Dale CMF55-series Resistor Distortion

[mediatechnology]

Turns out my revised concept doesn't work as well as what I originally tried.
In the revised concept common mode to differential conversion due to resistor mismatch swamps differential distortion created by the relative difference in resistors.
The CM rejection is also not enough to keep the A/D within its SFDR. (<-20 dBu; <-28 dBFS.)

Previous tests placed a passive fully-balanced and "floating" notch filter ahead of the +40 dB instrumentation amp.
(The +40dB INA loads the balanced dual Twin-T outputs with 100K bias resistors so in that sense its not truly floating.)
The notch filter itself doesn't affect the CM signal but it does filter out the 1 kHz differential signal arising from mismatch.
This gives it the ability to look deeper into the harmonics and provide gain without using up A/D SFDR on the 1 kHz fundamental not cancelled by the bridge.

The block diagram of what I previously used is this:



It can be argued that common mode to differential conversion for the oscillator's harmonics are also amplified by 40 dB less filter insertion loss and this is indeed the case.
But the oscillator's harmonics are low enough relative to the "bad" CMF-55 that all the other 0.1% resistors produce a clean spectrum and the CMF-55 doesn't.

The bottom line is I was able to use a bridge method to find which resistors to buy and which one to avoid after finding that the CMF-55 made THD 15 dB worse in the oscillator when no other resistors did.
The tester conformed what I originally found and I was able to fashion it with the ULDO-Nacho board in a manner of minutes using everything I had on hand.
I'm not going to spend much more time on this now that I have the answer I needed.

[SandyTodorov]

There are basics in measuring technique.
With an ADC with its own THD -120dB, you can measure with 10% accuracy THD no less than -100dB.
Using the bridge with a base frequency suppression of -60dB allows you to measure with this ADC already -160dB THD
Improving the demands on the generator, post-bridge amplifier and ADC with that 60dB with which the signal is suppressed.
Yes, the requirements for the amplifier to measure -160dB THD are no longer -180dB but only -120dB.
But there is no such integrated operational amplifier on the market, which at a gain of 100 has a THD of -120 dB.
Even more so with a balanced input and at 7.5k signal source resistance.

These are basic things though.
But if what you have done satisfies you, sleep well.
It is important that you are happy.

[mediatechnology]

But I'm not attempting to measure THD with any precision.
I'm attempting to find, and am finding, bad resistors by comparing them to good ones.
I don't need an incredible DR to find these in the CMF-55s that led me down this rabbit hole.

I recreated the original setup - it takes only a couple of minutes - and ran test on all the modern Dale CMF-55 10K 0.1% and 15K 0.1% resistors I had.

I used original 1984 10K 1% selected to 0.1% for the bottom two reference resistors and a third random MBB0207 which consistently test clean.
For a control I first measured 2X 1984 CMF-55 10K and 2X MBB0207.
The FFT showed THD was in the noise floor.

I then substituted a modern CMF-55 in one arm of the bridge.

This is what I found:

One of the "better" bad ones:


One of the worst.


And now one of the "golden ones"...


Only the resistor changed in the test.

Based on the six 10K 0.1% CMF-55 I had, one third are unusable.

I repeated this test at least three times and its repeatable.

I then ran tests on the CMF-55 15K 0.1% and only 3 out of 15 are clean.

I can't fault the measurement technique if shows such huge changes when only the DUT is changed.
These resistors are bad enough that a large DR isn't needed.
And I've yet to find a single bad Vishay MBB0207 or Holsworthy Y1RB that isn't in the noise floor.
The CMF-55s - or at least most of them - are the clear culprit.

[SandyTodorov]

The fact that there may be bad resistors does not mean that during the measurement one should not check what the capabilities of the measurement are.
What appears as pictures is not clear.
You say the bridge signal is +20db after the bridge is reduced by -60db and then amplified by +40db
So the ADC input should become 0db (1V)
And in your picture it is -60dB

[mediatechnology]

The fact that there may be bad resistors does not mean that during the measurement one should not check what the capabilities of the measurement are.
What appears as pictures is not clear.
You say the bridge signal is +20db after the bridge is reduced by -60db and then amplified by +40db
So the ADC input should become 0db (1V)
And in your picture it is -60dB

No, not entirely correct and yes I have checked it in many other ways you are not aware of.

