Is a 0.8 difference in dynamic range audible?

[bearcatsandor]

I have some otherwise itentical albums in both PCM and DSD. The difference bewteen the two is a dynamic range of about 0.8 dB, favoring the DSD every time. I have very good hearing (for a 50 year-old anyway), so am I likely to be able to hear that difference in dynamic range over isolating, closed headphones? The DSD can be a bit of a hassle, and I'm deciding which one to keep, but I'll take quality over convenience every time. There's a "purist" part of me that wants to keep the DSD as the original format, but it was likely converted to PCM anyway so meh on that.

Anyway, what's the threshold for hearing differences in dynamic range?

 

[staticV3]

I have very good hearing (for a 50 year-old anyway), so am I likely to be able to hear that difference in dynamic range?

You tell us
https://www.foobar2000.org/components/view/foo_abx

 

[bearcatsandor]

You tell us
https://www.foobar2000.org/components/view/foo_abx

To make a long story short: I would, but I'm runing a wayland server in Linux, and WINE isn't there yet.

 

[bearcatsandor]

Well, I found an on-line test, and I can easily hear 1 dB differences: Current score: 10/10 (100%) — Confidence : 99.99%
I tried 0.5 dB differences: Current score: 7/10 (70%) — Confidence : 98.03%

For the 0.5 dB i really had to focus, and I'm in a noisy environment so I donno. Does that mean that 0.8 dB of dynamic range would matter to me?

edit: I fogot to link to the test. https://www.audiocheck.net/blindtests_level.php?lvl=0.5

 

[digitalfrost]

To make a long story short: I would, but I'm runing a wayland server in Linux, and WINE isn't there yet.

Well that's no problem sir, here you go: https://lacinato.com/cm/software/othersoft/abx

e: Ah I see you beat me to it. Anyway. My link contains very nice multi platform ABX software.

Well, I found an on-line test, and I can easily hear 1 dB differences: Current score: 10/10 (100%) — Confidence : 99.99%
I tried 0.5 dB differences: Current score: 7/10 (70%) — Confidence : 98.03%

For the 0.5 dB i really had to focus, and I'm in a noisy environment so I donno. Does that mean that 0.8 dB of dynamic range would matter to me?

Are you sure this is the only difference? What about the frequency response? It could simply be two different masters.

 

[danadam]

I have some otherwise itentical albums in both PCM and DSD. The difference bewteen the two is a dynamic range of about 0.8 dB,

How is this dynamic range measured?

 

[artburda]

Well, I found an on-line test, and I can easily hear 1 dB differences: Current score: 10/10 (100%) — Confidence : 99.99%
I tried 0.5 dB differences: Current score: 7/10 (70%) — Confidence : 98.03%

For the 0.5 dB i really had to focus, and I'm in a noisy environment so I donno. Does that mean that 0.8 dB of dynamic range would matter to me?

are you comparing here a 1dB and 0.5dB difference in volume? Or in dynamic range?

 

[bearcatsandor]

Well that's no problem sir, here you go: https://lacinato.com/cm/software/othersoft/abx

e: Ah I see you beat me to it. Anyway. My link contains very nice multi platform ABX software.


Are you sure this is the only difference? What about the frequency response? It could simply be two different masters.

It's a rip of the same SACD layer. One is DSD, the other is PCM. Measuring the same track via ffmpeg -af drmeter, the PCM has an overall dynamic range of 11.1 and the DSD has an overall dynamic range of 11.7. Is that likely to be an audable difference? Does anyone know what the threshold is? To answer @artburda's question, the test was measuring volume, using https://www.audiocheck.net/blindtests_level.php?lvl=0.5

 

[Curvature]

Do you mean level or DNR/SNR?

 

[bearcatsandor]

Do you mean level or DNR/SNR?

Honestly, I don't know. The sparse manual is https://ffmpeg.org/ffmpeg-filters.html#drmeter and the output is:
[Parsed_drmeter_0 @ 0x5629abde1540] Channel 1: DR: 11.3
[Parsed_drmeter_0 @ 0x5629abde1540] Channel 2: DR: 12.2
[Parsed_drmeter_0 @ 0x5629abde1540] Overall DR: 11.7

What it's worth it agrees with what I've seen spit out by foobar, but ffmpeg gives me the decimal places, which is what made me wonder in the first place.

 

[DVDdoug]

It could simply be two different masters.

Right. There could be other differences, it could be a different mix or even a different recording, etc. The dynamic range may not be the most important thing.

If you A/B to decide which one you prefer, if you think you are hearing a difference, do your best to level match. If one is louder you'll probably get the impression that the louder one is more dynamic, or just "better sounding".

via ffmpeg -af drmeter

I don't know how that works. Dynamic range is "complicated". There are short-term dynamics (like drum hits) and longer term dynamics (like a song that starts-out quiet and ends loud) and everything in-between and there is no perfect way to measure it.

The crest factor is commonly used, which is the peak-to-RMS ratio. That's probably what ffmpeg is doing and that's probably OK when comparing DSD and PCM but if you compress to MP3 or cut a vinyl record the wave shape changes making some peaks higher and some lower (without changing the sound of the dynamics). The new higher peaks give the MP3 or vinyl a higher crest factor, making the dynamic range "measure better".

There is an EBU R128 LRA (loudness range) standard which actually tries to measure the loudness differences. I forget how it works but maybe something like comparing the loudest 20% with the quietest 20%. This seems like a better way to do it, but like I said there is no perfect measure of dynamic range of program material.

 

[617]

Greater dynamic range is not necessarily better. Just my experience.

