RE: [802.3ae] A question about D3.4/ 47, 48
Rich,
> -----Original Message-----
> From: Rich Taborek [mailto:rtaborek@xxxxxxxxxxxxx]
> Sent: Thursday, November 29, 2001 2:03 PM
> Cc: HSSG (E-mail)
> Subject: Re: [802.3ae] A Questuin about D3.4/ 47, 48
>
>
>
> Boaz,
>
> Please review the jitter tolerance mask explanatory material
> in Annex G
> of the Fibre Channel - Methodology for Jitter Specification.
> Here's the
> link to that document:
> ftp://ftp.t11.org/t11/member/fc/jitter_meth/99-151v2.pdf
In FC-MJS page 28 Figure 9, the max value for the amplitude of the
sinusoidal jitter is 1.5 UI. Note that the mask defines only the amplitude,
and not the phase of the jitter. When the phase of the Sinusoidal jitter
applied to a lane is exactly the opposite phase of the sinusoidal jitter
applied to another lane, the peak-to-peak jitter is translated to skew
between those two lanes, when the frequency of the jitter is low.
(BTW, being accurate here, the MJS mask do not define the jitter in
[0,42Khz])
However, while 1.5 UI is not too big, and can be absorbed or represented by
the line "PMA Rx" in the skew budget table D3.4/table 48-5, the value of
additional 8.5 UI skew is quite big and should be expressed there some how.
>
> All relatively low frequencies, jitter orders of magnitude
> above 22 kHz
> is completely tracked out by the CDR unit associated with a XAUI
> receiver on each lane independently.
This is exactly the reason that peak-to-peak sinusoidal jitter amplitude
generates a skew between lanes. Each of the CDRs tracking them
independently, and if Jitter applied with 180 degree phase shift, you get a
skew of 8.5 UI, as each of the CDRs take its lane to th eopposite direction.
Boaz
>
> Best Regards,
> Rich
>
> --
>
> Boaz Shahar wrote:
> >
> > Rich,
> > I do not think it has something to do with the 100ppm
> drift. Drift is
> > frequency deviation between two different clock domains,
> while jitter is the
> > deviation of a certain signal edge location from its
> nominal location,
> > regardless its frequency or other clock domain frequency.
> >
> > I think that the interpretation of jitter is that
> Jitter=The deviation of a
> > certain bit edge from its nominal location. So, saying
> Jitter of 8.5 UI in
> > frequency 0 is allowed, as in figure 47-5, is saying that a
> certain point
> > may be misslocated by 8.5 UI. This can happen even with DRIFT=0.
> >
> > For instance, suppose there is no skew at all, but lane 0
> is jittering to
> > the right by 4 bits, and lane 1 to the left by 4.5 bits,
> and this happens in
> > jitter frequency=0, that is, in a very slow way. This
> implies that this
> > situation is almost constant. So there is is a skew of 8.5
> bits between lane
> > 0 and lane 1 although if there was not any jitter, the skew
> would be 0.
> >
> > Otherwise, can somebody explain the meaning of the
> Sinusoidal jitter mask in
> > figure 47-5? What is happening there in the interval [0,22Khz]?
> >
> > Boaz
> >
> > > -----Original Message-----
> > > From: Rich Taborek [mailto:rtaborek@xxxxxxxxxxxxx]
> > > Sent: Thursday, November 29, 2001 3:46 AM
> > > To: Boaz Shahar
> > > Cc: HSSG (E-mail)
> > > Subject: Re: [802.3ae] A Questuin about D3.4/ 47, 48
> > >
> > >
> > > What??? 8B/10B has a max run length of 5. This translates
> to a lowest
> > > frequency component of 312.5 MHz. This is slightly higher
> > > than 22 KHz.
> > >
> > > The "slanted" portion of Figure 47-5 is the low frequency mask and
> > > corresponds to the +/-100 ppm XAUI clock tolerance. At
> the really low
> > > frequencies, I believe that the 8.5 UI corresponds to the
> > > number of bits
> > > that would have to be buffered in the case that clock tolerance
> > > compensation is performed for a packet length equivalent to
> > > 22 kHz. The
> > > 8.5 UI and the slanted line itself has no relevance if
> clock tolerance
> > > compensation is not performed. The 8.5 UI is only relevant on
> > > a per lane
> > > basis and has no significance lane to lane. Therefore, the 41
> > > bit deskew
> > > in Table 48-5 holds.
> > >
> > > Best Regards,
> > > Rich
> > >
> > > --
> > >
> > > Boaz Shahar wrote:
> > > >
> > > > In clause 47, 47.3.4.6 and figure 47-5, the sinusoidal
> > > jitter is 8.5 UI in
> > > > very low frequency (Interval [0,22Khz]). This means that
> > > there is additional
> > > > skew of 8.5 UI between lanes in the XAUI. That is included
> > > in Table 48-5
> > > > (Skew Budget)? In other words, while doing de-skewing, one
> > > should consider
> > > > 41+8.5 as the max deskew situation or just 41?
> > > > Thx.,
> > > > Boaz
>
> ---------------------------------------------------------
> Richard Taborek Sr. Intel Corporation
> XAUI Sherpa Intel Communications Group
> 3101 Jay Street, Suite 110 Optical Group Marketing
> Santa Clara, CA 95054 Santa Clara Design Center
> 408-496-3423 JAY1-101
> Cell: 408-832-3957 mailto:rich.taborek@xxxxxxxxx
> Fax: 408-486-9783 http://www.intel.com
>