The 24-Bit Delusion

Al M. · Mar 28, 2017

Link:
https://mojoaudiofiles.wordpress.com/the-24-bit-delusion/

rbbert · Mar 28, 2017

I think this is an "infommercial"? For one thing, he says no existing DAC's have better than 20 bit resolution, but FWIW Stereophile has tested and measured quite a few that have between 23 and 24 bit resolution (delta-sigma and dcs "ring"), and even Schitt's best R2R DAC gets between 21 and 22 bit resolution. And even skimming it I noticed a few other half-truths as well.

Al M. · Mar 28, 2017

rbbert said:
I think this is an "infommercial"? For one thing, he says no existing DAC's have better than 20 bit resolution, but FWIW Stereophile has tested and measured quite a few that have between 23 and 24 bit resolution (delta-sigma and dcs "ring"), and even Schitt's best R2R DAC gets between 21 and 22 bit resolution. And even skimming it I noticed a few other half-truths as well.

You mention Schiit's best DAC (the Yggdrasil).

Here is from the Computer Audiophile review of the DAC):

The "Yggdrasil is the world’s only closed-form multibit DAC, delivering 21 bits of resolution with no guessing anywhere in the digital or analog path." According to Schiit Audio. Let's dive into that statement a bit. Many audio enthusiasts will immediately see the 21 bit number assume this DAC is inferior to other DACs that claim 24 or even 32 bits of resolution. Several manufacturers today advertise the fact that their DACs feature multiple 32 bit DAC chips per channel. Making a judgement on a DAC's superiority or inferiority based on the number of bits advertised is foolish. For example, a 24 bit DAC has a theoretical maximum SNR of 144 dB, but the best current DACs can only obtain an SNR of 124 dB or 21 bits due to the noise floor of the components. In addition, human hearing has a dynamic range of about 120-130 dB. What's more, DACs have what's called Equivalent Number of Bits (ENOB) to signify the actual resolution of the DAC. A closer look at many 32 bit DACs reveals they actually have an ENOB of 19.5. Can you see why making judgements about DACs based on specifications is ridiculous?

Readers may be asking themselves, what happens when I play a 24 bit recording on the Yggdrasil if it only supports 21 bits? The reality is that 24 bit recordings don't have 24 bits of resolution / information. It's possible to select 24 or even 32 bits as the output resolution for the Yggdrasil in Audio Midi. The truth is that it doesn't matter on any DAC. Note 1: Vinyl playback has about 12 bits of resolution, CD has 16 bits.

(End quote.)

Elsewhere, on their website, they say:

Q: But the Arglebargle has like twelve 32-bit DACs in it! Yours only has 21 bits! Hell, that’s not a full 24 bits even! What about my 24-bit recordings?

A: If your 24 bit recordings actually have 24 bits of resolution, we’ll eat a hat. And those "32-bit" DACs? Well, they have this measurement known as “equivalent number of bits.” This means, in English, how many bits of resolution they really have. And that number is 19.5. And 21 is better than 19.5, in all the math books we know. (End quote.)

***

So they support the argument in the article that I linked to, that in the real world you can't get better than 20 or 21 bits.

Al M. · Mar 28, 2017

rbbert said:
I think this is an "infommercial"? For one thing, he says no existing DAC's have better than 20 bit resolution, but FWIW Stereophile has tested and measured quite a few that have between 23 and 24 bit resolution (delta-sigma and dcs "ring"), and even Schitt's best R2R DAC gets between 21 and 22 bit resolution. And even skimming it I noticed a few other half-truths as well.

I think you are confusing the ability to decode 24 bits with the resolution at the output of the DAC. From the article, which clearly makes this distinction:

Using the lowest noise power supplies, the most sophisticated grounding, and the most sophisticated resonance control currently available in a digital-to-analog converter, you can’t resolve the least significant bit on a 20-bit recording.

In reality, there are no DACs in the world that are capable of discerning greater than an 20-bit resolution. So any company that claims greater than 20-bit resolution from their DAC is simply full of ****. Oh they can decode 24-bits, because 24-bits does exist in software, but the output from their DAC has less than 20-bits of resolution and dynamic range.

Of course that doesn’t even account for the significant amount of distortion added by signal cables, amplification, and speakers, all of which would not allow resolving even an 18-bit recording.

Talking about "full of ****": Schiit Audio broadly agrees with the author, see above.

(BTW now, upon typing this post, I understand why Schiit spells their name like this: it evades the **** detector in computer programs like also the one governing this forum...)

rbbert · Mar 28, 2017

Take a look at Stereophile reviews and measurements of dcs, Benchmark, Auralic and some other DAC's. I'm pretty sure the S/N ratios are around 140 dB, which is between 23 and 24 bits. And the last time I checked, 21 bits is also more than 20 bits. Instead of assuming I don't know what I'm talking about, why not assume that the author of your linked article might be trying to sell you something, then see what happens from there?

