Investigation Into Effects Of PC load On DAC Analogue Output

[botrytis]

20 minutes ago, Summit said:

 

Are you serious? Noise, jitter, FN, RFI, EMI etc etc.

 

I think the point is, there wasn't any of that in these measurements. Not saying it can be but believe that music is so special that it can be interrupted by farts of unicorns - OK a little sarcasm here - the point is, I work on instrumentation that 100X more data than ever comes from music and also the changes in voltages we look at are very very small. A PPB, signal, from a Mass Spec, is around 0.0000005 V. If these instruments have no problems with noise etc. (and we are dealing with huge magnet systems also and vacuum pumps, etc) why should music have such issues?

 

Jitter was explained - it is an also-ran argument.

 

I am just asking an honest question.

[Summit]

3 minutes ago, idiot_savant said:

Right, so where is this pollution on the measurements provided? @March Audio hasn’t taken any outrageous measures, so maybe we’re worrying about nothing?

 

When you say signal is polluted with jitter etc do you mean the samples are corrupted somehow? Delivered at the wrong time?

 

I’m genuinely interested - “signal” could be the actual audio samples, or the USB packets, or the output from the DAC itself

 

 

your friendly neighbourhood idiot 

 

Sorry am I'm not interested and find it a waste of time to discuss with people who mean that samples have to be corrupted to affect sound quality. 

[botrytis]

1 minute ago, Summit said:

 

Sorry am I'm not interested and find it a waste of time to discuss with people who mean that samples have to be corrupted to affect sound quality. 

 

You are missing the point. You say that the data is corrupted by jitter. Measurement? or is it you hearing it and saying, 'AH, that's jitter'. I am not trying to be sarcastic at all, just trying to understand is all. You need to be more specific in your example. 

 

[idiot_savant]

Erm,

 

If you’re not interested, why post?

 

This sub-forum is the objective forum. If I can’t work out what “signal” means from your statement, and I am an idiot, if you can’t help me out, what can I do?

 

 

your friendly neighbourhood idiot 

[Racerxnet]

6 minutes ago, idiot_savant said:

Erm,

 

If you’re not interested, why post?

 

This sub-forum is the objective forum. If I can’t work out what “signal” means from your statement, and I am an idiot, if you can’t help me out, what can I do?

 

 

your friendly neighbourhood idiot 

All right, the idiot made me chuckle. Carry on. 

[idiot_savant]

Ok, 

 

thought experiment time - we’ve got the designer of the best DAC in the world, but he’s obsessed about EMC/RFI/jitter, so the only interface he provides is punched cards. 
so, he sets up his system, so it  starts  reading in red book audio into a buffer when he presses start, and there’s a button, let’s call it “end of file” that says we can start playback. 
so, he loads up “wish you were here”, waits about 7 hours and puts in his ear defenders until the last card is read. 
he presses “end of file”. 
the DAC starts playing. Was the fact there was someone playing a trumpet in the room in the previous 7 hours make a difference?

 

*edit* euro 2020 has started, so this analogy not quite as thought through as I’d have liked

 

your friendly neighbourhood idiot 

[idiot_savant]

Wish you were here, punched card official photo - see how happy we are to be free of jitter and RFI

[Racerxnet]

12 minutes ago, idiot_savant said:

Wish you were here, punched card official photo - see how happy we are to be free of jitter and RFI

She probably has the jitters from thinking about having to stick 40K of them in the crack.

[March Audio]

9 hours ago, Summit said:

 

Are you absolutely sure that it is indeed desirable that a DAC is very insensitive to noise on the signal input? Serious and fundamental question.

 

Context: There is always a great risk that in order to become completely insensitive to noise, extra circuits such as galvanic insulation are needed. Some think that step analogue filters, GI etc. can have a devastated effect on SQ and that it is bigger than the advantage they have on reducing noise, which often are already very low, at least if feed from a low noise source. The negative effects is caused because those filter/GI etc. function by preventing electrons from flowing directly between two points. Typical the sound can sound less direct, less exiting, less transparent and softer.

 

Best not to pollute in the first place, instead of trying to fix it later .

Well the first point to appreciate is that you won't stop a pc from being "noisy".  Multiple clocks, buses, processes and so on.  These don't stop.

 

None of the measures that mitigate this are on the analogue output.  They have no negative impact on SQ.

[manueljenkin]

13 hours ago, idiot_savant said:

@PeterSt - maybe it would be helpful if you could describe the interface used within your tests? i.e. how is the DAC connected and powered? I believe ( happy to be corrected on this ) that @March Audio is using a DAC that is connected via USB with an asynchronous style interface, and is powered separately from the PC.