 

[Rja4000]

Well, I found an on-line test, and I can easily hear 1 dB differences: Current score: 10/10 (100%) — Confidence : 99.99%
I tried 0.5 dB differences: Current score: 7/10 (70%) — Confidence : 98.03%

For the 0.5 dB i really had to focus, and I'm in a noisy environment so I donno. Does that mean that 0.8 dB of dynamic range would matter to me?

edit: I fogot to link to the test. https://www.audiocheck.net/blindtests_level.php?lvl=0.5

This test seems to check your ability to identify a change of level.
That does not check your ability to detect a tiny difference in dynamic range (ie noise level, basically).

I guess the sensitivity to noise level difference varies with noise level itself, rather than in absolute terms.

You may perfectly hear 1dB difference of noise level if noise is around -40dB, while you probably won't hear a 10dB noise level difference at -120dB.

So, to answer your original question :
I don't think 0.8dB of dynamic range on CD quality music can be detected.

Can you hear any noise at all out of your speakers when no music is played and your level is set for your maximum listening level ?
Do you hear it more if you increase the level by 1 or 2 or 3dBs ?

EDIT: Read my other post for an answer more to the point.

 

[dpuopolo]

I have some otherwise itentical albums in both PCM and DSD. The difference bewteen the two is a dynamic range of about 0.8 dB, favoring the DSD every time. I have very good hearing (for a 50 year-old anyway), so am I likely to be able to hear that difference in dynamic range over isolating, closed headphones? The DSD can be a bit of a hassle, and I'm deciding which one to keep, but I'll take quality over convenience every time. There's a "purist" part of me that wants to keep the DSD as the original format, but it was likely converted to PCM anyway so meh on that.

Anyway, what's the threshold for hearing differences in dynamic range?

1 db is considered the smallest sound increment that humans can differentiate.

 

[Rja4000]

It's a rip of the same SACD layer. One is DSD, the other is PCM. Measuring the same track via ffmpeg -af drmeter, the PCM has an overall dynamic range of 11.1 and the DSD has an overall dynamic range of 11.7. Is that likely to be an audable difference? Does anyone know what the threshold is?

Ok, so I misunderstood what you meant by "dynamic range".

I have no clue what "11.7" or "11.1" means here, but that should be an indication of how much dynamic compression is applied to the recording.

We should understand what the figure means (how it's computed) first.

EDIT: I found some basic explanation here:

How to use the TT 'Dynamic Range' Meter - and where to get it - Production Advice

Find out how loud your mix is and make informed decisions about compression and limiting. Choose to make your mixes punchy, loud and competitive.

productionadvice.co.uk

So this value seems to indicate the crest factor of the music. (Which has little to do with what we call dynamic range usually).

Difficult to know what could be the threshold here. As explained in the text, that is probably very dependent on the frequency content of the music.

You will probably hear the difference in level if you set both at same listening level (which then means you align for the peaks). But I'm not sure you'll hear any difference if you align the levels for average loudness (same average RMS level or LUFS - watch this excellent video for more explanation about these units)

(I usually compute crest factor in my measurements, to make sure I don't overload. This value varies with the signal form: a square wave has 0dB CF, a sine wave 3dB, Audio Precision Multitone Amir is using 9.7dB, less optimized multitone rather around 12dB...)

 

[Mnyb]

Try to convert the DSD rip to PCM and then compare ?

I’ve find the notion of ripping the same SACD I bit dubious the CD layer and DSD layer may not be then same ?

 

[Rja4000]

Try to convert the DSD rip to PCM and then compare ?

I’ve find the notion of ripping the same SACD I bit dubious the CD layer and DSD layer may not be then same ?

And check both files for average LUFS
(I think ffmpeg is able to measure it)
Then lower level of the loudest file by the difference before comparing.

 

[solderdude]

1 db is considered the smallest sound increment that humans can differentiate.

Nope... it isn't 1dB is simply a ratio (12%) and level differences of 0.5dB (6%) can be differentiated when switched directly.

I have some otherwise itentical albums in both PCM and DSD. The difference between the two is a dynamic range of about 0.8 dB

Anyway, what's the threshold for hearing differences in dynamic range?

Could be a mastering difference and does not have to be a 1:1 transfer of the studio master for the same album.
There can also be big enough level differences making a direct comparison difficult.

It may be best to convert either the DSD to PCM or the PCM to DSD and compare those first to see if you can measure any DR differences between the converted and original.
If you can't then most likely the masters were not exactly the same. Assuming the DAC outputs the exact same output level for DSD and PCM you can even AB those.

 

[Rja4000]

Depending how clever the DR figure computing is, that may even be some silence added at the end of the track. A few seconds could cause this difference if taken into account.
I think LUFS measurement standard uses a minimum threshold to avoid this kind of "cheating".

That could also be ffmpeg DSD conversion that screws it.
That's why it would be better to compare on 2 same format PCM files.

 

[voodooless]

Could be a mastering difference and does not have to be a 1:1 transfer of the studio master for the same album.

It could even be a DSD to PCM conversation thing in ffmpeg that led to the .8 dB difference. It must convert DSD to do any processing on it. There are many options for this conversion available as well. Might well be that with a bit of tweaking, you’ll get the same result. If the actual playback is not though the same conversion, all bets are basically off.

It may be best to convert either the DSD to PCM or the PCM to DSD and compare those first to see if you can measure any DR differences between the converted and original.

Convert both to same rate PCM, level match, and then through it though something like deltawave. That would give some indication of the actual differences.

If you think the DSD version is the better one, and you don’t want the hassle, just convert it to PCM once, and play that.