Don't get me wrong, though, 120 dB S/N (or dynamic range) is plenty. 24 bit is mostly about the extra flexibility and resolution in recording and mastering, not an absolute need or benefit from 144 dB dynamic range.

Al M. · Mar 28, 2017

rbbert said:
Take a look at Stereophile reviews and measurements of dcs, Benchmark, Auralic and some other DAC's. I'm pretty sure the S/N ratios are around 140 dB, which is between 23 and 24 bits. And the last time I checked, 21 bits is also more than 20 bits. Instead of assuming I don't know what I'm talking about, why not assume that the author of your linked article might be trying to sell you something, then see what happens from there?

Why should I assume the author wants to sell me something? His DAC, according to his website, "converts 24-bit 192KHz high-resolution music files", so it's not as if he is making a lower-bit DAC and then tries to sell me that as being just as good as hi-res. No, he is talking as a digital engineer, and Mike Moffat from Schiit, and of Theta Digital fame (or do you suggest that he does not know what he's talking about?), seems to agree with him.

It's a technical limitation that starts with the power supply. From the article:

Based on a 2.5V output on a DAC (higher than average), below are the voltages power supply noise must be below in order to hear the LSB:

16-bit LSB noise floor voltage = 76uV
18-bit LSB noise floor voltage = 19uV
20-bit LSB noise floor voltage = 4.75uV
24-bit LSB noise floor voltage = 0.3uV
For a reference, a common LM317 regulator, the quality used in most commercial electronics, has about 150uV peak-to-peak noise, and the world’s lowest noise power supplies (we’re talking NASA, not audiophile) have about 5uV of peak-to-peak noise. That means even with the most sophisticated linear power supplies or batteries available today, 20-bit is theoretically the highest playback resolution and dynamic range possible.

***

rbbert said:
Don't get me wrong, though, 120 dB S/N (or dynamic range) is plenty. 24 bit is mostly about the extra flexibility and resolution in recording and mastering, not an absolute need or benefit from 144 dB dynamic range.

Yes, as the article says:

Then why use these insanely high sampling rates and bit depths? The reason is that higher resolution digital formats minimize quantization errors and quantization noise when editing, mixing, and mastering the recording in a studio environment. These higher resolution digital formats truly only exist in a software and are not capable of existing in actual sound reproduction.

Yuri Korzunov · Mar 28, 2017

rbbert said:
24 bit is mostly about the extra flexibility and resolution in recording and mastering, not an absolute need or benefit from 144 dB dynamic range.

For recording 24 bit is enough. For mastering float point is more suitable. Because integer formats too limited by math. As example, 0.1 dB gain correction.

For playback dynamic range of 16 bit is calculated by 6 dB per bit. But it is not fully correct, in my opinion. Because high and low level signals have different signal/noise ratio.

Of course, here we again stumble about "threshold of audibility".

amirm · Mar 28, 2017

rbbert said:
Take a look at Stereophile reviews and measurements of dcs, Benchmark, Auralic and some other DAC's. I'm pretty sure the S/N ratios are around 140 dB, which is between 23 and 24 bits.

If you mean measurements like this DCS Vivaldi DAC:

The actual noise floor is not -140+ db. Oversampling is used in the measurements resulting in much lower measured noise floor. Without it we would be seeing the ADC noise of the measurement system itself!

The next test they do is a better judge of that with this text that goes with it:

"With a dithered 1kHz tone at –90dBFS, increasing the bit depth from 16 (fig.12, cyan and magenta traces) to 24 (blue and red) dropped the noise floor by 24dB, indicating that the Vivaldi DAC has at least 20-bit resolution, which is the state of the art.
Read more at http://www.stereophile.com/content/dcs-vivaldi-digital-playback-system-measurements#f6btlsLLrVvIUcTE.99"

There is also the issue of linearity. It is not sufficient to output something but rather, it needs to be accurate representation of those bits.

The article does have some technical errors but his overall message is sound. In analysis of I have done so far of high resolution recordings I can't find evidence of true 24 dynamic range. I will be testing more but for now, what he says is true in my opinion.

Also, in listening tests I find people can't tell 16 bit and 15 bit apart let alone these types of bit depths.

Ken Newton · Mar 28, 2017

These are interesting technical questions, however, I feel they more serve to highlight the larger issue of digital audio having having unquestionably excellent traditional performance figures, yet too often failing to deliver a commensurate subjective performance. To be more specific, to me, digital typically sounds wideband, high in dynamic range, and seemingly low in distortion. Where I find digital to fall down is in long term listening fatigue, often manifesting as boredom with the music (a cardinal sin). The onset of this fatgue is gradual these days, yet it still seems to rear it's ugly head much more quickly and frequently than it does with vinyl, for example.

