Beta Test: DISTORT - audibility of distortions

[pkane]

Was going to download version 7 and saw it was up to 8. There is a problem with 8. It is leaving H2 and H3 at -300 db no matter what settings I use. Appears to only be distorting H4. Sounds that way too. At the settings shown it is nearly clean sounding.

I spent some time testing with smaller FFT sizes. v1.0.9 is available as the result.

It should behave better than v.8 with smaller FFTs. I still recommend keeping the FFT size at 64k, if possible, but the differences should not be nearly as big as they were previously:

FFT Size=64k:

And FFT Size=16k:

 

[Theriverlethe]

H2 at -90 dB

H3 at -98 dB

H4 at -107 dB

H5 at -107 dB

The three files better with Bypass.

But my modded AV Marantz SR4500 is the bottleneck, I would need an amplifier with better specifications, so better we not take it for valid, I think.

https://www.soundandvision.com/content/marantz-sr4500-av-receiver-ht-labs-measures

View attachment 41024

My improvements have focused on reducing noise and interference (higher SNR -> better sound), so it is assumed that the distortion will be the same as the original. This week I made other internal improvement, improving sound depth (usually I listen to in near field) -> perhaps I had also modified the H2 phase?

- The End -

You realize this is meaningless with a sighted comparison, right?

 

[pkane]

You realize this is meaningless with a sighted comparison, right?

I'm going to add a blind comparator to Distort, as the distortion is too obvious by looking at the charts

 

[MediumRare]

Something that's now possible with DISTORT is to reproduce the effect of the harmonic distortion of many of the on-line measured devices without access the physical device.
...
Assuming I correctly programmed various harmonic levels, these settings should make the file sound very similar to what I would hear coming out of the physical device... Let me know if you can hear the effect of the tube

I've also read that part of the appeal of the tube/vinyl sound is the added analogue compression. Would this reproduce that "benefit" as well - or is that a whole different kettle of fish?

 

[PierreV]

I'm going to add a blind comparator to Distort, as the distortion is too obvious by looking at the charts

Or even a small game - pick a harmonic spray, play the tune and ask the user to guess the respective values (or more simply guess the applied profile blindly). I suspect this will objectively kill any claim that an anally (sorry ) specified harmonic profile is better than another. But I am ready to be surprised as my musical ear must rank in the bottom percentile...

 

[maty]

By the way, the other harmonic profile that I had thought to emulate, and that makes no sense with my Marantz, is the LJM L12-2. Maybe with the future PURIFI 1ET400A kit...

https://www.audiosciencereview.com/...s-a-amplifier-off-aliexpress.3372/post-269889

PS: I know of one ASR member who is really liking.

 

[Theriverlethe]

I'm going to add a blind comparator to Distort, as the distortion is too obvious by looking at the charts

And watch the H2 JND magically shoot up to -30dB...

 

[pkane]

I've also read that part of the appeal of the tube/vinyl sound is the added analogue compression. Would this reproduce that "benefit" as well - or is that a whole different kettle of fish?

Compression is just another effect of the non-linear transfer curve, which is the same mechanism for creating the harmonic distortion and IMD. So, yes, you can reproduce it with DISTORT.

 

[Theriverlethe]

Compression is just another effect of the non-linear transfer curve, which is the same mechanism for creating the harmonic distortion and IMD. So, yes, you can reproduce it with DISTORT.

Can you add a mode that shows the decibel axis as logarithmic? I think people have a poor conception of what they claim to be able to hear.

 

[digitalfrost]

Added support for DirectSound playback (@digitalfrost please check!)

Using version 1.0.9:

Playback works, in the sense that I don't get an error, and it appears to be playing, but I get no actual output to the card, no matter if I run it at 48khz (where it would need to resample) or at 44khz ("native" sampling rate).

See

https://imgur.com/a/PqGxyMf

You can see in TotalMix there's nothing at all.

 

[Blumlein 88]

I spent some time testing with smaller FFT sizes. v1.0.9 is available as the result.

