Thread Links Date Links
Thread Prev Thread Next Thread Index Date Prev Date Next Date Index

[10GBT] FW: [10GBT] PAM8 and PAM12 sys time domain analysis



Title: Message
Sailesh,

In the studied PAM8 and PAM12 systems the Tx launch level was
fixed at 1 Vp (popular with many phy vendors) before the Tx
filter. So, the actual tx power at MDI is reduced due to the
filtering effects.

Alternatively, the Tx power could be maintained at a const. level
at MDI -- this is irrespective of a Tx filtering scheme.

Pls refer to ungerboeck_1_0704.pdf, slide 22 -- the Tx pwr is
5 dBm at the LPF output. To achieve this objective, one needs the
PAM level at app. 1.3 Vp. Due to PAM pdf transformation via high
order filtering, the LPF output peaking will be higher than 1.3 V:
1.64 Vp has been observed over a 67 kBaud time interval for the
above quoted example.

Hope, this clarifies.

Regards,
Albert



sailesh rao wrote:
Albert,

As you know, the PAM12 proposal (powell_1_0704.pdf, slide 6) used a 3rd
order Butterworth low-pass transmit filter with a 3dB point at 206.25MHz.
This filter has an attenuation of 18dB at fs/2=412.5MHz, which dwarfs the
attenuation due to the magnetics that you are showing. The SNR margin loss
due to this transmit filter is actually only around 1dB.

Therefore, I'm not sure of your calculations without having access to the
channel+magnetics attenuation curve that you are using. Can you please share
the equational form of this curve so that I can understand what you are
getting at?

In any case, please note that with the reduced symbol rate of 952.381Ms/s,
the PAM8 system will gain at least 1dB of additional SNR margin in Model 3,
which should improve the situation considerably in your simulations.

Regards,
Sailesh Rao.
srao@phyten.com

From: Albert Vareljian <albertv@IEEE.ORG>
Subject: Re: [10GBT] PAM8 and PAM12 sys time domain analysis
Date: Thu, 5 Aug 2004 10:44:34 -0700

Sailesh,

When the line transformers (specified roughly as per
presentations the group has seen so far) are included in the
analysis -- the channel shows additional ~5.6 dB of loss at
500 MHz. See attached graph.

This, combined with Tx 1 Vp launch constraint at the IC, but
not at MDI, accounts for the bulk of SNR losses w.r.t. "ideal"
Class E channel Salz SNR.

The actual SNR loss in the time-domain bench vs. its frequency
domain reference proves to be below 1 dB -- this would be a very
hard target to achieve for any practical h/w implementation of
10GBASE-T.

Regards,

Albert






sailesh rao wrote:

Albert,

I assumed that the reference to Model 3 in your report included ANEXT
with a
64.5dB intercept and other worst-case impairments, as agreed upon in the
task force.

If there was no ANEXT or residual Echo/NEXT/FEXT in your simulations,
then I
calculate the implementation loss in your simulations to be at least
3.6dB
for PAM8 and at least 4.0dB for PAM12.

I don't think we should be contemplating such implementations for
10GBASE-T.

Regards,
Sailesh Rao.
srao@phyten.com

From: Albert Vareljian <albertv@IEEE.ORG>
Subject: Re: [10GBT] PAM8 and PAM12 sys time domain analysis
Date: Fri, 30 Jul 2004 00:20:13 -0700

Hi Jose,

No ANEXT or other impairments except -140dBm/Hz were employed in sims
covered in the report.

Adding ANEXT would seem to be the next logical step. However, correctly
modeling ANEXT may prove a bit tricky.

As we already discussed on IEEE floor -- our agreed ANEXT models
are specified only in terms of the frequency domain magnitude (no
phase). So, the time-domain implementation has been left open up
to the user...

There could be many interpretations as to how one arrives at a
reasonably behaved time-domain ANEXT TF and its excitation method.
Our analysis indicates that end results in the system could vary
significantly on the case by case basis, depending on the methodology
used to model time-domain ANEXT behavior.

Based on the above, it may be helpful if the group agrees on and adopts
some "uniquely" defined causal, scalable time-domain capable model
for ANEXT that could be used for system qualification. One possible
example of ANEXT TF implementation in s-domain (usable in time- and
frequency- sims) is illustrated in the attachment.


Regards,

Albert




Jose Tellado wrote:

Hi Albert,

Thank you for your detailed time-domain report, I have a couple of
simple questions on the simulation assumptions.

Have you included the effects of ANEXT in these simulations? If so,
what
approved PHY channel model (1-4) would this approximate?

Did you include other receiver impairments such as residual EC/NX/FX or
did you lump all these effect into the -140dBm/Hz noise?

Regards,
Jose Tellado



-----Original Message-----
Sent: Monday, July 26, 2004 7:57 PM
Subject: [10GBT] PAM8 and PAM12 sys time domain analysis

Hi All,

Pls find attached pdf report on PAM8 and PAM12 systems time domain
simulation and comparative analysis.

Regards,

Albert Vareljian



<< ANEXT_Fig.doc >>