Would genuinely 24-bits of resolution (144dB SNR) provide any subjective musical improvement versus 20-bits (120dB SNR), or even 16-bits (96dB SNR)? Is this why digital is sometimes disappointing with music? Good vinyl has what, less than 80dB SNR? I just don't see that dynamic range is, musically speaking, anywhere near an issue for consumer digital playback. This parameter seems like the kind of objective spec-manship that once was common for amplifier THD figures.

GaryProtein · Mar 28, 2017

.
I think the best thing about 24 bit CDs and SACDs is because they charge so much more for them, the buyers' expectations are higher, so they usually have better engineers making the master than the morons who make them for redbook recordings.

GaryProtein · Mar 28, 2017

Ken Newton said:
. . . .For example, would genuinely 24-bits of resolution (144dB SNR) provide any subjective musical improvement versus 20-bits (120dB SNR), or even 16-bits (96dB SNR)? Is this why digital is sometimes disappointing with music? Good vinyl has what, less than 80dB SNR? I just don't see that dynamic range is, musically speaking, anywhere near an issue for consumer digital playback. This parameter seems like the kind of objective spec-manship that once was common for amplifier THD figures.

The day vinyl gets ANYWHERE close to 80dB S/N ration, I'll think about buying a high end turntable but not for too long!

Those incredibly high S/N ratios are NOT necessary for music to sound good or great.

I agree with all your points.

I defy anyone who has been listening to sounds at 120 dB to even be able to hear anything at 40 dB several minutes later until the ear re-accommodates to normal volume levels, so the need for those S/N ratios are really moot.

rbbert · Mar 28, 2017

Ken Newton said:
These are interesting technical questions, however, I feel they more serve to highlight the larger issue of digital audio having having unquestionably excellent traditional performance figures, yet too often failing to deliver a commensurate subjective performance.

For example, would genuinely 24-bits of resolution (144dB SNR) provide any subjective musical improvement versus 20-bits (120dB SNR), or even 16-bits (96dB SNR)? Is this why digital is sometimes disappointing with music? Good vinyl has what, less than 80dB SNR? I just don't see that dynamic range is, musically speaking, anywhere near an issue for consumer digital playback. This parameter seems like the kind of objective spec-manship that once was common for amplifier THD figures.

Good points, although as you know dynamic range and s/n ratio for analog tape is not necessarily the same thing, since tape records and plays signals well below the noise floor, as can digital (to a much lesser extent) by using dithering. Still, it's likely that a dynamic range of 120 dB is "best", since live musical signals can have peaks that loud (although it is true that 0 dB on the lower end is more challenging to achieve). And I suspect most of us agree that what happens in the digital domain is not the biggest obstacle to good sound, but rather in ADC and DAC.

Al M. · Mar 28, 2017

amirm said:
The actual noise floor is not -140+ db. Oversampling is used in the measurements resulting in much lower measured noise floor. Without it we would be seeing the ADC noise of the measurement system itself!

The next test they do is a better judge of that with this text that goes with it:

"With a dithered 1kHz tone at –90dBFS, increasing the bit depth from 16 (fig.12, cyan and magenta traces) to 24 (blue and red) dropped the noise floor by 24dB, indicating that the Vivaldi DAC has at least 20-bit resolution, which is the state of the art.
Read more at http://www.stereophile.com/content/dcs-vivaldi-digital-playback-system-measurements#f6btlsLLrVvIUcTE.99"

There is also the issue of linearity. It is not sufficient to output something but rather, it needs to be accurate representation of those bits.

The article does have some technical errors but his overall message is sound. In analysis of I have done so far of high resolution recordings I can't find evidence of true 24 dynamic range. I will be testing more but for now, what he says is true in my opinion.

Also, in listening tests I find people can't tell 16 bit and 15 bit apart let alone these types of bit depths.

Thanks, Amir, informative technical comments as always!

caesar · Mar 28, 2017

Gentlemen, interesting conversation! How do considerations here change in the case of digitized vinyl? For example, take a Rolling Stones record from late 60s or early 70s, that has been touched by man only once or twice, and utilize top notch equipment to digitize the record (and additionally eliminate Stylus microphics)? The guy does everything to 24 / 192, and it sounds fabulous (also very few ticks and pops).

Ken Newton · Mar 28, 2017

caesar said:
Gentlemen, interesting conversation! How do considerations here change in the case of digitized vinyl?

I suppose, that the most obvious effect on the digital channel is that the groove noise will act as very strong dither. Further, and, as has been documented elsewhere, low frequency groove noise will be perceived as additional hall ambience.

Yuri Korzunov · Mar 29, 2017

amirm said:
If you mean measurements like this DCS Vivaldi DAC:
....
The actual noise floor is not -140+ db. Oversampling is used in the measurements resulting in much lower measured noise floor. Without it we would be seeing the ADC noise of the measurement system itself!

