Measured evidence that bit perfect playback software alters the analog output of DAC's?

[Paul R]

My goal is to document and understand what kind of changes in sound quality are caused by playback software. Then one could avoid or optimize the various factors for best playback quality.

 

Have you considered that some of this might well be proprietary information? Perhaps you would be better served to talk in private with Peter St., Rob Robinson, Miska, John from Amarra, Damien from Audirvana+ and so on.

 

I rather expect they have already done this work, and could, if they were so inclined, help you repeat it for the public domain.

 

-Paul

[esldude]

How do you know? Did you try measuring it with different computer and DAC combinations? But in any case it is at least five times more expensive than V-Link192...

 

V-Link192 is USB bus-powered, so it is somewhat likely to have different jitter patterns with different computers (since it's the clock-master). DACs that have built-in USB and don't have galvanic isolation tend to be sensitive too. But good thing is that V-Link192 has isolating output transformers, while the non-192 V-Link doesn't.

 

 

 

Obviously it would be bad idea for me to start posting any measurements "my player vs competitors"... So I stick to my own things.

 

 

 

It can be configured for bit-perfect output too, but it is not intended to be used like that. It would waste 75% of the code in there...

 

Yes I tested multiple computers and three DACs. The Audiophileo is bus powered as well. Obviously there are minor differences between the DACs. The computer or software with a given DAC did not seem to matter. But maybe there is something different that needs measuring or some converters are more effected.

[esldude]

Hi Dennis, a question about difference testing as you implement it:

 

How do you distinguish between error in measuring temporal response that should be compensated for (as you mention in the first quote above) and actual transient/temporal response of the equipment that should not be compensated for?

 

I intersperse sawtooth waves as their harmonic structure will show if residual differences are due to timing or not. If there are transient differences they would show as well as any other signal difference. Another method I use less often due to it being tedious is to dump sample values into a spreadsheet and compare each sample to find any patterns sticking out.

[esldude]

Have you considered that some of this might well be proprietary information? Perhaps you would be better served to talk in private with Peter St., Rob Robinson, Miska, John from Amarra, Damien from Audirvana+ and so on.

 

I rather expect they have already done this work, and could, if they were so inclined, help you repeat it for the public domain.

 

-Paul

 

Yes the propriatary nature of it occured to me. Not sure why they would disclose such privately, but I have not asked. I would think if they had measured results they might advertise them unless doing so also would disclose what they are doing.

[Paul R]

I am assuming that they have found something different to focus on than the common measurements. I don't know that for sure, but I can see several ways software could influence typical DAC measurements.

 

Like, these Rega Dac measurements courtesy of Stereophile.

 

-Paul

 

 

Yes the propriatary nature of it occured to me. Not sure why they would disclose such privately, but I have not asked. I would think if they had measured results they might advertise them unless doing so also would disclose what they are doing.

[Miska]

Yes I tested multiple computers and three DACs.

 

How did you measure it? So far my main methods have been to measure noise floor variations while playing back dithered silence (you cannot play back plain digital silence because in most cases it engages DAC's output mute). Second way is with the normal J-test signal and observe any differences.

 

By the way, can you check if Audiophilleo's BNC output ground is connected to the USB input ground? When the output is correctly transformer isolated they are not connected. Like in V-Link192 coax output. The balanced (AES) output of V-Link192 has pin 1 connected to USB ground which is problem unless used with floating cable.

 

But in many cases performance with transformer isolated S/PDIF is much more consistent than DAC's built-in USB input unless the USB input is galvanically isolated (pretty rare so far).

[esldude]

How did you measure it? So far my main methods have been to measure noise floor variations while playing back dithered silence (you cannot play back plain digital silence because in most cases it engages DAC's output mute). Second way is with the normal J-test signal and observe any differences.

 

By the way, can you check if Audiophilleo's BNC output ground is connected to the USB input ground? When the output is correctly transformer isolated they are not connected. Like in V-Link192 coax output. The balanced (AES) output of V-Link192 has pin 1 connected to USB ground which is problem unless used with floating cable.

 

But in many cases performance with transformer isolated S/PDIF is much more consistent than DAC's built-in USB input unless the USB input is galvanically isolated (pretty rare so far).

 

Yes, I tested with dithered silence. With very low level tones, high level tones, the Jtest and music. The music I mostly differenced when for instance switching from one software player to the other or from one computer to the other. The residuals are down near the silence noise floor. Now I record the output to a TC Impact Twin AD unit. So it isn't exactly lab quality equipment. I also have tested with an MAudio sound card (which isn't as good as the Impact Twin). I also have a Tact unit with an AD converter built into it. A Wolfson based AD I believe.

