Reverse engineering a black box circuit?

[Andrew_K]

Hi All,

I'm looking for advice or resources (textbooks, articles, programs, etc) on how to characterize and reverse engineer a "black box" circuit.

Not just get measured parameters, but actually get a transfer function (and equations, so poles and zeros), and come up with an actual schematic.


----

In my particular case, I am trying to create a schematic from a semi "black box" circuit. I do have a few known values and a rough skeleton of what the circuit is.

The circuit is a filter of some kind. I believe it is a 2-cascaded RC low pass. The resistor values are known.



There are two metal disks that are connected to the 2 inner nodes of the resistors. There is a plastic shell and spacer that (I assume) are to form capacitors.



As for things I've tried:

1) I tried taking it apart and measuring in parts.

I removed the middle wire and measured the capacitance between the metal disks, as well as each metal disk to shell.
Using the 3 measured values and their respective parallel/series combinations, I solved for the 3 capacitors individually.
This actually gave me a negative capacitance between the two disks.

2) I tried using a network analyzer to get 2-port S parameters. I used these to calculate the Z-parameters, as well as the voltage transfer function.



This plot shows the log sweep and lin sweep (magnitude)

Note that this circuit was *NOT* at 50 ohm characteristic freq. However, the impedance values calculated seemed (mostly) reasonable. The VNA also has a function to measure thru series/shunt impedance and port impedances.

I used rather long alligator clips to connect. And I'm also not sure if my fixturing and cal process were very good. (Speaking of, does anyone have any good resources or books on the theory of fixturing and calibrating a VNA??)

When I did a log-sweep on the analyzer, I got reasonable results, but I also did not get the resolution I wanted for lower frequencies.

When I did a linear sweep, the results were very unstable at sub 100kHz. I'm scratching my head a bit here. Somehow the transfer function gain is >1 at many points. My hypothesis is that the sweep was too fast, or possibly, there is resonance at these lower frequencies.

[Benta]

Could you post a picture of the thing itself? It looks like something I once designed.