Hi Amir,

This DAC have about -130 dB noise floor. It is very good for modern DACs.

P.S. Thank you for shared measurements.

DaveC · Mar 29, 2017

Ken Newton said:
These are interesting technical questions, however, I feel they more serve to highlight the larger issue of digital audio having having unquestionably excellent traditional performance figures, yet too often failing to deliver a commensurate subjective performance. To be more specific, to me, digital typically sounds wideband, high in dynamic range, and seemingly low in distortion. Where I find digital to fall down is in long term listening fatigue, often manifesting as boredom with the music (a cardinal sin). The onset of this fatgue is gradual these days, yet it still seems to rear it's ugly head much more quickly and frequently than it does with vinyl, for example.

Would genuinely 24-bits of resolution (144dB SNR) provide any subjective musical improvement versus 20-bits (120dB SNR), or even 16-bits (96dB SNR)? Is this why digital is sometimes disappointing with music? Good vinyl has what, less than 80dB SNR? I just don't see that dynamic range is, musically speaking, anywhere near an issue for consumer digital playback. This parameter seems like the kind of objective spec-manship that once was common for amplifier THD figures.

Exactly, imo this is where higher resolution seems to help, especially DSD. Not sure it really has much to do with resolution as upsampling redbook to DSD provides a lot of the same benefits, but in comparing I do think DSD has a more relaxed and less fatiguing sound.

Al M. · Mar 29, 2017

DaveC said:
Exactly, imo this is where higher resolution seems to help, especially DSD. Not sure it really has much to do with resolution as upsampling redbook to DSD provides a lot of the same benefits, but in comparing I do think DSD has a more relaxed and less fatiguing sound.

Fortunately I, and it seems many others, are not affected by the long-term listening fatigue issue. I can listen to digital for hours on end, and, if the music is good, be as excited at the end as at the beginning. I am never 'bored' with digital, ever. Of course, physical fatigue is a different matter, but that can rear its ugly head also when I listen to vinyl. I don't see any correlation there.

And yes, I listen to 16/44.1 material exclusively (and as PCM, without DSD conversion). That's where all the music is. Also, I still fail to see a consistent benefit of hi-res, and I am not interested in computer audio either.

I do sympathize with those who are affected by digital fatigue. On the other hand, I do get bored with an overly 'polished', 'relaxed' and 'inoffensive' sound, a sound that many audiophiles seem to prefer. Live music doesn't sound polished, relaxed and inoffensive either, so I want no part of that.

DaveC · Mar 29, 2017

Al M. said:
Fortunately I, and it seems many others, are not affected by the long-term listening fatigue issue. I can listen to digital for hours on end, and, if the music is good, be as excited at the end as at the beginning. I am never 'bored' with digital, ever. Of course, physical fatigue is a different matter, but that can rear its ugly head also when I listen to vinyl. I don't see any correlation there.

Luckily it's not an issue for me either, but I have owned DACs in the past that were fatiguing. I think DACs and server programs like HQ Player are getting much better and many modern DACs provide fatigue-free or much less fatiguing sound than in the past. Personally, I upsample everything to DSD... it works for me and makes redbook sound very good.

Joe Whip · Mar 29, 2017

Al M. said:
Fortunately I, and it seems many others, are not affected by the long-term listening fatigue issue. I can listen to digital for hours on end, and, if the music is good, be as excited at the end as at the beginning. I am never 'bored' with digital, ever. Of course, physical fatigue is a different matter, but that can rear its ugly head also when I listen to vinyl. I don't see any correlation there.

And yes, I listen to 16/44.1 material exclusively (and as PCM, without DSD conversion). That's where all the music is. Also, I still fail to see a consistent benefit of hi-res, and I am not interested in computer audio either.

I do sympathize with those who are affected by digital fatigue. On the other hand, I do get bored with an overly 'polished', 'relaxed' and 'inoffensive' sound, a sound that many audiophiles seem to prefer. Live music doesn't sound polished, relaxed and inoffensive either, so I want no part of that.

+100 Al. I buy music in DSD if I can't get it any other way and convert it to PCM. I know there are plenty of DSD fans here but I am not one as I find DSD to be too rolled off and smooth for my taste. Not all DSD but I find far to many to have a sameness in overall presentation. Whatever floats your boat. The great thing now is there are many options to get the flavor you want.

The 24-Bit Delusion

VIP/Donor

Well-Known Member

VIP/Donor

VIP/Donor

Well-Known Member

VIP/Donor

Member

Banned

Well-Known Member

VIP/Donor

VIP/Donor

Well-Known Member

VIP/Donor

Well-Known Member

Well-Known Member

Member

Industry Expert

VIP/Donor

Industry Expert

Well-Known Member

Similar threads