Best single signal for comparing sound quality

[Blumlein 88]

What would be the best single signal of a few seconds for comparing sound quality between two conditions. Two conditions being two amps, or DACs or any such devices or between playback processing. Something you could use blind so you only need a few seconds of it.

My nomination is pink noise.

I can think of a few criticisms of that, but lets hear yours.

More importantly does someone have an idea of something better than pink noise?

 

[watchnerd]

Brown noise

 

[Blumlein 88]

Brown noise

Okay, why brown instead of pink? Because it sounds more 'natural'?

 

[watchnerd]

Okay, why brown instead of pink? Because it sounds more 'natural'?

For me, yes.

Also, I can listen to it for extended durations at high volume without trying to run away, which gives me more time to focus on what I'm hearing.

Violet noise, in contrast, makes me immediately run away / turn down the volume.

 

[RayDunzl]

https://en.wikipedia.org/wiki/Colors_of_noise

 

[DonH56]

I'd vote for pink noise mainly because it is very widely available. Besides, just turned 58, my HF hearing is low enough that pink doesn't bother me the way it used to. Note pink rolls off at 3 dB/octave (1/f noise), brown at 6 dB/octave (1/f^2), IIRC.

Personally I like to use pink noise plus a two-tone test since the two-tone test more readily reveals distortion the pink noise can hide. It is also a little less sensitive to listener bias since it is not such a broad-spectrum test.

All IME/IMO - Don

 

[Blumlein 88]

I'd vote for pink noise mainly because it is very widely available. Besides, just turned 58, my HF hearing is low enough that pink doesn't bother me the way it used to. Note pink rolls off at 3 dB/octave (1/f noise), brown at 6 dB/octave (1/f^2), IIRC.

Personally I like to use pink noise plus a two-tone test since the two-tone test more readily reveals distortion the pink noise can hide. It is also a little less sensitive to listener bias since it is not such a broad-spectrum test.

All IME/IMO - Don

So Don which two tone test do you have in mind. Is it the 19+20 khz IMD test, in which case most of us more experienced gentlemen will only hear the distortion products or do you have another in mind? I can see the value of that, but might fear for the life of my tweeters if run over and over for a listening test.

 

[Blumlein 88]

Listened to a little brownian noise vs pink noise. It would be easier on the ear for an extended test session. Not sure if it will show differences quite as well though it is pretty clear it will do better than music. Anyone have any good idea or info about whether pink noise would be more discriminating than brownian noise?

As for violet noise, just say no. Even blue noise is too much for me thanks.

For those who didn't read Ray's link or notice what Don posted, blue noise is the inverse of pink noise. Rising response of 3 db/octave and violet is inverse brownian with rising response at 6 db/octave.

 

[DonH56]

So Don which two tone test do you have in mind. Is it the 19+20 khz IMD test, in which case most of us more experienced gentlemen will only hear the distortion products or do you have another in mind? I can see the value of that, but might fear for the life of my tweeters if run over and over for a listening test.

No, well it is for IMD testing, but for listening tests I use something like 1 kHz + 2 kHz or usually something more closely spaced (like a major third in musical terms) and in the 440 Hz (tuning A for an orchestra) to maybe 2~3 kHz range. We are more sensitive to the upper frequencies but they can be annoying and I want to be able to hear the difference tones (reference previous comment about my hearing rolling off, though in fact this has been my choice for ages). I might use C5 on a piano at about 523 Hz mixed with the E above at 659 Hz, or C6 and E6 (1046.5/1318.5 Hz). I'd be listening mainly for second- and third-order products. It is useful to choose notes based on a piano because you can sit down at a piano and plink out the fundamental and difference tones (keys) to know what to listen for. I used to have a test CD with two tones at various points so I could listen for woofer, midrange, and tweeter distortion.

 

[RayDunzl]

More importantly does someone have an idea of something better than pink noise?

I made a track with all 88 fundamentals of the Piano...

Once in a while I remember to play it.

Oh, SoundCloud is still working:

https://soundcloud.com/ray-880875693/piano-notes-88

As output by the JBLs at the Liznin position:

 

[DonH56]

Actually, a major third sounds better if the third is about twenty cents flat, so I use 1046.5 Hz (C) and 1308.125 Hz (E). Then listen for dissonance.

Ray, interesting idea, would be a good multitone or NPR test, but for a simple listening test I prefer just two tones to make it easy to hear what ain't supposed to be there. Be interesting to compare your "all piano keys (APK)" test vs. white or pink noise.

Pink (or brown, or whatever) noise allows one (or maybe just me) to easily compare differences between two speakers (or whatever) but it is hard to tell which is "right". That is IME/IMO the biggest drawback of pink noise for listening tests; you need a reference in mind to compare the "quality" of two sounds (components etc.) and it is too easy to fool my ears into thinking something is "better" when it is really just "different" and more in line with my own preference.

 

[RayDunzl]

I use Pink as a measuring tool...

If listening to it, I don't find it tells me much, unless there is a gross problem, unless I have some sliders to bring different bands out of and just back into audibility.

