Investigating a full linear PSU for the XXHE PC and NOS1

[acg]

Glad you have it working Akremer!!  I am also glad that you notice a sound improvement.  Could you please describe the changes?  The picoPSU won't be causing the restarts unless it is faulty...I have used them before to good effect with no problems.

Peter, this is a reasonably common setup:  a relatively large, low cost LPS feeding a picoPSU into the motherboard and then smaller LPS's for the peripherals such as the SSD/HDD etc..  Depending on the version of the picoPSU, the 12V may be passed through to the mobo unscathed but the other rails are all derived via switching regulators, which sort of defeats the purpose to a certain extent, but it still is should be a step forward.

Anthony

[Arjan]

Hi Anthony,
About the SQ: overall more rest. We were able to play the music louder then before.
The soundstage is deeper! And more details can be heard.
The notes seems to be longer, and the voices are really nice. The SQ is really a step forward.

Before I used the same bequiet atx psu as used in the xxhighend pc.

I also switched (later on) from the Paul Pang usb to the Silverstone (isolated) usb. I do not have issues with Ssss's. They are there but really depending on the recordings and voice. Maybe later I will make the changes to the Paul Pang and try that card again. but for now I like the Silverstone.

I am also thinking about buying the Paul Pang picopsu, to see the improvement on a standard picopsu.

regards,
Arjan

[manisandher]

I think it's probably worth noting that what Peter seems to have done is eliminate the influence of the USB interface. The NOS1 will still be heavily influenced by both platform and airborne vibrations. Maybe electrical noise via the AC mains will still play a part? If so, isolating the PC's SMPS from the NOS1's AC power supply may still be beneficial.

Mani.

[acg]

I think it's probably worth noting that what Peter seems to have done is eliminate the influence of the USB interface. The NOS1 will still be heavily influenced by both platform and airborne vibrations. Maybe electrical noise via the AC mains will still play a part? If so, isolating the PC's SMPS from the NOS1's AC power supply may still be beneficial.

Mani.

I have had this thought about the LPS also Mani.

[PeterSt]

Mani,

Maybe electrical noise via the AC mains will still play a part? If so, isolating the PC's SMPS from the NOS1's AC power supply may still be beneficial.

It well can be but at this moment I don't know how to prove that.

If you look at the excursion just before 18KHz - that's my dreaded PWM heating pump (not heat pump, which is something else). To the left of that the first next peak at almost the same level is the freezer. Both are on a totally separate mains ring (incl. different earth pin). Nothing, really nothing can take that out because it comes through air and sneaks into all, including the other mains ring(s) with different earth pins. You see it here at -140dB because it was "tweaked" that and this is all about the DAC's grounding (say guiding away).

If we look at e.g. 65Khz then I don't even know what that is. Same for the other spuria on the right hand side. It could be PCs, it could be anything. And then to think this sees 96Khz only, while beyond that there's quite some more, and especialy the very high frequencies may show nasty things. But next problem : how to measure *that* because no high bandwidth analyzer will show noise that deep. Merely think 120dB down max and if you envision that to have happened in the picture above, well, all would be completely spuria free !
Ehm ...

Next is that we don't even know how for example the still present stuff at -125dB as the worst (the 65KHz) will be able to influence for example the oscillators. Or how it may be able to the D/A conversion itself with its utmost small currents (as explained elsewhere).

As you can see I try to approach it from the measurement level and not from the audible level; I regard this latter too difficult anyway; it must measure good first, and now there's thus a small conflict.
Also, all already starts with good measurement. Like here you see the lower frequency sh*t which is from plain lousy measurement or the maybe inability to lead away the probe lead sufficiently from the power supply.; all it tells is that it won't be worse, but it doesn't tell a thing about how much better it can be. Of course, when behind the wheel this is fairly easy to check (move the probe lead and see if something changes) but now notice that each measurement has a purpose and really requires preparation and approach. Nice example of this is the 1GHz analog measurement which superbly shows how I can't even press the print screen button on that scope because I create a field myself and the beam moves already by that (you can see that by the misalignment of the Delta-T cursors compared to the beam it should embrace (and which it did).

This was all a fairly long talk about "all is moot anyway" because too much out of out control.
And of course, when we exchange our freezer motors with one that doesn't spit like all undoubtedly do we are under control more. We could end up in a bunker in the middle of nowhere without fridges and all, but we would have warm beers in there and no dishwasher to clean the glasses. This latter is OK because no beer glass is to be cleaned in a dishwasher, but still.
But only then we can be sort of guaranteed that stopping a PC from spitting will help, just because it will be the last bit of it.
But will it influence anywhere ? Maybe we go too far ?

You don't know it much, but in the not-isolated situation we would see a fairly high spike at 8KHz of USB noise. I forgot by now, but towards -100dB in above picture. From something like that we can at least reason out that it will influence the oscillators to begin with. My pump I could not see, which only happens now because the noise is more down. So now that will influence ? Well, don't think that without checking, because behind what you see here is another ultra low noise supply which is thus fed first by what you see (white trace) and of which I can expect that nothing any more shows but I will only know when I measure *that*. This will count for any PC spitting too, and everything else.

But yes, that is what I can do more at this moment; measure the supply noise to the oscillator(s); I'm fairly confident that nothing will change to my pump's visibility (because that is always a matter of guiding out and I don't know how to do that better) but possibly we can learn something from the remainder. And then still not easy, because who says if I put up the same measurement now, it shows the same; can be a matter of the neighbors just finished with the dishwasher.

Peter

[PeterSt]

But yes, that is what I can do more at this moment; measure the supply noise to the oscillator(s); I'm fairly confident that nothing will change to my pump's visibility (because that is always a matter of guiding out and I don't know how to do that better) but possibly we can learn something from the remainder.

