2 illustrates the clock management used with an external RGMII interface. All
clocks illustrated have a frequency of 125 MHz. The gtx_clkclock must be provided to
the GEMAC core. This is a high-quality clock that satisfies IEEE 802.3-2005 requirements. It
is expected that this clock will be derived from an external oscillator and connected into the
to a DCM from where phase-shifted clock signals are generated for use in the RGMII
transmitter logic. The zero phase-shifted clock is used as the input gtx_clkto the
GEMAC core.
The receiver clock, rgmii_rxcis usually derived from a different clock source to
gtx_clk. In this case, rgmii_rxcwill be received through an IBUFG. This clock is routed
into a DCM where it is used to generate phase-shifted clock signals for use in the RGMII
receiver logic. A fixed phase-shift value is applied to the DCM to meet RGMII setup and
used as the input gmii_rx_clkto the GEMAC core.
1-Gigabit Ethernet
MAC Core
BUFG
DCM
BUFG
DCM
CLK_0
gtx_clk
gmii_rx_clk
CLK_0
IBUFG
IBUFG
gtx_clk
rgmii_rxc
BUFG
CLK_90
RGMII Rx Logic
RGMII Tx Logic
Figure 10-2: Clock Management with External RGMII
Multiple Cores
With External GMII
Figure 10-3 illustrates how to share clock resources across multiple instantiations of the
core when using the optional GMII. gtx_clkmay be shared between multiple cores as
illustrated, resulting in a common transmitter clock domain across the device.
A common receiver clock domain is usually not possible as each core will receive an
independent receiver clock from the PHY attached to the other end of the GMII. As
110
1-Gigabit Ethernet MAC v8.5 User Guide
UG144 April 24, 2009
Download from Www.Somanuals.com. All Manuals Search And Download.
-- DISCONTINUED PRODUCT --
R
Multiple Cores
Note: Although not illustrated, if the optional Management Interface is used, host_clkcan also be
shared between cores.
1-Gigabit Ethernet MAC
BUFG
IBUFG
IBUFG
BUFG
gmii_rx_clk1
gtx_clk
gmii_rx_clk
gtx_clk
1-Gigabit Ethernet MAC
BUFG
IBUFG
gmii_rx_clk2
gtx_clk
gmii_rx_clk
Figure 10-3: Clock Management Logic with External GMII (Multiple Cores)
With RGMII
Figure 10-4 illustrates sharing clock resources across multiple instantiations of the core
using the optional RGMII. gtx_clkmay be shared between multiple cores as illustrated,
resulting in a common transmitter clock domain across the device.
As a general rule, a common receiver clock domain is not possible. Each core receives an
independent receiver clock from the PHY attached to the other end of the RGMII—as
Note: Although not illustrated, if the optional Management Interface is used, host_clkcan also be
shared between cores.
1-Gigabit Ethernet MAC v8.5 User Guide
111
UG144 April 24, 2009
Download from Www.Somanuals.com. All Manuals Search And Download.
-- DISCONTINUED PRODUCT --
R
Chapter 10: Clocking and Resetting
BUFG
DCM
CLK_0
IBUFG
gtx_clk
BUFG
CLK_90
1-Gigabit Ethernet MAC Core
gtx_clk gmii_rx_clk
BUFG
DCM
CLK_0
IBUFG
rgmii_rxc1
RGMII Tx Logic
RGMII Rx Logic
BUFG
1-Gigabit Ethernet MAC Core
gtx_clk gmii_rx_clk
DCM
CLK_0
IBUFG
rgmii_rxc2
RGMII Rx Logic
RGMII Tx Logic
Figure 10-4: Clock Management Logic with External RGMII (Multiple Cores)
Reset Conditions
Internally, the core is divided up into clock/reset domains that group together elements
with common clock and reset signals. The reset circuitry for one of these domains is
the design.
reset
Configuration reset
Core Registers
PRE
D Q
PRE
D Q
PRE
D Q
PRE
D Q
'0'
C
C
C
C
PRE
PRE
PRE
FDP
FDP
FDP
FDP
Clock
Figure 10-5: Reset Circuit for a Single Clock/reset Domain
112
1-Gigabit Ethernet MAC v8.5 User Guide
UG144 April 24, 2009
Download from Www.Somanuals.com. All Manuals Search And Download.
-- DISCONTINUED PRODUCT --
R
Chapter 11
Interfacing to Other Cores
Ethernet 1000Base-X PCS/PMA or SGMII Core
The GEMAC core can be integrated in a single device with the Ethernet 1000BASE-X
PCS/PMA or SGMII core to extend core functionality to provide the following:
1000BASE-X Physical Coding Sublayer (PCS) logic designed to the IEEE 802.3 specification
with either:
•
1000BASE-X Physical Medium Attachment (PMA) using a device-specific RocketIO™
transceiver.
♦
♦
Virtex®-5 SXT and LXT devices use RocketIO GTP transceivers
Virtex-5 FXT devices use RocketIO GTX transceivers
After the first introduction of a device-specific RocketIO transceiver, all subsequent
references use the generic term RocketIO transceiver.
•
•
1000BASE-X parallel Ten-Bit-Interface (TBI) for connection to external SERDES.
Alternatively, the Ethernet 1000BASE-X PCS/PMA or SGMII core can function as a
GMII to Serial-GMII (SGMII) bridge, meaning that this can be used to provide the
GEMAC core with an SGMII for serial connection to appropriate PHYs.
