Beta Test: DISTORT - audibility of distortions

[pkane]

Its is viable to do the 32-tone multitone test on speakers like in DAC/AMPs?

Sure, why not? You may need to be careful with levels, possibly apply a pink tilt to the multitone to not burn out the tweeter.

 

[Aurvandil]

It is the phase and group delay that I would be interested in dealing with. I have heard different opinions both in discussions on forums and in articles by reputable people.

 

[atmasphere]

Is there a version available for Linux?
Any chance of distortion vs frequency be added?

 

[BKDad]

I looked for this information, but the search terms I used probably were not the right ones, because I didn't find the answer. I apologize in advance.

Jitter on the conversion clocks in mixed signal applications usually is a combination of amplitude modulation and phase modulation. They are oscillators, after all.

Does Distort add amplitude modulation, phase modulation, or both? I'd like to investigate the effects of phase modulation off the clock at low frequencies.

Thanks!

 

[pkane]

Is there a version available for Linux?
Any chance of distortion vs frequency be added?

Somehow I didn't get the notifications about these posts, sorry about the late reply! But yes, I did add this and produced some results that showed the effect and a few A/B tests.

 

[pkane]

I looked for this information, but the search terms I used probably were not the right ones, because I didn't find the answer. I apologize in advance.

Jitter on the conversion clocks in mixed signal applications usually is a combination of amplitude modulation and phase modulation. They are oscillators, after all.

Does Distort add amplitude modulation, phase modulation, or both? I'd like to investigate the effects of phase modulation off the clock at low frequencies.

Thanks!

Jitter is a time-domain effect that impacts the timing of samples in a recording or reproduction. It can be viewed as amplitude modulation of the original signal with some other, undesirable signal, that may be correlated or uncorrelated with the the real content. DISTORT provides the ability to define the corrupting signal as a set of individual frequencies and amplitudes, as noise, or as a correlated signal at some specific amplitude. The corrupting signal phase is currently not controllable through DISTORT, although would be easy to add.

 

[BKDad]

Jitter is a time-domain effect that impacts the timing of samples in a recording or reproduction. It can be viewed as amplitude modulation of the original signal with some other, undesirable signal, that may be correlated or uncorrelated with the the real content. DISTORT provides the ability to define the corrupting signal as a set of individual frequencies and amplitudes, as noise, or as a correlated signal at some specific amplitude. The corrupting signal phase is currently not controllable through DISTORT, although would be easy to add.

Well, "phase noise" on conversion clocks can be both amplitude related as well as phase related. For example, a slow frequency drift over time would be phase related and not necessarily show up as sidebands on each converted tone over a single FFT "sweep" time. If it's easy, or you're plain curious about it, adding that capability to DISTORT might be revealing. Emphasis on might.

I do know that loudspeaker cabinets physically moving, even a little, has an effect on the perceived sound. Could electrical modulation of the waveform phase cause something similar? Dunno, and I haven't found much research on this. I could be bad researcher, though.

 

[pkane]

Well, "phase noise" on conversion clocks can be both amplitude related as well as phase related. For example, a slow frequency drift over time would be phase related and not necessarily show up as sidebands on each converted tone over a single FFT "sweep" time. If it's easy, or you're plain curious about it, adding that capability to DISTORT might be revealing. Emphasis on might.

I do know that loudspeaker cabinets physically moving, even a little, has an effect on the perceived sound. Could electrical modulation of the waveform phase cause something similar? Dunno, and I haven't found much research on this. I could be bad researcher, though.

The goal of jitter simulation was to reproduce "standard" audiophile-type jitter that is commonly used to explain why digital doesn't sound like analog and why real audiophiles have to invest in $5k external clocks to beat the dreaded jitter effects or to invest in mega-buck D/A converters with oven-controlled oscillators. What you're describing is more like wow or flutter type of an effect, and is a large issue with vinyl/LP playback and mostly non-existent in digital.

 

[BKDad]

What you're describing is more like wow or flutter type of an effect, and is a large issue with vinyl/LP playback and mostly non-existent in digital.

Is it?

I had some friends who got very interested in clocks. They tested a zillion clocks, both commercial and designed by them, using lots of different crystals. They found that they could hear differences in perceived sound. They couldn't detect any measurable differences in the audio testing or with standard RF testing of the oscillators. It drove them (more) crazy over at least a year.

Eventually, they purchased a test system that used cross correlation to measure Allan variance down to hundredths of a Hz. They found that the best sounding clocks had much lower phase noise below 1 Hz than the others. All else was pretty much the same.

Keep in mind that they did the audibility testing first and were perplexed by what they heard so they then did more extensive testing. So, no biases upfront and the test equipment had no biases, either.

This was not DBT or any of that, but they were convinced.