So, did that. But funny ... nothing changed much apart from just that pump.

This is all what's left of the pump (17700 HZ) and in the left the freezer.

Here the clock supply noise (this is 3.3V positive) :

And might someone recognize something in these details (possibly regarding SPSU's) :

So that is all I have for unrecognized anomalies.

Peter

[acg]

Peter,

Can you please confirm just what you did to remove the pump noise from your noise plot?

Cheers,

Anthony

[acg]

And might someone recognize something in these details (possibly regarding SPSU's) :

Roughly 50Hz harmonics?  You guys in Europe have 50Hz mains don't you?  SMPS for various appliances.

[PeterSt]

Roughly 50Hz harmonics?

Anthony, hmm ... I did not even see that. But is there any logic in why only there ? So all is quiet, expect for that little area.

And of course in the lower region. Here :

Beyond that it stops. Except for that small area ...


Peter

[PeterSt]

Can you please confirm just what you did to remove the pump noise from your noise plot?

Funny man ...

Nothing. Or nothing that I know of. But remember, this is a supply fed with the other supply from the screen shot of a couple of days back (few posts up now).

What I expected is that the 50Hz would be better rejected (say another time "filtered"), but that did not happen. So I guess this is too much transformer (AC) influence or pickup indeed. This time I took care (to some extend) that no probes were routed over "doubtful" places, but I don't think I saw a difference.

Notice that there is more to it than just "applying" a very low noise supply and that it is all related to how much headroom you give the regulator (that's what it is) to flatten the "ripple" (can we call it that ?). So envision (and correct me wherever I look suspicious) a voltage wave which is e.g. 7Volts but now in RMS terms. So this wave is no nice sine at all, which is what we are doing this in the first place. Now, the extensions of the wave are sort of shaped off until the wave becomes a nice sine. So, it is not so that the deep dips are are filled up with new "voltage". It is only shaping. So, if the dips are too deep then dips remain. I think I can even see that on the noise base for what it's *under side" is concerned (see the "ripple").

Now, when the input voltage is 7 and the output is 5, then maybe the dropout is too low (too few). Thus, make that 8V and the dips rise as well, and when (virtually) the dips stay above the 5V level, the 8V can now be shaped to 5V neatly.

The whole thing is fairly complex because all is related. So before you know the transformer is too small (to enable the e.g. 8V DC) and the larger one implies too much heat on the AC/DC towards/in the main PSU. So there actually the same happens ("shaping") and what's thrown out is heat. Similarly, now having a nice 8V to shape into 5V also takes additional heat. And as you already know Anthony, this heat can be of big concern because before you know it no PCB with sufficient cooling copper will fit anywhere, nor will heat sinks. So all is a matter of feasible balance and the main issue is always heat.
And heat again implies (thermal) noise so that already makes it worse again.

So this is my reasoning about the 50Hz (or 60Hz, won't matter) with of course my remark that this is all fairly new and no idiot should be looking for such low noise in the first place. As I told earlier - when the noise would be all over at -120dB to begin with, you wouldn't see 50Hz either.
Of course I still wonder (or am challenged/intrigued) what would happen when the whole AC to DC business would be at several meters distance. Just to be 100% sure it is not pickup.

Thanks,
Peter

[acg]

So this is my reasoning about the 50Hz (or 60Hz, won't matter) with of course my remark that this is all fairly new and no idiot should be looking for such low noise in the first place. As I told earlier - when the noise would be all over at -120dB to begin with, you wouldn't see 50Hz either.
Of course I still wonder (or am challenged/intrigued) what would happen when the whole AC to DC business would be at several meters distance. Just to be 100% sure it is not pickup.

Yes, this was one of the things that I thought could be important in the ATX LPS project.  The design had matured to a stage where all the AC to DC stuff was to be done in an external box along with pre-regulation (in DC of course) and then inside the computer the final regulation was to occur so that the antennae effect of the longer DC wires could be reduced and so that the final regulation occurred right at the load.

Anthony

[Arjan]

Hi,
While waiting for a NOS1a in the very near future I did a final linear psu upgrade on my XXHE pc. And I am surprised about the stepup in SQ.

As you see in my profile I am not using a NOS1 but a AP2 + purepower usb - spdif converter to a NAD m51.
For my pc I use a Paul pang picoPsu feeded from a linear lab psu. But I removed the silverwire of the picoPsu, because I do not like silverwire, and replaced it with better 'copper' wire. The external esata hdd together with the pc fans are powered by 2 separate linear lab psu's.
Next there is the usb card, it is a Silverstone card but with a twist, I replaced the clock with a Paul pang clock. I feed the usb card with 2 linear psu's. The 3.3v from a linear lab psu. But, and that works remarkably good, for the 5v I use a aqvox linear psu. It was used for clean usb power but I put a molex connector on it to connect to the modified usb card.
This last upgrade, the aqvox on the usb card, made a significant difference. I think the ripple of the psu is <0.1mV.

This has raised the bar for the next to come NOS1a. But it needs 5 linear psu's for the XXHE pc to get this far!
I expect that with the NOS1a I can detach all the wires and psu's from the pc and go back to a normal decent atx psu. We will see......

Regards, Arjan

[PeterSt]

Great stuff Arjan !

It is strange ... I think I am not the only one who hopes that these kind of tweaks still keep on working for the better ...


Yes, sometimes I think "what have I done".
But then if you hear it ... 

Peter

[acg]

Hi Peter,

Given your recent efforts with super low noise clocks in the NOS1a and Nick going nuts with clocks here there and everywhere in the PC, perhaps it is time for me to resurrect this project.  Maybe this will help with the next rounds of advancement in computer audio.

Cheers,

Anthony

[Scroobius]

Mani has a linear PS on order for his PC..........