A description of the latest available IP Update containing the Ethernet 1000BASE-X
PCS/PMA or SGMII core and instructions for obtaining the IP Update can be found on the
Ethernet 1000BASE-X PCS/PMA or SGMII product page. A full description of the Ethernet
1000BASE-X PCS/PMA or SGMII core is outside the scope of this document.
1-Gigabit Ethernet MAC v8.5 User Guide
113
UG144 April 24, 2009
Download from Www.Somanuals.com. All Manuals Search And Download.
-- DISCONTINUED PRODUCT --
Chapter 11: Interfacing to Other Cores
R
Integration to Provide 1000BASE-X PCS with TBI
Figure 11-1 illustrates the connections and clock management logic required to interface
the GEMAC core to the Ethernet 1000BASE-X PCS/PMA or SGMII core (when used in
1000BASE-X mode with the parallel TBI). It depicts the following:
•
•
Direct internal connections are made between the GMII interfaces of the two cores.
If the GEMAC has been built with the optional management logic, the MDIO port can
be connected up to that of the Ethernet 1000BASE-X PCS/PMA or SGMII core to
access its embedded configuration and status registers. See “Using the Optional
•
Due to the embedded Receiver Elastic Buffer in the Ethernet 1000BASE-X PCS/PMA
or SGMII core, the entire GMII is synchronous to a single clock domain. For this
reason, gtx_clkis used as the 125 MHz reference clock for both cores and the
transmitter and receiver logic of the GEMAC core now operate in the same clock
domain. This allows clock crossing constraints between the gtx_clk and
gmii_rx_clk clock domains to be removed from the GEMAC user constraints file
IOB LOGIC
IBUFG
BUFG
gtx_clk
IPAD
gtx_clk_bufg (125 MHz)
component_name_block
(Block Level from example design)
Ethernet 1000BASE-X
PCS/PMA or SGMII
LogiCORE
1-Gigabit Ethernet
MAC
LogiCORE
gtx_clk
gtx_clk
gmii_rx_clk
gmii_txd[7:0]
gmii_tx_en
gmii_txd[7:0]
gmii_tx_en
gmii_tx_er
gmii_tx_er
TBI
gmii_rxd[7:0]
gmii_rx_dv
gmii_rx_er
gmii_rxd[7:0]
gmii_rx_dv
gmii_rx_er
mdc
mdc
mdio_in
mdio_out
mdio_tri
mdio_in
mdio_out
mdio_tri
no
connection
Figure 11-1: 1-Gigabit Ethernet MAC Extended to Include 1000BASE-X PCS with TBI
114
1-Gigabit Ethernet MAC v8.5 User Guide
UG144 April 24, 2009
Download from Www.Somanuals.com. All Manuals Search And Download.
-- DISCONTINUED PRODUCT --
Ethernet 1000Base-X PCS/PMA or SGMII Core
R
Integration to Provide 1000BASE-X PCS and PMA using a RocketIO
Transceiver
Virtex-4 Devices
Figure 11-2 illustrates the connections and clock management logic required to interface
the GEMAC core to the Ethernet 1000BASE-X PCS/PMA or SGMII core (when used in
1000BASE-X mode with PMA using the device-specific RocketIO MGT transceiver.
Virtex-4
GT11CLK_MGT
brefclkp
(125 MHz)
IPAD
MGTCLKP
MGTCLKN
IPAD
brefclkn
(125 MHz)
synclk1
(125MHz)
BUFG
SYNCLK1OUT
component_name_block
(Block Level from example design)
userclk2
(125 MHz)
Ethernet 1000BASE-X
PCS/PMA or SGMII
LogiCORE
1-Gigabit Ethernet
MAC
Virtex-4
GT11
RocketIO
LogiCORE
gtx_clk
TXOUTCLK1
userclk
userclk2
gmii_rx_clk
REFCLK1
‘0’
‘0’
TXUSRCLK
TXUSRCLK2
RXUSRCLK
RXUSRCLK2
gmii_txd[7:0]
gmii_tx_en
gmii_txd[7:0]
gmii_tx_en
gmii_tx_er
gmii_tx_er
gmii_rxd[7:0]
gmii_rx_dv
gmii_rx_er
gmii_rxd[7:0]
gmii_rx_dv
gmii_rx_er
RocketIO I/F
mdc
mdc
mdio_in
mdio_out
mdio_tri
mdio_in
mdio_out
mdio_tri
no
connection
Figure 11-2: 1-Gigabit Ethernet MAC Extended to Include 1000BASE-X PCS and PMA
using a RocketIO Transceiver
1-Gigabit Ethernet MAC v8.5 User Guide
115
UG144 April 24, 2009
Download from Www.Somanuals.com. All Manuals Search And Download.
-- DISCONTINUED PRODUCT --
Chapter 11: Interfacing to Other Cores
R
Figure 11-2 illustrates the following:
•
Direct internal connections are made between the GMII interfaces between the two
cores.
•
If the GEMAC has been generated with the optional Management Interface, the
MDIO port can be connected up to that of the Ethernet 1000BASE-X PCS/PMA or
SGMII core to access its embedded configuration and status registers. See “Using the
•
Due to the embedded Receiver Elastic Buffer in the Ethernet 1000BASE-X PCS/PMA
or SGMII core, the entire GMII is synchronous to a single clock domain. For this
reason, userclk2is used as the 125 MHz reference clock for both cores and the
transmitter and receiver logic of the GEMAC core now operate in the same clock
domain. This allows clock crossing constraints between the gtx_clk and
116
1-Gigabit Ethernet MAC v8.5 User Guide
UG144 April 24, 2009
Download from Www.Somanuals.com. All Manuals Search And Download.