 

The Audiophileo does not use transformers. Rather than transformers according to the designer he uses differential ECL chips. The power is from TDK Lambda power regeneration chips. It is galvanicaly isolated input to output. If his approach is that much better at isolating the DAC from upstream noise issues, maybe others should use it. If I get a chance, I may swap my Audiophilleo with a friend's Vlink so I can bring it home and see how it does by comparison.

[esldude]

I am assuming that they have found something different to focus on than the common measurements. I don't know that for sure, but I can see several ways software could influence typical DAC measurements.

 

Like, these Rega Dac measurements courtesy of Stereophile.

 

-Paul

 

If you are referring to the different filters, I think Miska's software does that at a much more sophisticated level. But doing that digitally in software means we are no longer bit perfect. Does not mean it isn't worthwhile or an improvement it just isn't quite the question I have. And it wouldn't be the answer for why these bit perfect players sound different.

[Paul R]

No I was referring to them as common measurements.

 

Software can alter the timing of transmitted information, alter the buffering, and in some cases, set parameters in hardware drivers, or even replace hardware drivers with custom versions. On the other paw, I am not sure what one might measure to determine that, save perhaps for a protocol analyzer.

 

I am enjoying this thread, though I have little hope that much will come of it. We have been down this road before and it would seem, that dog does not hunt!

 

-Paul

 

If you are referring to the different filters, I think Miska's software does that at a much more sophisticated level. But doing that digitally in software means we are no longer bit perfect. Does not mean it isn't worthwhile or an improvement it just isn't quite the question I have. And it wouldn't be the answer for why these bit perfect players sound different.

[esldude]

No I was referring to them as common measurements.

 

Software can alter the timing of transmitted information, alter the buffering, and in some cases, set parameters in hardware drivers, or even replace hardware drivers with custom versions. On the other paw, I am not sure what one might measure to determine that, save perhaps for a protocol analyzer.

 

I am enjoying this thread, though I have little hope that much will come of it. We have been down this road before and it would seem, that dog does not hunt!

 

-Paul

 

Well, I agree. But it is interesting why the dog does not hunt.

 

If changes in the protocol also change the output, again seems it would be measurable. Miska is showing noise changes. I agree with an earlier post that even what he shows would appear to be inaudible. But a first step is to show there is a difference. Audibility could be worked out as a separate question.

 

That brings up another question I have, as often such software is said to change the noise of the computer environment. I have a hard time seeing how that effect would be very consistent in particulars. Each different collection of hardware has different noise characteristics even when running identical OS.

[acg]

 

That brings up another question I have, as often such software is said to change the noise of the computer environment. I have a hard time seeing how that effect would be very consistent in particulars. Each different collection of hardware has different noise characteristics even when running identical OS.

 

True. That is why some software lets you tune those hardware influencing parameters yourself. For example XXHE, for a few different hardware influencing settings, has different recommendations based on the number of cores in your CPU, whether Hyperthreading is turned on, and whether the CPU is AMD or Intel.

[Superdad]

That brings up another question I have, as often such software is said to change the noise of the computer environment. I have a hard time seeing how that effect would be very consistent in particulars. Each different collection of hardware has different noise characteristics even when running identical OS.

 

Sure, but if software code changes, OS driver work-arounds, OS minimization tweaks, and running OS/player/music via "quieter" interfaces all reduce "activity" in one system, why would they not reduce them in a similar way in most all other similar systems? Anecdotal reports all across various computer audio forums show people reporting positive results with the above sorts of optimization. So while actual measurable/audible results will vary from system to system, there clear is some consistency of benefit. I realize this is not the proof you are seeking, but I do think it is relevant to your above question.

 

By the way, based on techniques and ideas that my pal John Swenson has--as well as hearing of a very few interfaces which are proving more immune to upstream elements--I do think there is great possibility of reducing susceptibility of the USB (and even S/PDIF) interface. As long as the cure is not worse than the disease (e.g. some methods of galvanic isolation, such as those ADuM devices add a lot of jitter themselves).

 

Cheers,

ALEX C.

[Superdad]

True. That is why some software lets you tune those hardware influencing parameters yourself. For example XXHE, for a few different hardware influencing settings, has different recommendations based on the number of cores in your CPU, whether Hyperthreading is turned on, and whether the CPU is AMD or Intel.

 

ACG: I am trying to get an old utility called CoolBook to work under Mavericks. It allows for under-clocking and voltage reduction. I am curious if it might have a positive effect.

