Notch filter recommendations

[electronic eel]

Hi,

I'm planning to build a DAC board and want to be able to do some measurements on it. What I want do measure is THD-1, THD-20 and IM3 (19kHz + 20kHz).

As the expected values for these measurements are in the same ballpark as the best available ADCs or sound cards, I need to suppress the fundamental signals, so that the dynamic range of my measurement system can be used to measure just the distortion and not the fundamental input tones.

As the input signals are digital and converted by my DAC, I don't think common techniques for amp measurements, like the Cordell distortion magnifier, will work. So I think the way to go is notch filtering the fundamental tones.

For the measurements above I'd need at least 3 notches: 1 kHz, 19 kHz, 20 kHz and these must be able to be combined to filter 19 & 20 kHz at the same time. They should be at least about 40 dB deep.

I've seen some fixed frequency diy projects especially for 1 kHz. But of course a fully tunable dual-notch would be more flexible.

What filters can you recommend?

I'd prefer either finished products or diy projects with a ready-to-order pcb or kit. I have no problem with a little weekend soldering project, but prefer not to develop everything on my own because that would distract me from the DAC project.

 

[ghg]

Well, do a web and/or diyaudio search for twin-t notch filters,
there are passive and active ones,
they perform well.


For IMD, I just found this, but never build it:

eejournal.ktu.lt/index.php/elt/article/download/695/906

BR

Gary

 

[electronic eel]

Thank you very much for posting the link. I couldn't find any info about such a double notch for IM3 measurements. So I'll probably end up building something like in this article.

Or does anyone else have other suggestions for filtering the two fundamental tones of an IM3 test signal?

 

[kevinkr]

You could also consider a moderately high Q LPF rather than the notch filters particularly if you don't care about the sum. (Which is way above the audible frequency range)

Another option if you do care might be to sum the responses of an LPF and HPF, but delays and phase shifts might make this problematic. (Say 16kHz LPF and 22kHz HPF) You could also just not sum them at all and splice the measurements afterwards.

I might be all wet, just thinking out loud here.

 

[electronic eel]

You could also consider a moderately high Q LPF rather than the notch filters particularly if you don't care about the sum. (Which is way above the audible frequency range)

You mean a brick-wall like LPF?

If I use 19kHz and 20kHz as test tones, I want to measure e.g. the 3rd order product at 18kHz. So 18kHz shouldn't be attenuated too much.

I don't care about the sum as in using an AC voltmeter to do the measurement. I plan to use a good sound card & FFT to analyze the results in the freq domain.