Sine Wave distortion measurements

Way back when I used a piece of HP equipment to measure distortion of a sine wave. It was fun to use, but long gone.

How would a hobbiest measure distortion of a sine wave? It can look good on a oscope and still be pretty bad.

One thought is to take the waveform in question, filter it though several low pass filters to create a reference (without phase shifting it), invert it, and null it with op amps. What is left is the distortion.

Anyone seen a home brew setup to do this? With all the audiophiles out there I suspect it exists.

I don't know how you could run a signal through several low pass filters and not get at least SOME phase shift, just due to propagation delays alone.

[eta]
Just a thought - wind yourself a broadband 1:1 isolation transformer using a quality ferrite toroid. Use the output of the transformer to drive the y-axis of your o-scope, and the input signal for the x-axis. If there is no distortion and no losses, you should get a perfect circle on the 'scope.

Any imperfections in the circle being distortion?

It is often overlooked that to get meaningful measurements on low value distortion you require a very low distortion signal.

The "poor man's" way is to use spot frequencies. Both the generator and the filter can be finely tuned, twinTEE being the frequency defining network of choice, although I have seen the Wein used as well.

The old HP equipment (don't remember it's model number) you adjusted for a in the meter, then flipped the switch to automatic, and it would tune itself to the waveform. My shop used it to check the quality of amplifiers in milspec equipment.

I think it was a precursor to a selective voltmeter.

Bill_Marsden said:

Way back when I used a piece of HP equipment to measure distortion of a sine wave. It was fun to use, but long gone.

How would a hobbiest measure distortion of a sine wave? It can look good on a oscope and still be pretty bad.

One thought is to take the waveform in question, filter it though several low pass filters to create a reference (without phase shifting it), invert it, and null it with op amps. What is left is the distortion.

Anyone seen a home brew setup to do this? With all the audiophiles out there I suspect it exists.

Click to expand...

Howdy!

The traditional distortion analyzer was pretty straightforward. It had a NULL circuit for notching out the fundamental frequency (a modified Wheatstone Bridge). You'd then measure alll the remaining power, which was all the harmonics, if any. Although you couldn't separate the diffferent harmonics, it was usually THD you were looking for.

With FFT analyzers, you can get a lot more detail, of course.


If you have an idea ahead of time as to what the harmonic content might be, you could also use a null-type analyzer on those particular harmonics and eliminate them, one at a time.


Eric

I might add that with a trained eye and a good oscilloscope, you CAN see harmonic distortion down to about 1%.


eric

For a signal generator, I don't know. For an amplifier, try an oscilloscope channel A at input, channel B at output, invert B if not inverted output, and select A+B.
Adjust channel gain for minimum amplitude reading at maximum magnification. It may need adjusting delay setting. Will yield a qualitative waveform, not a % reading.

The thing to do to eyeball it in a scope is to use a dual channel.
Input the signal input to the amp into one channel
Take the output from the amp through a sharp filter to remove this frequency and amplify what's left 1000 or 10,000 times in the second channel.

You can see the distortion this way.

Sorry, no dual channels here. X and Y maybe, but have I mentioned how old my oscope is?

Another thought occurs, the PC sound card oscope can also be used as a spectrum analyzer, so I've heard.

You can feed any signal (even a square-wave) into a switched-capacitor lowpass filter IC and get a low distortion sine-wave output.

My extremely low distortion sine-wave generator uses an over-sampled stepped sine-wave made from a CD4018 and a few resistors and an 8th-order Butterworth lowpass filter IC. My distortion analyser uses a switched-capacitor notch filter IC that uses the same clock as the generator so they track.

Bill_Marsden said:

Sorry, no dual channels here. X and Y maybe, but have I mentioned how old my oscope is?

Another thought occurs, the PC sound card oscope can also be used as a spectrum analyzer, so I've heard.

Click to expand...

Yes, the FFT analyzer built into GoldWave is really very good. And Goldwave is FREE!


You can also use a ramp generator to check linearity...but you need to know if your scope is linear first! (Old rubber yardstick syndrome!)

eric

Audioguru said:

You can feed any signal (even a square-wave) into a switched-capacitor lowpass filter IC and get a low distortion sine-wave output.

My extremely low distortion sine-wave generator uses an over-sampled stepped sine-wave made from a CD4018 and a few resistors and an 8th-order Butterworth lowpass filter IC. My distortion analyser uses a switched-capacitor notch filter IC that uses the same clock as the generator so they track.

Click to expand...

OK, this I know, but how do you measure the numbers?

What's driving this is simple enough, how do you know when you have a good design?

Made ya'll think at least. I'm not sure I would buy that old HP equipment if it was available, but it would be tempting.

Sorry, no dual channels here. X and Y maybe

Click to expand...

Well it doesn't have to be dual. You can just measure one, then the other.

Rectify and filter the input sine-wave. It is the reference.
Adjust the rectified and filtered output level so its DC signal is the same level as the input DC reference signal.
Then rectify and filter the output with the fundamental notched out. Its level is the percentage of noise plus distortion.

I would try recording it into your PC sound card which is 16bit, then look for some free software to analyse it. It would be easy enough to write a VB application etc to analyse the sound data, wave files are very easy to read, but I would be surprised if there is not already a free app out there to do it.

Probably. I like AudioGuru's notch idea, it is simple, and I'm betting that is how HP did it.

The spectrum analyzer would also be an interesting approach, though getting numbers out of it would be hard. It would let you tweak a wave shaper for the best waveform.

I wouldn't bother recording it, you can do this stuff real time if I'm not mistaken. I think the sound card makes a questionable oscope, but it definately has its uses.

It should be remembered that most THD instruments actually record THD + Noise.

Switched capacitor filters have two difficulties that have to be overcome at the expense of circuit complexity. This is not a problem for an IC manufacturer, but extra work for a one-off designer.

These problems are:

1) Keeping the switching noise out of the measurement.

2) Achieving appropriate impedance levels presented to the source.

I still maintain that for the average amateur analog based spot frequency measurement is the way to go.

studiot said:

Switched capacitor filters have two difficulties that have to be overcome at the expense of circuit complexity. This is not a problem for an IC manufacturer, but extra work for a one-off designer.

These problems are:
1) Keeping the switching noise out of the measurement.
2) Achieving appropriate impedance levels presented to the source.

Click to expand...

The switching noise of modern switched capacitor filter ICs is very low and is 100 times the cutoff frequency so is easily filtered out.

An opamp buffer isolates the fairly low and changing-with-frequency input impedance of a switched capacitor filter IC.

Yes Guru I agree both those points are true.

All I said was you have to take them into account if you use an SCF.

Incidentally you need a special op amp to do the buffering so as not to introduce distortion from your equipment.