Skip to main content

Notice

Please note that most of the software linked on this forum is likely to be safe to use. If you are unsure, feel free to ask in the relevant topics, or send a private message to an administrator or moderator. To help curb the problems of false positives, or in the event that you do find actual malware, you can contribute through the article linked here.
Topic: Has anyone ever successfully ABXed 16-bit vs 24-bit at the same sample rate? (Read 1583 times) previous topic - next topic
0 Members and 1 Guest are viewing this topic.

Has anyone ever successfully ABXed 16-bit vs 24-bit at the same sample rate?

I've searched the forum and there seem to be some old threads about ABXing RedBook spec audio vs 96+ kHz 24bit "HD", but of course, it's known that higher sample rates can induce inter-modulation distortion, so, I'm interested to know if anyone ever managed to demonstrate being able to differentiate 24-bit or more from 16-bit with proper noise-shaped dither but with both samples at the same non-HD rate, be it 44.1kHz or 48kHz.
I'm having a small debate over this in a chat and I was tempted to say nobody has ever done it, but I'm not certain about that, though I suspect that to be the case.
I know I can't hear it but this person claims that the difference between 16 and 24 is "very audible", and refers hearing it between an "iPhone/iPad" and "Bose PC Companion 5" without understanding what was causing it untit, as it turns out, it was found that the later is 16-bit only or now way could be found to make it do 24. Mentions are made about reverb in some tracks being clearer at 24 and smudgier at 16...

For starters it's a comparison from two entirely different devices, which is not a good start for a claim on a PCM format. The source material is WAV files rendered from MIDI at 48 kHz 32-bit integer PCM with a custom rendering engine. Ruling out placebo, and given what I understand about the issue, I suspect there are more differences in the signal path to account for the alleged audible difference: Hidden or Unaccounted for DSP filtering on he iThings perhaps? Bad bit-depth reduction method in the player, or driver, or the OS mixer itself? bad 16-bit reconstruction on the DAC?
I'd say this is hardly conclusive evidence to throw away all the accumulated knowledge on limits of human auditory perception.

So, has anyone ever successfully told apart 16-bit from 24-bit on a proper ABX test?

Re: Has anyone ever successfully ABXed 16-bit vs 24-bit at the same sample rate?

Reply #1
People "have problems" differentiating 16 bit lossless vs lossy audio of high enough bitrate.
I've not heard about properly set-up ABX test where people were able to hear difference between 16 and 24 bit files.
Error 404; signature server not available.

Re: Has anyone ever successfully ABXed 16-bit vs 24-bit at the same sample rate?

Reply #2
Sure it is if the signal is quiet enough and you turn up loud enough. Like we kids would do to the fade-out of a track, only to get the start of the next at overfull volume, that was fun and our parents were killjoys.

But at the levels you want to listen to music

* Meyer and Moran: "Audibility of a CD-Standard A/D/A Loop Inserted into High-Resolution Audio Playback"
Take a SACD/DVD-A player. Play back with or without an extra A/D/A chain limited to 16/44.1. Result: this bottleneck was inaudible unless you would behave like we kids did. The HA TOS technically forbids me linking to a non-paid version of the paper nor telling you what search engine I used to find it.
Meyer and Moran considered SACD/DVD-A, so they tested more than the 16 vs 24 bits - also the sampling rates.
A "criticism" of M&M was that they didn't really check whether the SACD/DVD-A contained truly high-resolution material in the first place. A null result could be out of the high-resolution material being simply upsampled, and (fraudulently) sold as high resolution. IIRC, Bob Stuart of Meridian voiced that criticism. Bob Stuart would later go on to market that very fraud under the "MQA" name.

* It's been eight years since Archimago reached out to some internet forums to gather blind testers for such a thing. Result: predictably nothing. http://archimago.blogspot.com/2014/06/24-bit-vs-16-bit-audio-test-part-ii.html
But not all about that test was boring: Archimago also asked respondents to rate their confidence in being able to tell. On the three samples, the confident respondents were 50 percent / 48.8 percent / 44 percent correct.

* And an absolutely non-blind but yet striking test posted here at HA. The late Arnold B. Krueger posted a stupidly simple "high resolution listening test": https://hydrogenaud.io/index.php/topic,115718.msg954778.html#msg954778
He spoke out "This is a high resolution listening test", repeated it attenuated ... and results: if listening at realistic volumes, nobody would hear the bottom of the CDDA range.

(Once upon a time, Arnie made a test where the transition between formats could be heard with foo_abx: Even if you couldn't tell A from B, the transition A->B would sound different from B->A. I don't remember whether it was delay or click, but an unfortunate gefundnes Fressen for the hirezophile crowd.)

Re: Has anyone ever successfully ABXed 16-bit vs 24-bit at the same sample rate?

