• 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.

Is big better? (A data into DAC musing)

Jimbob54

Master Contributor
Forum Donor
Joined
Oct 25, 2019
Messages
7,281
Likes
8,633
by “native” I mean the rate presented by the “software transport”. I’m assuming that (for example) the audio stream is encoded as say 24/192 from Tidal, it’s then either sent to the DAC which converts it at “native” 24/192 or by using OSX Audio MIDI Setup (or the likes of an Mscaler or HQplayer??) can then be “forced” the DAC to process/output to 32/768

I dont think thats how it works. The DAC will oversample whatever PCM content it receives to whatever it oversamples at. You can either set your OS to up (or down) sample anything from your player to a set rate, or to allow the software to pass native direct to the DAC.

Upsampling (in software on your PC/MAC/phone) and oversampling (in the DAC to do what it does) are not the same. I dont think the last half of the last sentence is correct. The DAC screen reports what it is fed, not what it is outputting. It outputs analog.
 
OP
Headchef

Headchef

Active Member
Joined
May 1, 2021
Messages
144
Likes
39
Location
In a jar, by the door
I dont think thats how it works. The DAC will oversample whatever PCM content it receives to whatever it oversamples at. You can either set your OS to up (or down) sample anything from your player to a set rate, or to allow the software to pass native direct to the DAC.

Upsampling (in software on your PC/MAC/phone) and oversampling (in the DAC to do what it does) are not the same. I dont think the last half of the last sentence is correct. The DAC screen reports what it is fed, not what it is outputting. It outputs analog.

you’re quite right, sorry half watching the “lawn wiff waff final” and dealing with our mysteriously poorly cat!

I think it’s a matter of the Audio MIDI set up enables you to select the upsampling rate set which is the carried out by the DAC’s XMOS.
 
OP
Headchef

Headchef

Active Member
Joined
May 1, 2021
Messages
144
Likes
39
Location
In a jar, by the door
You see mathematical justification for things like the chord scaler but they just ignore post raw DAC conversion analog filters conveniently. Higher sample rates also means more noise pretty much always at some intermediate point in the conversion. High sample rates are balanced against better or simpler analog filtering so again comes down to specific implementation. Typical best raw DAC performance is near 24/96 or 24/192, but that is typical not a stake in the sand.

thats interesting , so the upscaling from say 24/192 to 32/768k is more likely to introduce noise rather than afford any benefits from the higher resolution filters?
 

Jimbob54

Master Contributor
Forum Donor
Joined
Oct 25, 2019
Messages
7,281
Likes
8,633
you’re quite right, sorry half watching the “lawn wiff waff final” and dealing with our mysteriously poorly cat!

I think it’s a matter of the Audio MIDI set up enables you to select the upsampling rate set which is the carried out by the DAC’s XMOS.

I may be very wrong but I think the mac settings in Audio midi tell the Mac what to upsample to (or not) to pass to the DAC, not what is done on the DAC (XMOS or otherwise). The "hardware" referred to I assume is on the Mac, not with reference to the DAC hardware. @JohnYang1997 may be able to offer us some insight as to what happens in the XMOS chip before the DAC chip in any DAC, not just the Aune with reference to the Apple Audio Midi settings.
 
OP
Headchef

Headchef

Active Member
Joined
May 1, 2021
Messages
144
Likes
39
Location
In a jar, by the door
I may be very wrong but I think the mac settings in Audio midi tell the Mac what to upsample to (or not) to pass to the DAC, not what is done on the DAC (XMOS or otherwise). The "hardware" referred to I assume is on the Mac, not with reference to the DAC hardware. @JohnYang1997 may be able to offer us some insight as to what happens in the XMOS chip before the DAC chip in any DAC, not just the Aune with reference to the Apple Audio Midi settings.

I was assuming that by selecting the DAC that you effectively assign it as the primary audio device over the internal sound card/dac, it would seem pointless to use an external dac otherwise surely? I’m guessing by manually setting the encoding to say 32/768 your effectively telling the OS to instruct the dac to process the incoming file via the DAC’s upscaling processor?
 

Jimbob54

Master Contributor
Forum Donor
Joined
Oct 25, 2019
Messages
7,281
Likes
8,633
I was assuming that by selecting the DAC that you effectively assign it as the primary audio device over the internal sound card/dac, it would seem pointless to use an external dac otherwise surely? I’m guessing by manually setting the encoding to say 32/768 your effectively telling the OS to instruct the dac to process the incoming file via the DAC’s upscaling processor?
Yes, the converter of digital to analog. That's not the same as what digital signal leaves the computer via USB, spdif etc. As Vincent said in his first reply, unless the player software bypasses the OS and sends native files direct to the DAC, any options at OS level will be what the computer does.
 
Last edited:

Jim Matthews

Addicted to Fun and Learning
Forum Donor
Joined
Mar 25, 2021
Messages
973
Likes
1,114
Location
Taxachusetts
have a read up on the noise/distortion figures of the EAR / Yoshino 859 ;)
Those figures do not appear widely distributed.

Your playback system acuity is limited by the noise floor of the least capable component. The ProAc two way (as a design) has also been improved apon in the intervening decades.

