CMRR measurement, am I doing it right?

[ballsystemlord]

Hello,
I followed Dave's video here: https://youtu.be/vDe_BHvRpks
And arrived at the following formula:
(20 * log10( RMS_diff_probe / input )) + dB
dB == 20 for 10x probe and dB == 40 for 100x probe.

I then applied it to a differential probe I got and I achieved a much higher than expected CMRR number, so I thought I'd ask here.

Attached is a screen shot of my test setup and the HT8100's (100MHz by Hantek) results. I also am including a screen shot of my PVP2350 passive probes's CMRR for comparison. I'm using an MSO5354. I'm feeding a 1MHz signal into the scope through a 50ohm termination. The differential probe is a 50X probe. Just out of curiosity, I turned on the FFT to see where the noise was coming from.

My calculation is as follows:
(20*log(0.018349/2.4508))+30 == -67.8919dB
The CMRR on the datasheet says >= 50dB at 1MHz.

NOTE: The scope is set to "high res" mode, meaning, it's doing averaging or over sampling to get a more accurate result. Therefore, it's BW is limited to 55MHz in the screen shots. If you turn off "high res" mode the RMS noise increases by about 10mV.

EDIT: The FFT is set to 70dB offset and 5dB/div. It's span is 500Hz to 1.2Mhz. My thinking was to try and see if the 1Mhz signal was leaking through (and it's not).

Am I doing the CMRR measurement right?

Thanks!

[CosteC]

Can you describe more setup you used? CMRR for passive, single ended probe puzzles me.

Did you used FFT to measure CMRR?

[tszaboo]

Can you please draw a quick sketch about your test setup, including where you are inserting the common mode signal?
This needs special investigations of the ground signals, it's very easy to make a loop.

[ballsystemlord]

Here's the drawing of how I connected up the probe. It's pretty much identical to Dave's setup. I have coax to a tee which goes into port 1 through a 50ohm feed-through resistor. The other side of the tee goes into a BNC->DMM adapter. Both leads of the differential probe are connected to one terminal of the adapter. I just unplug the diff probe and connect up the passive one identically.

Thanks!

[ballsystemlord]

Here's a picture of the feed-through terminator.

[Randy222]

Here's a picture of the feed-through terminator.

Well, that's a cheap feed-through.

Terminate the female end with a 1% or 0.5% carbon resistor that matches the Z of scope. Then use a VNA attached to the male side of the feed-through to see what that thing looks like on a Smith chart across the freq (or band) you are testing with

[ballsystemlord]

Here's a picture of the feed-through terminator.

Well, that's a cheap feed-through.

Terminate the female end with a 1% or 0.5% carbon resistor that matches the Z of scope. Then use a VNA attached to the male side of the feed-through to see what that thing looks like on a Smith chart across the freq (or band) you are testing with

I'm a student. When I've saved up enough for a VNA, I'll get one. I really didn't see the need for a more expensive feed-through. I'm never going beyond 350MHz with this scope.

[Randy222]

Here's a picture of the feed-through terminator.

Well, that's a cheap feed-through.

Terminate the female end with a 1% or 0.5% carbon resistor that matches the Z of scope. Then use a VNA attached to the male side of the feed-through to see what that thing looks like on a Smith chart across the freq (or band) you are testing with

I'm a student. When I've saved up enough for a VNA, I'll get one. I really didn't see the need for a more expensive feed-through. I'm never going beyond 350MHz with this scope.

LiteVNA (nano vna h4) is good for up to near 1.5GHz. You can get that for around $70(US) $94(CAD) $63(EUR) $53(GBP).
I think it's a tool you must have in your toolbox.

I think maybe the feedthrough is throwing off reading on thye scope (maybe).

[Randy222]

Here's the 50ohm feedthrough on my VNA at various bands, open and then with a 1meg terminator I made to get close to my scope input.

On low freq the open feedthrough is around 42-43 ohms.
On hi freq things get non-usable. The P57 is no good beyond ~800MHz.

Does this impact your test setup? I don't know, you will need to determine that.

Note: my LiteVNA64 is only spec'd for 50kHz -6.3GHz. Probably best (perhaps) to ignore the chart from 25kHz-50kHz.

[ballsystemlord]

Wow! Thanks so much! I never expected you to go the extra mile to test my setup.

