Could 8-bit be enough for carrying all the information in a given piece of music?

[GrimSurfer]

So, that was with a noise floor ~50dB down, now imagine 16Bit which is ~100dB down. With audible differences getting exponentially more difficult as bit-depth increases, the answer to OP’s question should be clear.

Now, one case Amir made in other threads is that “music” thresholds vary as the music varies, so we should instead aim for absolute thresholds, which is ~116dB. I agree that this is good form, but is indeed overkill.

I like overkill, if for no other reason that it takes away any rational reason to second guess the outcome afterwards.

This is important in audio, which is so riddled with subjectivity and pockmarked with pseudoscience that finding the truth can be like untying a Gordian knot.

 

[j_j]

Ok, look, if one purely removes redundancy, one can get to ~60% or so of the original bit depth losslessly.

So why try for 50%?

 

[amirm]

It the difference between a rather dumb versus intelligent scheme.

 

[Tks]

I like overkill, if for no other reason that it takes away any rational reason to second guess the outcome afterwards.

This is important in audio, which is so riddled with subjectivity and pockmarked with pseudoscience that finding the truth can be like untying a Gordian knot.

Same here, but I like it equally as much for the aspect of it being a flex piece of the engineers demonstrating their capabilities, rather than buying devices where my money goes to fund more audiophile myth marketing. Some people find beauty in strictly build materials (like precious metals). Me, not so much - there is an elegance to something performing effortlessly under duress that I find attractive. Sorta like Starquad cables if we're going to take the simplest/cheapest example of where this beauty in engineering design can be found, or something like the Benchmark/THX789 amps with boatloads of power that don't sacrifice much of any other aspects. Or the RME DAC for just pretty much everything.

 

[Wombat]

Believe it or not, there is virtually no academic research on the threshold of audibility of different bit depths. (With sample rates, there is some research but limited consensus, but with bit depths, there actually is almost nothing out there.) Reiss did a review of all of the published research in 2016 along both axes; you might find it interesting: http://www.aes.org/tmpFiles/elib/20191025/18296.pdf

It is plausible that 8 bit, properly dithered, would be enough (or close to enough) to be audibly transparent for most people. We just don't have a lot of good published data either way.

Amateurs should read this before posting their home brewed loose impressions.

 

[GrimSurfer]

Ok, look, if one purely removes redundancy, one can get to ~60% or so of the original bit depth losslessly.

So why try for 50%?

After a while, the engineering compromises start to remind one of a 3rd world bridge or a game of Janga where the winner is the second last one to remove a structural element until the thing collapses.

 

[Wombat]

Deregulation in the '1st world': https://www.msn.com/en-au/news/aust...erties-across-australia-are-unsafe/ar-AADpfKI

'HOAs' are often owned by the developers/builders interests.

We need reasonable standards and their enforcement, even in audio.

 

[Frank Dernie]

First, the title is a little provocative because of the phrase "all the information".

Second, the focus on musical signal ignores the issue of noise. A dynamic range of 48 dB would bring noise well into the audible band, particularly in the middle frequencies. Some of this might be masked by the room or signal, so may not be immediately apparent. But it is noise nonetheless and would contribute of fatigue and irritation over time.

The last point is something that doesn't get factored into ABX tests. The sound clips are small and frequently switched. It can, however, be picked up in other ways. Prolonged use at home, for instance, which isn't affected as much by bias because there is no "B sample" against which to compare "A".

You really ought to have read and understood what I wrote before writing this...

I've tested this on myself anecdotally and I was not surprised that I can tell the difference between 44/16 and 44/8 with or without noise shaping. I've got fairly young ears and very clean measuring gear, so I might be an outlier.

I want to go back and see how low I can drop the bitrate before it's audibly transparent with the original. I wouldn't be surprised if 9 or 10 bits is transparent to me. I also wonder what the lowest sampling rate would be in order to make noise-shaped 8-bit transparent for me.

You need the 96 kHz sample rate so that the noise shaping can shift the noise into the part between 22kHz and 48kHz so that you can't hear it. 8-bid is too noisy without noise shaping, as I pointed out, so this won't work with a 44.1 or 48 sampling rate.

 

[Frank Dernie]

The point I was trying to make was simply that even quite dynamic music has pretty well all the musical information within 8 bits. If you listen to rock music less is probably required.
I have zero interest in movies so had not considered how important the wider dynamic range is for them, but it clearly will be, so if you are into home theatre and studio derived special effects a full 120dB of dynamic range would be splendid, I imagine.

But for music, particularly recorded live, it has never been IME.

 

[KSTR]

IMHO it doesn't make sense to compare bit depths with techniques that turn the PCM signal into PWM or any other scheme that uses fast toggling to fake intermediate bit levels. I mean, back in the day we used the 1-bit driven onboard speaker of the IBM-PC to output something which resembled music and speech pretty OK when a low pass filter was used to reduce the overall bandwidth and the switching noise of the PWM signal.
So, I would only compare true PCM streams at 44.1/48kHz with only one bit of white dither applied.

 

[Frank Dernie]

IMHO it doesn't make sense to compare bit depths with techniques that turn the PCM signal into PWM or any other scheme that uses fast toggling to fake intermediate bit levels. I mean, back in the day we used the 1-bit driven onboard speaker of the IBM-PC to output something which resembled music and speech pretty OK when a low pass filter was used to reduce the overall bandwidth and the switching noise of the PWM signal.
So, I would only compare true PCM streams at 44.1/48kHz with only one bit of white dither applied.