-- DISCONTINUED PRODUCT --
Ethernet 1000Base-X PCS/PMA or SGMII Core
R
Virtex-5 LXT and SXT Devices
Figure 11-3 illustrates the connections and clock management logic required to interface
the GEMAC core to the Ethernet 1000BASE-X PCS/PMA or SGMII core (when used in
1000BASE-X mode with PMA using the device-specific RocketIO transceiver).
IBUFGDS
brefclkp
IPAD
IPAD
brefclkn
clkin
(125MHz)
BUFG
component_name_block
(Block Level from example design)
userclk2
(125 MHz)
Ethernet 1000BASE-X
PCS/PMA or SGMII
LogiCORE
1-Gigabit Ethernet
MAC
Virtex-5
GTP
RocketIO
LogiCORE
gtx_clk
TXOUTCLK0
gmii_rx_clk
CLKIN
TXUSRCLK0
TXUSRCLK20
RXUSRCLK0
RXUSRCLK20
gmii_txd[7:0]
gmii_txd[7:0]
gmii_tx_en
gmii_tx_en
userclk
gmii_tx_er
userclk2
gmii_tx_er
gmii_rxd[7:0]
gmii_rx_dv
gmii_rx_er
gmii_rxd[7:0]
gmii_rx_dv
gmii_rx_er
RocketIO I/F
mdc
mdc
mdio_in
mdio_out
mdio_tri
mdio_in
mdio_out
mdio_tri
no
connection
Figure 11-3: 1-Gigabit Ethernet MAC Extended to Include 1000BASE-X PCS and PMA
using a RocketIO transceiver
Figure 11-3 illustrates the following:
•
Direct internal connections are made between the GMII interfaces between the two
cores.
•
If the GEMAC has been generated with the optional Management Interface, the
MDIO port can be connected to that of the Ethernet 1000BASE-X PCS/PMA or SGMII
core to access its embedded configuration and status registers. See “Using the
1-Gigabit Ethernet MAC v8.5 User Guide
117
UG144 April 24, 2009
Download from Www.Somanuals.com. All Manuals Search And Download.
-- DISCONTINUED PRODUCT --
Chapter 11: Interfacing to Other Cores
R
•
Due to the embedded Receiver Elastic Buffer in the Ethernet 1000BASE-X PCS/PMA
or SGMII core, the entire GMII is synchronous to a single clock domain. For this
reason, userclk2is used as the 125 MHz reference clock for both cores and the
transmitter and receiver logic of the GEMAC core now operate in the same clock
domain. This allows clock crossing constraints between the gtx_clk and
Virtex-5 FXT Devices
Figure 11-4 illustrates the connections and clock management logic required to interface
the GEMAC core to the Ethernet 1000BASE-X PCS/PMA or SGMII core (when used in
1000BASE-X mode with PMA using the device-specific RocketIO transceiver).
DCM
BUFG
BUFG
userclk2 (125MHz)
userclk (62.5MHz)
IBUFGDS
CLKIN CLK0
brefclkp
IPAD
FB
IPAD
brefclkn
clkin
(125MHz)
CLKDV
component_name_block
(Block Level from example design)
Ethernet 1000BASE-X
PCS/PMA or SGMII
LogiCORE
1-Gigabit Ethernet
MAC
Virtex-5
GTX
RocketIO
LogiCORE
gtx_clk
REFCLKOUT
gmii_rx_clk
CLKIN
TXUSRCLK0
TXUSRCLK20
RXUSRCLK0
RXUSRCLK20
gmii_txd[7:0]
gmii_txd[7:0]
gmii_tx_en
gmii_tx_en
userclk
gmii_tx_er
userclk2
gmii_tx_er
gmii_rxd[7:0]
gmii_rx_dv
gmii_rx_er
gmii_rxd[7:0]
gmii_rx_dv
gmii_rx_er
RocketIO I/F
mdc
mdc
mdio_in
mdio_out
mdio_tri
mdio_in
mdio_out
mdio_tri
no
connection
Figure 11-4: 1-Gigabit Ethernet MAC Extended to Include 1000BASE-X PCS and PMA
using the RocketIO transceiver
118
1-Gigabit Ethernet MAC v8.5 User Guide
UG144 April 24, 2009
Download from Www.Somanuals.com. All Manuals Search And Download.
-- DISCONTINUED PRODUCT --
R
Ethernet Statistics Core
Features of this configuration include:
•
Direct internal connections are made between the GMII interfaces between the two
cores.
•
If the GEMAC has been generated with the optional Management Interface, the
MDIO port can be connected up to that of the Ethernet 1000BASE-X PCS/PMA or
SGMII core to access its embedded configuration and status registers. See “Using the
•
Due to the embedded Receiver Elastic Buffer in the Ethernet 1000BASE-X PCS/PMA
or SGMII core, the entire GMII is synchronous to a single clock domain. For this
reason, userclk2is used as the 125 MHz reference clock for both cores and the
transmitter and receiver logic of the GEMAC core now operate in the same clock
domain. This allows clock crossing constraints between the gtx_clk and
Integration to Provide SGMII Functionality
The connections between the two cores to provide SGMII functionality are identical to the
connections required for 1000BASE-X PCS and PMA using the device-specific RocketIO
transceiver. The only difference is that the Ethernet 1000BASE-X PCS/PMA or SGMII core
is generated with the SGMII option. See “Integration to Provide 1000BASE-X PCS and
PMA using a RocketIO Transceiver” for a description of SGMII integration.
