Power Supply Noise measurement

[PeterSt]

Here are two screenshots of the very low power supply noise as how it can be achieved today with the NOS1 (yours too). First screenshot shows the positive rail, second one shows the negative rail.

Watch out : the red trace is the normal system output noise (measured at the output terminals) and the white trace is the supply noise (in this case to the DAC board).

The supply noise (thus white trace) was measured under normal load (of in this case the DAC board).

For anomalies seen, note that this comes from a test DAC full with measurement wires (antennas).

For those really interested, mind the negative rail which is as good as the positive rail. This is not so easy to achieve.

Positive rail

Negative rail


Separate supplies are provided to a. the DAC board (pos/neg), b. the gain stage (pos/neg), c. the clock section (pos) and d. the USB interface (pos).

Peter

[acg]

Hi Peter,

Are you able to give an indication on the improvement in noise that you have achieved with the new power supplies?

Cheers,

Anthony

PS:  Great to see this coming out!!

[PeterSt]

Hey Anthony,

PS:  Great to see this coming out!!

Haha, yes, you know I was working on that (with you working in similar areas). But yes, after what ? 3 or 4 months on this little subject only and two versions of it, it worked out completely as expected;

I honestly don't recall what the previous measurements were (never took screenshots of that either) but depending on where it was measured (USB, clocks themselves, DAC sestion postitive vs. negative) I recall something like 10s of mV (USB could have been over 100mv) to 3mV or so. Don't pin me down on this.
(I could measure an "old" NOS1 though)
This now all is under 2uV up to a theoretical 1.5uV RMS. "Theoretical" because the measurement wires (antenneas) catch 50Hz (as the mains is over here) which make it wobble somewhat (you can see that in the left hand side of the screenshots). Not sure how to measre the supply noise without the measurement wires. (all is on PCBs).
Also see below snapshot. White trace again is the supply noise ves red the normal output. You can see the "wobbling" which without measurement wires is not there. Peak to peak of the white trace is around 7uV.

Btw, I am not sure whether you meant this kind of improvement ?

Regards,
Peter

[acg]

Hey Anthony,

PS:  Great to see this coming out!!

Haha, yes, you know I was working on that (with you working in similar areas). But yes, after what ? 3 or 4 months on this little subject only and two versions of it, it worked out completely as expected;

Yes I did know what you were up to (and I have a good idea of what you have yet to come!) and I must say Peter that I still feel privileged that you kept me in the loop during the processes of throwing ideas around, designing the upgrades and the even the initial results.  Here I am, a guy from the other side of the world that you have never met but we bounced ideas around, I learnt a hell of a lot, and you placed a lot of trust in me.  Thanks for that...it was one of the things that I will remember for a long, long time.

Hey Anthony,

PS:  Great to see this coming out!!

I honestly don't recall what the previous measurements were (never took screenshots of that either) but depending on where it was measured (USB, clocks themselves, DAC sestion postitive vs. negative) I recall something like 10s of mV (USB could have been over 100mv) to 3mV or so. Don't pin me down on this.
(I could measure an "old" NOS1 though)
This now all is under 2uV up to a theoretical 1.5uV RMS. "Theoretical" because the measurement wires (antenneas) catch 50Hz (as the mains is over here) which make it wobble somewhat (you can see that in the left hand side of the screenshots). Not sure how to measre the supply noise without the measurement wires. (all is on PCBs).
Also see below snapshot. White trace again is the supply noise ves red the normal output. You can see the "wobbling" which without measurement wires is not there. Peak to peak of the white trace is around 7uV.

Btw, I am not sure whether you meant this kind of improvement ?

Regards,
Peter

The old "the act of measuring changes the results" conundrum.  I don't have anything to suggest about getting around that, but it is good that you are able to identify it.

My (ambiguous) question about the kind of improvement was basically about comparing say the -155dBFS at the output with the new power supplies and say -1??dBFS for the currently available version of the NOS1.  At these sorts of numbers I would imagine that the noise improvement at the outputs would be barely audible (if at all) but of course things are never so simple because you have separately powered items within the NOS1 that we're not individually powered before eg. clocks, gain stage.

BTW:  when can we get our hands on this???

[acg]

The other thing that has probably happened with the use of cutting edge voltage regulators is that the total bandwidth of the noise that you have under control is much greater now as well...I notice that the two graphs from the first post go out to 90MHz with those excellent results.

[PeterSt]

I notice that the two graphs from the first post go out to 90MHz with those excellent results.

Haha, maybe, but then I need to hook up another analyzer first. So, 96 KHz this shows (but there is no reason for ultra high frequency spuria unless sourced from my own house (like you see my ever and old PWM heating pump close to 18KHz).