If you're after more dynamic range in playback: lower the noise floor and deploy more efficient transducers.

https://www.audiosciencereview.com/forum/index.php?threads/kef-r3-speaker-review.12021/
 

audio2design

Major Contributor
Joined
Nov 29, 2020
Messages
1,308
Likes
1,230
You see mathematical justification for things like the chord scaler but they just ignore post raw DAC conversion analog filters conveniently. Higher sample rates also means more noise pretty much always at some intermediate point in the conversion. High sample rates are balanced against better or simpler analog filtering so again comes down to specific implementation. Typical best raw DAC performance is near 24/96 or 24/192, but that is typical not a stake in the sand. At higher input rates the analog performance starts to degrade from higher total noise from switching transients mainly.
 
OP
Headchef

Headchef

Active Member
Joined
May 1, 2021
Messages
144
Likes
39
Location
In a jar, by the door
Those figures do not appear widely distributed.

Your playback system acuity is limited by the noise floor of the least capable component. The ProAc two way (as a design) has also been improved apon in the intervening decades.

If you're after more dynamic range in playback: lower the noise floor and deploy more efficient transducers.

https://www.audiosciencereview.com/forum/index.php?threads/kef-r3-speaker-review.12021/

like I say I’m actu very happy with the sound that comes out of my system, it impresses me every day which is something I’d ever managed to achieve in the past. I cannot fault the ProAc’s they sound magnificen, so all that is both irrelevant and off topic. This is a question as to ”is it better to “force” a DAC’s settings via OS or is it better to just let whatever software transport you have play at its “native” rate? (I’m not including Roon’s DSP or other EQ)
 
OP
Headchef

Headchef

Active Member
Joined
May 1, 2021
Messages
144
Likes
39
Location
In a jar, by the door
You see mathematical justification for things like the chord scaler but they just ignore post raw DAC conversion analog filters conveniently. Higher sample rates also means more noise pretty much always at some intermediate point in the conversion. High sample rates are balanced against better or simpler analog filtering so again comes down to specific implementation. Typical best raw DAC performance is near 24/96 or 24/192, but that is typical not a stake in the sand. At higher input rates the analog performance starts to degrade from higher total noise from switching transients mainly.

if we could move away from the scaler and back to the optimal rates….

why are 24/96-24/192 optimal?

do you know what’s actually happening in the scenario that I’m referring to? Is the computer simply decompressing the file then sending the unencoded data into the DAC which is then converting it into analog at either the “native” source set rate or sending the decompressed file to the DAC with the “forced“ instruction to upsample and the convert to analog?
 

Jimbob54

Master Contributor
Forum Donor
Joined
Oct 25, 2019
Messages
7,281
Likes
8,633
if we could move away from the scaler and back to the optimal rates….

why are 24/96-24/192 optimal?

do you know what’s actually happening in the scenario that I’m referring to? Is the computer simply decompressing the file then sending the unencoded data into the DAC which is then converting it into analog at either the “native” source set rate or sending the decompressed file to the DAC with the “forced“ instruction to upsample and the convert to analog?

Upsampling doesnt happen in the DAC for PCM- oversampling does. MQA gets tricky as I think then the further unfolds can happen either in the decoder/ interface chip pre DAC chip, or on the DAC chip before conversion in newer models. But if you are playing a PCM file , its either being sent at native rates or upsampled in the source device pre transfer . Modern DS DACs will then oversample that to whatever rate they operate at (say 32/384) and convert. The discussions linked previously focus on whether there is any merit in upsampling in the source device from, say, 16/44.1 to something higher in terms of bit depth and sample rate (all the way up to perhaps 32/384 in this example). But unless you know different, that optional upsample happens in the source machine and passed to the DAC interface chip (the XMOS in the Aune case) then into the DAC itself - or it passes at native file rates - both will be oversampled to what ever the DAC operates at.

See the last comment and others in this thread https://www.forum.rme-audio.de/viewtopic.php?id=28996

And this practically ancient article https://www.audioholics.com/audio-technologies/upsampling-vs-oversampling-for-digital-audio
 
Last edited:
OP
Headchef

Headchef

Active Member
Joined
May 1, 2021
Messages
144
Likes
39
Location
In a jar, by the door
This is intriguing, for all the testing and measuring that is done in reviews it seems there’s actually a fundamental mystery as to how this all works and what is actually the optimal way to run all this clever technology. I feel that with every step towards understanding it there are two steps left or right most of which offer signage in other directions usually marked either “subjective route”, “listener dependent“ or “your mileage may vary” :D
 

audio2design

Major Contributor
Joined
Nov 29, 2020
Messages
1,308
Likes
1,230
why are 24/96-24/192 optimal?

Balanced between internal compute power, I/O interface speed, etc. required for upsampling and/or Delta sigma conversion while keeping noise low and high enough end result for effective analog filters. It creeps up with shrinking process geometry as processing steps require less power and hence usually generate less noise
 
OP
Headchef

Headchef

Active Member
Joined
May 1, 2021
Messages
144
Likes
39
Location
In a jar, by the door
Those figures do not appear widely distributed.

