• WANTED: Happy members who like to discuss audio and other topics related to our interest. Desire to learn and share knowledge of science required as is 20 years of participation in forums (not all true). Come here to have fun, be ready to be teased and not take online life too seriously. We now measure and review equipment for free! Click here for details.

Does DSD sound better than PCM?

Status
Not open for further replies.

March Audio

Major Contributor
Manufacturer
Patreon Donor
Joined
Mar 1, 2016
Messages
4,091
Likes
4,090
Location
Albany Western Australia
And there was the presumption on the forum before that that music production pro's and companies and wrong and digital is better.

Not my claim, I just said that I understand their point and why that technically can very well be valid.
"Are you suggesting that the level of noise and distortion (relative to the signal) introduced by 32-bit floating are comparable in level to the noise and distortion (relative to the signal) introduced by the best analogue consoles and units?"

"Error -> distortion & noise. Yes"

"Do you have any evidence ?"

"Yes, plenty.

YOUR claim.

Now, please back up your claim with all that evidence.
 

March Audio

Major Contributor
Manufacturer
Patreon Donor
Joined
Mar 1, 2016
Messages
4,091
Likes
4,090
Location
Albany Western Australia
Because I know many DSP algorithms where value of 32-bit floating point Epsilon for example will kick your nuts. It is just pure math. Not too hard to hit such cases for example with IIR. And with FIR too. Combine number of such limit cases in the processing pipeline and you errors really begin to grow. And to know the final error amount you need to know all the plugins and all their parameters set up by the user.
Now now, you need to stop that. You are making up stories to justify your claim. You would obviously need to compare to an equivalent analogue process and device.
 

SIY

Technical Expert
Technical Expert
Joined
Apr 6, 2018
Messages
3,142
Likes
5,799
Location
Phoenix, AZ
My Earthworks M30 is specified flat from 3 Hz to 30 kHz. The higher model, the M50, is spec'd from 3 Hz to 50 kHz. There are other specialty mics from other companies with significantly higher frequency range (and prices to match).

https://earthworksaudio.com/microphones/m-series/
FWIW, I reviewed the M50 and indeed their 50 kHz claim is correct. I used it to characterize some ultrasonic transducers.

Irrelevant to music recording, of course, but useful for other purposes.
 

andreasmaaan

Major Contributor
Joined
Jun 19, 2018
Messages
3,517
Likes
2,790
And there was the presumption on the forum before that that music production pro's and companies and wrong and digital is better.
Let’s be a bit more careful about how we frame this discussion. I was at pains to say neither was “better”. The claim was simply that digital was higher fidelity, i.e. produces lower noise and distortion.

Since you are making the extraordinary claim that analogue is capable of higher fidelity (is that actually what you’re suggesting btw? I’m still not 100% sure), I think it is you who should provide the supporting evidence.

However, I will take up your challenge when I have some time, and use a DAW (Ableton Live, which BTW is not renowned especially for its audio quality) to mix together 32 separate channels of discrete tones to see what noise and distortion is produced in the mixdown.

I’m confident we’ll see lower noise and distortion than that demonstrated by anything in the analogue domain, but am certainly open to being proved wrong :)

PS: I completely agree with you that we should not make the assumption that all digital is always perfect - but that’s a separate question, and far less controversial than the claim you seem to be making.
 

Miska

Active Member
Joined
Feb 20, 2019
Messages
169
Likes
149
Now now, you need to stop that. You are making up stories to justify your claim. You would obviously need to compare to an equivalent analogue process and device.
Yes, that was my point. Analogue processes don't suffer from this effect which is of artificial digital kind. That's why I said you would need to compare same production work flow for example through 88D and 88RS. Then you could begin to claim if the pro's are right or wrong about this.

But before you do that, don't assume digital is perfect and the professionals using analog consoles are wrong, delusional, don't understand anything or just like distortion. Because you may very well be wrong in your assumption regarding digital paths superiority. It is not granted. Digital certainly has lot of it's own problems that need quite a bit of effort to deal with and are especially tricky in loosely controlled environment like DAW.
 