It should behave better than v.8 with smaller FFTs. I still recommend keeping the FFT size at 64k, if possible, but the differences should not be nearly as big as they were previously:

FFT Size=64k:
View attachment 41131

And FFT Size=16k:
View attachment 41133

Yes I had been experimenting with small FFTs in the previous version and didn't change it back. Was using 16 k. The new version 9 does work at 16 k, but it sounds cleaner than I think it should vs the 64 k FFT. I've switched back to 64k FFT. Maybe you shouldn't go below 32 k on that part.

 

[pkane]

Yes I had been experimenting with small FFTs in the previous version and didn't change it back. Was using 16 k. The new version 9 does work at 16 k, but it sounds cleaner than I think it should vs the 64 k FFT. I've switched back to 64k FFT. Maybe you shouldn't go below 32 k on that part.

Well, that tells me I do need a blind comparator in DISTORT

FFTs are used for display only and to compute THD and harmonic content. They are not used for playback or to distort a file. At all

 

[pkane]

Using version 1.0.9:

Playback works, in the sense that I don't get an error, and it appears to be playing, but I get no actual output to the card, no matter if I run it at 48khz (where it would need to resample) or at 44khz ("native" sampling rate).

See

https://imgur.com/a/PqGxyMf

You can see in TotalMix there's nothing at all.

Strange. This is using some very basic standard Windows sound output and works fine with the drivers I have tested. Let me see if I missed anything obvious. If not, I may need to build a special version for you to produce a more detailed log so we can figure out why it doesn’t play.

 

[pkane]

Distort v1.0.10 is now available.

Changes include:

• Option to chose the number of generated harmonics, from 1 to 500
• 20Hz and 32 multi-tone test signals

Setting more harmonic levels becomes important if the chosen harmonic profile is high-amplitude or very slow decaying. The default was 40 in all the previous versions. This can now be adjusted to up to 500. Things will get much slower with more harmonics being added, so I recommend doing testing/initial evaluation with the default for 40-100 harmonics, then add up to 500 before generating the output file.

Example of why this may be needed:

Slow-decay harmonic spray starting with a 50Hz tone ends abruptly, after 40 harmonics in the previous versions:

Now, setting number of harmonics to 500, the result looks like this:

In addition, testing with a 32 tone multi-tone test signal (switching FFT window to Dirichlet -- same as the rectangular window):

 

[pkane]

Can you add a mode that shows the decibel axis as logarithmic? I think people have a poor conception of what they claim to be able to hear.

Isn't the dB scale already logarithmic?

 

[Theriverlethe]

Isn't the dB scale already logarithmic?

Yes, but it’s graphically represented as linear in order to show useful information at very low levels. Someone who isn’t very familiar with what they’re looking at might be led to think that a spurious tone at -50dB is half as loud, when it’s actually 1/32.

 

[pkane]

Yes, but it’s graphically represented as linear in order to show useful information at very low levels. Someone who isn’t very familiar with what they’re looking at might be led to think that a spurious tone at -50dB is half as loud, when it’s actually 1/32.

So you are asking to add a linear display in addition to logarithmic?

Log scale (for example, in dB) represents the way we hear sound. That's the main reason it's used. This is a "linear" scale as far as human hearing is concerned. For example, 40dB sounds twice as loud as 20dB to us, while the actual (linear) sound level is ten times greater.

 

[Theriverlethe]

So you are asking to add a linear display in addition to logarithmic?

Log scale (for example, in dB) represents the way we hear sound. That's the main reason it's used. This is a "linear" scale as far as human hearing is concerned. For example, 40dB sounds twice as loud as 20dB to us, while the actual (linear) sound level is ten times greater.

This is incorrect. While a 20dB increase represents a 100x increase in power, it is perceived as (roughly) 4x louder. So, I guess I’m asking for a “perceptually linear” scale rather than an actual linear scale.

 

[pkane]

This is incorrect. While a 20dB increase represents a 100x increase in power, it is perceived as (roughly) 4x louder. So, I guess I’m asking for a “perceptually linear” scale rather than an actual linear scale.

You are right, 20dB is 4x the louder, not 2x as I wrote originally.

 

[pkane]

So, I guess I’m asking for a “perceptually linear” scale rather than an actual linear scale.

OK, so something like a sone scale instead of dB?