 

If your tests were using, for example a synchronous interface, or USB in adaptive mode, or the DAC is somehow powered by the PC, this might give us some insight?

 

 

your friendly neighbourhood idiot

 

I have run these tools on usb asynchronous mode dacs, and I've had friends who ran pc optimization softwares on things like rme adi2, ifi micro bl (both of which have galvanic isolation afaik) and still heard differences.

[March Audio]

6 hours ago, Summit said:

 

Most manufactures are indeed trying to stop noise from getting both in and out of the DAC using filters and galvanic isolation. GI can be done with transformers or optocouplers, but it’s not perfect since there is always very small amount of capacitance between two sides of galvanic isolation which allowing very high frequencies to pass thru. Another (IMO better) way is to reduce noise on the source.

How would you know if you have reduced the noise on the source?

[manueljenkin]

On 6/10/2021 at 10:42 AM, March Audio said:

First test is to simply use an FFT to look at the noise and distortion in an unloaded situation.  

Fft is an averaging algorithm. And it's long been proven that human hearing works on time-frequency basis and not mere frequency domain approximations. So this misses out coverage and cannot be used for any kind of conclusion.

 

On 6/10/2021 at 10:42 AM, March Audio said:

As mentioned the above test will not see transient noises, for that we need to look at the waveform.

Sorry, real world ADC aren't perfect and sine sweeps again don't correlate with music signals for performance analysis.

[March Audio]

13 minutes ago, manueljenkin said:

I have run these tools on usb asynchronous mode dacs, and I've had friends who ran pc optimization softwares on things like rme adi2, ifi micro bl (both of which have galvanic isolation afaik) and still heard differences.

Which could be down to any number of reasons.

 

This is an objective thread.  "I hear it therefore it is" statements have no credibility here.

[March Audio]

5 minutes ago, manueljenkin said:

Fft is an averaging algorithm. And it's long been proven that human hearing works on time-frequency basis and not mere frequency domain approximations. So this misses out coverage and cannot be used for any kind of conclusion.

 

Sorry, real world ADC aren't perfect and sine sweeps again don't correlate with music signals for performance analysis.

You should have read the OP in more detail.

 

That's precisely why I used pkanes Deltawave to look at the time domain.  This also found no significant difference beyond random noise.

 

BTW FFTS are perfectly valid for looking for certain types of noise.

 

Please explain why using a sine signal would obscure noise problems generated by a pc?

[opus101]

6 hours ago, idiot_savant said:

Right, so where is this pollution on the measurements provided?

 

Common-mode noise is very unlikely to show up on a differential-mode measurement. Unless in the circuit is a mechanism for CM to DM conversion or the AP lacks decent CM rejection (it doesn't).

[Summit]

7 hours ago, March Audio said:

How would you know if you have reduced the noise on the source?

 

By measure the level of noise or blind test. Noise can easily be measured btw. It's the affect different form of low level noise have on different equipment and SQ that is not established. 

[Summit]

7 hours ago, March Audio said:

Well the first point to appreciate is that you won't stop a pc from being "noisy".  Multiple clocks, buses, processes and so on.  These don't stop.

 

None of the measures that mitigate this are on the analogue output.  They have no negative impact on SQ.

 

Please provade evidence to support your claim that it's impossible to "stop a PC from being "noisy".

 

Please provade evidence to support that source noise have no negative impact on SQ.

 

You state "PCs are noisy, that's a given.  It's the way they are." If that's the case (and I don't say it is) wouldn't it be better to use another type of source than a PC?

 

 

[Summit]

23 hours ago, March Audio said:

Can DACs be affected by whats happening up stream?  Yes.  Already discussed ground loops, noise currents etc.  However good dacs are not sensitive to this. So its a secondary effect, not that of identical data sounding different.

 

How does this statment correlate with your other posts and who will decide which DAC that is good?

[idiot_savant]

Right, so the point of this thread ( I believe )  is there are certain people who reckon you have to go through many, many hoops to get good playback from a PC. Typically those same people claim that even though you *need* to go through these hoops you’ll get better sound than from anything else. When pressed as to what these hoops solve, hands start getting waved, and typically some nebulous “noise” is being generated somewhere that you need to reduce. Being Objective type people, all this hand waving and nebulous claims makes us unhappy. 
So, we assume that the “noise”  is electrical rather than acoustic, and look for it influencing the output of a DAC on a non-tweaked system. And can’t find it.

