Help me sort out USB noise and a ground loop in a rather convoluted setup

[snowsurfer]

How can optical be noisy?

I have absolutely no idea, haven't even looked into it. Tried a couple of different cables from my PC's motherboard to my SMSL SU-8 and with both it was terrible, with very noticeable noise of clearly digital nature. I have no answers though. I also don't really care since I got USB working fine.

 

[Julf]

I have absolutely no idea, haven't even looked into it. Tried a couple of different cables from my PC's motherboard to my SMSL SU-8 and with both it was terrible, with very noticeable noise of clearly digital nature. I have no answers though. I also don't really care since I got USB working fine.

All I can tell you is that an optical connection provides total, perfect galvanic (electrical) isolation. There is no way it would conduct (electrical) noise.

 

[Eirikur]

All I can tell you is that an optical connection provides total, perfect galvanic (electrical) isolation. There is no way it would conduct (electrical) noise.

True... unless that noise is injected at the source (electric->optical).

By chance I experimented today with the S/PDIF coax input of an old Gigabyte main board I have, and found this to be extremely noisy! At first it seemed that the perfect sine wave was received perfectly, but a null-test against a locally generated sine showed significant noise. Further testing showed this noise to present even without any connection to the pins to give me a S/N ratio of 90.3dB.

 

[snowsurfer]

True... unless that noise is injected at the source (electric->optical).

By chance I experimented today with the S/PDIF coax input of an old Gigabyte main board I have, and found this to be extremely noisy! At first it seemed that the perfect sine wave was received perfectly, but a null-test against a locally generated sine showed significant noise. Further testing showed this noise to present even without any connection to the pins to give me a S/N ratio of about 88dB.

I have a nagging feeling it is something inside my PC, probably my motherboard, that is causing the noise, but as I said I have a fine USB connection so I'm good, but I am curious and when I finally get my mitts on a Ryzen 3950X this winter and build my new system I will give it a quick test to see whether it is still happening.

 

[Julf]

True... unless that noise is injected at the source (electric->optical).

By modifying the audio data?

By chance I experimented today with the S/PDIF coax input of an old Gigabyte main board I have, and found this to be extremely noisy! At first it seemed that the perfect sine wave was received perfectly, but a null-test against a locally generated sine showed significant noise. Further testing showed this noise to present even without any connection to the pins to give me a S/N ratio of about 88dB.

But that is noise on the digital connection, right? Not actual audio.

 

[Julf]

I have a nagging feeling it is something inside my PC, probably my motherboard, that is causing the noise.

How would that noise get to your DAC?

 

[snowsurfer]

How would that noise get to your DAC?

As I said, I have no answers, I do not know, what I know is that it is completely unlistenable. These are just my uneducated guesses.

 

[Julf]

As I said, I have no answers, I do not know, what I know is that it is completely unlistenable. These are just my uneducated guesses.

And what I am saying is that an optical connection does not conduct electricity. Thus if your system is completely unlistenable, the problem is not noise in your source, but something seriously wrong in your setup. Changing hardware is not the solution.

 

[snowsurfer]

And what I am saying is that an optical connection does not conduct electricity. Thus if your system is completely unlistenable, the problem is not noise in your source, but something seriously wrong in your setup. Changing hardware is not the solution.

It only happens on the optical connection though. And I am not changing anything due to this, it's just that the time for upgrading has come. I'm not even looking for a solution, to be honest. I only noticed because I gave optical a shot (thinking that it was impossible that it would be this bad, yes) while I was trying to solve USB ground loop problems.

 

[Julf]

It only happens on the optical connection though. And I am not changing anything due to this, it's just that the time for upgrading has come. I'm not even looking for a solution, to be honest. I only noticed because I gave optical a shot (thinking that it was impossible that it would be this bad, yes) while I was trying to solve USB ground loop problems.

Is the noise hiss/hum, or crackles/dropouts? In the first case, your DAC handles optical badly (can the inputs be adjusted separately?), in the latter case you have a marginal optical cable that causes dropouts/data loss.