Ethernet Statistics Core
The Ethernet Statistics core can be integrated in a single device with the 1-Gigabit Ethernet
MAC core. Using the Ethernet Statistics core lets you
•
Count statistics based on the rx_statistics_vectorand
tx_statistics_vectoroutputs from the MAC
•
•
Select 32-bit or 64-bit counters
Specify which statistics are counted and have precise control of the conditions under
which the counters are incremented
•
Read statistics optionally through the host interface of the GEMAC or independently
of the MAC
A description of the latest available IP Update containing the Ethernet Statistics core and
instructions on obtaining the IP update can be found on the Ethernet Statistics product
page.
Connecting the Ethernet Statistics Core to Provide Statistics Gathering
The Ethernet Statistics core interfaces directly to the 1-Gigabit Ethernet MAC core. The
Ethernet Statistics core takes in the tx_statistics_vectorand
rx_statistics_vectoras inputs. Statistics values gathered can then be read out
through the Statistics core Management Interface, that can be shared with the MAC
Management Interface.
1-Gigabit Ethernet MAC v8.5 User Guide
119
UG144 April 24, 2009
Download from Www.Somanuals.com. All Manuals Search And Download.
-- DISCONTINUED PRODUCT --
Chapter 11: Interfacing to Other Cores
R
Figure 11-5 illustrates connecting the Ethernet Statistics core to the MAC.
LogiCORE 1-Gigabit
Ethernet Statistics
Example Design
Ethernet MAC
tx_clk
gtx_clk
tx_statistics_vector[31:0]
tx_statistics_valid
tx_statistics_vector[31:0]
tx_statistics_valid
rx_clk
gmii_rx_clk
rx_statistics_vector[27:0]
rx_statistics_valid
rx_statistics_vector[27:0]
rx_statistics_valid
host_clk
host_clk
host_miim_sel
host_req
host_miim_sel
host_req
host_addr[8:0]
host_addr[9]
host_wr_data[31:0]
host_rd_data[31:0]
host_addr[8:0]
host_addr[9]
host_rd_data[31:0]
host_clk
host_miim_sel
host_req
host_addr[8:0]
host_addr[9]
host_wr_data[31:0]
OR
S
I1
host_rd_data_mac[31:0]
host_rd_data[31:0]
I0
host_rd_data_stats[31:0]
Figure 11-5: Interfacing the Ethernet Statistics to the 1-Gigabit Ethernet MAC
120
1-Gigabit Ethernet MAC v8.5 User Guide
UG144 April 24, 2009
Download from Www.Somanuals.com. All Manuals Search And Download.
-- DISCONTINUED PRODUCT --
R
Ethernet Statistics Core
The management interfaces of the two cores can be shared by avoiding bus conflict, as
follows:
•
Selecting a different address range for the statistics to that of the MAC configuration
registers. This is achieved by setting host_addr[9]to logic 0 when reading from the
statistics and logic 1 when writing and reading to the MAC configuration registers.
•
Using the host_miim_sel signal to differentiate between a statistical counter read
and a MAC initiated MDIO transaction. This is achieved by setting host_miim_sel
to logic 0 for a statistical counter read and logic 1 for a MAC initiated MDIO
transaction.
Table 11-1 describes the type of host transactions that occur if the host interface is shared
Table 11-1: Management Interface Transaction Types
Transaction
Configuration
host_miim_sel
host_addr[9]
0
1
0
1
X
0
MIIM access
Statistics Read
1-Gigabit Ethernet MAC v8.5 User Guide
121
UG144 April 24, 2009
Download from Www.Somanuals.com. All Manuals Search And Download.
-- DISCONTINUED PRODUCT --
R
Chapter 11: Interfacing to Other Cores
122
1-Gigabit Ethernet MAC v8.5 User Guide
UG144 April 24, 2009
Download from Www.Somanuals.com. All Manuals Search And Download.
-- DISCONTINUED PRODUCT --
R
Chapter 12
Implementing Your Design
This chapter describes how to simulate and implement your design containing the
GEMAC core.
Pre-implementation Simulation
A unit delay structural model of the GEMAC core netlist is provided as a CORE
Generator™ output file. This can be used for simulation of the block in the design phase of
the project.
Using the Simulation Model
For information about setting up your simulator to use the pre-implemented model, see
the Xilinx Synthesis and Verification Design Guide, included in your Xilinx software
installation.
The unit delay structural model of the GEMAC core can be found in the CORE Generator
project directory. Details of the CORE Generator outputs can be found in the 1-Gigabit
Ethernet MAC Getting Started Guide.
VHDL
•
<component_name>.vhd
Verilog
<component_name>.v
•
Synthesis
XST—VHDL
A component and instantiation template for the core named <component_name>.vhois
provided in the CORE Generator project directory. Use this to help instance the GEMAC
core into your VHDL source.
After your entire design is complete, create:
•
An XST project file top_level_module_name.prjlisting all the user source code
files.
•
An XST script file top_level_module_name.scrcontaining your required
synthesis options.
1-Gigabit Ethernet MAC v8.5 User Guide
123
UG144 April 24, 2009
Download from Www.Somanuals.com. All Manuals Search And Download.
-- DISCONTINUED PRODUCT --
Chapter 12: Implementing Your Design
R
To synthesize the design, run:
$ xst -ifn top_level_module_name.scr
See the XST User Guide for more information on creating project and synthesis script files,
and running the xstprogram.
