RE: [SI-LIST] : Frequency or time domain for component characteri zation

About this list Date view Thread view Subject view Author view

From: George Borkowicz (jbor@nortelnetworks.com)
Date: Tue Feb 29 2000 - 11:57:47 PST


> What are the issues if you would like to characterize a
> circuit at baseband, when using a VNWA? Seems to me that
> there are significant issues in dealing with the DC and low
> frequency components.
>
> Anyone have experience in correlating frequency domain measurements
> with baseband pulse performance?
>
> regards,
>
> scott
>
        --------------------------------->

        Scott,

        Actually I think the problems start at high frequencies. These are
strictly practical calibration issues due to cable imperfections, connector
wear instrument differences, human factors etc. In many cases when the S
parameters were part of more demanding component spec (at CATV frequencies
i.e. 5 - 1000 MHz) we had to bring the vendor AND HIS SETUP to our lab to
prove his performance claims and correlate with our results.

        At low frequencies (below 10 MHz) precision is not much of an issue.
However, some other practicalities may get you, particularly when you want
differential results (baluns, limited frequency ranges, common mode
rejection, etc.). Also, if you sweep down to really low frequencies (5 Hz)
the measurement time becomes quite long even with an adaptive bandwidth
filter.

        As someone mentioned, nothing beats collecting model data with NA
when measurement noise is an issue AND your system is linear (otherwise the
inverse process is invalid). But then, on nonlinear systems you can take
advantage of low noise level to reduce your stimulus to stay in a linear
range (for amplitude dependent nonlinearities) and still get sufficient
dynamic range. Long time ago I used to work in the electrochemical area
where these kind of system identification measurements are bread and butter.

        One issue to watch is model overbuild (too many components). In
overdesign, the solver may not converge to the absolute minimum fit.
Functional approximations and careful pole/zero (or equivalent)
analysis/synthesis usually do much better than cut and try methods.

        Also, component sensitivities may be different in time and frequency
domains. The fact that the frequency response of the model fits nicely to
data does not automatically guarantee that time response will do the same.

**** To unsubscribe from si-list or si-list-digest: send e-mail to majordomo@silab.eng.sun.com. In the BODY of message put: UNSUBSCRIBE si-list or UNSUBSCRIBE si-list-digest, for more help, put HELP.
si-list archives are accessible at http://www.qsl.net/wb6tpu
****


About this list Date view Thread view Subject view Author view

This archive was generated by hypermail 2b29 : Thu Apr 20 2000 - 11:35:15 PDT