FreeLists / si-list / [SI-LIST] hapice simulation of Z_tdr

["Dmitriev-Zdorov, Vladimir" <vladimir_dmitriev-zdorov@xxxxxxxxxx>]

Hi ZHENGGANG,

Unfortunately, this does not seem as a correct approach.
By definition, you need to measure Z(t) in time domain by applying the
unit step current and measure the voltage response. In frequency domain,
you can reverse the ratio by swapping nominator and denominator.
However, this operation does not correspond to swapping nom/denom in
time domain.

What you do is good for measuring conductance profile. Apply voltage
step, measure the current response. Then, go into frequency domain by
FFT. You will also need to divide on 1/s to get the conductance. Then,
inverse it and get impedance in frequency domain. After that, if you
need time domain profile, apply IFFT - you will get Dirac response. If
you need impedance step response just integrate this in time domain.

But if you simulate in HSPICE there should be no problem applying
current step and measure voltage.

Vladimir

=20
>Date: Fri, 31 Aug 2007 15:50:40 -0700
>From: "ZHENGGANG CHENG" <zhenggang.cheng@xxxxxxxxx>
>Subject: [SI-LIST] hapice simulation of Z_tdr
>
>Hi all,
>Thanks for your clear explanations!
>
>I have another interesting question about my hspice simulation deck for
>simulating the TDR Z (differential) profile (vs time):
>
>The Z profile in the sim deck is defined as (V(t)_p - V(t)_n)/I(t).
V(t)_p
>is probed at the point right after source impedance (50ohms) for
positive
>lane, and V(t)_n is probed at the point right after source impedance
> (50ohms) for negative lane. The source is a step voltage source
connecting p
>and n lanes.
>
>It seems, sometimes my simulated Z  profile can be very close to
>measurements, but sometimes it is not, especially for DUT with many
large
>discontinuities inside.
>
>I am wondering if  (V(t)_p - V(t)_n)/I(t) is sufficient and correct to
>represent the real TDR Z profile?
>
>Look forward to your explanations. Thanks.
>
>ZG
>
>2007/8/31, Loyer, Jeff <jeff.loyer@xxxxxxxxx>:
>
>  If the S-parameters are identical, then the TDR/TDT results will
likewise
> be identical (assuming proper risetime to match the frequency limit of
the
> S-parameters).
>
> But you have to remember that the phase of S-parameters are just as
> critical as the magnitude (in evaluating the circuit).  Often we only
> focus on the magnitude, and forget the phase.
>
> Another caveat is that we tend to concentrate on S21, which is the
> frequency equivalent of TDT.  This can be very similar for different
> networks, even though S11 (and TDR) are dramatically different.
>
> I believe these 2 factors add up to make it appear that very
differentcircuits have the same S-parameters: you
> might only be looking at S21 magnitude, when you should be looking at
the
> magnitude and phase of S11, and phase of S21 also.
>
> You might look at some work Cherry Wakayama did, comparing
S-parameters
> derived from VNA vs. TDR/TDT measurements (see link below).  There
were
> several very different structures, and the S-parameters varied
> accordingly, especially when you look at phase and S11, not just the
magnitude
> of S21.
>
> *
>
http://www.tek.com/products/oscilloscopes/sampling/interconnect_analysis
/customer_papers/vna_tdr_correlation.pdf
>
*<http://www.tek.com/products/oscilloscopes/sampling/interconnect_analys
is/customer_papers/vna_tdr_correlation.pdf>
>
> Unfortunately, we didn't include the original TDR waveforms in the
paper,
> but they contained all kinds of various "bumps" in them before being
> converted to S-parameters.
>
> P.S. - I'd also like to thank Vladimir for his great explanation.
>
> Jeff Loyer
>
> -----Original Message-----
> From: si-list-bounce@xxxxxxxxxxxxx
[mailto:si-list-bounce@xxxxxxxxxxxxx<si-list-bounce@xxxxxxxxxxxxx>]
> On Behalf Of ZHENGGANG CHENG
> Sent: Thursday, August 30, 2007 4:15 PM
> To: vladimir_dmitriev-zdorov@xxxxxxxxxx; si-list@xxxxxxxxxxxxx
> Subject: [SI-LIST] Re: FW: TDR S-parameter and correlation
>
>  >
>
> > Vladimir,
>
> Thanks for your detailed explanations.
>
> I still have a question, since many topologyies can produce the same
>
> S-parameters, the real TDR of these equivalent circuits with different
>
> topologies would still be the same?  TDR can see the characteristic
>
> impedance discontinuities inside the circuits in terms of delay/flight
> time.
>
> I thought the different topologies would give different TDR restults.
>
> ZG
>
> -----Original Message-----
>
> > From: si-list-bounce@xxxxxxxxxxxxx
[mailto:si-list-bounce@xxxxxxxxxxxxx<si-list-bounce@xxxxxxxxxxxxx>
> ]
>
> > On Behalf Of Dmitriev-Zdorov, Vladimir
>
> > Sent: Thursday, August 30, 2007 3:03 PM
>
> > To: si-list@xxxxxxxxxxxxx
>
> > Subject: [SI-LIST] TDR S-parameter and correlation
>
> >
>
> >
>
> > Hi ZHENGGANG,
>
> >
>
> >
>
> >
>
> > As I understand, the issue is about "general type" model, not just a
>