These were not fancy audiophile clocks either. They all were typical commercial oscillators of various kinds and, as I said, ones they built. Even within the ones they built, different crystals from the same vendor and same lot had very different low frequency phase noise performance.

Anyway, I'd personally like to explore what this might be about. Adding a varactor modulated by an external voltage to a clock crystal in a DAC might allow the same thing, but it's less controlled. And way more tedious.

I can also tell you that in telecommunications systems, low frequency wander is a thing...

 

[pkane]

Is it?

I had some friends who got very interested in clocks. They tested a zillion clocks, both commercial and designed by them, using lots of different crystals. They found that they could hear differences in perceived sound. They couldn't detect any measurable differences in the audio testing or with standard RF testing of the oscillators. It drove them (more) crazy over at least a year.

Eventually, they purchased a test system that used cross correlation to measure Allan variance down to hundredths of a Hz. They found that the best sounding clocks had much lower phase noise below 1 Hz than the others. All else was pretty much the same.

Keep in mind that they did the audibility testing first and were perplexed by what they heard so they then did more extensive testing. So, no biases upfront and the test equipment had no biases, either.

This was not DBT or any of that, but they were convinced.

These were not fancy audiophile clocks either. They all were typical commercial oscillators of various kinds and, as I said, ones they built. Even within the ones they built, different crystals from the same vendor and same lot had very different low frequency phase noise performance.

Anyway, I'd personally like to explore what this might be about. Adding a varactor modulated by an external voltage to a clock crystal in a DAC might allow the same thing, but it's less controlled. And way more tedious.

I can also tell you that in telecommunications systems, low frequency wander is a thing...

The good thing is there are tools that let you measure clock effects. For example, Multitone has a plot for measuring sub-1Hz effects (the LP measurement is designed specifically for measuring flutter and wow in a recorded signal). DeltaWave measures jitter and clock drift (and includes tools that do much better than a simple cross-correlation).

I have gone down the path of seeking precise clocking, but for a very different hobby - astrophotography. This is where precise timing is much more critical than in audio. Using oven-controlled GPS-disciplined oscillators is a lot of fun. But audible in a D/A converter? Nah

In an answer to your friends, DBTs exist for the simple reason that bias is not something that can be discounted or eliminated or even trained-out. It is omni-present in all perception. Hearing is possibly one of the worst offenders in that respect. Expectation bias is just one of many hundreds of types of biases that one has to be concerned about when trying to reach objective conclusions about perception. Here's a pretty good list to start with:

 

[BKDad]

The only response I have with regard to the test my friends performed was that they set out looking for better clocks and/or better oscillator circuits and ways to optimize their performance. If they had any preconceived notions, it was that they could use better transistors, topologies, power supply circuits, and so on to improve things in a meaningful way. They didn't find that, but did observe audible performance differences that were reproducible for them. They then set out to find out what could be causing this. That ultimately led them to the low frequency phase noise thing.

But, anyway, so many people have told me that things are already better than they could otherwise possibly be that I am going to drop this kind of investigation in audio. Maybe consider astrophotography or something.

Thanks for humoring me.

 

[pkane]

Maybe consider astrophotography or something.

Can highly recommend it. Technically very challenging (not inexpensive, and with its own share of snake oil providers) but very rewarding.

 

[Hayabusa]

This was not DBT or any of that, but they were convinced.

so sad they omitted the chance to make real audio headline shocking news by proving this is audible...

 

[atmasphere]

Somehow I didn't get the notifications about these posts, sorry about the late reply! But yes, I did add this and produced some results that showed the effect and a few A/B tests.

Is it possible to show how something like the GFA565 monostrapped at power levels greater than those of the testing of that amp?

The reason I ask is that when you look at the spectrum of distortion, you see the 5th and higher ordered harmonics down quite a lot compared to the lower orders. I would predict, based on seeing that, that the amp would 'sound' fairly smooth. Usually when I've heard 'brightness' in solid state amps, its been at higher power levels. I do have to point out that at my age, I'm not hearing nearly what I could 40 years ago. A lot of solid state amps I'm hearing these days don't seem to 'sound' all that bright...

 

[pkane]

Is it possible to show how something like the GFA565 monostrapped at power levels greater than those of the testing of that amp?

The reason I ask is that when you look at the spectrum of distortion, you see the 5th and higher ordered harmonics down quite a lot compared to the lower orders. I would predict, based on seeing that, that the amp would 'sound' fairly smooth. Usually when I've heard 'brightness' in solid state amps, its been at higher power levels. I do have to point out that at my age, I'm not hearing nearly what I could 40 years ago. A lot of solid state amps I'm hearing these days don't seem to 'sound' all that bright...

In order to simulate larger levels of distortion at higher levels, I'd need to see some measurements at that power output. I can then adjust the settings in DISTORT to reproduce the same type of distortion, including any variation with frequency.