• 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?

Don Hills

Addicted to Fun and Learning
Joined
Mar 1, 2016
Messages
606
Likes
246
Location
Wellington, New Zealand
It may not mean a lot in playback. But when doing multi-track noise floor adds up fast. But then if we do 16-bit out of 24-bit master that involves dithering ...
OK so far...

... noose floor elevated ...
True, but misleading. Dithering adds noise, but if it's audible in any sane scenario all you've done is demonstrate your incompetence.

... content gets compressed a lot
True, content gets compressed a lot these days. But it's not done by the 24 to 16 conversion. It's done by the heavy hand of the engineer on the knobs of the compressor.
 

Frank Dernie

Major Contributor
Patreon Donor
Joined
Mar 24, 2016
Messages
1,174
Likes
1,169
Location
Oxfordshire
I'm not sure why u guys keep telling that more bits is inaudible. Bits are just bits. I understand that u use absolute scale translating bit into dBs. But! Bits can be used for more precision, not necessarily more dynamic range. I'd say use fixed DR, and split it to 24 or 32 bits. to main topic: DSD must die!
You are mistaken. The ONLY thing more bits does is give more dynamic range. It can NOT be used for more precision since it does not influence precision.
This is one of the myths (there are many) which has been perpetuated leading non-technically minded people to draw false conclusions.
 

bennetng

Senior Member
Joined
Nov 15, 2017
Messages
464
Likes
311
Well yes 24 bit is good for everything, but not a big deal for final distribution. Dithered 16 bit can reproduce to -120 dbFS or in some types of dither more. That is good enough it isn't a limitation for playback. And even if used for mixing, mastering it could be much better than your old classic recordings done on reel to reel tape.
Also DAWs do multitrack stuff in 32/64 bit float, dithering is not used due to the fact that floating point arithmetic doesn't have a fixed noise floor. For instance 32-bit float has 25 bits of integer precision but it can handle +/- 770dBs of range in each single sample. The only step involves dithering is at the master bus when export to fixed point formats.
 

Gomjab

Member
Patreon Donor
Joined
Jan 3, 2019
Messages
27
Likes
31
Again it is not an advantage of DSD. They can use the same (no editing) approach even if they use PCM, if they are so confident about their performing and recording skills.
I agree it is not a technical advantage but a side effect. Binaural recordings like those from Chesky also benefit from the creative constraint of the binaural “head”.

I agree that an engineer could use similar techniques recording PCM however it is much safer for a recording engineer to mic everything and have the ability to fix things in post, if necessary, with all the wonderful tools available in a modern DAW. Most recording engineers seem reluctant (understandably) to give up that level of fine grained control over the final product.

My post was just an observation. You’ll also notice it ends with a wink....

:)
 

RayDunzl

Major Contributor
Central Scrutinizer
Joined
Mar 9, 2016
Messages
6,220
Likes
2,266
Location
Riverview, FL
it can handle +/- 770dBs of range
OMG...

*feverishly extends the Shoutometer...

-768dBfs 35,992,521,229,531,800,000,000
-774dBfs 71,985,042,459,063,500,000,000

-770dBfs is like the difference of somebody shouting at you from 1 meter

...and someone shouting at you somewhere between 35 and 71 septillion light years away.
 
Last edited:

bennetng

Senior Member
Joined
Nov 15, 2017
Messages
464
Likes
311
OMG...

*feverishly extends the Shoutometer...

-768dBfs 35,992,521,229,531,800,000,000
-774dBfs 71,985,042,459,063,500,000,000

-770dBfs is like the difference of somebody shouting at you from 1 meter

...and someone shouting at you somewhere between 35 and 71 septillion light years away.
It seems that Audacity cannot zoom to this level to show the differences of fixed point vs floating point, but I attached the audio files so that you can normalize them to 0dBFS and see/listen for yourself.

That's why dithering is not used during floating point operation.

32float.png


32fixed.png
 

Attachments

Gomjab

Member
Patreon Donor
Joined
Jan 3, 2019
Messages
27
Likes
31
OMG...

*feverishly extends the Shoutometer...

-768dBfs 35,992,521,229,531,800,000,000
-774dBfs 71,985,042,459,063,500,000,000

-770dBfs is like the difference of somebody shouting at you from 1 meter

...and someone shouting at you somewhere between 35 and 71 septillion light years away.
According to this site the largest hydrogen bomb ever detonated created an SPL of 282 dB

http://www.makeitlouder.com/Decibel Level Chart.txt

I do understand the advantage of higher bit depth for internal processing in a DAW but it obviously isn’t needed for playback.
 

andymok

Active Member
Joined
Sep 14, 2018
Messages
113
Likes
33
Do stuff like mic, preamp, analog gain staging and mixing chain before digitizing have anything to do with digitizing format and digital mixing procedure?
Jes you guys are so mean.

Does "largest hydrogen bomb ever detonated created an SPL of 282 dB " has any to do with thread?

Does replying a post has anything to do with you?
 

Gomjab

Member
Patreon Donor
Joined
Jan 3, 2019
Messages
27
Likes
31
Jes you guys are so mean.

Does "largest hydrogen bomb ever detonated created an SPL of 282 dB " has any to do with thread?

Does replying a post has anything to do with you?
Just that for playback you don’t need much more bit depth than 16. I gave the bomb example just to give scale to what the 32/64 bit float format being discussed would be like if there were a way to playback something with that much dynamic range.
 

Dismayed

Active Member
Joined
Jan 2, 2018
Messages
195
Likes
109
Location
Boston, MA

graz_lag

Addicted to Fun and Learning
Joined
Oct 13, 2018
Messages
795
Likes
584
Location
Le Mans, France
I prefer the PCM. Up to 768 kHz.
It is a very good attitude of yours to keep your ears well trained, so that they will be ready to decode the 1,536kHz SR ...
 

FrantzM

Addicted to Fun and Learning
Patreon Donor
Joined
Mar 12, 2016
Messages
789
Likes
359

graz_lag

Addicted to Fun and Learning
Joined
Oct 13, 2018
Messages
795
Likes
584
Location
Le Mans, France
1546983533425.png
 

Gomjab

Member
Patreon Donor
Joined
Jan 3, 2019
Messages
27
Likes
31
Last edited:

Gomjab

Member
Patreon Donor
Joined
Jan 3, 2019
Messages
27
Likes
31
Hopefully you can experience native DSD playback one of these days and let us know if you think it sounds any different than converted to PCM.
My SU-8 from Massdrop arrived this weekend so I did get a chance to listen to some sample DSD files from Blue Coast. The sample sounded great. But then the 96 kHz PCM version sounded just as great to my ears. The supplied samples were DSD128, DSD256, 96 kHz FLAC, and 44.1 WAV.

I think I’ll just stick to PCM for now.

The good news is I’m happy with the SU-8 paired with my THX AAA 789!
 
Joined
Jan 21, 2019
Messages
13
Likes
3
DSD has zero benefits. 24Bit has lower noise floor than we can hear (even dithered 16Bit), and 44.1kHz covers all frequencies we can hear, and modern filters are good enough that any decent DAC should not audibly affect frequencies <20kHz.

People saying improved transient response are uninformed; by Nyquist, as long as the sampling rate is >=2x the frequencies you want to cover, it can reproduce it 100% identically.
Might be true for frequency as in single tones, but definitely not w.r.t. time differences or also called time-smear. Regarding Nyquist, on the Nyquist frequency the distortion is maximal, if memory serves me well, 50% or 100%, need to look into that stuff again.
Same as in photography, need of a slanted edge to determine resolution, which is sort of time smear.
 

Similar threads

Top Bottom