[esldude]

Sure, but if software code changes, OS driver work-arounds, OS minimization tweaks, and running OS/player/music via "quieter" interfaces all reduce "activity" in one system, why would they not reduce them in a similar way in most all other similar systems? Anecdotal reports all across various computer audio forums show people reporting positive results with the above sorts of optimization. So while actual measurable/audible results will vary from system to system, there clear is some consistency of benefit. I realize this is not the proof you are seeking, but I do think it is relevant to your above question.

 

By the way, based on techniques and ideas that my pal John Swenson has--as well as hearing of a very few interfaces which are proving more immune to upstream elements--I do think there is great possibility of reducing susceptibility of the USB (and even S/PDIF) interface. As long as the cure is not worse than the disease (e.g. some methods of galvanic isolation, such as those ADuM devices add a lot of jitter themselves).

 

Cheers,

ALEX C.

 

Spend some time with an AM pocket radio around a few computers. While the idea of reducing noise can be reducing noise in any case the types and frequencies of noise are just very variable. So a type noise leaking from one machine were it audible in its effects downstream could be reduced and make the sound change. Another machine may have nearly none of that same noise or it may come out in very different frequencies. The change that worked on the first machine, might have no discernible effect on the second. There are hundreds of such combinations or more.

[Miska]

That brings up another question I have, as often such software is said to change the noise of the computer environment. I have a hard time seeing how that effect would be very consistent in particulars. Each different collection of hardware has different noise characteristics even when running identical OS.

 

I would say it all winds down to hardware always. Power supplies and particular components. Noisiest components in normal computer are display adapter and HDD/SDD drives. Software behavior of course has some effect on these too. For example on Vista or 7 you can tame some of the GPU activity by disabling Aero. Windows 8 reduced the composition load to some extent with lighter theme.

 

But if you take a bit-perfect player and compare it's activity to all other things going on all the time in normal Windows or OS X, it's negligibly small. Moving a mouse pointer around usually generates more activity than the music playback. And even more so because it keeps GPU busy too. But this is one of the things you can measure, play back again dithered silence, run a realtime spectrogram measurement of the output and move mouse pointer around and pause for a while and observe any changes in background noise. If there are any, you have hardware to fix... (you can amplify this my dragging a large window around instead of just mouse pointer)

 

On my past MacBook Pro, if I played dithered silence with sensitive headphones connected to the headphone out I could hear the GPU and HDD activity. If you use integrated "sound card" the software activity also has much more impact than if you use for example USB DAC with galvanic isolation... So we got back to the hardware again.

 

If you want to have some slightly related fun reading, go here:

http://www.cs.tau.ac.il/~tromer/acoustic/

(it also demonstrates and emphasizes why you should have galvanic isolation for USB ground)

[Superdad]

Spend some time with an AM pocket radio around a few computers. While the idea of reducing noise can be reducing noise in any case the types and frequencies of noise are just very variable. So a type noise leaking from one machine were it audible in its effects downstream could be reduced and make the sound change. Another machine may have nearly none of that same noise or it may come out in very different frequencies. The change that worked on the first machine, might have no discernible effect on the second. There are hundreds of such combinations or more.

 

Okay, but I don't think we are always talking about general radiated EMI and RFI emissions as the "noise." Activity on a given internal bus (Firewire, PCI, USB, memory, etc.) generates a lot of logic switching, and anything CMOS generates tiny spikes. Maybe those things will modulate an AM radio, but they may be swamped by overall emissions, so obviously that won't be a fine enough instrument.

 

 

An anecdote related to what Miska was saying about even OS window/mousing movement creating noise: For the few weeks after I pulled the internal SMPS from my 2010 Mac mini until we built and hooked up the Swenson linear a couple of weekends ago, I had been using an industrial 13.8V/10A Astron linear. Its magnetics hummed whenever the load changed. My room is so quiet that when I screen-share-controlled the mini I could modulate the hum by moving my mouse or a wind around. And I knew when Audirvana had finished with its SRC/buffering because the supply would get quieter! (So glad to be rid of that noisy beast! Our new one is dead silent.)

 

So aside from any generated "noise" from various processes, I think that transient power supply current draw may be a big factor in some of this.

[esldude]

Okay, but I don't think we are always talking about general radiated EMI and RFI emissions as the "noise." Activity on a given internal bus (Firewire, PCI, USB, memory, etc.) generates a lot of logic switching, and anything CMOS generates tiny spikes. Maybe those things will modulate an AM radio, but they may be swamped by overall emissions, so obviously that won't be a fine enough instrument.