XST—Verilog
A module declaration for the GEMAC core is provided in the CORE Generator project
directory:
<component_name>/example_design/<component_name>_mod.v
Use this module to help instance the GEMAC core into your Verilog source.
After your entire design is complete, create:
•
•
An XST project file top_level_module_name.prjlisting all the user source code
files. Make sure you include
%XILINX%/verilog/src/iSE/unisim_comp.v
and
<component_name>/example_design/component_name_mod.v
as the first two files in the project list.
An XST script file top_level_module_name.scrcontaining your required
synthesis options.
To synthesize the design, run:
$ xst -ifn top_level_module_name.scr
See the XST User Guide for more information on creating project and synthesis script files,
and running the xstprogram.
Implementation
Generating the Xilinx Netlist
To generate the Xilinx netlist, the ngdbuild tools are used to translate and merge the
individual design netlists into a single design database, the NGD file. Also merged at this
stage is the UCF for the design.
An example of the ngdbuild command is:
$ ngdbuild -sd path_to_core_netlist -sd path_to_user_synth_results \
-uc top_level_module_name.ucf top_level_module_name
Mapping the Design
To map the logic gates of the user design netlist into the CLBs and IOBs of the FPGA, run
the mapcommand. The mapcommand writes out a physical design to an NCD file. An
example of the mapcommand is:
$ map -o top_level_module_name_map.ncd top_level_module_name.ngd \
top_level_module_name.pcf
124
1-Gigabit Ethernet MAC v8.5 User Guide
UG144 April 24, 2009
Download from Www.Somanuals.com. All Manuals Search And Download.
-- DISCONTINUED PRODUCT --
R
Post-Implementation Simulation
Placing-and-Routing the Design
Execute the par command to place-and-route your design logic components (mapped
physical logic cells) contained within an NCD file in accordance with the layout and timing
requirements specified within the PCF file. The par command outputs the placed and
routed physical design to an NCD file.
An example of the par command is:
$ par top_level_module_name_map.ncd top_level_module_name.ncd \
top_level_module_name.pcf
Static Timing Analysis
Execute the trace command to evaluate timing closure on a design and create a Timing
Report file (TWR) derived from static timing analysis of the Physical Design file (NCD).
The analysis is typically based on constraints included in the optional PCF file.
An example of the trcecommand is:
$ trce -o top_level_module_name.twr top_level_module_name.ncd \
top_level_module_name.pcf
Generating a Bitstream
To create the configuration bitstream (BIT) file based on the contents of a physical
implementation file (NCD), the bitgencommand must be executed. The BIT file defines
the behavior of the programmed FPGA. An example of the bitgencommand is:
$ bitgen -w top_level_module_name.ncd
Post-Implementation Simulation
The purpose of post-implementation simulation is to verify that the design as
implemented in the FPGA works as expected.
Generating a Simulation Model
Run the netgen command to generate a chip-level simulation netlist for your design.
VHDL
$ netgen -sim -ofmt vhdl -ngm top_level_module_name_map.ngm \
-tm netlist top_level_module_name.ncd \
top_level_module_name_postimp.vhd
Verilog
$ netgen -sim -ofmt verilog -ngm top_level_module_name_map.ngm \
-tm netlist top_level_module_name.ncd \
top_level_module_name_postimp.v
1-Gigabit Ethernet MAC v8.5 User Guide
125
UG144 April 24, 2009
Download from Www.Somanuals.com. All Manuals Search And Download.
-- DISCONTINUED PRODUCT --
Chapter 12: Implementing Your Design
R
Using the Model
For information about setting up your simulator to use the pre-implemented model, see
the Xilinx Synthesis and Verification Design Guide included in your Xilinx software
installation.
Other Implementation Information
For more information about using the Xilinx implementation tool flow, including
command line switches and options, see the Xilinx ISE® software manuals.
126
1-Gigabit Ethernet MAC v8.5 User Guide
UG144 April 24, 2009
Download from Www.Somanuals.com. All Manuals Search And Download.
-- DISCONTINUED PRODUCT --
R
Appendix A
Using the Client-Side FIFO
The example design provided with the GEMAC core contains a FIFO used on the client-
side of the core. The source code for the FIFO is provided, and may be used and adjusted
for user applications.
The 10 Mbps/100 Mbps/1 Gbps Ethernet FIFO is designed for use with the GEMAC and
Tri-Mode Ethernet MAC (TEMAC) cores. The FIFO directly interfaces to the MAC client
interface providing a buffer between the MAC and the user’s logic. The FIFO implements
a direct connection to other LocalLink modules or the user’s logic.
The 10 Mbps/100 Mbps/1 Gbps Ethernet FIFO consists of independent transmit and
a typical implementation.
LocalLink Interface
Client Interface
GMII/RGMII
Xilinx FPGA
10 Mbps,
100 Mbps,
1 Gbps
1-Gigabit Ethernet MAC
Core
Ethernet FIFO
1000BASE-T
PHY
Tx
FIFO
Twisted
Copper
Pair
IOBs
Switch or
Router
Client
GMII
I/F
Rx
FIFO
Figure A-1: Typical 10 Mbps/100 Mbps/ 1 Gbps Ethernet FIFO Implementation
1-Gigabit Ethernet MAC v8.5 User Guide
127
UG144 April 24, 2009
Download from Www.Somanuals.com. All Manuals Search And Download.