Miska

Active Member
Joined
Feb 20, 2019
Messages
169
Likes
149
Let’s be a bit more careful about how we frame this discussion. I was at pains to say neither was “better”. The claim was simply that digital was higher fidelity, i.e. produces lower noise and distortion.
And I said it depends on what goes on in your digital and analog paths and how exactly those are implemented. It is not possible to give blanket statements like "digital has higher fidelity".

Since you are making the extraordinary claim that analogue is capable of higher fidelity (is that actually what you’re suggesting btw? I’m still not 100% sure), I think it is you who should provide the supporting evidence.
I fail to see anything extraordinary in it, since neither one is defined in any way. For sure, any semi-decent $100 analog mixing console is better fidelity than the 8/16-bit fixed point mixing done in trackers for example.

Do normalized equal level mix of four 16-bit samples in integer and you have lost three bits of precision.

However, I will take up your challenge when I have some time, and use a DAW (Ableton Live, which BTW is not renowned especially for its audio quality) to mix together 32 separate channels of discrete tones to see what noise and distortion is produced in the mixdown.
Again, mixing is extremely simple DSP process and doesn't demonstrate well the potential problems which appear in more complex algorithms and scenarios that you encounter in real world cases. You need to start playing with typical production things like IIR EQ, reverb FIRs, compressors and such, but bunch of those in chain on every of those channels and then mix the final result.

In addition you typically have at least two chained steps with something like 24-bit middle pipeline. Recording/mixing stage (from a desk) to mastering stage (possibly a desk or DAW).

However, software DAWs running on computers are also likely using better resolution than digital mixing desks and other digital hardware products.
 

March Audio

Major Contributor
Manufacturer
Patreon Donor
Joined
Mar 1, 2016
Messages
4,091
Likes
4,090
Location
Albany Western Australia
Yes, that was my point. Analogue processes don't suffer from this effect which is of artificial digital kind. That's why I said you would need to compare same production work flow for example through 88D and 88RS. Then you could begin to claim if the pro's are right or wrong about this.

But before you do that, don't assume digital is perfect and the professionals using analog consoles are wrong, delusional, don't understand anything or just like distortion. Because you may very well be wrong in your assumption regarding digital paths superiority. It is not granted. Digital certainly has lot of it's own problems that need quite a bit of effort to deal with and are especially tricky in loosely controlled environment like DAW.
Eh. Noise and distortion. Measurable and quantifiable. Use SFDR if you like.

Your claim was that YOU had plenty of evidence, so stop the diversionary tactics talking nebulously about "professionals" and "digital not being perfect" and show us your evidence.

We are all quite aware that digital isn't perfect but that clearly wasn't the question.

What do you think the sum noise and distortion of mixing 32 channels in the analogue domain is?
 

March Audio

Major Contributor
Manufacturer
Patreon Donor
Joined
Mar 1, 2016
Messages
4,091
Likes
4,090
Location
Albany Western Australia
And I said it depends on what goes on in your digital and analog paths and how exactly those are implemented. It is not possible to give blanket statements like "digital has higher fidelity".



I fail to see anything extraordinary in it, since neither one is defined in any way. For sure, any semi-decent $100 analog mixing console is better fidelity than the 8/16-bit fixed point mixing done in trackers for example.

Do normalized equal level mix of four 16-bit samples in integer and you have lost three bits of precision.



Again, mixing is extremely simple DSP process and doesn't demonstrate well the potential problems which appear in more complex algorithms and scenarios that you encounter in real world cases. You need to start playing with typical production things like IIR EQ, reverb FIRs, compressors and such, but bunch of those in chain on every of those channels and then mix the final result.

In addition you typically have at least two chained steps with something like 24-bit middle pipeline. Recording/mixing stage (from a desk) to mastering stage (possibly a desk or DAW).

