Topping D10s noise via Mac mini

[Thomas_A]

@Thomas_A is this a full scale signal you’re looking at and if so what are the output voltages in those 3 cases?

Should have measured it but it was -3 dB in the REW output, not 0 dB

 

[BKDad]

One last question from me...

Have you tried this combination:

Mac Mini > USB Isolator > DAC > ADC > Battery powered MacBook.

I'm pretty sure you have, but I'm not sure if that's what you're showing in your most recent plots.

OK - I lied. One more. In a situation as above ^^^, have you tried powering the DAC side of the isolator with a battery?

I think it's hard to find a way to correlate the harmonics of the AC mains with anything that is in the data stream (the bits part) going to the DAC. The AC mains doesn't magically have a component associated with it that translates to the bit pattern needed to give those responses in the converted bit stream. That suggests that this is something getting into some analog part of the system. It could be that the ADC is converting common mode signals into differential mode at its analog input. The CMRR at the input of an E1DA is about 60-65 dB.

 

[antcollinet]

One last question from me...

Have you tried this combination:

Mac Mini > USB Isolator > DAC > ADC > Battery powered MacBook.

I'm pretty sure you have, but I'm not sure if that's what you're showing in your most recent plots.

OK - I lied. One more. In a situation as above ^^^, have you tried powering the DAC side of the isolator with a battery?

I think it's hard to find a way to correlate the harmonics of the AC mains with anything that is in the data stream (the bits part) going to the DAC. The AC mains doesn't magically have a component associated with it that translates to the bit pattern needed to give those responses in the converted bit stream. That suggests that this is something getting into some analog part of the system. It could be that the ADC is converting common mode signals into differential mode at its analog input. The CMRR at the input of an E1DA is about 60-65 dB.

Its going to be much worse than this since the connection is not balanced.

 

[Sokel]

Like I say - connect your PC (as noisy as you like) to your audio system with Toslink - and it will be as though it is not there - except for the pure sweet music.

The problem is that Toslink does not meet certain preferences* like extreme sample rates and DSD over DSD64.Part from DAC's evolution is that sample rates so excluding them is not a solution

* no better fidelity or other benefit is inclined,just preference which is everyones right.

 

[BKDad]

Its going to be much worse than this since the connection is not balanced.

Yeah, maybe. If the ADC is truly floating with respect to any "ground" (horrible term) or inter-chassis connections, whether it's balanced or not doesn't matter. I mean, balanced with respect to what?

However...

Most all balanced inputs are wildly susceptible to the effects of impedance imbalance between the two polarities of the incoming signal with regard to CMRR. With balanced connections, often - not always - the two polarities are close in impedance relative to any common connection. But, with unbalanced outputs going into a balanced input, that's usually not so. I think the output impedance of the D10s is around 20 Ohms based on reports I've read. (Haven't measured it myself.) So, the source impedance for the positive polarity is 20 Ohms relative to pin 1 of the XLR connection you have and close to ZERO Ohms for the negative polarity.

There's ways around that, some exotic and pretty great, some not as much.

If it was me, and it certainly isn't, I'd try two things.

1. Disconnect pin 1 at the XLR connector. See what happens. This floats the ADC.

2. Try adding a 20 Ohm resistor between the ground connection at the D10s RCA and the wires going to pin 3 of the XLR connector, as shown in the drawing your referenced. (The blue wire.) This balances the impedances sourced to each balanced input polarity at the ADC. See what happens.

Note that in neither case are you changing what's being to the DAC. You can try one or the other. They each have pros and cons associated with them.

I'm not saying that either of these are perfect solutions, but they will give you an idea if the problem is common mode to differential mode signal conversion at the ADC.

This, of course, raises an interesting question relative to your actual music listening system. What kind of common mode rejection does your preamp (or whatever is being fed by the DAC) have? You could measure that at the output of that device the same way you're measuring the DAC, at least when you get that part working properly. Same for the power amplifier. I never see measurements of that sort of thing at the system level.

 

[gvl]

Should have measured it but it was -3 dB in the REW output, not 0 dB

I mean in Volts, you need a voltmeter to measure it.

 

[Thomas_A]

I mean in Volts, you need a voltmeter to measure it.

Yes. I should have measured it, I can check tomorrow.

 

[antcollinet]

The problem is that Toslink does not meet certain preferences* like extreme sample rates and DSD over DSD64.Part from DAC's evolution is that sample rates so excluding them is not a solution

* no better fidelity or other benefit is inclined,just preference which is everyones right.

My preference is for a perfectly noise free connection - which does bring better fidelity.

And education is a really good way to eliminate preferences for things that can only sound identical. Knowledge leads to understanding. Understanding leads to peace.

