Possible to use difference math of DSO as distortion analyzer?

[perfknee]

Warning, this may be considered a somewhat newbie question.

My understanding of what a distortion analyzer does is it basically notches out the original sine wave signal, so that what's left is just the distortion.

Could a modern DSO (e.g. Rigol 1052e or Instek 1062a) that has math functions be used to do the same thing without purchasing a distortion analyzer? I'm thinking take a low noise audio generator, and split the signal to feed the circuit being tested and one input channel of the DSO. Then feed the output of the circuit being tested into the other channel of the DSO, set it to subtract the two signals, and voila we can look at the distortion by itself, and even do FFT analysis on it.

Thinking about this a little more, I suppose there might be issues with the 8-bit sampling in the scope introducing quantization errors (distortion). And there might be a need to precision-adjust the levels so that we don't end up measuring level mismatches.

Has anyone tried this?

[macboy]

It's a good question actually. Unfortunately it is not that simple.

One problem is phase shift. If you introduce just 1 degree of phase shift between input and output, then you will end up with a residual signal after subtracting, at a level of about -40 dB (1%). I suppose you could try to insert a phase-shifting circuit, but then that circuit could introduce even more distortion and noise, and you would need to constantly tune the phase shift. And you can forget about using anything other than a sine wave input since the phase shift will be different at different frequencies and you would not be able to fix that. Very old distortion analyzers with a manually tuned notch had a similar problem... the notch would drift out of tune while in the process of taking the measurement (a multi-step process) which would lead to a significant source of error.

This still might be useful to look at the spectrum of the distortion. You can remove most of the fundamental in the subtraction, and do an FFT on the residual. Ignoring the spike at the fundamental frequency, you could then observe the relative levels of the harmonics.

[jcx]

for audio you can use better quality soundcards - although most cheap motherboad chipsets today are good to much better than -80 dB noise and distortion

Audio DiffMaker is free sw for very accurate .wav audio differencing with time, fr, sample rate compensation
Audio DiffMaker

RMAA is another free audio test sw which is good for quick loopback testing of your PC/soundcard to see its limitations

[tomchr]

To measure distortion, there's two fundamental approaches.

1) Notch out the fundamental and measure the residual. This gives you THD+N.
2) Perform an FFT on the measured signal. This will give you the THD (no +N).

Most digital scopes use an FFT. You can get software that'll use a computer soundcard to do the same.

If you want to go the first route, I suggest looking at state-variable tracking filters. That's a bit of a research project, but in the end all you'll need for test equipment is a true RMS voltmeter. You can look at some of the existing designs. Something like the HP 8903A or 8903B audio analyzer would be a good place to start. You can get the service manual with full schematics on Agilent's website.

~Tom

[perfknee]

Thanks for the very enlightening responses. I've been up late last night researching standalone distortion analyzers (#1 from the last post), and it looks like buying a reliable one even used gets up towards $1000 (not to say there aren't bargains that people luck into here and there). That's way out of my just-starting-out hobbyist budget.

I actually don't have a need for a distortion analyzer right now, but I just thought while I'm in the process of buying a scope and associated gear I might want to find one that comes along with something that could be used for distortion analysis later on down the road when I actually need it.

The Instek 1062a scope that I just ordered sounds like it won't do it well, but I suppose if I find a situation where I really need distortion analysis I can beg or borrow someone's PC based scope with built in distortion analysis.

[AndrewT]

read about Cordell's distortion magnifier.

[tomchr]

I've actually made a bit of a hobby out of buying broken distortion analyzers and fixing them. I run them through the full adjustment procedure and performance testing to verify the repair. I sell them for around $600. I don't have one right now, but I'll post them in the Swap Meet forum when I do.

~Tom

[Ronnie4416]

Originally Posted by perfknee

Could a modern DSO (e.g. Rigol 1052e or Instek 1062a)

So you decided to go for the Instek? I think Instek 1062A is a good basic scope for hobbyist but most forum members at EEVblog EEVblog Electronics Community Forum - Index choose Rigol over the Instek, maybe there is something on the 1052E that is missing on 1062A?

[benb]

These DSO's only use 8 bit A/D's (at least one of these, this one shows the spec here Instek GDS-1062A Oscilloscope Digital Storage 60MHz Instek GDS-1062A GDS1062A GDS 1062A), so with the FFT you're not really going to see anything (such as a distortion harmonic) below circa 48dB of the fullscale input. A modern 24-bit computer interface with free software will beat that bigtime.

[cliffyk]

^+1, the 8-bit ADCs of most DSOs is a very limiting factor in doing distortion or any sort of signal level analysis on audio signals.

At very best the level of harmonics and noise would have to be above -48dB to be visible at all, making the measurable THD+N floor a rather high 0.4%. In practice 0.8% would be a more practical limit.

OTH a caution re: using a sound card is that the sample frequencies are rather low, and that many consumer level cards/USB devices have 20-22kHz brick wall low pass filters on the inputs. These limit the useful fundamental test frequencies to 10kHz or less, as even at 10kHz only 1 harmonic will make it through the input filter. At frequencies >10kHz no harmonics would get past the brick wall.

Some high-end/pro level cards and USB devices do have switchable filters, at a bit less than the Nyquist frequency, depending on the selected sample rate. These can extend the useful fundamental frequency range.