[PeterSt]

Anthony,

Ok, first off, YES, all is about how "we" can spar which may challenge personal me. So honestly, while a lot of the best guys in this forum are always and ever busy with tweaking the NOS1('s interface) for the better at least exactly that challenged me to keep on thinking and thinking how to do it. Well, not known to all (but to those same fine people Yes) is that ever back I indeed already thought and thought and that e.g. that Adnaco interface already was part of my equation. But it couldn't work for theories (in my mind) as the people found out later. Same (and/or similar) with transformer coupling of which a few things were physically tried.

While the above seems quite unrelated it is not as you will see quite soon.

Wen you came about things went into a different direction or at least were approached from a different angle; The Linear Power Supply for the PC. Btw it is worth while noticing that you (Anthony) are (or were) not the only one working on that. I myself was to some smaller extend, Juan was to a bigger extend, and some other guy (not much posting lately) was to a real life extend. But as you all know I sort of "cancelled" that, knowing what I was really working on. But ... with possibility to fail.

More back to the merits of your post, yes we nicely exchanged some data about how to make that LPSU work with the lowest noise possible and behind the scenes more (Aussies) work was being performed. Say all in parallel and with the best possible solution in mind (and with the best minds to begin with).

In the very end, and this is a kind of sad thing, when I would succeed on my little secret project, possibly Anthony's work (and of others) would be without reason. This turned out to be so ...
But first this :

My (ambiguous) question about the kind of improvement was basically about comparing say the -155dBFS at the output with the new power supplies and say -1??dBFS for the currently available version of the NOS1.  At these sorts of numbers I would imagine that the noise improvement at the outputs would be barely audible (if at all)

It depends. So to answer this question : Normally I don't think this can be audible. And, do notice that I did not really see an improvement on the output of the DAC regarding this noise, or 1dB better is to be regarded an improvement (so 1dB less noise sure is visible). But this is how or why I also showed the system output noise, which, say, "strangely" is worse than the supply noise. Now notice that this just as well keeps on being so when looked at it the other way around : former NOS1 showed higher supply noise than system output noise.
It just isn't about that ...

In between the lines : We can well say that *because* it isn't about that, the supply noise was easily allowed to be higher (than it is now) by me. But mind you, this is from 4-5 years ago and besides that this was not about the USB interface.

USB interface ? aha ...

It is one huge complex thing, and with this as a base all comes together for solutions. Before we hop over to that and the real merit ...

BTW:  when can we get our hands on this???

Today. And with the notice in advance that this is (or has been) one of the major challenges and how to get all done within the boundaries of the existing design.
Ready ?

Here : Jitter.

Peter

[acg]

I notice that the two graphs from the first post go out to 90MHz with those excellent results.

Haha, maybe, but then I need to hook up another analyzer first. So, 96 KHz this shows (but there is no reason for ultra high frequency spuria unless sourced from my own house (like you see my ever and old PWM heating pump close to 18KHz).

Sorry, thought a "K" and typed an "M".

[hochopeper]

Peter,

What is your FS set to in these measurements? (might be more meaningful for comparisons in say dBV)

What happens to the noise floor when you put a test tone say 1kHz through rather than silence?


Cheers,
Chris

[PeterSt]

Hi Chris,

I always measure in dBFS; I could easily look if I would be on location which I am not at this moment. But easy enough : full scale is just that for signal (so -0dBFS) and what you see is relative to that. -0dBFS in my case is 2.25VRMS.

Let me know if you really still need to know the dBV (or any unit for that matter).

What happens to the noise floor when you put a test tone say 1kHz through rather than silence?

Totally nothing zilch (signal at -0dBFS). I checked that as the first. Nothing "moves" anywhere. Uhm, for the supply noise (white trace) that is. Checked 18KHz as well because that could be better visible above the field because of the more "wild" area at the left hand side (1KHz). But still nothing.

Regards,
Peter

[hochopeper]

Hi Peter,

No need to worry about dBV, you've given the important bit of missing info, 2.25V = 0dBFS was all I really needed to convert the above numbers to absolute voltages. I was more interested in what happens to red plot when signal plays, wouldn't expect white to change. Was trying to compare this with the measurements earlier this year that you showed in the NOS1-USB thread.

Chris

[PeterSt]

Chris, Ah. I think I told (somewhere) that nothing much changed to that except for a small 1dB difference for the better. This is when idle.

When playing 16 bit material the noise or that (red trace) rises with 20dB which is inherent to the upsampling and say "the quantization noise". This goes along with the level of the output signal and when attenuated 20dBFS the output noise is at the idle level (see pictures).
Such thing does not happen at all with 24 bit material (so also not 44100). Then output noise then stays at the idle level also at -0dBFS.

Additionally - and just saying - the noise occurring with 16 bit material can digitally be filtered out (in advance of the DAC, thus in-PC). This of course tells that indeed it is not related to the electronics - also not the D/A chips or anything in that area; it's just the 16 bits (actually also without upsampling but the main point is that the upsampling which also expands to 24 bits can not cover for that phenomenon).

Peter