 

[snowsurfer]

Is the noise hiss/hum, or crackles/dropouts? In the first case, your DAC handles optical badly (can the inputs be adjusted separately?), in the latter case you have a marginal optical cable that causes dropouts/data loss.

It is definitely crackles, not hiss/hum. The crackles are not particularly loud but they are quite constant and extremely distracting. The cables I tested are very old and not in great cosmetic condition, but I thought optical cables either worked or not? They were stored in boxes all bunched up and beneath a lot of other stuff though. I can order a cheap cable from Amazon with my next order just to give it a shot. The DAC is an SMSL SU-8, I don't think inputs can be adjusted separately.

 

[Eirikur]

By modifying the audio data?

Yes, it is filled with spikes of 90.3dB, they are actually very accurately exactly the same.

But that is noise on the digital connection, right? Not actual audio.

The noise is always present, and it does screw up actual audio I captured.
My 1kHz sine wave was littered with this noise.

Some pictures, using 192kHz as input sampling rate (I tried 44.1, 28, 96 and 192 -> all the same)

Top is the left channel of the SPDIF input, bottom is the same with 90.3dB gain

Note that I captured this without an actual input source, but as said, it is exactly the same when inputting an SPDIF stream with null values. Zoomed in it looks like this:

No need to debug my system as I don't use it for audio, but I think it illustrates the level of noise that may be present on the PC side on the direct digital interface - it almost seems deliberate.

 

[Julf]

It is definitely crackles, not hiss/hum. The crackles are not particularly loud but they are quite constant and extremely distacting. The cables I tested are very old and not in great cosmetic condition, but I thought optical cables either worked or not? They were stored in boxes all bunched up and beneath a lot of other stuff though. I can order a cheap cable from Amazon with my next order just to give it a shot. The DAC is an SMSL SU-8, I don't think inputs can be adjusted separately.

Optical cables can work badly - they can cause data errors and data loss. The result is not a subtle deterioration of sound quality, but clearly audible crackling and dropouts.

 

[Julf]

Yes, it is filled with spikes of 90.3dB, they are actually very accurately exactly the same.

Actually changing "0"s and "1"s?

The noise is always present, and it does screw up actual audio I captured.

So those captures are from the audio output, not the S/PDIF input?

 

[Eirikur]

Actually changing "0"s and "1"s?
So those captures are from the audio output, not the S/PDIF input?

Yes indeed, actually changing 0 to 1 here and there, but the mechanism is unclear to me.
As you can see there are no volume spikes, and all noise is suspiciously exactly the same amplitude...

Input only, I have no way to reliably capture the output!

In any case, measuring this on the input gives me little hope for the output, whether coax or optical.
This may be relevant to the problem of to OP: it is clearly possible that the PC screws up the digital signal of the internal sound card. I would not expect this to happen over the USB interface.

 

[Julf]

Yes indeed, actually changing 0 to 1 here and there.

How did you determine that?

 

[Eirikur]

How did you determine that?

See the graphs. The zero points are actually zero.
And note that it is probably not in the signal itself, as I see the same result with or without input signal!

 

[Julf]

See the graphs. The zero points are actually zero.
And note that it is probably not in the signal itself, as I see the same result with or without input signal!

Right, so it is not changing the audio data - what you are seeing is analog noise, right?

 

[Eirikur]

Right, so it is not changing the audio data - what you are seeing is analog noise, right?

Really don't know but it shouldn't be: S/PDIF is "TTL" voltages captured and turned into bits - except for the capturing process itself (jitter and missed bits) there should not be any analog processing.

I'm baffled and really caught out by the result.

 

[Julf]

Really don't know but it shouldn't be: S/PDIF is "TTL" voltages captured and turned into bits - except for the capturing process itself (jitter and missed bits) there should not be any analog processing.

Indeed. I am baffled too. Just to be clear - what your graph shows is the audio as received by the S/PDIF input, with input data that is totally silent (or even with no input signal)?