> > segment of T-line.
>
> >
>
> >
>
> >
>
> > The answer to your first question is YES provided that (a) DUT's
>
> > S-parameters were measured correctly, (b) the equivalent circuit is
>
> > correctly built from S-parameters and (c) the way you assign ports
in
>
> > your equivalent circuits is consistent with how the measurements was
>
> > done.
>
> >
>
> >
>
> >
>
> > Topology cannot be uniquely derived from S (or Y/Z/G/H... etc.)
>
> > parameters unless you have it predefined, like in case of T-line.
There
>
> > could be many circuits with quite different topology producing the
same
>
> > S/Y/Z... parameters.
>
> >
>
> >
>
> >
>
> > Yes, S-parameters only characterize the model from outside ports,
but
>
> > this is exactly what other pieces of your design see from this
model:
>
> > this is sufficient to use the model in many simulation procedures,
>
> > unless you are interested in voltages/currents inside the model
itself.
>
> >
>
> >
>
> >
>
> > The problems above named (b) and (c) are sometimes not well
understood.
>
> >
>
> >
>
> >
>
> > First of all, any measured S-parameters (think of touchstone file)
>
> > contain the data in a limited frequency range, while any equivalent
>
> > circuit is 'defined' from DC to infinite frequency. Hence, they
cannot
>
> > be equivalent if S-parameters do not cover sufficiently wide range
of
>
> > frequencies, presumably from the lowest frequency where they start
>
> > changing up to the highest frequency where they stop changing and
>
> > approach to constant level. In all other cases, be prepared that the
>
> > circuit does not accurately capture the model behavior at very low
and
>
> > high frequencies. Building equivalent circuit requires rational
>
> > polynomial fitting. This may be done by different tools with
different
>
> > accuracy. In many cases, passivity enforcement is required on the
post
>
> > fit stage to prevent unstable model behavior, especially if the
upper
>
> > frequency in touchstone data is not sufficient. "Passivation" adds
some
>
> > inaccuracy to the fitted model. Typically, representing the
>
> > poles/residues with circuit elements does not bring much error.
These
>
> > are main sources of discrepancy we may have between given
S-parameters
>
> > and equivalent circuit. Plus, time domain simulation of the
equivalent
>
> > circuit, performed with finite resolution, adds LTE (local
truncation
>
> > error).
>
> >
>
> >
>
> >
>
> > The problem (c) may sometimes be insidious. Imagine the S-parameters
of
>
> > DUT were measured for two ports, without a 'common' ground.
Essentially,
>
> > these measurements only characterize how the wave may propagate
between
>
> > these two ports. Then, an equivalent circuit was built with four
>
> > external nodes making two ports. Now, someone use this model not
only by
>
> > apply input to the first port and measure output at the second, but
also
>
> > making arbitrary connections between any of external nodes of this
>
> > model. The result: behavior inconsistent with the original device.
(Same
>
> > also possible with common ground). That is, we need to make sure the
>
> > model is used the same way the measurements were done for
S-parameters.
>
> >
>
> >
>
> >
>
> > Vladimir
>
> >
>
> >
>
> >
>
> >
>
> >
>
> >
>
> >
>
> > Msg: #15 in digest
>
> >
>
> > Date: Thu, 30 Aug 2007 12:30:19 -0700
>
> >
>
> > From: "ZHENGGANG CHENG" <zhenggang.cheng@xxxxxxxxx>
>
> >
>
> > Subject: [SI-LIST] TDR S-parameter and correlation
>
> >
>
> >
>
> >
>
> > Hi,
>
> >
>
> > My question is:
>
> >
>
> >
>
> >
>
> > If we TDR  (assuming the TDR method is 100% correct) the equivalent
>
> > circuit converted using S-parameter, will the result is exactly the
same
>
> > as real TDR of the same DUT? (assume the converting error and
bandwidth
>
> > are not issues) Assume this DUT has many large discontinuities
inside.
>
> >
>
> >
>
> >
>
> > To me, the real TDR can distinguish all the discontinuities inside a
>
> > DUT; however, the S-parameter is only the characterization at the
ports
>
> > rather than inside. Will two equivalent circuits give the same
>
> > S-parameter but have two different topologies?
>
> >
>
> >
>
> >
>
> > Look froward to your replies.
>
> >
>
> >
>
> >
>
> > Thanks,
>
> >
>
> >
>
> >
>
> > ZG
>
> >
>
> >
>
> >
>
> >
>
> >
>
> >
>
> >
>

------------------------------------------------------------------
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:
http://www.freelists.org/webpage/si-list

For help:
si-list-request@xxxxxxxxxxxxx with 'help' in the Subject field


List technical documents are available at:
                http://www.si-list.net

List archives are viewable at:     
                http://www.freelists.org/archives/si-list
or at our remote archives:
                http://groups.yahoo.com/group/si-list/messages
Old (prior to June 6, 2001) list archives are viewable at:
                http://www.qsl.net/wb6tpu