In other news, I reasoned, "Maybe I set the differential probe to 500X instead of 50X, thus throwing off my readings by 20dB." I walked over to my test setup, and the probe is set to 500X! I hurried over to my video, from whence I extracted those photos you see above, and the probe is set to 50X as it aught to have been. So, that's not the culprit. I'm relieved. You wouldn't believe how many times I, and other people I know, make dumb mistakes like that.

[Randy222]

Wow! Thanks so much! I never expected you to go the extra mile to test my setup.

I did not test your setup. I only put that feedthrough on my VNA.
1/2 the battle in testing is knowing how all the pieces in the test setup work as individual components, and when all connected together. If the signal entering the scope channel is the thing that determines results, you need to make sure you understand how that input signal changed from actual source, then add in your adjusting factors accordingly. Some test gear might allow you to add those adjusting factors (like an offset) into the gear, others will not and you will need to do it manually.

Golden rule - when a number is way off from expectation, it's either the DUT, or the test setup/config.

[ballsystemlord]

Wow! Thanks so much! I never expected you to go the extra mile to test my setup.

I did not test your setup. I only put that feedthrough on my VNA.

When I re-read what I had wrote before posting I thought, "You should say, 'part of my setup.'" But I decided otherwise because a small exaggeration would better convey my feelings on the matter (or so I thought). I'm saddened to see that it just led to misunderstanding. I'll be more careful from now on.

1/2 the battle in testing is knowing how all the pieces in the test setup work as individual components, and when all connected together. If the signal entering the scope channel is the thing that determines results, you need to make sure you understand how that input signal changed from actual source, then add in your adjusting factors accordingly. Some test gear might allow you to add those adjusting factors (like an offset) into the gear, others will not and you will need to do it manually.

Golden rule - when a number is way off from expectation, it's either the DUT, or the test setup/config.

In my own case, the setup is so simple I don't really see how anything other than the terminator would be affecting my results. That is to say, my BNC tee and BNC cable. Mind, I'm not disagreeing with you here nor trying to be difficult. I just lack the experience and know how to understand and isolate what I'd need to discover any flaw in the setup. Granted, if I had different equipment I could do A/B testing, but I don't.

[Randy222]

Wow! Thanks so much! I never expected you to go the extra mile to test my setup.

I did not test your setup. I only put that feedthrough on my VNA.

When I re-read what I had wrote before posting I thought, "You should say, 'part of my setup.'" But I decided otherwise because a small exaggeration would better convey my feelings on the matter (or so I thought). I'm saddened to see that it just led to misunderstanding. I'll be more careful from now on.

1/2 the battle in testing is knowing how all the pieces in the test setup work as individual components, and when all connected together. If the signal entering the scope channel is the thing that determines results, you need to make sure you understand how that input signal changed from actual source, then add in your adjusting factors accordingly. Some test gear might allow you to add those adjusting factors (like an offset) into the gear, others will not and you will need to do it manually.

Golden rule - when a number is way off from expectation, it's either the DUT, or the test setup/config.

In my own case, the setup is so simple I don't really see how anything other than the terminator would be affecting my results. That is to say, my BNC tee and BNC cable. Mind, I'm not disagreeing with you here nor trying to be difficult. I just lack the experience and know how to understand and isolate what I'd need to discover any flaw in the setup. Granted, if I had different equipment I could do A/B testing, but I don't.

This is why having something like a VNA and SA would be beneficial, you could then characterizes your BNC T and the feedthrough, then adjust your readings accordingly within reason. If the adjusting makes no sense then perhaps CMRR is just horrible and your tests validate that.

Keep studying it.

[ballsystemlord]

Hello,
I followed Dave's video here: https://youtu.be/vDe_BHvRpks
And arrived at the following formula:
(20 * log10( RMS_diff_probe / input )) + dB
dB == 20 for 10x probe and dB == 40 for 100x probe.

...

My calculation is as follows:
(20*log(0.018349/2.4508))+30 == -67.8919dB
The CMRR on the datasheet says >= 50dB at 1MHz.
...

Serendipity!
I was testing my setup again, and used a TI-85 calculator instead of my PC's. Now I get about 12dB of CMRR. I then plugged the same numbers into my PC's calc but wrote "log10(X/Y)" and got the same result as the TI. I'm surprised no one else noticed!

Thanks for your help guys!

PS: Now I need to see if I can return this very under performing product.
PPS: My PC's calc is doing log1, in case anyone's curious.
EDIT: PPPS: Because I was used to working with the TI, which actually doesn't accept the syntax "log10(X/Y)", I didn't think to try that with my PC's calculator. It's not that I was lazy or something.