Thread Links Date Links
Thread Prev Thread Next Thread Index Date Prev Date Next Date Index

RE: Long distance links



Title: RE: Long distance links

Now I'm confused and my brain is starting to hurt.  Why would the delay be dependent on the link length if we're only doing full duplex?  I must be missing something in that calculation. 

My understanding is that if 802.3 MAC was designed to have some form of pacing mechanism to guarantee a data rate of less than or equal to 9.58464 Gb/s for the WAN PHY, then the WAN PHY would only require enough buffering to compensate for the insertion of header information.  If that's the case, then we'd only have to compensate for one row of an OC-192 frame (assuming that bytes will be received during the transmission of a header and stored until header transmission is complete, and the FIFO will be emptied before the next row) which would be equivalent to storing 640 bytes (17,280 byte-columns * (9.95328 Gbps - 9.58464 Gbps) / 9.95328 Gbps).  Because the instantaneous drain rate (9.95328 Gb/s) is always greater than the average fill rate (9.58464 Gb/s), the FIFO only requires storing the transmission bytes during times it cannot transmit plus a few extra bytes for synchronization of control signals.

On the receive side, the FIFO only has to be able to store the maximum size frame in case the last byte of the frame doesn't occur until after the reception of the overhead.  This FIFO could also be optimized, but that is implementation specific.

Thanks,
Brad

Brad Booth
bbooth@xxxxxxxxxx
Level One Communications, Austin Design Center
(512) 407-2135 office
(512) 589-4438 cellular