Hunting for noise

[AlainGr]

Hi Nick,

I am sitting on your shoulder If this experiment turns for the better, I will sure need a little course on soldering, but also exactly where to solder it

Alain

[PeterSt]

Nick,

Coen was ahead of me, but/and reading your post about the Black Gates I wanted to say the very same, but with another proposal :

First of all, to me it doesn't make much sense to put in Black Gates there. So, why do it in the first place but then with the focus on people not having them around. Thus, I'd say that any el cheapo cap will do the job here.

Then, still assumed you are looking at HF noise here, it could be a fairly moot thing to "see" that improvement was achieved. I think Coen said the same (with maybe different words), but the "pattern" will change over time (aging of the caps).

In the far end - but this is a very different story - you look at HF noise and I don't think you have any influence on it, and while possibly today for the better, tomorrow for the very worse (again the frequency changing thing).
Let's keep in mind that this is all about jitter (as far as I can see), so if you can tell me how this is going to influence jitter for the better ... I am all ears. Of course I am a bit nasty here because the answer is quite impossible to give (I think). So, I won't say there's no improvement possible here, but I do say that we won't be able to see what is for the better. And I mean : never think that less noise of high frequency is for the better. And if anything, I'd say it is for the worse, as long as noise remains (which it undoubtedly will). Higher frequency noise will imply jitter more randomly. The higher the better. Maybe it is not easy to agree upon this, but might we after all, then we also agree about this all not being so useful (up to working out for the worse).

If possible, I'd like to see the technical merits (somehow !) from the comparison between the copper connection and the glass connection. And then, sadly, with the original setup of the oscillators, which you changed quite a bit ...
Yes, sadly, because this is your concidental situation.
So whatever you come up with, at least to me it can't tell it all because it is not my situation. Nor anyone's else ... This, btw, also includes the being active of one oscillator only, which I dedicate much difference to begin with. Good stuff (for sure), but we all don't have that ...

In the far (far) end the only thing with real merit is to measure jitter (its characteristics) at the output. And here it stops I'm afraid (I am not able to do that myself).

Is your noise at -144dB completely out of the way now ? if so I say nothing. But if not, ... we don't have that either (no matter how logically it is per your own (means of) proving it).

Too many variables Nick ! I wish it were different !!
Peter

[Nick]

Peter hi,

Don’t get too hung up on what the “noise” related objective are, these tweaks do aim at noise but they are also aimed at a wider target. In addition to the -144db breakthrough and reduction of noise entering the DAC from the PC but I have always had it in mind to take a detailed look at the NOS1 with a view to seeing if it can be tweaked for an even more enjoyable sound. The hunt for noise in this post and looking at the DAC generally have come together here. My motive is primarily to do this for myself to enjoy but if this can help the community gain more enjoyment I am happy to share findings.

Too many variables Nick ! I wish it were different !!

So far I have provided no real information on what I have done so proberbly not the best time for this comment due to the absence of information (now addressed below).

So whatever you come up with, at least to me it can't tell it all because it is not my situation. Nor anyone's else ... This, btw, also includes the being active of one oscillator only, which I dedicate much difference to begin with. Good stuff (for sure), but we all don't have that ...

I understand your point about a common config but remember my perception of SQ change is relative to these “baseline” mods already being in place. If needed I can place the standard NOS clocks that I have back in my NOS for comparison but Paul has spent several hours listening to my system as it currently is and could probably take a view on the recent changes, I think the NOS1 tweaks make a fundamental difference so I don’t think it will be hard to spot the changes at all for Paul.

In the far (far) end the only thing with real merit is to measure jitter (its characteristics) at the output. And here it stops I'm afraid (I am not able to do that myself).

I agree, the NOS tweaks are all about improving Jitter performance by improving supply rail conditions. Neither of us have the equipment to measure jitter (not many do given the cost of the kit needed), however I’v worked on many devices over the years (USB cards, DACs, CD Players Transports and Audio interfaces etc) almost without exception they have responded to PSU bypass tuning measures used here so I am not especially surprised by what is happening.

Below I have given information / my rational to target the tweeks but I have intentionally not given details of how to apply them to avoid problems with folks getting in to trouble trying to applying them.

So the tweeks to the NOS1 USB board and the DAC board and the rational applied.


The focus has been on selecting locations in the digital processing chain that are likely to benefit from improved power supply. The chain considered consists of USB decoder => FPGA => FPGA => DACs (I am excluding the Anadco in this post). We both are both probably comfortable linking this approach with jitter reduction. There is an element of trial here but it not entirely a “hit and miss” process rather it is based on experience informed by the information you kindly provided.