 

And to those people who claim FFTs are poor for detecting noise, or “average it out” I’d advise them to do some research. FFTs are poor for glitch detection, but incredibly sensitive for just about everything else, but do need a bit of interpreting

 
PCs are noisy due to switching noise - every time something is turned on, or off, it generates a tiny amount of noise - the problem is a PC typically has millions if not billions of things all turning on and off at mind boggling speeds, some at the same time, some not. ( so your CPU and GPU will generally have independent clocks ). 
 

Nobody is trying to tell you which DAC is good, or even which DAC sounds the best, you have to draw your own conclusions 

 

 

your friendly neighbourhood idiot 

[PeterSt]

On 6/10/2021 at 7:12 AM, March Audio said:

DAC and measurement ADC is a RME ADI2 pro FSR BE. 

 

I have been thinking about the virtues of this, given by the fact that I could possibly do all myself just the same. So I only now checked and see what could be happening. Thus:

 

The DAC and the ADC is one and the same box, Yes ?

[March Audio]

2 hours ago, Summit said:

 

By measure the level of noise or blind test. Noise can easily be measured btw. It's the affect different form of low level noise have on different equipment and SQ that is not established. 

So you have measured this noise?

[March Audio]

2 hours ago, Summit said:

 

Please provade evidence to support your claim that it's impossible to "stop a PC from being "noisy".

 

Please provade evidence to support that source noise have no negative impact on SQ.

 

You state "PCs are noisy, that's a given.  It's the way they are." If that's the case (and I don't say it is) wouldn't it be better to use another type of source than a PC?

 

 

No evidence is needed.  The assertion by others  is that pc activity makes noise. 

 

Considering that PCs don't stop operating, the busses keep whizzing away, the disks keep being accessed, the memory keeps being refreshed and accessed, the CPU speed and load keeps going up and down and dozens of processes are still running.

 

Therefore the noise that people make claims about is still being generated.  However I can put a spectrum analyser to show you if you want.

 

You just need to look at the OP to see there is no negative affect with a decent dac.

[March Audio]

1 hour ago, PeterSt said:

 

I have been thinking about the virtues of this, given by the fact that I could possibly do all myself just the same. So I only now checked and see what could be happening. Thus:

 

The DAC and the ADC is one and the same box, Yes ?

On the first test yes, obviously not with the second test using a Gustard x16.

[PeterSt]

3 minutes ago, March Audio said:

obviously not with the second test using a Gustard x16.

 

Ah, OK.

I see that you planned some more testing with that DAC - if I see correctly, could you do it at  ~ -3dBFS ?

I think I saw you writing an in between the lines reason to have that way more low (something with noise IIRC), but wouldn't it be better to have all full scale ? at least that should give others a more normal reference (like where the noise line really is).

N.b.: I came to this because I did not trust the (low) noise line of the RME. But since this is with loop-back I now understand. Still it is not realistic within the normal playback environment. That is, *if* that matters at all for these tests.

 

OffTopic: How does the RME actually sound through your ears ?

[March Audio]

1 hour ago, idiot_savant said:

Right, so the point of this thread ( I believe )  is there are certain people who reckon you have to go through many, many hoops to get good playback from a PC. Typically those same people claim that even though you *need* to go through these hoops you’ll get better sound than from anything else. When pressed as to what these hoops solve, hands start getting waved, and typically some nebulous “noise” is being generated somewhere that you need to reduce. Being Objective type people, all this hand waving and nebulous claims makes us unhappy. 
So, we assume that the “noise”  is electrical rather than acoustic, and look for it influencing the output of a DAC on a non-tweaked system. And can’t find it.

 

And to those people who claim FFTs are poor for detecting noise, or “average it out” I’d advise them to do some research. FFTs are poor for glitch detection, but incredibly sensitive for just about everything else, but do need a bit of interpreting

 
PCs are noisy due to switching noise - every time something is turned on, or off, it generates a tiny amount of noise - the problem is a PC typically has millions if not billions of things all turning on and off at mind boggling speeds, some at the same time, some not. ( so your CPU and GPU will generally have independent clocks ). 
 

Nobody is trying to tell you which DAC is good, or even which DAC sounds the best, you have to draw your own conclusions 

 

 

your friendly neighbourhood idiot 

Indeed it's all hand waving about "noise", the possible causes, the possible mechanisms and precisely zero evidence to show this.

 

This is why I'm analysing dac outputs to see what's actually happening.

 

Regarding FFTs they are superb for digging into noise floors for seeing spurious tones, distortion etc.

 

However I also used planes deltawave software to look at the time domain at sample level.