True, it isn't any practical use, or suggestion, for actual practical technique but an observation that more than 8-bits of dynamic range is pretty well not needed to capture the musical information. It wasn't just recognisably the same music it was indistinguishable from the full recording to everybody in the room to the extent that many of the people listening were angry and thought they were being tricked since a lot of them were sure even 16 bits wasn't enough, and even more thought LPs had more musical information than digital.

I hadn't thought about home theatre myself, sorry, since I have no interest, but clearly for studio generated sound tracks a full 120dB range may well be there, then 8-bit can't contain all the -sound- information then, obviously.

 

[KSTR]

Well, it depends on the music. Metallica's "Death Magnetic" sure remains enjoyable with 8 bits or even less, raw uncompressed solo piano much less so....

 

[restorer-john]

but an observation that more than 8-bits of dynamic range is pretty well not needed to capture the musical information.

At the end of the day, it's an interesting discussion, Frank. Some people appear to think it may be an attack on their castle built of bits, but it's all fun.

Would you ever go back to 8bit 256 colour PC displays? But do we really need 24 bit (16.7Million) colours?

 

[Cosmik]

NICAM did the following: if you started with a 16-bit signal and were aiming to get to 8-bit, you split the signal into 1ms windows and if any window will fit into 8 bits you send it unchanged. Otherwise you shift everything in it rightwards until the window fits into 8 bits; you send the window of data plus a single value that represents the size of the shift.

At the other end you do the reverse. It's a lossy codec but very, very, simple. Most people would be happy with 8-bit sound using this method.

 

[yikky900]

No go try making 8-bit copies of anything Industrial & metal sounds bad, 512kb/s wavpack hybrid with DNS sounds way better. This pretty much a very basic lossy compression since AAC aim to be fully transparent at 192kb/s by using advanced perceptual models.

 

[Pluto]

A dynamic range of 48 dB would bring noise well into the audible band, particularly in the middle frequencies

Ahah! But the trick of noise shaping is that we move most of the noise out of the middle band into an area where the ear is far less fussy. I assure you, it is possible to see the meters hovering around the -30dB area yet hear hardly any noise of consequence.

[noise]... may not be immediately apparent. But it is noise nonetheless and would contribute of fatigue and irritation over time.

I have to say that I don't readily subscribe to the view, waved by some audiopiles like a flag, implying that flaws become more apparent over time. Your moods and functioning of your hearing may change (hearing is notoriously dependent upon your day to day nasal state etc.). Audio played via a physically stable medium does not change – so you either hear a flaw or you do not; ability to hear faults (and recognise them for what they are) as well as some of the more more subtle qualities inherent in a great recording, is largely a matter of training and practice rather than the passage of time.

 

[Fluffy]

I have to say that I don't readily subscribe to the view, waved by some audiopiles like a flag, implying that flaws become more apparent over time. Your moods and functioning of your hearing may change (hearing is notoriously dependent upon your day to day nasal state etc.). Audio played via a physically stable medium does not change – so you either hear a flaw or you do not; ability to hear faults (and recognise them for what they are) as well as some of the more more subtle qualities inherent in a great recording, is largely a matter of training and practice rather than the passage of time.

I do agree with the notion that flaws become more irritating over time. It's not about audibility, it's about mental fatigue. If you can hear, say, an unwanted hum or resonance, it might not annoy you at first, but over time it erodes your enjoyment of the music, as your ability to ignore it is reduced. For example, in headphones it happened to me a lot that in the first couple of listening I just notice the good parts, but as I used them more the positive qualities become less novel, and the negative qualities take over my attention.

Of course, if the noise is not audible at all, there's no reason you will be bothered by it. The things that become bothering over time are the ones that are just on the edge of audibility – they don't stand out initially, but as you learn to expect hearing them, they become more prominent and irritating.

 

[digicidal]

At the end of the day, it's an interesting discussion, Frank. Some people appear to think it may be an attack on their castle built of bits, but it's all fun.

Would you ever go back to 8bit 256 colour PC displays? But do we really need 24 bit (16.7Million) colours?

No I wouldn't... on either... I demand 10bit's per color for the full 1.07Billion! With HDR and full-array backlighting!
Edit: (I will accept a "meager" 144Hz refresh rate however).

 

[Fluffy]

Actually we do need more than 10 bit color. The 8 bit color standard is limited by the displays we are using, which is also why HDR is much better. Not that 8 bit per channel looks "bad", but if you aspire to simulate realistic colors and contrast ratios, you need to go way higher.

 

[ayane]

You need the 96 kHz sample rate so that the noise shaping can shift the noise into the part between 22kHz and 48kHz so that you can't hear it. 8-bid is too noisy without noise shaping, as I pointed out, so this won't work with a 44.1 or 48 sampling rate.

Why 96? Why not 88.2 or even 75? I want to find the lowest sampling rate that works well enough for 8-bit such that it's audibly transparent to me.

I also want to work from the other way and find the lowest bit rate that will be audibly transparent at 44.1 kHz. I don't think 16-bit is necessary for audible transparency if clever noise shaping is used.