I guess the starting point has been working out what could be done to address the I2S related noise that we discussed earlier in the this post but also there has been focus on what might help sound quality.

The NOS tweeks and their rational.


Observations

The NOS USB interface has a single regulated 3.2 volt supply to power 5 IC devices which presents a challenge getting a responsive low noise supply to all the important supply pins of the ICs (eg supply pins that power key points in the digital signal train). In the past I’v have experienced improvements in sound quality all the way up to placement of 7 separate shunt supplies to power various rails on just 3 active IC devices within a DAC, probably overkill here, but you get the picture.
There is a total of 570uf of capacitance at the point that the 3.3v PSU connects to the USB board. I think that this large cap may present a relatively “slow” response to transient power demands and will dominate speed that the 3.3v rail reacts.

The level of rail decoupling capacitance on the USB board is on the low side of what I’v seen before on “audio” boards. The board ICs have a dozen or more supply rail pins on the 5 devices these are decoupled by a few caps generally of 4.7uf on the 3.3v rail. The 2.5v and 1.2v regulators have 4.7uf each in their output rails. There are also 100nf decoupling caps on many of the power pins. I guess this is not that unusual but I have found many times that more good quality capacitance in the right places can help significantly.

On the DAC board the FPGA (which receives and distributes the I2S to the DACs) has a an SMD 3.3v regulator, I cannot determine the exact amount of decoupling capacitance on the 3.3v rail as it splits to the FPGA power pins but the physical size of the SMD caps make me think it may not be a lot.

So NOS changes:

USB Board

1)   Remove the 470uf silmic and remain the 100u cap (I know this much capacitance may have been placed here for another purposes)
2)   Place 33uf of capacitance at C1xx which I think may be the point that the USB and FPGA chip rails fork. (Significant)
3)   Place 33uf of additional 3.3v rail bypass capacitance on the supply pin of the quadrant of the FPGA which is programmed to support I2S output processing. (Significant)
4)   Place 33uf of additional 3.3v rail bypass capacitance on the supply pin of the quadrant of the FPGA which processes the Audio oscillator inputs and distributes timing to the rest of the FPGA. (Significant)
5)   Add 33uf of additional bypass capacitance to the outputs of the regulators generating the 2.5v and 1.2v rails.

Only one of the targeted USB board cap placements i had in mind turned out not to improve sound quality and was reversed.

DAC Board

1)   Place 33uf of additional capacitance on the output of the 3.3v regulator that feed the FPGA on the board. This makes a very significant difference to SQ however I intend to tune the value further as I think 33uf is just a little to much in this location. The basis for this is a very slight drop in the remarkable top end sparkle that the USB board changes provide.

EDIT 19/09/2013 I dropped the cap from 33uf to 4,7 then to no additional capacitance. After careful listening the sound is best with no changes to the board here so I would not suggest anything here now.

That’s it….

First of all, to me it doesn't make much sense to put in Black Gates there. So, why do it in the first place but then with the focus on people not having them around. Thus, I'd say that any el cheapo cap will do the job here.

Regards the capacitors used, I happened to have Black Gates to hand. I do not want to heat stress and damage my PCBs so I do not want to be soldering and re-soldering to SMD pads and components so I went with the best caps I have which were the Black Gates. I have used Black Gates N Types may times for digital component PSU bypassing and got excellent results. I’m familiar with the likely SQ changes the Black Gates will give as they run in.

I really don’t want to get hung up on capacitor brands and types. My view is that the changes in SQ are so profound that I really think just adding the additional capacitance at these locations is likely to be much more important than the brand of cap used. The tweaks also happen to be cheap and relatively simple to implement, that ticks a lot of boxes for me. 

I am very happy for Paul and Mani to come have listen and compare my DAC with these tweaks at mine or in their systems if they are interested. This will give a second opinion on effectiveness of the changes.

Best to get the changes and rational on the table so we can have an informed discussion both knowing the details. Happy to discuss how all this might be addressing jitter but ultimately theory linking PSU noise to jitter spectra and our ability to measure jitter may run out quickly meaning that we may have to listen, trust our ears, and trust experience (uncomfortable though this may be). After all, were the current 4.7uf and 100nf decoupling caps on the USB board selected specifically for their contribution to reproduced sound ? I’m guessing they were more likely selected following recommended values in the Data sheet for the devices / good practice for general decoupling, and since Audio is a niche application for FPGAs and USB decoders etc the recommendations probably do not really conceder what works best in an audio application. Just a thought….

