Re: Test patterns ?
Mike,
I believe that CJTPAT for 8B/10B is functionally equivalent to PRBS with
CID with/without A1/A2. I just termed the latter patterns "CJTPAT-like".
I just didn't have the foggiest idea of the makeup of the CJTPAT
equivalent for SONET. I'm not proposing scrambling the CJTPAT pattern or
anything like that. Does this make sense or am I just throwing confusion
on the fire?
Best Regards,
Rich
--
Mike Dudek wrote:
>
> I am in general agreement with Peter's comments about a specific pattern for
> testing OMA and agree that 4 (or 8) 1's followed by 4 (or 8) O'a would be
> better than the 5 used for the 8B/10B codes. I don't think a CJTPAT like
> sequence is needed for the serial streams due to the scrambling used. The
> probability of this type of sequence occuring is infinitesimally low, and if
> it does occur the same sequence will not occur on a re-transmission. (I don't
> think this is the case for the 8B/10B system). I think the general conclusion
> of the jitter ad hoc was however that something more stressful than a PRBS was
> needed for jitter testing. A PRBS with CID (approx 66-72 bits) for the LAN
> phy, and a PRBS with CID and A1/A2 inserted for the WAN phy was my proposal.
> (Note that the PRBS with CID and A1/A2 inserted is I believe a SONET test
> pattern.)
>
> Rich Taborek wrote:
>
> > Peter,
> >
> > Good responses, thank you. I was just doing a sanity check. On your last
> > point, the Fibre Channel CJTPAT is exactly the type of pattern which is
> > less "BER friendly" than PRBS and is useful for stressing the receiver.
> > Among other patterns, it contains longer than average durations of long
> > sequences of ones and zeros. I suggest developing such patterns for
> > serial PMD use.
> >
> > Happy Holidays,
> > Rich
> >
> > --
> >
> > Peter Öhlén wrote:
> > >
> > > Rich and others,
> > >
> > > It is true that 4 or 5 consecutive digits is considerably less than the
> > > maximum 66 that can occur using 64B/66B coding. However, my rationale
> > > for the 1GHz square wave test pattern for extinction ratio and OMA is
> > > that I think it is good to decouple extinction ratio and OMA from the
> > > vertical eye closure penalty. If a long run length square wave (eg.
> > > 66x"0"+"66x"1") was used, baseline wander would be part of OMA or
> > > extinction ratio. I think it is better to have baseline wander be a part
> > > of the eye closure, but I am open for other arguments.
> > >
> > > 4, 5 or 8 ?? For the 10G serial PMDs, I think that it is easier to
> > > implement something that is a power of 2. While 5 would still be simple,
> > > 4 (or 8) is simpler.
> > >
> > > I can see that PRBS is very "BER-friendly" compared to some kind of real
> > > traffic. Is the scrambled serial 10GbE traffic of one those cases? I
> > > thought the PRBS patterns were quite representative of scrambled data.
> > > Even if this is true I can see the need to have other features in a test
> > > pattern, e.g. long sequences (> approx. 31) of ones or zeros which
> > > repeat more frequently than they would in a PRBS pattern and/or make a
> > > CID test.
> > >
> > > /Peter
> > >
> > > > -----Original Message-----
> > > > From: Rich Taborek [mailto:rtaborek@xxxxxxxxxxxxx]
> > > > Sent: den 27 december 2000 09:55
> > > > To: 'Serial PMD Ad Hoc Reflector'
> > > > Subject: Re: Test patterns ?
> > > >
> > > > Peter,
> > > >
> > > > Just some "food for thought comments":
> > > >
> > > > Öhlén Peter wrote:
> > > > >
> > > > > We need a few test patterns for serial PMDs, and I think it
> > > > > would be good to start listing the different measurements
> > > > > which need specific patterns, and what kind of patterns
> > > > > that would be good to use. In 1GbE 5 different test
> > > > > patterns seem to be defined, and a few of them are based on
> > > > > 8b/10b code groups which I don't think is a good idea for
> > > > > 10G serial PMDs.
> > > >
> > > > > 1. For extinction ratio and OMA measurements I think a
> > > > > repeating "11110000" pattern (4x"1" + 4x"0") would be a
> > > > > good idea. (Better than 5x"1" + 5x"0" as in 1G.)
> > > >
> > > > For 8B/10B, the extreme maximum run length is 5 and the
> > > > typical maximum run length is 4. For 64B/66B, the extreme
> > > > maximum run length is 66 and the typical maximum run length
> > > > is >>4. Since 64B/66B is employed for all 10G serial PMDs,
> > > > why would a repeating "11110000" pattern, equivalent to
> > > > a 1.25 GHz square wave, be "a good idea" for extinction ratio
> > > > and OMA measurements? Why would it be any better than a
> > > > "1111100000" pattern? Is it representative in any way of the
> > > > very wide dynamic range of 64B/66B signaling (relative to
> > > > the narrow dynamic range of 8B/10B)?
> > > >
> > > > > 2. For BER measurements I we should use something that is
> > > > > at least very similar to PRBS. This could be a true 2^n-1
> > > > > PRBS, or some pattern that is framed and scrambled by the
> > > > > PCS (and for the WAN PHY, also by the WIS). Right now I
> > > > > don't really see the full implications of choosing one
> > > > > over the other.
> > > >
> > > > PRBS is typically a very BER "friendly" sequence. I would
> > > > think that the pattern should be a bit more stressful.
> > > > Possibly something along the lines of a CJTPAT tailored to
> > > > 64B/66B code.
> > > >
> > > > > 3. Do we need a CID (consecutive identical digits) test,
> > > > > and what is the number of consecutive digits that is needed ?
> > > > > CID=66 can occur, but very infrequently.
> > > >
> > > > > Also, other patterns may be needed, especially for the jitter test.
> > > >
> > > > The 8B/10B patterns in 1000BASE-X include high frequency, low
> > > > frequency, mixed frequency and CRPAT. It is likely that CJTPAT will
> > > > be added to or substituted for CRPAT for 10GBASE-X.
> > > >
> > > > > ================================================================
> > > > > OPTILLION --- High speed optical transceivers
> > > > > Peter Öhlen Phone: +46 8 477 41 56
> > > > > Kronborgsgränd 9 MPh: +46 70 181 52 05
> > > > > S-164 87 Kista Fax: +46 8 477 41 51
> > > > > http://www.optillion.com mailto:peter.ohlen@xxxxxxxxxxxxx
-------------------------------------------------------
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