What is being shown in the FFT is the notch filter/INA/CMRR output.
The differential component has a notch filter which you do not account for.

In summary:

The bridge is driven at the top and bottom by +26 dBu.
There is 20 dBu across the test and reference resistors.

The differential voltage across the center of the bridge is small due to mismatch and in the case of the bad CMF-55s there is added distortion.
The differential voltage across the bridge is then notch filtered at 1 kHz by 40-60 dB before the +40 dB INA stage.
So it's +26 dBu drive, approx -60 dB in the bridge with 0.1% resistors, -40 to -60 dB attenuation in the notch and then +40 dB gain.
So the 1 kHz differential component sees the notch filter which reduces the overall level at 1 kHz 40-60 dB, about 9 dB at 2K and about 5 dB at 3KHz.

The common mode voltage across the center of the bridge is essentially driven to 0V a little less, or a little more owing to slight 0.1% resistor mismatch.
The common mode insertion loss of the floating notch filter is essentially 0 dB. Not quite but close.
The common mode gain of the INA is "1."
The output of the INA has a cross-coupled differential stage which rejects common mode about 80 dB or so.
So there's at least 80 dB of CM attenuation in the final CMRR stage.

You didn't account for the notch filter attenuation...

And on another note I checked about 34 Vishay MBB0207 or Holsworthy Y1RB and NONE of them had HD above the noise floor.

[SandyTodorov]

That's why I asked you to show the diagram by which you make this measurement, so that it is clear what exactly you are doing.
Instead, you send me schematics from the Internet, and different and without specific values ​​such as levels and gain.

You put different measuring resistors in the bridge and the -60db level doesn't budge, even though they are the same, they have differences in the frames of these 0.1%, so you need a trimmer for additional adjustment of the bridge to make its output the same with all resistors.

[mediatechnology]

"I send you schematics from the internet?"

Isn't this entire thread being sent to you by the internet?

So what in the hell are you talking about?
I drew that block diagram yesterday for this thread. That's my name on the drawing.

You put different measuring resistors in the bridge and the -60db level doesn't budge

Actually if you look it does budge by a few dB.

The attenuation level of the 1 kHz isn't particularly significant to the measurement.
Whether it's -50 dB in the notch, -60 dB or -70 dB we're not measuring the 1 kHz residual.
I'm only looking at the spurs at HD2, HD3 etc. which, except for insertion loss, aren't affected by the notch depth or degree of null.

No a trim is not needed to measure 0.1% resistors to get enough attenuation to prevent the INA and A/D from seeing large amounts of 1 kHz level and contributing their own harmonics of 1 kHz to the measurement.
If I were measuring 1% resistors it might be needed.

I also don't need to calculate the THD of the bad Dale CMF-55 because I can look at all the spurs and tell that I don't want to use them.

[mediatechnology]

This just was posted over at DIYAudio https://www.diyaudio.com/community/thre ... 4/page-504

Media Technology: I received the parts today
First quick check on a 15K resistor Distortion was high (off scale on the linear readout) and the excess noise index (Quan-tech 315C) was 20 dB higher than the Vishay foil. Really bad.
I will do a more detailed report later.

Good to have some confirmation from 1audio who is using the Radiometer CLT1 Component Linearity Tester

[john12ax7]

This is quite surprising, CMF is supposedly high quality. Bad batch maybe? Years ago saw a noise study and the MBB was one of the better ones, good to see your tests also confirm their quality. Worth it over generic cheap metal films imo.

[mediatechnology]

I recently used the Vishay MBB and some Holsworthy in the assembled ULDO-Nacho and FMCP and they were trouble-free.

Some of the CM-55 in both 10 and 15K did test good in the same date code lots but they were a minority so I'm wondering if they have a trimming issue that's not cleanly cutting spirals. If it were materials or annealing seems like it would affect all of them.