Regards,
Nick

Ps Anadco observations to follow.

EDITS were to grammar and one of the tweaked cap values was given as 3.3 uf not 33 uf as intended.

EDITS 19/09/2013 remove change to DAC board after listening test see above.

[PeterSt]

Thank you for (explicitly) sharing that Nick ...

Peter

[PeterSt]

Nick, I now realize that I wanted to say "thank you" for the Edits in there.

Peter

PS: But editing a post so much later isn't the best idea I think. But no worries.

[Nick]

PS: But editing a post so much later isn't the best idea I think. But no worries.

Peter,

Sorry, I agree it was a late update to the post. I did think about a separate post but I was also keen to make sure that the information was all in one place. In future if a similar situation comes up I will put a short post on the thread to say a change is made and edit the main material.

By way of an update, the USB card tweaks have been auditioned by another individual and I hope to get further subjective feedback soon. Feedback on the effect of the tweaks is the same as my comments about their effect.

I think there is something significant for sound quality here that could be worth a look. I will PM you some pics showing the detailed changes in case you would like to try the tweaks out for yourself (they are reversible and take about an hour to apply). Given the nature of the changes I strongly suspect that 80% of the "benefits" will be available using "normal" capacitors which would be very nice of it turns out to be the case. I would test this myself but I am not keen to stress my USB board with all of the de-soldering and re-soldering as I only have one to work with (or break  ).

Cheers,

Nick.

[PeterSt]

I can always send you 10 or so with a good discount Nick.
Haha.

But, interested of course ...
Peter

[Nick]

I can always send you 10 or so with a good discount Nick.

Thanks for the offer  I'll try to keep a steady hand with the soldering iron though haha

Nick.

[Nick]

I wrote part 1 of this post a few days ago and was going to post in the USB Cables.... Again thread but decided against it. Then after sleeping on the thoughts below decided to try something else which may add weight to the ideas on the effect of USB power supply noise put forwards below. Part 2 of the post deals with observations today.

Part 1

 I had a play about with isolating the 5v pin of my USB cable last night. Its something I had a go at quite a while ago with a diy platted silver usb cable and got "good" results at the time.

http://www.phasure.com/index.php?topic=1977.msg20341#msg20341

I tried "cutting" the USB 5v supply (using PVC tape over the pin in the USB socket),  and following this then went on to try replacing the PC 5v supply with regulated 5v and 3.3v supplied derived from inside the NOS1.

I though I would post some observations on sound changes and ideas about possible factors contributing to the changes.

First sound quality.

1) With 5v PC USB connected.

Boarder sound stage, fuller but possibly slightly less damped bass richer sound and tone.

2) Without 5v PC USB connection (USB power pin taped over)

Sound stage shrinks slightly but is sharper in terms of pinpointing individual sounds. Sound is generally tighter but slightly thinner, tone is not so convincing. Bass is better damped but the room ambiance form low frequency is not there possibly giving the sense that the sound stage has reduced in size.

3) With 5v PC USB disconnected but NOS generated 5v or 3v supply provided to the USB power input.

Sound is smooth but loses high frequency information which results in sound stage shrinking and a loss of perceived presence. This is very similar to the presentation of the Anadco fibre interface or the X-sound PCIe capacitance card.



My preference overall was easily listening with the PCs 5v USB connection in place. This is not what I expected. 



Some thoughts on what may be happening.

What does the 5v supply do within the NOS ? two things (and similar for other interfaces):-

• First it is used to pull up a bus enable signal into the USB chip (this does not appear to be needed because the NOS works fine without any 5v supply).
• Second and possibly important here is that the 5v usb supply connects within the NOS (and may other usb interfaces) into an anti surge protection chip that protects the d+ d- data lines of the USB interface connection from surge conditions on the USB connection.
  

The surge protection chip is a diode network that connects the d+ d- to the +5v USB supply and signal ground of the NOS. What all of this might mean is that conditions on the +5v supply wherever it comes from could link noise to the d+ d- signal lines ???

What is interesting though is that sound seemed better with the PC USB 5v line connected. This is what lead to the test replacing PC USB 5v supply with the NOS's own Internal supply voltages to see if a "low noise" 5v supply would have less influence on the d+ d- lines and signal ground. As reported above the sound was smooth and similar to having the 5v USB connection cut completely but still my preference was for using the PC USB 5v connection.