 

 

An anecdote related to what Miska was saying about even OS window/mousing movement creating noise: For the few weeks after I pulled the internal SMPS from my 2010 Mac mini until we built and hooked up the Swenson linear a couple of weekends ago, I had been using an industrial 13.8V/10A Astron linear. Its magnetics hummed whenever the load changed. My room is so quiet that when I screen-share-controlled the mini I could modulate the hum by moving my mouse or a wind around. And I knew when Audirvana had finished with its SRC/buffering because the supply would get quieter! (So glad to be rid of that noisy beast! Our new one is dead silent.)

 

Well yes, but then I seem to get it back and forth. Talk about buss noise, and someone will say EMI. Talk about EMI and someone will say buss noise. BTW, you can usually find a frequency to keep tabs on your mouse movement with an AM radio. It will nearly go quiet, move the mouse and it picks up the noise from that activity. And of course I know an AM radio isn't a precision instrument in this use.

 

That is why I would like some good measurements of what happens to the analog signal, I don't even need to know if it is buss noise, EMI or some combination just to see a relationship between computer loading/activity and sound output. Now the next step, figuring out how minimize it then you will need to know more.

[Superdad]

If you want to have some slightly related fun reading, go here:

Acoustic cryptanalysis

(it also demonstrates and emphasizes why you should have galvanic isolation for USB ground)

 

Wow Miska, that's crazy! Also gives credence to the concern about the ever-pernicious ground plane noise.

[esldude]

Wow Miska, that's crazy! Also gives credence to the concern about the ever-pernicious ground plane noise.

 

Yeah, I agree with all that Miska has said above. And yet, my particular device doesn't seem bothered by it. Obviously lots of them are. I have done things like play silence or play a single low level tone, and run movies, or do video re-encoding or other intensive tasks. Didn't get anything from it. Increasingly seems like any serious person needs to do some basic measuring of dithered silence and see how much their computer is bleeding through.

 

As to using on board sound cards, yeah, almost every computer I have used throws out garbage related to activity. Another interesting thing I observed with one was activities causing more audible grunge on the sound card when a hard drive was at 95% capacity. I assume lots more thrashing by the drive. But it made mouse movements and re-drawing screens cause audible trills or other modulated tones out the speakers. Replaced with a bigger drive that wasn't full, it all went away.

[Daudio]

Increasingly seems like any serious person needs to do some basic measuring of dithered silence and see how much their computer is bleeding through.

 

Well, Mr. Serious, time for you to get busy !! Alex and Miska have given you some nice phenomena that you could set up tests to quantize. I mean that's the kind of thing you are always asking about, right ?

 

 

I observed with one was activities causing more audible grunge on the sound card when a hard drive was at 95% capacity. I assume lots more thrashing by the drive. But it made mouse movements and re-drawing screens cause audible trills or other modulated tones out the speakers.

 

When I've had HD's getting too full the whole machines performance suffered from being much too busy. All that activity is going to put more load on the PSU, and generate a lot more noise. An idle system is a quiet one, ah ?

[esldude]

Well, Mr. Serious, time for you to get busy !! Alex and Miska have given you some nice phenomena that you could set up tests to quantize. I mean that's the kind of thing you are always asking about, right ?

 

 

 

In case you missed it, and it looks like you did. I have done pretty much those tests with different equipment. And gotten different result. Leading one to think some stuff is better designed than others. Big surprise huh?

 

I also said I would get a different SPDIF unit and see what results it gave. Still we are mainly talking basic noise pollution. And one that may not be related so much to a given software player as to given hardware.

[Jud]

 

In case you missed it, and it looks like you did. I have done pretty much those tests with different equipment. And gotten different result. Leading one to think some stuff is better designed than others. Big surprise huh?

 

I also said I would get a different SPDIF unit and see what results it gave. Still we are mainly talking basic noise pollution. And one that may not be related so much to a given software player as to given hardware.

 

I'm thinking the noise in the power out of the wall may vary quite a bit from home to home as well. Lots of variables.

[esldude]

I'm thinking the noise in the power out of the wall may vary quite a bit from home to home as well. Lots of variables.

 

Jud,

 

I sent you a couple PM's. Not exactly on this topic though.

[rennq]

Hey esldude- If you're so "curious" why don't you go back to school and get an actual EE and CS degree instead of constantly demand everyone explain everything to you?

[Boris75]

Hey esldude- If you're so "curious" why don't you go back to school and get an actual EE and CS degree instead of constantly demand everyone explain everything to you?

 

Why such aggressiveness?