 

[antcollinet]

Yeah, maybe. If the ADC is truly floating with respect to any "ground" (horrible term) or inter-chassis connections, whether it's balanced or not doesn't matter. I mean, balanced with respect to what?

Balanced with respect to impedance.

Most all balanced inputs are wildly susceptible to the effects of impedance imbalance between the two polarities of the incoming signal with regard to CMRR. With balanced connections, often - not always - the two polarities are close in impedance relative to any common connection.

Exactly - in fact a balanced connection is not balanced unless it is impedance balanced - that is what balanced means. If the impdedance is not balanced, then it is just differential. Without the impedance balancing you don't get the CMRR.

My view is that all the noise problems here are due to common mode differential between DAC and ADC (wherever that is coming from - probably capacitive coupling between PSU or other electronics and chassis/device ground) and the unbalanced connection being unable to reject this.

Your 2 suggestions are exactly what I would also suggest. Though I'd adapt the 20ohm to whatever the rated output impedance of the D10S is (if you haven't already done this.)

I would add a third suggestion. (which I've already given). Ground the DAC, Ground the ADC, then try grounding them both (or ground them together) - using a low impedance (flat copper braid) connection.

 

[Sokel]

My preference is for a perfectly noise free connection - which does bring better fidelity.

And education is a really good way to eliminate preferences for things that can only sound identical. Knowledge leads to understanding. Understanding leads to peace.

Agreed but the choice of which way to connect shouldn't be a users problem but one engineers should face according to the use case.
And since they offer such a connection (which they base a good chunk of their advertising and also takes a good chunk of the cost of the device) they should make sure to make it immune against ANY sensible use case.

That's how I see it as a user and if that means they have to implement isolators,filters,nuclear technology or magic that's their problem,not mine.
There's a joke in my country about stuff that need certain conditions to operate as advertised

 

[BKDad]

My preference is for a perfectly noise free connection - which does bring better fidelity.

And education is a really good way to eliminate preferences for things that can only sound identical. Knowledge leads to understanding. Understanding leads to peace.

Toslink certainly is not electrically conductive, which absolutely eliminates a common. mode current path. All to the good.

Whether it's noise free and without other compromises, well, I guess it's up to the user to decide if those compromises are limiting in some way.

 

[BKDad]

Agreed but the choice of which way to connect shouldn't be a users problem but one engineers should face according to the use case.
And since they offer such a connection (which they base a good chunk of their advertising and also takes a good chunk of the cost of the device) they should make sure to make it immune against ANY sensible use case.

That's how I see it as a user and if that means they have to implement isolators,filters,nuclear technology or magic that's their problem,not mine.
There's a joke in my country about stuff that need certain conditions to operate as advertised

While I agree with you in a broad sense, the unfortunate parts are that these additional performance enhancements don't come for free - which nobody likes - and it seems like there's zero agreement about their usefulness. Measurements or not.

 

[antcollinet]

Agreed but the choice of which way to connect shouldn't be a users problem but one engineers should face according to the use case.
And since they offer such a connection (which they base a good chunk of their advertising and also takes a good chunk of the cost of the device) they should make sure to make it immune against ANY sensible use case.

That's how I see it as a user and if that means they have to implement isolators,filters,nuclear technology or magic that's their problem,not mine.
There's a joke in my country about stuff that need certain conditions to operate as advertised

Sure - they can do that. But then people wouldn't pay the resulting price, and those companies risk failure.

The people would buy the cheaper kit that doesn't have all that built in isolation. And they mostly wouldn't care because almost no-one is measuring inaudible levels of noise, so they simply wouldn't know they were not getting perfection. In fact in most cases they are getting (audible) perfection.

Those who set up systems with ground loops and unbalanced connections** with the resulting audible noise will complain that it doesn't meet spec. And round and round we go.


**actually universal use of balanced connections would be the best way to go. It doesn't have to cost that much and would eliminate almost all ground noise problems. We should just outlaw RCA.

 

[antcollinet]

Toslink certainly is not electrically conductive, which absolutely eliminates a common. mode current path. All to the good.

Whether it's noise free and without other compromises, well, I guess it's up to the user to decide if those compromises are limiting in some way.

Back to education again.

 

[BKDad]

Back to education again.

Well, my education and experience is that Toslink has noise and optical reflection challenges. The associated electronics also add various instabilities and jitter. It doesn't have to be that way, of course, but Toslink was designed to be cheap with cost being the driving design considerations. I wish that wasn't so, but it is. For reasons you stated yourself.

In addition, the whole SPDIF idea where the data is buried within the clock is a compromise. The system never was intended to be high quality - just a test port.

 

[BKDad]

**actually universal use of balanced connections would be the best way to go. It doesn't have to cost that much and would eliminate almost all ground noise problems. We should just outlaw RCA.

I'd vote for that. Not a perfect solution, but a really great start.