-- DISCONTINUED PRODUCT --
Appendix A: Using the Client-Side FIFO
R
Interfaces
Transmit FIFO
Table A-1 describes the transmit FIFO client interface. For more information on the MAC
Table A-1: Transmit FIFO Client Interface
Clock
Domain
Signal
Direction
Description
tx_clk
Input
Input
Input
N/A
Transmit clock used by MAC.
Synchronous reset.
tx_reset
tx_clk
tx_clk
tx_enable
Clock enable for tx_clk. Tie to logic 1
when using GEMAC.
tx_data[7:0]
Output
tx_clk
Data presented t o MAC for
transmission.
tx_data_valid
tx_ack
Output
Input
tx_clk
tx_clk
tx_clk
tx_clk
Valid signal for data.
Ack signal from MAC.
Underrun signal to MAC.
tx_underrun
tx_collision
Output
Input
Collision indication from MAC. Tie
to logic 0 when using GEMAC.
tx_retransmit
Input
tx_clk
Retransmit request from MAC. Tie to
logic 0 when using GEMAC.
Table A-2 describes the transmit FIFO LocalLink interface. For more information on the
Table A-2: Transmit FIFO LocalLink Interface
Clock
Domain
Signal
tx_ll_clock
Direction
Description
Input
Input
N/A
Write clock for LocalLink interface
tx_ll_reset
tx_ll_clock Synchronous reset
tx_ll_data_in[7:0]
tx_ll_sof_in_n
Input
tx_ll_clock Write data to be sent to transmitter
tx_ll_clock Start of frame indicator
tx_ll_clock End of frame indicator
tx_ll_clock Source ready indicator
tx_ll_clock Destination ready indicator
tx_ll_clock FIFO memory status
Input
tx_ll_eof_in_n
tx_ll_src_rdy_in_n
tx_ll_dst_rdy_out_n
tx_fifo_status[3:0]
tx_overflow
Input
Input
Output
Output
Output
tx_ll_clock Overflow signal indicates when a
frame has been dropped in the FIFO
128
1-Gigabit Ethernet MAC v8.5 User Guide
UG144 April 24, 2009
Download from Www.Somanuals.com. All Manuals Search And Download.
-- DISCONTINUED PRODUCT --
R
Interfaces
Receive FIFO
Table A-3 describes the receive FIFO client interface. For more information on the MAC
Table A-3: Receive FIFO Client Interface
Clock
Domain
Signal
Direction
Description
rx_clk
Input
Input
Input
N/A
Receive clock used by MAC
Synchronous reset
rx_reset
rx_clk
rx_clk
rx_enable
Clock enable for rx_clk, tie to logic 1
when using GEMAC
rx_data[7:0]
Input
Input
Input
rx_clk
rx_clk
rx_clk
Data received from MAC
Valid signal for data
rx_data_valid
rx_good_frame
Indicates if frame is valid and should
be accepted by client
rx_bad_frame
rx_overflow
Input
rx_clk
rx_clk
Indicates if frame is invalid and
should be dropped by the FIFO
Output
Overflow signal indicates when a
frame has been dropped in the FIFO
Table A-4 describes the receive FIFO LocalLink interface. For more information on the
Table A-4: Receive FIFO LocalLink Interface
Clock
Domain
Signal
rx_ll_clock
Direction
Description
Input
Input
N/A
Read clock for LocalLink interface
rx_ll_reset
rx_ll_clock Synchronous reset
rx_ll_data_out[7:0]
rx_ll_sof_out_n
rx_ll_eof_out_n
rx_ll_src_rdy_out_n
rx_ll_dst_rdy_in_n
rx_fifo_status[3:0]
Output
Output
Output
Output
Input
rx_ll_clock Data read from FIFO
rx_ll_clock Start of frame indicator
rx_ll_clock End of frame indicator
rx_ll_clock Source ready indicator
rx_ll_clock Destination ready indicator
rx_ll_clock FIFO memory status
Output
1-Gigabit Ethernet MAC v8.5 User Guide
129
UG144 April 24, 2009
Download from Www.Somanuals.com. All Manuals Search And Download.
-- DISCONTINUED PRODUCT --
Appendix A: Using the Client-Side FIFO
R
Overview of LocalLink Interface
Data Flow
Data is transferred on the LocalLink interface from source to destination, with the flow
being governed by the four active low control signals sof_n, eof_n, src_rdy_n, and
dst_rdy_n. The flow of data is controlled by the src_rdy_nand dst_rdy_nsignals.
Only when these signals are asserted simultaneously is data transferred from source to
destination. The individual packet boundaries are marked by the sof_nand eof_n
signals. For more information on the LocalLink interface see to Xilinx Application Note
XAPP691, “Parameterizable LocalLink FIFO” available at
Figure A-2 shows the transfer of an 8-byte frame.
clock
data[7:0]
sof_n
0
1
2
3
4
5
6
7
eof_n
src_rdy_n
dst_rdy_n
Figure A-2: Frame Transfer across LocalLink Interface
dst_rdy_nsignals are used to control the flow of data across the interface.
clock
0
1
2
3
4
data[7:0]
sof_n
eof_n
src_rdy_n
dst_rdy_n
Figure A-3: Frame Transfer with Flow Control
130
1-Gigabit Ethernet MAC v8.5 User Guide
UG144 April 24, 2009
Download from Www.Somanuals.com. All Manuals Search And Download.
