[SI-LIST] Re: Peak Distortion Analysis
- From: "Dmitriev-Zdorov, Vladimir" <vladimir_dmitriev-zdorov@xxxxxxxxxx>
- To: "si-list@xxxxxxxxxxxxx" <si-list@xxxxxxxxxxxxx>
- Date: Mon, 25 Feb 2013 17:05:25 +0000
Hi Richard,
You touched a very interesting question about relationship between peak
distortion analysis (sometimes considered as finding the worst case pattern
producing the maximal eye stress) and statistical analysis.
Strictly speaking, the answer is that accurately and properly organized
statistical simulation should be able to consider all possibilities including
the ones coming from the worst pattern/xtalk etc. combinations, plus
non-deterministic impairments such as Tx/Rx jitter and noise. If statistical
analysis does all that correctly, then from practical point of view it is
preferable to PDA because it gives the realistic - not overly pessimistic -
eye/BER estimates.
We can turn above statement into the following: accurate statistical simulation
MUST INCLUDE the elements of PDA, implicitly or explicitly. And, accurate
statistical analysis by its 'complexity' exceeds PDA, because not only it
provides the eye margins, it should also provide the probabilities of all
possible sampling locations on the Voltage-UI plot.
However, - and we should make it very clear - it is very difficult to make
statistical analysis pair with PDA and its modifications in some special cases.
Even PDA could be a non-trivial problem.
One simple case which both classical PDA and statistical simulation easily
handle is completely unconstrained uncorrelated bit pattern. In this case, the
'worst' bit input/xtalk combinations are included into ISI PDF 'automatically'.
But, any non-triviality, such as band limited pattern (including 8b10b
encoding), asymmetry between rising/falling edge transitions, or data-dependent
jitter where Tx phase jitter depends on the number of preceding bits - all
those cases, and worse their fancy combination makes peak distortion analysis
and also accurate statistical simulation very difficult and almost impossible.
Imagine 8b10b input pattern which is combined with strict data-dependent jitter
(this jitter cannot be considered statistically, independently from the input)
in presence of asymmetry in the rising/falling transitions. What is the input
that PDA would need to find?
We in Hyperlynx have several unique solutions for such special cases: finding
the worst case input pattern and hence maximally tight voltage/timing margins
for
- 8b10b input pattern
- regular uncorrelated input when rising/falling transitions are not
symmetrical (typically, when Tx produces considerable common mode component and
the channel provides some skew that converts part of this common mode into
differential)
- 8b10b input with edge asymmetry
- stiff data-dependent Tx jitter given by a table of bit combinations and
corresponding phase jitter.
For those and others we have patent protected solutions targeting the worst
input and maximal eye stressing, and the elements of these solutions are also
embedded into the statistical analysis.
Be aware that not all tools support those things in statistical analysis. For
example, try to perform statistical simulation with 'uncorrelated' bit pattern
and 8b10b pattern as an input. Does the tool allow such distinction, don't you
have identical results in both cases? You shouldn't.
Vladimir
> -----Original Message-----
> From: si-list-bounce@xxxxxxxxxxxxx [mailto:si-list-bounce@xxxxxxxxxxxxx] On
> Behalf Of Richard Allred
> Sent: Saturday, February 23, 2013 9:03 PM
> To: si-list@xxxxxxxxxxxxx
> Subject: [SI-LIST] Peak Distortion Analysis
> Greetings,
> A few years ago I actively used peak distortion analysis (PDA) to estimate
> the worse case timing and voltage margin from a simulated
> channel pulse response. More recently I have totally moved away from PDA
> toward using statistical type analysis which directly
> estimates >BER performance.
> I am wondering if there are people out there who actively use peak distortion
> analysis. Is it still useful to anyone? Is there a place for
> PDA in the signal integrity engineer's toolbox still?
> Thanks,
--
> Richard Allred
------------------------------------------------------------------
To unsubscribe from si-list:
si-list-request@xxxxxxxxxxxxx with 'unsubscribe' in the Subject field
or to administer your membership from a web page, go to:
//www.freelists.org/webpage/si-list
For help:
si-list-request@xxxxxxxxxxxxx with 'help' in the Subject field
List forum is accessible at:
http://tech.groups.yahoo.com/group/si-list
List archives are viewable at:
//www.freelists.org/archives/si-list
Old (prior to June 6, 2001) list archives are viewable at:
http://www.qsl.net/wb6tpu
Other related posts:
