RE: RF/LF
Boaz,
Of course it is okay. I was explaining what changed in regard to LF/RF
generation.
One of the changes is in regard to clock compensation. The PCS used to only
delete idles for clock compensation. Now it will delete idles or Sequence
ordered sets (they aren't called Pulse Ordered Sets any more so if you want
to use an acronym it would be SOS, but normally we either write it out or
use the notation ||Q|| because ||S|| is already assigned to Start). One
other nit. In the 10GBASE-R PCS Receive process, an LF that hasn't passed
through an XGXS may look like continuous sequence ordered sets or it may
look like sequence ordered sets plus a few idles inserted for clock
compensation.
Regards,
Pat
-----Original Message-----
From: Boaz Shahar [mailto:boazs@xxxxxxxxxxxx]
Sent: Wednesday, January 31, 2001 12:13 PM
To: 'THALER,PAT (A-Roseville,ex1)'; Boaz Shahar; Tim Warland; 802.3ae
Subject: RE: RF/LF
I think that it still ok: For Clock tolerance compensation you may remove
the LF-POS from the data stream (Or add Idle). And in the Receive process,
when 64/66 words are converted back to continuous POS, the PCS machine of
the PHY-XGXS will convert it to legal XAUI data stream, where the POS is
substituted after each /A/. Wasnt that the intention?
> -----Original Message-----
> From: THALER,PAT (A-Roseville,ex1) [mailto:pat_thaler@xxxxxxxxxxx]
> Sent: Tuesday, January 30, 2001 8:03 PM
> To: Boaz Shahar; THALER,PAT (A-Roseville,ex1); Tim Warland; 802.3ae
> Subject: RE: RF/LF
>
>
> Boaz,
>
> If sending continuous ordered sets, the 64B/66B PCS still
> needs to be able
> to do clock compensation. Otherwise it will have FIFO overruns.
>
> Pat
>
> -----Original Message-----
> From: Boaz Shahar [mailto:boazs@xxxxxxxxxxxx]
> Sent: Monday, January 29, 2001 11:02 PM
> To: 'THALER,PAT (A-Roseville,ex1)'; Boaz Shahar; Tim Warland; 802.3ae
> Subject: RE: RF/LF
>
>
> Pat,
> If the PHY-XGXS discovers LF fault on the Tx XAUI path and generate
> continuous POS, why cant the PCS just transfer it as is? In
> the worst case,
> the other side XAUI will transmit continuous POS (P+/P-) and even if
> SYNC/DESKEW is lost in the other side, its still interpreted
> as LF there.
> Boaz
> > -----Original Message-----
> > From: THALER,PAT (A-Roseville,ex1) [mailto:pat_thaler@xxxxxxxxxxx]
> > Sent: Monday, January 29, 2001 9:04 PM
> > To: Boaz Shahar; THALER,PAT (A-Roseville,ex1); Tim Warland; 802.3ae
> > Subject: RE: RF/LF
> >
> >
> > Boaz,
> >
> > That applies when there is no XAUI and it also applies when
> > the Phy XGXS
> > attached to the PCS is generating LF. The XGXS/10GBASE-X
> > receiver now sends
> > continuous LF to the XGMII interface when it detects a loss
> > of alignment.
> >
> > Pat
> >
> >
> > > -----Original Message-----
> > > From: THALER,PAT (A-Roseville,ex1) [mailto:pat_thaler@xxxxxxxxxxx]
> > > Sent: Wednesday, January 24, 2001 11:19 PM
> > > To: Tim Warland; 802.3ae
> > > Subject: RE: RF/LF
> > >
> > >
> > >
> > > For the change to the receive state machine, see comment 1159
> > > which removes
> > > the detection of |P| (which in d2.1 will be |Q|) which is the
> > > behavior Boaz
> > > mentions. The decision to send constant ordered sets on the
> > XGMII when
> > > detecting a local fault was made. Comment 267 was part of
> > > that, but the full
> > > resolution isn't in that comment's response. Also see comment
> > > 592 and 588.
> > > These resolutions also impact clause 49 because it will now
> > > need to be able
> > > to drop pulse ordered sets,
> >
> > You probably mean when there is no XAUI, right?
> >
> > Boaz
> >
> >
> >
>