-- DISCONTINUED PRODUCT --
R
Functional Operation
Functional Operation
Clock Requirements
The FIFO is designed to work with rx_clkand tx_clkrunning at MAC clock speeds up
to 125 MHz. The rx_ll_clockshould be no slower than the rx_clk. The tx_ll_clock
should be no slower than the clock on the transmitter client interface divided by 2. For this
reason, it is suggested that the rx_ll_clockand tx_ll_clockare always 125 MHz or
faster.
Receive FIFO
The receive FIFO is built around two Dual Port Block RAMs giving a memory capacity of
4096 bytes.
The receive FIFO takes data from the client interface of the GEMAC core and converts it
GEMAC receive client interface. If the frame is marked as good by the GEMAC, that frame
is presented on the LocalLink interface for reading by the user. If the frame is marked as
bad, it is dropped by the FIFO.
If the receive FIFO memory overflows, the frame currently being received is dropped,
regardless of whether it is a good or bad frame,. The signal rx_overflowis then asserted.
Situations in which the memory may overflow are:
•
The FIFO may overflow if the user is reading data from the FIFO at a slower data rate
than data is being written in from the MAC receiver.
•
The FIFO size of 4096 bytes limits the size of the frames that it can store without error.
If a frame is larger than 4000 bytes, the FIFO may overflow and data will be lost. For
this reason, it is recommended that the FIFO not be used with the GEMAC in jumbo
frame mode for frames larger than 4000 bytes.
Transmit FIFO
The transmit FIFO is built around two Dual Port block RAMs, giving a total memory
capacity of 4096 byes of frame data.
The transmit FIFO accepts frames in LocalLink format and stores them in block RAM for
transmission via the GEMAC. When a full frame has been written into the transmit FIFO,
the FIFO will present data to the MAC transmitter. On receiving the tx_acksignal from
the MAC, the rest of the frame is transmitted to the MAC.
VHDL
The generic FULL_DUPLEX_ONLYis provided to allow the removal of logic and
performance constraints necessary for half-duplex operation when using with the Xilinx
Tri-Mode Ethernet MAC core. This generic can always be set to true when the FIFO is used
with the GEMAC.
1-Gigabit Ethernet MAC v8.5 User Guide
131
UG144 April 24, 2009
Download from Www.Somanuals.com. All Manuals Search And Download.
-- DISCONTINUED PRODUCT --
Appendix A: Using the Client-Side FIFO
R
Verilog
The compiler directive FULL_DUPLEX_ONLYis defined to allow for removal of logic and
performance constraints that are necessary only in half-duplex operation, that is, when
using with the Tri-Mode Ethernet MAC core. This directive can always be defined when
the FIFO is used with the GEMAC.
The FIFO has two signal inputs specific to half-duplex operation, tx_collisionand
tx_retransmit. These signals are provided to make the FIFO compatible with both the
1-Gigabit Ethernet MAC and Tri-Mode Ethernet MAC cores, and should be tied to logic 0
when using the FIFO with the GEMAC core.
If the FIFO memory fills up, the dst_rdy_out_nsignal is used to halt the LocalLink
interface writing in data until space becomes available in the FIFO. If the FIFO memory
fills up but no frames are available for transmission (for example, if a frame larger than
4000 bytes is written into the FIFO), the FIFO may assert the tx_overflowsignal and
continue to accept the rest of the frame from the user. The overflow frame will be dropped
by the FIFO. This ensures that the LocalLink interface does not lock up. For this reason, it
is recommended that the FIFO not be used with the GEMAC in jumbo frame mode for
frames larger than 4000 bytes.
Expanding Maximum Frame Size
The transmit FIFO size is optimized to allow line rate transmission of maximum size
Ethernet frames at 1518 bytes in half-duplex operation.
When using the FIFO in full-duplex operation, the full block RAM capacity can be utilized
at all times. As a whole frame must be stored in the FIFO block RAM before being
presented to the MAC transmitter, the maximum size frame that can be handled is
determined by the memory capacity of the FIFO (in this case 4000 bytes).
Both transmit and receive FIFO sizes can be expanded by the user to handle larger frame
sizes. This can be done by instantiating further block RAMs into the FIFO design,
expanding the block RAM address signals, and adding the necessary control signals. The
HDL source files provide guidance in the comments on how to achieve this.
User Interface Data Width Conversion
Conversion of the user interface 8 bit data path to a 16, 32, 64 or 128 bit data path can be
made by connecting the LocalLink interface directly to the Parameterizable LocalLink
132
1-Gigabit Ethernet MAC v8.5 User Guide
UG144 April 24, 2009
Download from Www.Somanuals.com. All Manuals Search And Download.
-- DISCONTINUED PRODUCT --
R
Appendix B
Core Verification, Compliance, and
Interoperability
The GEMAC core has been verified with extensive simulation and hardware testing.
Verification by Simulation
A highly parameterizable transaction-based test bench (not part of the core deliverables)
was used to test the core. Tests include:
•
•
•
•
•
Register Access
MDIO Access
Frame Transmission and error handling
Frame Reception and error handling
Address Filtering
Hardware Verification
The GEMAC core has been tested in a variety of hardware test platforms at Xilinx to
include a variety of parameterizations, including the following.
The core has been tested with the Ethernet 1000BASE-X PCS/PMA or SGMII core from
built around these cores, including a back-end FIFO capable of performing a simple ping
function, and a test pattern generator. Software running on the embedded PowerPC®
processor was used to provide access to all configuration, status and statistical counter
registers. Version 3.0 of this core was taken to the University of New Hampshire
Interoperability Lab (UNH IOL) where conformance and interoperability testing was
performed.
