impedance matching for oscilloscope measurement

[xavierr]

Hello,

I have a problem during testing some IC. It has mosfet totem pole output (push-pull?). In ON state has 100Ohm resistance, in OFF state 100kOhm. Output capacitance is let say (not remember exactly now) 100pF.
I have the setup as in the picture - 500MHz oscilloscope and 10x probe. I'd like to connect the probe with IC pin by means of coaxial cable. What kind of matching should be used in the red rectangle region? Paralell 50Ohm resistor? How to make a simulation for that? Output signal is 5V square wave 50kHz. How to match in order to obtain the best fidelity of signal? Is the model I've done in LTSpice ok? I like to use 50Ohm cable (50Ohm is Z0, but what exactly impedance sees 50kHz square signal form IC side? - should I model Coax with L&C lumped parameters to get best simulation?)

 

[FvM]

I'd like to connect the probe with IC pin by means of coaxial cable.

Why? That's really a bad idea. The probe should be be attached directly to the test point.

In any case, that's no impedance matching problem. Operation principle of a high impedance probe is minimal loading of test circuit, not maximum power transfer.

The oscilloscope probe cable isn't a regular 50 ohms cable. It's a lossy cable with a resistance wire inner conductor and about 200 ohms characteristic impedance. See the Tektronix paper linked in this previous thread: https://www.edaboard.com/threads/181660/

A RLCG lossy line model should roughly model the probe cable, although it doesn't represent skin effect correctly. The 50 ohms cable can be best modeled as ideal transmission line.

Another thread about probe cable https://www.edaboard.com/threads/266702/

 

[xavierr]

Why? That's really a bad idea. The probe should be be attached directly to the test point.

Actually the probe is too short.
But theoretically there should be some good way of matching.. If not, I'm curious why.. Not enough wideband to cover square wave? That's for sure, but how to check/simulate the band of paralell R matching? How to assess how wide band should be covered in theory for such wave (maybe based on Rise Rime - 0.35/RT)?
For single harmonic signal such matching would also be problematic?
For single harmonic (theoretically) signal the matching should be both:
1.parallel resistor between coax and IC of the value that gives with IC output impedance 50 Ohm in parallel (R_match||R_IC_out = 50Ohm)?
2. 50 Ohm resistor between coax and probe to match 10meg of probe to 50 coax or 10meg series resistor to match coax to probe impedance?
Additional questions in this topic, just academic:
1. What is the output impedance of switching pin of IC if it switches from high to low state? There should be such parameter I think so. Not sure about that but isn't that IBIS model that require output impedance - what then we shut put there?
2. What is the impedance of coax cable at given frequency? Can we calculate this having no possibility to measure?
3. Not matching in such circuit gives bad results for 50kHz square signal? At 10kHz as well? 1kHz?

 

[FvM]

I have problems to suggest a reasonable alternative because your signal source model is already wrong. There's no actual signal source (voltage or current source) shown, and a totem pole output unlikely has 100k output impedance. Please try to sketch a reasonable model. If it's a commercial IC, tell the type.

Presumed you actually have a high impedance output, e.g. open drain with pull-up, the only way to probe it without completely distorting the original waveform is a an active high impedance probe (e.g. a wide band FET buffer). A regular passive probe does already considerable signal distortions, but it might be acceptable. And because it's cheap, it's the usual means even in cases where it better shouldn't be used.

If the output impedance is lower than 100k (a totem-pole output should be expected to be), a resistive divider might be the next best solution for good signal reproduction. Connect a 5k series resistance to 50 ohm cable and select 50 ohm oscilloscope input impedance. Gives a 100:1 wideband resistive probe, good up to full oscilloscope bandwidth if made correctly.