THD 0.0001%

On a sunny day (Tue, 27 Jul 2010 09:48:37 -0500) it happened Vladimir Vassilevsky wrote in :

How do you measure the distortion?

Surf around for papers on compound connections of OpAmps that dramatically improve GBW product. ...Jim Thompson

How did you perform the measurement?

Vladimir Vassilevsky a écrit :

Well, a few weeks ago I just measured levels just 20 times below that at

Seems impossible, but doing it the right way it's not that difficult... It's even deceptively simple once you've figured how to do it.

The real fun is to generate the signal and make sure that your measurement is not pure illusion :-)

Again, H2 only made that easier (a ladder of notch filter does the job pretty well). The funny thing was, for the first cut experiment, to tweak the RLC filter distortion with magnets to set the inductor (last stage) cycle mid point to a pure odd function this cancels any even component. A small NeB magnet at half a meter distance did the job but you had to be really careful to not blow the distortion up to the ppm levels :-) (RM8 cores)

In my case, that was really simple: there is an ADC with less then 1 ppm of distortion. The measurement is straightforward. The purpose of the filter is generate clean sine wave to verify the performance of ADC.

Actually, measuring low distortion levels is not too difficult if you have a good bandpass and same as good notch filter at the fundamental frequency.

Vladimir Vassilevsky DSP and Mixed Signal Design Consultant

=20

=20

t=20

e=20

At 0.00001% levels, SNR and signal coherence could represent a problem :-= )

:-)

I can easily bring the distortion down to the numeric accuracy floor by=20 digital correction, however I am suspicious about repeatability of this=20 kind of cheating.

That looks very professional and reliable. I wonder how would you make=20 it in production :-)

Why not an air core?

Vladimir Vassilevsky DSP and Mixed Signal Design Consultant

You can't tweek the permeability of air with a NeB magnet.

-- Bill Sloman, Nijmegen

On a sunny day (Tue, 27 Jul 2010 17:49:52 -0500) it happened Vladimir Vassilevsky wrote in :

I am not sure I complete grasp that sentence. Do you digitise the filter output, then do a FFT?

I know that, in the long ago past I did measurements on Eurovision sound lines using just that procedure, cannot remember the name of the instrument, but it was sort of a weekly ritual :-) Simple bit of math gave the exact distortion. Always curious how somebody else does it. You also needed a good signal [generator].

You just made me remember doing something similar using one of these:

Vladimir Vassilevsky a écrit :

Lock-ins easily deal with that.

You're right to be cautious. There's no reason at all for your DAC errors to be repeatable well below the LSB level.

Did I mention prod somewhere ?-)

It was a quick and dirty lash-up to do some tests down to the 10s of ppb level.

'coz I had the RM8 cores already wound and trimmed to the percent level and I had to find a way of doing the test as soon as I had the idea for the low distortion notch filter...