Re: [10GBT] Discussions on precoder decisions
George Zimmerman writes:
>
> Sailesh -
> What you're seeing is a response from the PAM-12 group to fill the holes
> [ ... ]
>
George,
Your points are well made, and I agree fully. In my view, PAM8 is and has
been a non-starter, for just the reasons you mention, and I think our
collective time would be more constructively spent dealing with those few
legitimate PAM12 issues that Sailesh and others have identified, and letting
the remaining ones -- which have already been refuted several times -- rest
in peace.
GZ writes:
>
> I believe that at this point in the disucssion, there is general
> agreement that the performance at the two baud rates is similar...
>
Exactly. And this is the same point that Scott, Jose, myself, and several
others (including yourself) have made numerous times since the July session.
>
> ...with one having better ANEXT noise margin and the other having
> increased constellation point spacing.
>
Even this seems to be a little generous. Regarding the level separation
analysis as it pertains to EM susceptibility, as far as we have been able
to tell, PAM12 has about a dB margin over PAM8 at 100m, they're almost
identical at 55m, and PAM8 has about a dB over PAM12 at 0m. And as Sanjay
and others have pointed out, it is difficult to imagine why, in the real
world, one should be concerned about susceptibility margin differential
between the two schemes at 0m, given that the margin at 100m will be
overwhelmingly the limiting case. If the margin at 100m is sufficient to
pass the tests, then it's hard to see why it should matter one way or the
other which scheme has a dB better margin at 0m: They should both be operating
robustly, with margin to spare.
Btw, the source of the discrepancy between the two views on this issue --
one factor leading to the conundrum-rich situation that Dan Dove remarked
on earlier (http://www.ieee802.org/3/10GBT/email/msg01010.html) -- is simply
due to diametrically opposing assumptions about the ground rules for
susceptibility testing. Sailesh believes that
On Tue, 31 Aug 2004 20:14:19 -0400 Sailesh Rao wrote:
>
> Regulatory EMI tests are not conducted with worst-case alien NEXT
> environments set up in the range.
>
and his analysis about level spacing in [rao_1_0704] appears to reflect
this assumption.
We, and others, believe that exactly the opposite is true:
On Mon, 6 Sep 2004 07:35:13 -0700 Hugh Barrass wrote:
>
> The tests are generally conducted under "worst case" conditions. That
> is, conditions where emissions are highest for EMC testing, or where
> equipment is most susceptible for immunity testing.
>
http://www.emctech.com.au/stand2.html writes:
>
> "The EUT must be arranged for normal operation and in the most
> sensitive mode, while subjecting it to radiated fields."
>
http://www.emctech.com.au/sample.html writes:
>
> "If an EUT has a variety of internal and external configurations, the
> type tests shall be done with one or more typical configurations that
> represents normal use."
http://www.cclab.com/susceptibility.htm writes:
>
> "The equipment under test should be it [sic] its normal mode and connected
> to any auxiliary equipment. The most susceptible modes of operation
> should be selected and the configuration varied to achieve the maximum
> susceptibility."
>
http://www.ce-mag.com/99ARG/Bjorklof109.html writes:
>
> "Immunity testing means that typical electromagnetic disturbances are
> applied to all relevant ports while the apparatus is in operation."
>
and given our assumption, we think that the appropriate analysis should
concentrate on the post-equalizer situation at the slicer input, not
the pulse height/level separation analysis in rao_1_0704. Jose's add-on
slides from his July presentation (tellado_1_0704) made just this point.
Regards,
Glenn Golden
Principal Engineer
Teranetics, Inc.
ggolden@teranetics.com