However, software DAWs running on computers are also likely using better resolution than digital mixing desks and other digital hardware products.
.... Oh dear, now all sorts of limitations and conditions are being stated to try and make your claim plausible.

What about chained analogue processes? Noise and distortion free of course ;)
 
Last edited:

Miska

Active Member
Joined
Feb 20, 2019
Messages
169
Likes
149
Your claim was that YOU had plenty of evidence, so stop the diversionary tactics talking nebulously about "professionals" and "digital not being perfect" and show us your evidence.

We are all quite aware that digital isn't perfect but that clearly wasn't the question.
The discussion I responded to was about music production through analog vs digital consoles. I'm not seeing horribly lot of evidence here regarding the original claim that pro's are stupid and wrong using analog consoles for doing production work and the digital way would be "higher fidelity". But nobody presented any evidence for the claim that digital would be "higher fidelity" in their work flows.

But I would be certainly interested to see a proof that digital is certainly "higher fidelity" in all music production work flows, so that one could actually make some claims about their choice of tools for doing their work.

I'm not sure what was the question. I'm not answering to any questions, just presenting my point of view / opinion like other people here.
 

Miska

Active Member
Joined
Feb 20, 2019
Messages
169
Likes
149
What about chained analogue processes?
They have different kinds of errors. I'm just not ready to swallow that 32-bit floating point produces less errors in all cases.

Then we have your favorite topic of which one is audibly less disturbing type of error.
 

March Audio

Major Contributor
Manufacturer
Patreon Donor
Joined
Mar 1, 2016
Messages
4,091
Likes
4,090
Location
Albany Western Australia
They have different kinds of errors. I'm just not ready to swallow that 32-bit floating point produces less errors in all cases.

Then we have your favorite topic of which one is audibly less disturbing type of error.
.

Are you suggesting that the level of noise and distortion (relative to the signal) introduced by 32-bit floating are comparable in level to the noise and distortion (relative to the signal) introduced by the best analogue consoles and units?"

"Error -> distortion & noise. Yes"


"Do you have any evidence ?"

"Yes, plenty.

It's clear that you have no such evidence. Your position is really to state that it is possible to have a high noise & distortion digital process. Well that's not news. However it's clear you have no basis to make the assertions you have. It's equally evident its possible to have a high noise & distortion analogue process.
 

Cortes

Active Member
Joined
Apr 13, 2019
Messages
118
Likes
119
They have different kinds of errors. I'm just not ready to swallow that 32-bit floating point produces less errors in all cases.

Then we have your favorite topic of which one is audibly less disturbing type of error.
There is not such a thing like "less errors in all cases" in any floating point arithmatic, be 32 bits, 64 bits, or 640000 bits. You can always build algorithms that amplify the round-off errors of representing irrational numbers. On the other hand, any normal coder knows how to manipalate properly real numbers, and is very, very unlikely that writes unstable algos. That your Intel/AMD computer or Kirin/Snapdragon cell phones keep working all the day is the proof of that.
 

March Audio

Major Contributor
Manufacturer
Patreon Donor
Joined
Mar 1, 2016
Messages
4,091
Likes
4,090
Location
Albany Western Australia
.

Then we have your favorite topic of which one is audibly less disturbing type of error.
My favourite topic? No, I just have a realistic view on audibility,. Can you provide some examples of tracks with audibly disturbing digital mixing issues? Ones that aren't part of a deliberate effect?
 

andreasmaaan

Major Contributor
Joined
Jun 19, 2018
Messages
3,517
Likes
2,790
They have different kinds of errors. I'm just not ready to swallow that 32-bit floating point produces less errors in all cases.
The claim is that high-quality digital is higher fidelity than high-quality analogue; it was never claimed that every single digital device is higher fidelity than every single analogue one.