It wouldn't have to be XLR, either. TRS and the 3.5 mm equivalent would be pretty convenient.

 

[antcollinet]

Well, my education and experience is that Toslink has noise and optical reflection challenges. The associated electronics also add various instabilities and jitter. It doesn't have to be that way, of course, but Toslink was designed to be cheap with cost being the driving design considerations. I wish that wasn't so, but it is. For reasons you stated yourself.

In addition, the whole SPDIF idea where the data is buried within the clock is a compromise. The system never was intended to be high quality - just a test port.

Noise? - as an interconnect being optical it can transfer zero noise. None. Absolutely none. If the transmitting device is the noisiest PC on the planet, and it is only connected to the audio system via Toslink, none of that noise will arrive at the other end.

Can the receiving electronics generate noise - sure but that is exactly like all the other circuits in the receiving device, and I'd suggest that the optical detector is amongst the least noisy.

Optical reflections, (and attenuation, and non linear optical paths spreading the pules - you missed those), they can only interfere with the ability of the reciever to correctly detect the bits, resulting in bit errors. But if you get those you'll know about it in the form of very very audible pops/clicks and crackles**. If you are not getting those, then you are not suffering any ill effects from those phenomena.

Finally jitter (generally a result of those reflections/nonlinear paths etc mentioned up ^ there) - the last refuge of the audiophile boogyman. Jitter is only an issue at the point of digital to analogue conversion. But the process used by a DAC to extract the clock from the digital stream, and sync the local DAC clock to that (PLL and/or ASRC) with associated buffering reject any jitter in the incoming stream to well below audible levels (at least in decent DACs from the last decade or two)

At the end of the day - Toslink is as close to a perfect interconnect as you need for two channel audio applications - regardless of what was intended.


(** I have a sound blaster Omni 5.1 with an optical output. It seems to be unable to transmit reliably at 24/96. It works for a few minutes when the crackling starts - which then builds until it completely drowns out the music. My solution when I need to use that interface is simply to set the rate to 24/48. More than enough for audible perfection.)

 

[BKDad]

Noise? - as an interconnect being optical it can transfer zero noise. None. Absolutely none. If the transmitting device is the noisiest PC on the planet, and it is only connected to the audio system via Toslink, none of that noise will arrive at the other end.

Can the receiving electronics generate noise - sure but that is exactly like all the other circuits in the receiving device, and I'd suggest that the optical detector is amongst the least noisy.

Optical reflections, (and attenuation, and non linear optical paths spreading the pules - you missed those), they can only interfere with the ability of the reciever to correctly detect the bits, resulting in bit errors. But if you get those you'll know about it in the form of very very audible pops/clicks and crackles**. If you are not getting those, then you are not suffering any ill effects from those phenomena.

Finally jitter (generally a result of those reflections/nonlinear paths etc mentioned up ^ there) - the last refuge of the audiophile boogyman. Jitter is only an issue at the point of digital to analogue conversion. But the process used by a DAC to extract the clock from the digital stream, and sync the local DAC clock to that (PLL and/or ASRC) with associated buffering reject any jitter in the incoming stream to well below audible levels (at least in decent DACs from the last decade or two)

At the end of the day - Toslink is as close to a perfect interconnect as you need for two channel audio applications - regardless of what was intended.


(** I have a sound blaster Omni 5.1 with an optical output. It seems to be unable to transmit reliably at 24/96. It works for a few minutes when the crackling starts - which then builds until it completely drowns out the music. My solution when I need to use that interface is simply to set the rate to 24/48. More than enough for audible perfection.)

I disagree with much of that, both in theory and in practice. Much of my disagreement is based on decades of designing and working with fiber optic transmission systems in telecommunications.

But, this audio stuff is a hobby for me. Whatever works for you gets my hearty approval. Not that you need my approval, of course, but I'm still supportive.

 

[Thomas_A]

I mean in Volts, you need a voltmeter to measure it.

Measured today; 1.97 Vrms from the D10s.

 

[Thomas_A]

One last question from me...

Have you tried this combination:

Mac Mini > USB Isolator > DAC > ADC > Battery powered MacBook.

I'm pretty sure you have, but I'm not sure if that's what you're showing in your most recent plots.

OK - I lied. One more. In a situation as above ^^^, have you tried powering the DAC side of the isolator with a battery?

I think it's hard to find a way to correlate the harmonics of the AC mains with anything that is in the data stream (the bits part) going to the DAC. The AC mains doesn't magically have a component associated with it that translates to the bit pattern needed to give those responses in the converted bit stream. That suggests that this is something getting into some analog part of the system. It could be that the ADC is converting common mode signals into differential mode at its analog input. The CMRR at the input of an E1DA is about 60-65 dB.

I have tried the first one. The isolator with a battery - no.

Agreed that there is something strange going on.