The core has been tested with an external 1000BASE-T PHY device. The MAC was
connected to the external PHY device using GMII, RGMII, and SGMII (in conjunction with
the Ethernet 1000BASE-X PCS/PMA or SGMII core).
The core has been tested with the Ethernet Statistics core from Xilinx, following the
1-Gigabit Ethernet MAC v8.5 User Guide
133
UG144 April 24, 2009
Download from Www.Somanuals.com. All Manuals Search And Download.
-- DISCONTINUED PRODUCT --
R
Appendix B: Core Verification, Compliance, and Interoperability
134
1-Gigabit Ethernet MAC v8.5 User Guide
UG144 April 24, 2009
Download from Www.Somanuals.com. All Manuals Search And Download.
-- DISCONTINUED PRODUCT --
R
Appendix C
Calculating DCM Phase-Shifting
DCM Phase-Shifting
A DCM is used in the receiver clock path to meet the input setup and hold requirements
with an external GMII implementation in Spartan®-3, Spartan-3E, Spartan-3A and
Virtex®-4 devices (see “Spartan-3, Spartan-3E, Spartan-3A and Virtex-4 Devices,” on page
63.)
In these cases, a fixed phase-shift offset is applied to the receiver clock DCM to skew the
clock. This performs static alignment by using the receiver clock DCM to shift the internal
version of the receiver clock such that its edges are centered on the data eye at the IOB DDR
flip-flops. The ability to shift the internal clock in small increments is critical for sampling
high-speed source synchronous signals such as RGMII. For statically aligned systems, the
DCM output clock phase offset (as set by the phase shift value) is a critical part of the
system, as is the requirement that the PCB is designed with precise delay and impedance-
matching for all the GMII/RGMII receiver data bus and control signals.
You must determine the best DCM setting (phase-shift) to ensure that the target system has
the maximum system margin to perform across voltage, temperature, and process
(multiple chips) variations. Testing the system to determine the best DCM phase-shift
setting has the added advantage of providing a benchmark of the system margin based on
the UI (unit interval or bit time).
System margin is defined as the following:
System Margin (ps) = UI(ps) * (working phase-shift range/128)
Finding the Ideal Phase-Shift
Xilinx cannot recommend a singular phase-shift value that is effective across all hardware
platforms, and does not recommend attempting to determine the phase-shift setting
empirically. In addition to the clock-to-data phase relationship, other factors such as
package flight time (package skew) and clock routing delays (internal to the device) affect
the clock-to-data relationship at the sample point (in the IOB) and are difficult to
characterize.
Xilinx recommends extensive investigation of the phase-shift setting during hardware
integration and debugging. The phase-shift settings provided in the example design UCF
are placeholders, and work successfully in back-annotated simulation of the example
design.
1-Gigabit Ethernet MAC v8.5 User Guide
135
UG144 April 24, 2009
Download from Www.Somanuals.com. All Manuals Search And Download.
-- DISCONTINUED PRODUCT --
Appendix C: Calculating DCM Phase-Shifting
R
Perform a complete sweep of phase-shift settings during your initial system test. Use only
positive (0 to 255) phase-shift settings, and use a test range that covers a range of no less
than 128, corresponding to a total 180 degrees of clock offset. This does not imply that 128
phase-shift values must be tested; increments of 4 (52, 56, 60, etc.) correspond to roughly
one DCM tap, and consequently provide an appropriate step size. Additionally, it is not
necessary to characterize areas outside the working phase-shift range.
At the edge of the operating phase-shift range, system behavior changes dramatically. In
eight phase-shift settings or less, the system can transition from no errors to exhibiting
errors. Checking the operational edge at a step size of two (on more than one board) refines
the typical operational phase-shift range. Once the range is determined, choose the
average of the high and low working phase-shift values as the default. During the
production test, Xilinx recommends that you re-examine the working range at corner case
operating conditions to determine whether any final adjustments to the final phase-shift
setting are needed.
You can use the FPGA Editor to generate the required test file set instead of resorting to
multiple PAR runs. Performing the test on design files that differ only in phase-shift setting
prevents other variables from affecting the test results. FPGA Editor operations can even
be scripted further, reducing the effort needed to perform this characterization.
136
1-Gigabit Ethernet MAC v8.5 User Guide
UG144 April 24, 2009
Download from Www.Somanuals.com. All Manuals Search And Download.
-- DISCONTINUED PRODUCT --
R
Appendix D
Core Latency
Transmit Path Latency
As measured from a data octet accepted on tx_data[7:0]of the transmitter client-side
interface, until that data octet appears on gmii_txd[7:0]of the physical side GMII style
interface, the latency through the core in the transmit direction is 9 clock periods of
gtx_clk.
Receive Path Latency
As measured from a data octet accepted on gmii_rxd[7:0]of the physical side GMII
style interface, until that data octet appears on rx_data[7:0]of the receiver client-side
interface, the latency through the core in the receive direction is 9 clock periods of
gmii_rx_clk.
1-Gigabit Ethernet MAC v8.5 User Guide
137
UG144 April 24, 2009
Download from Www.Somanuals.com. All Manuals Search And Download.
-- DISCONTINUED PRODUCT --
R
Appendix D: Core Latency
138
1-Gigabit Ethernet MAC v8.5 User Guide
UG144 April 24, 2009
Download from Www.Somanuals.com. All Manuals Search And Download.
|