Re: XAUI/XGXS Question
The 39 bit maximum skew figure is deemed to be worst case and is already less
than 1/4 the Align code-group (/A/) spacing.
Skip column processing for an Ethernet link including two XAUI may result in the
removal of two /R/ columns, reducing /A/ spacing to 140 bits. The same removal
would not be required again for more than 13 maximum sized frames.
In summary, a 160-bit spacing is overkill and more than adequate. There is no
reason to increase the minimum /A/ spacing.
Boaz Shahar wrote:
> If MaxSkew=40 Bits and all /R/ columns are removed, then the size of the
> IPG is 8 characters (=80 bits) which is exactly twice the skew. In this
> case, there is ambiguity. Isn't it more safe to increase the gap between two
> /A/ columns to 32 (In fact any even number bigger than 16 will be fine)
> Best Regards
> > -----Original Message-----
> > From: Rich Taborek [mailto:rtaborek@xxxxxxxxxxxxx]
> > Sent: Monday, May 15, 2000 4:52 AM
> > To: HSSG
> > Subject: Re: XAUI/XGXS Question
> > Boaz,
> > The worst case XAUI/XGXS lane-to-lane skew anticipated is
> > approximately 39 bits.
> > Skew sources included in this figure are described in:
> > http://grouper.ieee.org/groups/802/3/ae/public/mar00/taborek_1
> > _0300.pdf, slide
> > 16. The same slide indicates that the minimum distance
> > between adjacent /A/'s is
> > 160-bits. Since /A/ spacing is more than 4 times the worst
> > case skew, there is
> > no ambiguity in determining which which /A/ code-groups are related.
> > The current proposal accounts for even the case of removal of
> > all /R/ columns.
> > Performing deskew on the column including the Start-of-Frame
> > delimiter requires
> > that both MAC devices transmit frames. This is not required
> > for the current
> > proposal since /A/s are present in the IPG and in the absence
> > of packets.
> > Best Regards,
> > Rich
> > --
> > Boaz Shahar wrote:
> > >
> > > Hi Rich,
> > > Recalling from your March presentation that the Max skew
> > between two lanes
> > > is 8 Characters (80 bit), and that the distance between two
> > adjacent /A/'s
> > > is 16 Characters.
> > >
> > > 1.So, in this case, how should one know Which /A/ is this?
> > (That is, if a
> > > certain lane exceeds the other in 8 characters or delayed
> > after the other in
> > > 8 characters it looks the same)
> > >
> > > 2.Suppose some /R/'s were dropped in the middle due to
> > clock tolerance
> > > compensation. The result is that the distance between two
> > adjacent A's is
> > > smaller than 16 characters. In this case its even more
> > problematic to
> > > distinguish between delayed lane to an exceeding one.
> > >
> > > Can it be solved by:
> > > 1.Enlarging the gap between /A/'s
> > > 2.Doing de-skew on the /S/ and /Dxx.y/ code-words
> > >
> > > Thank you,
> > > 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@xxxxxxxxxxx
Santa Clara, CA 95054 http://www.nSerial.com