![Xilinx Virtex-4 RocketIO Скачать руководство пользователя страница 124](http://html1.mh-extra.com/html/xilinx/virtex-4-rocketio/virtex-4-rocketio_user-manual_3383739124.webp)
124
Embedded Tri-Mode Ethernet MAC User Guide
UG074 (v2.2) February 22, 2010
Chapter 4:
Physical Interface
R
The FPGA RX Elastic Buffer Requirement
illustrates a simplified diagram of a common situation where the Ethernet
MAC in SGMII mode is interfaced to an external PHY device. Separate oscillator sources
are used for the FPGA and the external PHY. The Ethernet specification uses clock sources
with a 100 ppm tolerance. In
, the clock source for the PHY is slightly faster than
the clock source to the FPGA. Therefore, during frame reception, the RX elastic buffer
(implemented in the MGT in this example) starts to fill up.
Following frame reception in the interframe gap period, idles are removed from the
received datastream to return the RX elastic buffer to half-full occupancy. This task is
performed by the clock correction circuitry (see
, V
irtex-4 RocketIO Multi-Gigabit
Transceiver User Guide
).
Assuming separate clock sources, each with 100 ppm tolerance, the maximum frequency
difference between the two devices can be 200 ppm, which translates into a full clock
period difference every 5000 clock periods.
Relating this information to an Ethernet frame, there is a single byte of difference every
5000 bytes of received frame data, causing the RX elastic buffer to either fill or empty by an
occupancy of one.
The maximum Ethernet frame (non-jumbo) has a 1522-byte size for a VLAN frame:
•
At 1 Gb/s operation, this size translates into 1522 clock cycles
•
At 100 Mb/s operation, this size translates into 15220 clock cycles (because each byte
is repeated 10 times
•
At 10 Mb/s operation, this size translates into 152200 clock cycles (because each byte
is repeated 100 times)
Considering the 10 Mb/s case, 152200/5000 = 31 FIFO entries are needed in the elastic
buffer above and below the halfway point to insure that the buffer does not underflow or
overflow during frame reception. This assumes that frame reception begins when the
buffer is exactly half full.
Figure 4-17:
SGMII Implementation Using Separate Clock Sources
Ethernet MAC
MGT
RX
El
as
tic
B
u
ffer
S
GMII Link
FPGA
125 MHz + 100 ppm
125 MHz - 100 ppm
Twi
s
ted
Copper
P
a
ir
10BA
S
E-T
100BA
S
E-T
1000BA
S
E-T
PHY
TXP/TXN
RXP/RXN
UG074_3_81_012508
www.BDTIC.com/XILINX