RE: [802.3ae] WIS test Pattern PRBS Generator Polynomial
Bruce,
The 2**31 PRBS has been reinstated because there is some evidence, theory
and experimental, that the LAN test patterns in D4.0 are not always good
predictors of performance and so not appropriate for a conformance test,
though they have merit in diagnostic tests. The committee, led by the
editors, decided to make PRBS31 optional in WAN as well as LAN because they
could freeze clauses 49 and 50 this way, before any resolution of any
discussion of the relative merits of 2**23 and 2**31 in WAN test patterns -
in fact, this way maybe we don't actually need to discuss or resolve that
before publishing the standard. As 2**31 is optional its reinstatement
doesn't obsolete any existing silicon.
Tom's suggestion of a (non-normative, superseded by O.150 in case of
conflict) statement of the polynomial has the merit that the industry
continues to use just one PRBS23. By referring to O.150 normatively, if we
slip up in drafting our statement of the polynomial we will not have done
any damage. I can't see why the polynomial in O.150 would change. Clause
1.3, Normative references, says "At the time of publication, the editions
indicated were valid. All standards are subject to revision, and parties to
agreements based on this standard are encouraged to investigate the
possibility of applying the most recent editions of the standards indicated
below. Members of IEC and ISO maintain registers of currently valid
International Standards." By referring to O.150 rather than O.172 we can
take out the double indirection you complained of.
To editor Clause 1: there is a newer G.957, prepublished 10/00 and probably
finally approved by now.
Piers
> -----Original Message-----
> From: Nepple, Bruce [mailto:bnepple@xxxxxxxxxxxxxxxxxxx]
> Sent: 04 February 2002 20:49
> To: stds-802-3-hssg@xxxxxxxxxxxxxxxxxx
> Subject: RE: [802.3ae] WIS test Pattern PRBS Generator Polynomial
>
> I have no idea what concepts are being floated regarding the
> 2**31 PRBS,
> why it is required for 64/66 testing, and why it will be
> required for WIS, or
> whether it will be optional in all cases. I scanned the
> serial-PMD email and saw
> nothing definitive.
>
> Regarding the current implementation,
> I'm not sure why the PRBS generator polynomial is not
> specifically specified for
> 802.3ae. The idea that it be tied to a specification in
> another forum through
> two layers of documents seems impractical. And, O.150 says
> that the specified
> polynomial MAY be used to generate a 2**23-1 sequence.
>
> 802.3ae specifies that it be reset every frame, inverted
> every frame, but not
> what polynomial to use. Seems inconsistant. And, if ITU
> decides to recommend
> another polynomial, do we really want to change all WIS hardware?
>
> Bruce
>
> > -----Original Message-----
> > From: Tom Alexander [mailto:Tom_Alexander@xxxxxxxxxxxxxx]
> > Sent: Monday, February 04, 2002 12:29 PM
> > To: Nepple, Bruce; stds-802-3-hssg@xxxxxxxxxxxxxxxxxx
> > Subject: RE: [802.3ae] WIS test Pattern PRBS Generator Polynomial
> >
> >
> > Hi, Bruce,
> >
> > I think you have a point. To the best of my knowledge, the
> > polynomial to be used is as specified in O.150.
> >
> > I think it might be useful to the reader if we had a
> > (non-normative, superseded by O.150 in case of conflict)
> > statement that provided the polynomial itself as well as the
> > PRBS length. If you could submit such a comment against the
> > next draft, that would be very useful. Ditto for the 2**31 -
> > 1 polynomial.
> >
> > Cheers,
> >
> > - Tom
> >
> > -----Original Message-----
> > From: Nepple, Bruce [mailto:bnepple@xxxxxxxxxxxxxxxxxxx]
> > Sent: Monday, February 04, 2002 11:43 AM
> > To: stds-802-3-hssg@xxxxxxxxxxxxxxxxxx
> > Subject: RE: [802.3ae] WIS test Pattern PRBS Generator Polynomial
> >
> >
> > Ouch --- I missed that comment. Thanks for pointing it out.
> > My issue was that there are MANY 2**23-1 polynomials. Which one
> > specifically is preferred or required was not specified, and there
> > was no comment indicating that the choice was implementation
> > dependent or to be determined in the future.
> >
> > Bruce
> >
> > > -----Original Message-----
> > > From: Tom Alexander [mailto:Tom_Alexander@xxxxxxxxxxxxxx]
> > > Sent: Monday, February 04, 2002 9:36 AM
> > > To: Nepple, Bruce; stds-802-3-hssg@xxxxxxxxxxxxxxxxxx
> > > Subject: RE: [802.3ae] WIS test Pattern PRBS Generator Polynomial
> > >
> > >
> > > Bruce,
> > >
> > > The PRBS for the mixed-frequency test pattern indeed uses a
> > > 2**23 -1 polynomial. This is explicitly described in the WIS
> > > clause (see lines 39-41 and 47-48 of page 397, lines 9-13 of
> > > page 398, and Figure 50-13 (payload field) on page 398). All
> > > these refer to Draft 4.0.
> > >
> > > Note that in the next draft of the WIS spec, there is a new
> > > (optional) additional test pattern introduced, as per the
> > > resolution of Comment #72 against Draft 4.0; this pattern
> > > uses a 2**31 -1 polynomial.
> > >
> > > Thanks,
> > >
> > > - Tom Alexander
> > > WIS Scribe
> > >
> > >
> > > -----Original Message-----
> > > From: Nepple, Bruce [mailto:bnepple@xxxxxxxxxxxxxxxxxxx]
> > > Sent: Friday, February 01, 2002 9:36 PM
> > > To: stds-802-3-hssg@xxxxxxxxxxxxxxxxxx
> > > Subject: [802.3ae] WIS test Pattern PRBS Generator Polynomial
> > >
> > >
> > > I could find no definition of the WIS test Pattern PRBS
> > > generator polynomial in Clause 50 (or any other clause, for
> > > that matter). I did find a reference to O.172 and in O.172 I
> > > found a reference to O.150, and in O.150 I found a definition
> > > for a 2**23 -1 length PRBS. (Section 5.6). Is that the
> > > required generator for WIS? Is there a reason that it is not
> > > specified? Or, did I just manage to overlook it in the
> > > specification? (wouldn't be the first time).
> > >
> > > O.150 Section 5.6 specifies a 23 stage shift register with
> > > taps from stage 23 and stage 18 added (mod-2) and applied to
> > > the first stage. (I'd call it a x**23 + x**18 + 1 generator
> > > with input data = 0, but I'm not a math dude)
> > >
> > > Bruce
> > >
> >
>