Reply #3
Sure it is if the signal is quiet enough and you turn up loud enough.
Which is actually where you start comparing "less-than-16bit vs less-than-24bit".

Well written, Porcus.
SoX 48000 hz, passband 91%, aliasing enabled
Opus VBR 256 kbps

Re: Has anyone ever successfully ABXed 16-bit vs 24-bit at the same sample rate?

Reply #4
Had one used a significant & exponent format like in float, that wouldn't have been an issue either. But that is a bygone. CD became 16 bits - rather than 14, because Sony insisted on destroying the competitive advantage of Philips' already developed 14-bit DAC.
And the rest, as they say, is history. Sure for computing it is convenient to have two bytes rather than fourteen bits.

You could in principle have imagined a floating-point format, and you could imagine that also 12+4 could have removed the edge Philips had with their already-developed 14 bit DAC. Actually I recently learned, from a link in a posting here at HA, that a few years later, the Yamaha DX-7 synth used something equivalent to some 12+3 bit float format because 12-bit processing was cheap enough: https://ajxs.me/blog/Yamaha_DX7_Technical_Analysis.html

Re: Has anyone ever successfully ABXed 16-bit vs 24-bit at the same sample rate?

Reply #5
Excelent. Thank you @Porcus

I find it really interesting that even that uncontrolled crowd-sourced test organized by that Archimago blog couldn't find any statistical evidence of audibility considering that, under those conditions, many could have faked it, and yet, the golden ears didn't show up in the results.
Somehow I doubt there was a conspiracy among potential golden ears among that crow to fail their ABX tests on purpose and thus hide their super-human hearing.

Re: Has anyone ever successfully ABXed 16-bit vs 24-bit at the same sample rate?

Reply #6
So, has anyone ever successfully told apart 16-bit from 24-bit on a proper ABX test?
You also should add "on normal music with normal listening levels".
I did it on pure 1 kHz tone with level -90 dB and insane amplification on headphone amp.

Re: Has anyone ever successfully ABXed 16-bit vs 24-bit at the same sample rate?

Reply #7
So, has anyone ever successfully told apart 16-bit from 24-bit on a proper ABX test?
You also should add "on normal music with normal listening levels".
I did it on pure 1 kHz tone with level -90 dB and insane amplification on headphone amp.
Heh, I though "normal music and normal listening levels" would be a reasonable implication in the idea of a "proper ABX test" :)

Re: Has anyone ever successfully ABXed 16-bit vs 24-bit at the same sample rate?

Reply #8
with proper noise-shaped dither
Out of curiosity, which shape is proper? :-)


Re: Has anyone ever successfully ABXed 16-bit vs 24-bit at the same sample rate?

Reply #10
I have dilemma too on 24/48 vs 16/88.2
Higher bit or higher sampling quality is better

My conclusion so far is higher sampling is more important than higher bit  for storing content not for mixing / recording /rendering which 24bit is a must

I'm more than delighted if anyone could debunk


Re: Has anyone ever successfully ABXed 16-bit vs 24-bit at the same sample rate?

Reply #11
@lithoc Not that I think your question is worthless, but I fail to see the relation to the topic:
This is entirely about audibility, while your question is related to mixing vs listening.
Also, in regards to listening, high sampling rate is still useless, as you don't have dog ears.

Nevertheless, while I don't fully understand dither at a strictly technical level, it seems to me that, in addition to cancelling out harmonic distortion, it can also let you push the noisefloor of a PCM signal down below the "native" undithered noisefloor of a given bitdepth by adding sufficient total noise energy, so you can raise the noise level in less senstive bands and lower it in more sensitive ones. In terms of audibility I sounds like harmonic distortion may be completely inaudible at 24bit even in silent parts, so perhaps you can do away with dither for listening.
Given this relationship between shaped dither and noise floor, it superficially seems to me like you could use the supersonic bands of "hi-res" audio (hypersonic?) to add all your dither noise and thus lower your noisefloor dramatically in the audible range? Probably not very practical for any use case, even if it works like that.

 

Re: Has anyone ever successfully ABXed 16-bit vs 24-bit at the same sample rate?

Reply #12
I have dilemma too on 24/48 vs 16/88.2
Higher bit or higher sampling quality is better

Neither really matters, but 88.2 is especially pointless.

Re: Has anyone ever successfully ABXed 16-bit vs 24-bit at the same sample rate?

Reply #13
I have dilemma too on 24/48 vs 16/88.2
Higher bit or higher sampling quality is better
Neither on a properly mastered material. 16/44.1 is absolutely enough for everyone. I can say for myself that on some material I even can't ABX 16/32 vs 16/44.1.
SoX 48000 hz, passband 91%, aliasing enabled
Opus VBR 256 kbps