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

A Question about "XGMII Min Inter Packet Gap"




Hi Steve,

Has your proposal, "Pick IPG variation" with minimum IPG of 9 bytes
on XGMII been adopted (voted) by the 802.3-hssg. Is this the direction 
XGMII min IPG going?

Thanks
-Sanjeev Mahalawat


At 12:24 PM 07/18/2000 -0700, you wrote:
>
>Boaz,
>
>The random /A/ or /K/ in the first IPG column will statistically provide
>opportunities to obtain and maintain word sync and lane de-skew within a
>stream of packets.  It is more important to have an /R/ column appearing
>between packets deterministically because you need to know a worst case
>spacing between /R/'s in order to properly size clock tolerance adjustment
>buffers.
>
>The only way to deterministically guarantee /A/ and /K/ and /R/ between
>every packet is to require 3 full columns *following the column containing
>the /T/ character*.  This means minimum IPGs of 13-16 bytes.  You cannot get
>/A/ and /K/ and /R/ columns in a 12 byte IPG.
>
>Steve
>
>-----Original Message-----
>From: Boaz Shahar [mailto:boazs@mysticom.com]
>Sent: Monday, July 17, 2000 9:00 AM
>To: 'Brown, Ben [BAY:NHBED:DS48]'; 802.3ae
>Subject: RE: A Question about "Inter Packet Gap and SOP Lane alignment"
>
>
>
>Ben and Rich,
>Thank you for the explanations. However, there are some things which I'm not
>so sure about:
>
>1.Refering  the state diagram at the end, after transmitting /A/ or /K/ and
>/R/ columns, we randomize the distance to the next A (Which is uniformly
>distributed between 16 to 31).
>
>Now, if it happened to be /K/, /R/ columns, then the next /A/ could be a
>distance of 16 columns away (Or may never come, if a new packet starts
>before). And, on the contrary, if it was /A/, /R/ and the next packet starts
>immediately, then the next /K/ can be far away. Since the synchronization
>condition requires 3x/K/ per lane and then /A/ for de skewing, due to this
>randomization, if you turn on your receiver at the "middle", you don't know
>when you will be synchronized! Actually, your time for synchronization is a
>random variable.
>
>I think this is a disadvantage relative to the previous proposed /A/, /K/,
>/R/ minimal sequence. Here, you can bound the time for synchronization,
>avoiding the "Random Time" parameter.
>
>At least, if we MUST support this 12 bytes, I would suggest that the third
>column will always be "the other one", i.e. if the first is /A/, and there
>is a third column, it will always be /K/. And if the first is /K/, let the
>third always be /A/, and only then start the randomization process. 
>
>In addition, I saw that Rich mentioned again the minimal gap of /A/,/K/,/R/,
>so it still an option? (The MIN IPG is 12 to 16 in this case, instead of 9
>to 12, and the "time to sync" is deterministic.)
>
>Boaz
>
>
>
>
>
>> -----Original Message-----
>> From: Brown, Ben [BAY:NHBED:DS48] [mailto:bebrown@baynetworks.com]
>> Sent: Sunday, July 16, 2000 10:15 PM
>> To: 802.3ae
>> Subject: Re: A Question about "Inter Packet Gap and SOP Lane 
>> alignment"
>> 
>> 
>> 
>> 
>> Rich,
>> 
>> Allow me to clean up a few editing errors to avoid confusion:
>> 
>> Rich Taborek wrote:
>> > 
>> > 1) No change to EOP column which contains the first 1 to 4 
>> columns of the IPG.
>> > Note that the EOP counts towards IPG.
>> 
>> The EOP column contians the first 1 to 4 BYTES of the IPG.
>> 
>> > 2) Transmit /A/ or /R/ randomly while maintaining minimum 
>> 16 column /A/ spacing.
>> 
>> Transmit /A/ or /K/ randomly.
>> 
>> Enjoy,
>> Ben
>> 
>> Rich Taborek wrote:
>> > 
>> > Boaz,
>> > 
>> > Your question is answered in a presentation made at the La 
>> Jolla meeting:
>> > "8B/10B Idle Pattern for 12-byte IPG" by
>> > Rich Taborek and Don Alderrou of nSerial,
>> > 
>> http://grouper.ieee.org/groups/802/3/ae/public/jul00/taborek_1
>> _0700.pdf
>> > 
>> > This proposal modifies the fixed starting /A/K/R/ Idle 
>> pattern for all
>> > applications of 8B/10B encoded 4-lane streams was accepted 
>> into the set of
>> > baseline proposals along with the core XAUI/XGXS proposal. 
>> The specific change
>> > is as follows:
>> > 
>> > 1) No change to EOP column which contains the first 1 to 4 
>> columns of the IPG.
>> > Note that the EOP counts towards IPG.
>> > 2) Transmit /A/ or /R/ randomly while maintaining minimum 
>> 16 column /A/ spacing.
>> > 3) Transmit /R/
>> > 4) Transmit /K/, /R/ or /A/ randomly while maintaining 
>> minimum 16 column /A/
>> > spacing until SOP detected.
>> > 
>> > Note that this proposal works for IPGs as small as 9. 
>> Detailed state machines
>> > are included at the end of the PDF file.
>> > 
>> > Best Regards,
>> > Rich
>> > 
>> > --
>> > 
>> > Boaz Shahar wrote:
>> > >
>> > > Hello Steve,
>> > > I have a question in regards to your presentation "Inter 
>> Packet Gap..." in
>> > > the last meeting:
>> > >
>> > > You suggest to use only 2 columns of Idle as a minimum 
>> for IPG. In previous
>> > > proposals (e.g.
>> > > 
>> http://grouper.ieee.org/groups/802/3/ae/public/may00/taborek_2
>> _0500.pdf
>> > > slide 14) the suggested gap is a minimum of 3 
>> columns-/A/, /K/, and /R/. So:
>> > > Can you please say which of them should be eliminated, or 
>> define a procedure
>> > > which say which type of Idle columns should be inserted?
>> > >
>> > > Thanks,
>> > > Boaz
>> > 
>> > -------------------------------------------------------
>> > Richard Taborek Sr.                 Phone: 408-845-6102
>> > Chief Technology Officer             Cell: 408-832-3957
>> > nSerial Corporation                   Fax: 408-845-6114
>> > 2500-5 Augustine Dr.        mailto:rtaborek@nSerial.com
>> > Santa Clara, CA 95054            http://www.nSerial.com
>> 
>> 
>> -- 
>> -----------------------------------------
>> Benjamin Brown
>> Router Products Division
>> Nortel Networks
>> 1 Bedford Farms,
>> Kilton Road
>> Bedford, NH 03110
>> 603-629-3027 - Work
>> 603-624-4382 - Fax
>> 603-798-4115 - Home
>> bebrown@nortelnetworks.com
>> -----------------------------------------
>> 
>