Re: Annex 48A Draft 2.1 peer review
Tom,
Thanks for your comments, please see my responses below:
Tom Lindsay wrote:
>
> Rich - a few comments.
>
> 1. FYI - the low, high, and mixed frequency patterns may be turn out to informational only. Is that what your note implies? The jitter group is still debating this (I have an action to start an email thread on this issue). Relatedly, the jitter group is still wrestling with whether "CJTPAT" is appropriate for testing jitter output in addition to tolerance.
This should be OK. The Annex 48A title is "Jitter Test Patterns". A
separate but related Annex is 48B which should be titled "Jitter Test
Methodology". The latter annex should determine which test patterns are
applicable to test for jitter compliance.
> 2. The pattern I proposed includes the elements of FC's JTPAT, but is quite different than FC's CJTPAT. The relationship (and reference) of the proposed pattern back to MJS is pretty weak. The 2 patterns are quite a bit different, and I think we should not use "Continuous Jitter Tolerance Test Pattern" (which acronyms to CJTPAT, although in FC, the C stood for Compliant).
>
> So, to avoid this confusion with FC's CJTPAT, and to avoid the unresolved question about suitability for jitter output testing, I would suggest dropping reference to tolerance (T) and simply renaming the pattern to JPAT for the payload and CJPAT for the overall frame pattern. Other ideas?
OK. I'll rename it to CJPAT. Is CRPAT still OK with C=Continuous? It's
already in the Gigabit Ethernet standard (Annex 36A) this way.
> 3. The phase shifting in the CDR occurs during the long runs of 7Es and B5s. The EBs and F4s, etc. are what provide the abrupt phase jumps. So, I suggest relabeling the list as such.
OK.
> 4. This comment is more for the jitter group, but here it is anyway...
> The core of RPAT repeats a specific block of 12 bytes 16 times. This can hardly be called random. Vixel has also developed a pattern that is based on randomly picking values between 0 & 255 to build up the pattern. Indeed, it has much broader/uniform spectral content than than RPAT. Its only drawback is difficulty in documentation (can't use the word "repeat" 'cause it doesn't...). The pattern we use is for Fibre channel and is 2048 bytes including 6x IDLE, SOF-, Header/payload, CRC, and EOF+. I would be willing to make it available for standardization if interested.
I guess a lot depends on the application of CRPAT for jitter testing.
It's certainly more random than a packet of all 0's or 1's. I don't
believe that using the exact Fibre Channel pattern is a significantly
better replacement since I have to ask the question: Better for what?
Updated Annex 48A is attached.
Best Regards,
Rich
--
> Have fun.
>
> Tom LIndsay
> Vixel
> 425/806-4074
>
> Rich Taborek wrote:
>
> > Ladies and Gentlemen,
> >
> > Attached is the file I plan to submit for Draft 2.1 Annex 48A. This
> > annex contains jitter test patterns for 10GBASE-X, which addresses the
> > 10GBASE-LX4 PMD as well as XAUI. Please give it a quick once over and
> > return any comments to me. My deadline to submit it for D2.1 is this
> > coming Sunday.
> >
> > Separately, I'm planning to generate a template/outline of Annex 48B for
> > D2.1 This annex will address jitter test methodologies for 10GBASE-X.
> >
> > --
> >
> > Best Regards,
> > Rich
-------------------------------------------------------
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
anx48.pdf