Maybe my ear isn't smart enough to do it by itself.

 

[DonH56]

One big advantage of using pink noise for measurements is that it yields a flat frequency response on many analyzers (that are based on power-of-two frequency bins and such). White noise (equal energy per decade; pink is equal energy per octave) may look (and sound) like a rising HF curve. Pink thus tends to emulate in-room response better, though an argument for brown could be made.

Checked Ray's link; there are other colors of noise (natch) but that covers most of them. I mainly checked to make sure my memory of pink and brown was correct... Colored noise is found in many other areas, natch, including noise generators for sleep. I first used colored noise (of a very specialized type) in the 1980's to help dither (i.e. noise-decorrelation of) the ADC and DAC in a radar system to achieve greater dynamic range in-band at the cost of higher out-of-band noise floor.

 

[Blumlein 88]

I use Pink as a measuring tool...

If listening to it, I don't find it tells me much, unless there is a gross problem, unless I have some sliders to bring different bands out of and just back into audibility.

Maybe my ear isn't smart enough to do it by itself.

Most real sound differences in gear are frequency response differences. I find with pink noise very small frequency aberrations between two devices jump out starkly. If basic distortion is getting audible like 1 ro 2% you hear that as a different tilt to the sound.

I have a couple times listened and thought one sounded a little different than another. Put on pink noise and it was obvious there was a real difference. Then an FR measure would show one was there.

Now I really think good measures to dot the i's and cross the t's is better than listening evaluations.

 

[Blumlein 88]

I tried your piano keys Ray. It could be useful, but I think I would prefer pink noise or certainly brown due to the ear fatigue issue. I suppose one could take your file and put a pink or brownian tilt to the response.

 

[Blumlein 88]

No, well it is for IMD testing, but for listening tests I use something like 1 kHz + 2 kHz or usually something more closely spaced (like a major third in musical terms) and in the 440 Hz (tuning A for an orchestra) to maybe 2~3 kHz range. We are more sensitive to the upper frequencies but they can be annoying and I want to be able to hear the difference tones (reference previous comment about my hearing rolling off, though in fact this has been my choice for ages). I might use C5 on a piano at about 523 Hz mixed with the E above at 659 Hz, or C6 and E6 (1046.5/1318.5 Hz). I'd be listening mainly for second- and third-order products. It is useful to choose notes based on a piano because you can sit down at a piano and plink out the fundamental and difference tones (keys) to know what to listen for. I used to have a test CD with two tones at various points so I could listen for woofer, midrange, and tweeter distortion.

That is interesting. I tried it with steady tones and didn't care for it though it might work with quick switching. By not caring for it I just meant it wasn't a pleasant sound. I then tried creating two plucks with those frequencies. Rapid .2 second build up and .5 second log fade out. Sounded almost like a pluck. That might be useful and you might notice distortion at the loud portion fade away as level drops. I also tried it with longer 2 second fade outs and that might be even better.

 

[watchnerd]

I made a track with all 88 fundamentals of the Piano...

Once in a while I remember to play it.

Oh, SoundCloud is still working:

https://soundcloud.com/ray-880875693/piano-notes-88

As output by the JBLs at the Liznin position:

View attachment 8781

That sounds like scary organ chords from old horror movies.

 

[watchnerd]

I should have asked what we're testing.

This is what I use for testing turntables:

 

[DonH56]

That is interesting. I tried it with steady tones and didn't care for it though it might work with quick switching. By not caring for it I just meant it wasn't a pleasant sound. I then tried creating two plucks with those frequencies. Rapid .2 second build up and .5 second log fade out. Sounded almost like a pluck. That might be useful and you might notice distortion at the loud portion fade away as level drops. I also tried it with longer 2 second fade outs and that might be even better.

It is a major chord so if the frequencies are right will sound pleasant, but note the right frequencies are not quite the ones on the piano keys. A major third needs to be significantly lower in pitch to sound "right". The fifth needs to be a little higher to sound right. Ideally what you should hear are the two tones and nothing else; any buzzing etc. is likely due to intermodulation products.

Given f1, f2:
HD = 2f1, 3f1, etc and the same for f2
IMD2 = f1+f2, f1-f2 (sum and difference tones)
IMD3 = 2f1-f2, 2f2-f1, 2f1+f2, 2f2+f1
etc.

Even-order IMD adds very low and very high tones that can add low- and high-frequency "buzzing" that is a little annoying since the tones are not exactly harmonics.
Odd-order IMD adds difference tones close to the original but offset a little and not harmonically related so sound "bad".
As a rule, IIRC, we are more sensitive to IMD than to HD.

Edit: Found a picture: http://www.electronicdesign.com/sit...files/uploads/2013/10/1107_MakingWaves_F1.gif

HTH - Don

 

[j_j]

I would argue that no one signal suffices, either for listening or measurement. Time domain issues, frequency domain issues, linearity, digital artifacts, all require specific qualities in a test signal that exclude observing other problems. Yes, of course you could catenate a set of appropriate signals.