distortion measurement when input is distorted

[meconlen]

Until my test oscillator comes in (and I know it's working right) I'm using a HP 331A distortion analyzer with a signal generator from my computer.

If I measure the output from the computer (actually an Edirol 24/192 Firewire unit) I measure -47.32 dB noise. If I measure the output from the amplifier I measure -43.2 dB noise. Is there a way to estimate what the noise is from the amplifier, or should I just wait to get a good signal generator.

[EUVL]

You can subtract the distorted input signal from the output first and use a spectrum analyzer to see the distortion spectrum.

Search "Bob Cordell Distortion Magnifier".

I built one. Very useful piece of kit.


Patrick

[bob91343]

Since the distortion is only about 3 dB greater than the input, you can assume the same distortion for both the generator and the amplifier, or about -47 dB. (That's about 0.5%.) You can build a simple filter for one frequency from cheap parts to clean up the source so you can get a more accurate reading. In an old HP manual they show a bridged-T circuit with three capacitors, a resistor and an inductor that can be set to filter the second harmonic. If that doesn't help, maybe it's the third harmonic that is causing it.

Another approach is a low pass filter. It may be your analyzer is reading the sampling frequency, and that's really easy to filter. Try a simple RC rolloff and see if that reduces the source distortion.

[Speedskater]

Both of those values are rather poor. A portable CD player might approach -80 dB. Either you have a signal voltage level mismatch or you have a problem.
Have you tried testing it using RightMark Audio Analyzer?

[WithTarragon]

I think you need to be more specific in your original question.

The "floor" that you are referring to: 1) is this a noise floor or 2) is this the level of the distortion components below the signal. IOW, is this noise or is this distortion (non-linearities)

[meconlen]

I don't see harmonic distortion (which isn't to say it isn't there, it's just I'm using the FFT function of the scope, not a real analyzer).

On the other hand when I build a Bubba oscillator and put it on the 331A I can easily see harmonic distortion easily.

My thought was that if I have noise in the input I should be able to treat it as signal mathematically and back it out of the calculation for noise on the output, but I'm not sure that would be valid.

[WithTarragon]

Okay, we are one step closer.

The figure of -43 dB for your noise floor: what is this relative to 1) re: 1 volt rms or 2) re: the signal amplitude?

[meconlen]

so it turns out I was reading the manual on the 331 A wrong, so after several RTFM sessions I've got the following.

I was reading harmonic distortion, but wasn't adjusting the sensitivity properly to be able to see it on the scope. I would minimize the needle then turn the meter range up until it was on scale and left it, rather than continue to adjust. These measurements are, from the fine manual, "distortion measured is ... from fundamental."

Am I correct to assume this means from the voltage level of the output? Would this be P-P or RMS?

Now, for noise measurements it does not say what it is relative to, but I can only presume it's relative to the input as well, In this case I'm generating white noise that's coming out at 1.26 V (it's a headphone amplifier).

So now we have each case, I can measure harmonic distortion and noise floor properly according to the manual and measure it for both input and output.

Here comes the tricky question, the 331A has an output connection and it's indicated that full deflection of the needle is 100 mV. If I did my math right that's -20 dB from 1 V. Now, I can adjust hte meter range by 10 dB increments and view the output on the scope. My guess is that what I measure using the FFT on the scope should have the meter range adjustment added to it, but then the -20 dB removed, and that it's relative to the measured voltage.

The noise floor measurements now