Even if you’re not willing to provide measurements, perhaps you could give a ballpark figure for (unwanted) noise and distortion for a particular high-quality DAW using 32+ bit floating that you’ve previously tested?

I.e. name the DAW, describe the chain/processes, and specify the level of noise and distortion you’ve measured relative to the signal.

Also please tell me if the test I proposed in post #704 would be of interest in your view, and if not, what changes you would propose to try to reveal the worst outcomes of any errors, and how one might measure these.
 

Miska

Active Member
Joined
Feb 20, 2019
Messages
169
Likes
149
Even if you’re not willing to provide measurements, perhaps you could give a ballpark figure for (unwanted) noise and distortion for a particular high-quality DAW using 32+ bit floating that you’ve previously tested?
I don't measure consoles, but I know math problems 32-bit floating point creates, that's why I replied.

Also please tell me if the test I proposed in post #704 would be of interest in your view, and if not, what changes you would propose to try to reveal the worst outcomes of any errors, and how one might measure these.
I already said it is not representative of full work flow of creating modern commercial recording.

Can you provide some examples of tracks with audibly disturbing digital mixing issues?
Oh, that is easy, pretty much 90% of pop music released in past 20 years.

However it's clear you have no basis to make the assertions you have. It's equally evident its possible to have a high noise & distortion analogue process.
Original assertion (not mine) which I replied to was a blanket statement:
Certainly, digital mixes using 32 bit floating point processing (or even 24 bit for that matter) and then bounced using appropriate (readily available) dither and anti-aliasing will introduce less noise and distortion than a mix that has passed through an analogue console and effects units.
But of course that doesn't need to have any basis.


This forum is like going to an extremist church in attempt to have meaningful objective discussion about different religions. Ridiculous.
 

Miska

Active Member
Joined
Feb 20, 2019
Messages
169
Likes
149
There is not such a thing like "less errors in all cases" in any floating point arithmatic, be 32 bits, 64 bits, or 640000 bits. You can always build algorithms that amplify the round-off errors of representing irrational numbers. On the other hand, any normal coder knows how to manipalate properly real numbers, and is very, very unlikely that writes unstable algos. That your Intel/AMD computer or Kirin/Snapdragon cell phones keep working all the day is the proof of that.
Yeah, I worked 7 years for Nokia on creating mobile phones and especially voice and video calls for those and some other multimedia things. I also worked 7 years for Intel. And 20+ years on audio DSP algorithms. So maybe I have some kind of idea.

And I know enought that I would never let my music near DSP algos from those companies.
 

pkane

Addicted to Fun and Learning
Patreon Donor
Joined
Aug 18, 2017
Messages
786
Likes
1,066
Location
North-East US
Yeah, I worked 7 years for Nokia on creating mobile phones and especially voice and video calls for those and some other multimedia things. I also worked 7 years for Intel. And 20+ years on audio DSP algorithms. So maybe I have some kind of idea.
Floating point is a 'lossy format'. Fixed point is even more so. Whether or not these losses result in audible artifacts is not a question of numeric representation, as much as the details of the implementation of the algorithm. It's entirely possible to code most audio processing algorithms so that the losses are minimized. It is much, much easier to code them so that the losses are maximized, instead (if simply by not paying attention :))
 

bennetng

Addicted to Fun and Learning
Joined
Nov 15, 2017
Messages
592
Likes
455
In that famous website I found something like these:
Image1.png


I am not claiming audible differences based on these measurement. However, if you look at some relatively expensive software DAWs like Nuendo, the result is not as clean as Audacity either. Do these results have direct relationship with the advertised bit-depth of the audio engine? Do these results made existing Nuendo users switch to Audacity?

-----------------------
Speaking of ancient hardware DAW...
https://hydrogenaud.io/index.php/topic,114203.msg941970.html#msg941970

The thread above is not a discussion about mixing bit-depth vs fidelity, but some members here may find it interesting.
 
Status
Not open for further replies.

Similar threads

Top Bottom