Your playback system acuity is limited by the noise floor of the least capable component. The ProAc two way (as a design) has also been improved apon in the intervening decades.

If you're after more dynamic range in playback: lower the noise floor and deploy more efficient transducers.

https://www.audiosciencereview.com/forum/index.php?threads/kef-r3-speaker-review.12021/

see attached, that’s about the most detai I’ve been able to find however if you’re interested further it’s probably worth reading Tim’s biography here as there’s further discussion about the 859 and also it’s good to know what the chap was all about: https://www.hiendnews.gr/tim-de-paravicini-the-legend-1945-2020-by-jonas-sakkis/
 

Attachments

  • 752B7F0F-6604-4E5A-910E-21B9D0202532.jpeg
    752B7F0F-6604-4E5A-910E-21B9D0202532.jpeg
    249.1 KB · Views: 10

Jimbob54

Master Contributor
Forum Donor
Joined
Oct 25, 2019
Messages
7,281
Likes
8,633
This is intriguing, for all the testing and measuring that is done in reviews it seems there’s actually a fundamental mystery as to how this all works and what is actually the optimal way to run all this clever technology. I feel that with every step towards understanding it there are two steps left or right most of which offer signage in other directions usually marked either “subjective route”, “listener dependent“ or “your mileage may vary” :D
Not really, there are contributors here who understand it very much so both from theoretical and practical angles. Neither you nor I are they though.

You cant build them if you dont understand them. But you can certainly listen to them without the first clue about the science behind them. They are fundamentally plug and play devices from the consumer perspective- all the talk of up/oversampling , set up options etc wont turn a well measuring DAC bad and vice versa.
 
OP
Headchef

Headchef

Active Member
Joined
May 1, 2021
Messages
144
Likes
39
Location
In a jar, by the door
Balanced between internal compute power, I/O interface speed, etc. required for upsampling and/or Delta sigma conversion while keeping noise low and high enough end result for effective analog filters. It creeps up with shrinking process geometry as processing steps require less power and hence usually generate less noise

so the primary constraints are computing/processing power? Or is it that the conversion processes are yet to reach optimal capabilities?
 

audio2design

Major Contributor
Joined
Nov 29, 2020
Messages
1,308
Likes
1,230
so the primary constraints are computing/processing power? Or is it that the conversion processes are yet to reach optimal capabilities?

No constraints on computing / processing power, but computing / processing power comes at a cost, namely power draw, and devices switching from one state to another which creates noise. Internally to the DAC process, every time you switch a FET you inject some charge and that charge has some variability and that creates noise. The more switching the more noise, but also as you shrink processes and improve device performance the amount of injected charge per switching event goes down. Higher power draw = more heat which also creates thermally induced noise.
 
OP
Headchef

Headchef

Active Member
Joined
May 1, 2021
Messages
144
Likes
39
Location
In a jar, by the door
Older Mac's had a headphone out doubling as a Toslink out.
Is your Mac connected over Toslink or over the USB?
I assume USB.

Personally I prefer to play everything at its native sample rate simply because most DAC's are up- or over-sampling.
Hence if the PC does re-sampling and send this to the DAC, it will be resampled by the DAC again.
Resampling as done by OSX is not a crime, in fact it is pretty good https://www.thewelltemperedcomputer.com/SW/OSX/AudioMidi.htm

so this is an odd thing…

software transport: Apple Music (lossless)

computer: A) AppleTV B) MacBook Pro

connects: A) Optical B) USB

DAC Displaying: A) PCM 48 B) PCM 768

now assuming the streams are being streamed at X resolution then processed via the (computer) which determines what encoding the DAC interprets.

and considering that these differences should only be minor and arguably almost unnoticeable (apart from by cats and dogs).

The reality is that the sound is far superior via the MacBook, it’s unsurprising really given the computer processing comparables, and the difference in PCM conversion?

listening to Prince’s Sign o’ the times the soundstage is much wider, the instruments more defined and the actual sounds/notes are clearer with more timbre and texture. And yes whist distracting myself and the wife from the worry of our cat being at the vets I roped the poor Lass into helping me conduct blind tests, the difference is unmistakable.

and as we’re still awaiting news about the cats health I’ll try and get her to help me test it with the MacBook output set to PCM48 so it matches what’s being sent via the AppleTV (which sounds compressed)

curiouser and curiouser….
 
OP
Headchef

Headchef

Active Member
Joined
May 1, 2021
Messages
144
Likes
39
Location
In a jar, by the door
No constraints on computing / processing power, but computing / processing power comes at a cost, namely power draw, and devices switching from one state to another which creates noise. Internally to the DAC process, every time you switch a FET you inject some charge and that charge has some variability and that creates noise. The more switching the more noise, but also as you shrink processes and improve device performance the amount of injected charge per switching event goes down. Higher power draw = more heat which also creates thermally induced noise.

well that probably explains why the S8 runs hot compared to other DAC’s which I’ve used that tend to be cool to the touch. But then is noise figures in the review all came back as pretty good. Do you think DAC/Noise is subject to a sort of Moores Law?
 
Top Bottom