New distortion measurement method for audio amplifiers

[PMA]

After having performed a large number of spectral measurements based on single-tone spectral analysis I have decided to try more complex input signal. The reason is that there is not much correlation between single-tone spectral analysis results and listening preferences during listening tests. This is especially truth for low distortion components (THD of 0.001% order) where audible differences still exist.

First results are shown at

http://web.telecom.cz/macura/fmanalysis/freqdist.html

Your comments and suggestions would be appreciated.

 

[richie00boy]

Looks good.

In your opinion, would it be acceptable to play this signal from a 16 bit audio source such as .wav file or CD, then analyse the amp output? I can borrow a spectrum analyser, but not a modulating signal generator.

 

[PMA]

Looks good.

In your opinion, would it be acceptable to play this signal from a 16 bit audio source such as .wav file or CD, then analyse the amp output? I can borrow a spectrum analyser, but not a modulating signal generator.

The input signal is generated exactly the same way you are describing - by Cool Edit SW. It is better to use a PC, as CD brings new components above 22.05 kHz (already tested).

 

[audio-kraut]

To really go overboard - why not sample a one millisecond pink noise and send this through an amp and sample what comes out the other end. With digital methods it should be easy to identify all distoriton products.

 

[richie00boy]

I have Cooledit so that's good. How would using a PC be different to a CD? Doesn't the PC have a steep filter at 22.05kHz just like a CD player?

 

[PMA]

By now I have always measured new components above audio band on CD output, when trying 13+14kHz. Please have a look at:

http://web.telecom.cz/macura/cdspectrum.html

 

[ingrast]

Error correction amp toppology

PMA:

Though off topic, looking at your error correction amplifier topology, I wonder whether this has been treated before.

I was about to start a thread based on a similar hybrid topology only not error correcting like yours, but with a 20 dB gain output buffer.

Being rather new to the audio amplifier speciality, I assume the approach is not original either, but have not seen it more widely mentioned. I mean, moderate gain low distortion output stage driven by high performance audio OpAmp, and mild global NFB.

Simulation results with LTSpice and stock library elements are promising.

Rodolfo

 

[destroyer X]

High precision method of evaluating distortion

Put a non reactive, non inductive, 8 ohms load in both channel output.

From the positive output use 100 ohms in series with some good speaker

Increase the audio volume till you perceive some distortion.....measure the peak voltage in your scope and you will discover the maximum power, undistorted, the amplifier can produce for human ears.

Very simple and stupid..... perceptual measurement based on reality.

hehe, Pavel, i will not go to the place you are thinking now to send me... and i did not like that story that only one way ticket...without return Pavel!?

Carlos

 

[PMA]

Re: Error correction amp toppology

PMA:

Though off topic, looking at your error correction amplifier topology, I wonder whether this has been treated before.


Rodolfo

Yes, it was discussed here:

http://www.diyaudio.com/forums/showthread.php?postid=469442#post469442

Pavel

 

[jan.didden]

Pavel,

Those results you showed are quite dramatic. I gather that the 20W Class-A has global feedback and the erroro correction one hasn't?

Is there any correlation to listening test differences to this? Independent, blind testing of course, not by you.

Jan Didden

 

[Graham Maynard]

Hi Pavel,

On the surface these findings look conclusive, but the distortion floor must surely be related to;-

(a) digital noise,
(b) the level of NFB within the amplifiers,
(An amplifier with less NFB is not going to have as low an apparent noise (error) floor as one with more NFB, as might be afforded by an IC.
(c) source impedance, and input filtering.

Can we hear sub -100dB supersonics via loudspeakers ?

I have just posted another test method in the 'Darlington and Triple Darlington' thread. This can throw up some surprising results too.

Hi Destroyer, Hope your iron's still smokin'.


Cheers .......... Graham.

 

[sam9]

Put a non reactive, non inductive, 8 ohms load in both channel output. From the positive output use 100 ohms in series with some good speaker

Increase the audio volume till you perceive some distortion.....measure the peak voltage in your scope and you will discover the maximum power, undistorted, the amplifier can produce for human ear

I suspect the main thing demonstrated will be that human ears tolerate quite a bit more distortion than we assume. I'vee listened to amps with clip indicators where it took a surprising amount of clipping before I could hear anything unpleasent.

 

[Mike Gergen]

Pavel
Interesting. Have you tried this test on any of the opamp or GC based amps or preamps?
Thanks
Mike

 

[destroyer X]

Yes!, we are very tollerant to distortions, and more than that...

Our brain correct many errors too.

Also human perception have relations with feelings, with empathy.

Those percentuals....0.005 are only academic and menthal masturb......, as speaker distort more than that...so.... small numbers represents nothing!... those small numbers are "covered" by speakers distortions...are only Academic Exercises.

regards,

Carlos

 

[andy_c]

Nice work Pavel. Really good stuff!

A couple of questions:

1) I saw the conventional THD for the 20W class A amp, but couldn't find it for the error correction amp. Is the conventional THD data similar for the two amps?

2) I'm assuming the output of your sound card is being fed directly to the input of the amp in both cases. Is that correct? I guess your sound card is of very high quality.

3) Assuming the conventional THD is similar for the two amplifiers, what's your theory for why the difference in the two measurements?

Thanks

 

[john curl]

Good input, Pavel. Keep it up!

 

[SY]

Re: Yes!, we are very tollerant to distortions, and more than that...

Those percentuals....0.005 are only academic and menthal masturb......, as speaker distort more than that...so.... small numbers represents nothing!... those small numbers are "covered" by speakers distortions...are only Academic Exercises.

Well, there's a school of thought that believes that nearly anything measurable is possibly audible. John Curl would be in that school. So would Pavel. I think they're wrong, but what the heck do I know, they're engineers and I only make wine corks for a living.

As an academic and design exercise, this is really interesting stuff. I'd also like to see standard distortion data at a few fixed frequencies for both those two amps.

 

[john curl]

My problem is that I have to make SUCCESSFUL audio products for a living. This means that even if I already know how to make a 'LOW DISTORTION' amp or preamp as well as most people, some of my designs are still considered 'DOGS' to my friends, and they will not use them. What to do? Ignore input that I cannot 'prove' to you or SE? I don't think so.

 

[andy_c]

Re: Re: Yes!, we are very tollerant to distortions, and more than that...

I'd also like to see standard distortion data at a few fixed frequencies for both those two amps.

Here it is for the 20W amp http://web.telecom.cz/macura/mosfetpre12V.jpg. I couldn't find it for the error correction amp, but it could be my mistake for not looking hard enough.

 

[SY]

John, your main defect is that you don't use tubes, just that awful silicon stuff. Woof, woof!

Andy, that was my problem, I could find it for the 20W Class A but not for the "error correction" amp. I'd like to see if a combination of single tone tests are predictive.