This is a stab here, but an idea that might to explain why the PC usb 5v supply sounded better than either no 5v connection or a "low noise" 5v connection, could be as follows. The noise level on the 5v line, wherever it is taken from is probably going to influence conditions on the d+ d- lines at the anti surge device.  However it may actually be better from a jitter point of view to use this PC 5v USB. it might be that the d+ d- lines may be carrying similar noise spectrum to the PC 5v USB line which might present as common mode noise on all three USB lines and not impact the d+ d- data transmission so much as it reaches the usb decoder chip. Alternately when using low noise 5v supply (from the NOS internally) any noise carried on the regulated 5v supply will not have the same spectrum as noise on the data lines and as a result may have a greater impact on the d+ d- transmission to the USB decoder chip.

This is just an idea to provoke discussion, I'v probably has a bit too much to drink here last night when thinking about this but thought I would post the ideas. It would be good to get to the bottom of what causes the differences heard.

Part 2

So today I wondered what would happen if the 5v PC USB supply were bypassed (smoothed) with capacitors at the NOS end of the USB cable. The answer turned out to be quite a lot happens and some potentially good stuff at that.

I added a range of capacitance values between the 5v PC USB supply and the NOS USB card signal ground. Observations on sound are as follows (note the NOS already has 100nf in this location so the values stated are in addition to this).

2200uf
Sound is smooth loses presence base is controlled and strong but sound is unbalanced with highs rolled off. Just like the regulated 5v and 3.2v supplies, a pattern forming perhaps.

470uf
Overall balance improving but sill nowhere near not having a capacitor for presence etc.

10uf
Still losing the harshness of the unfiltered 5v USB PC supply but still also not well enough balanced or detailed enough.

400pf smother sparkling top end mids and lows sounding good but transients are just a little slower than an unfiltered connection

220pf now we are really getting there. Hash gone, presence intact, beguiling sparkling highs, transients intact !

I'll try a few more values round 220pf and listen more but this is sounding very nice. It's a tuning point rather than extremely large change of the USB pcb tweaks but very nice all the same. Seems to give the sense of transparency and focus of lifting the PC 5v supply whilst keeping the life scale and tone of having the PC 5v supply connected.

I will post more thoughts after some more listening.

My guess is that the value of cap that gives good results will vary from system to system because it may be linked to the noise spectrum coming from a particular PC.

Nick

UPDATE 25.10.13

Looking at a few PCIe USB 3 cards most seem to use a DC to DC converter to take the PCIe 3.3v bus voltage and convert this to the 5v rail that is carried by the USB lead. So the cards i have looked have a SMPS running at 300 to 450 kHz straight onto the diode clamp mentioned above that terminates the USB d+ d- signal lines most DACs. For most of us the supply will just sit idle (supplying no current) 'humming' away on the USB card. The tests above showed that bypassing this 5v supply with a clean supply was not great. But what happens of the USB's 5v rail has some resistive rater than capacitive load applied to it ?

I put a 1k potentiometer between the USB 5v rail and the signal ground on the NOS1 interface. Having the load really help sound here. Best sound for me is around 200 ohms of load so around 25ma of current. If you try this be carful not to go below about 75 ohms as you will start approaching the 500ma rating of your USB card or mobo usb port (by 10 ohms load) and may get a hot potentiometer.

Sound quality changes with the resistance but is better in my system with a resistance in place. At 200 ohms there is subtle but important change throughout the frequency spectrum. Sounds like a slight reduction in jitter.  Bass takes on a more tuneful deeper and percussive feel especially.

The resistor seems to work well for both pcie usb and mobo usb.

I was wondering of the effect might be about countering ground loop current between the PC and DAC. I know ATx supplies tend to elevate the pcs signal ground a few mv with respect to PE and in my case my amp signal ground is connected to PE which could be making a loop from PC to DAC to amp. Whatever the cause the resistor is defiantly staying in place for now 


Regards,

Nick.

 

[acg]

Nick, that is an awesome post.  Very much appreciated.  The more that I get into this computer audio thing the more I realise that I don't know.  Regarding your capacitor auditions, I had a mental picture of Peter doing the same thing when developing the NOS1.

Cheers,

Anthony

[PeterSt]

Haha Anthony, you don't want to know. But most of these things can just be measured. And otherwise much is (my) theory only. Point is though : all these combinations (and different values !) are hardly feasable to test. So I really don't do that, but apply the theories.

Peter

[PeterSt]

PS: For fun look at the main PSU in the NOS1. All ready for all sorts of cap combinations. And not by accident. Nobody started working on that as far as I know ...

[PeterSt]

Hey Nick,

it might be that the d+ d- lines may be carrying similar noise spectrum to the PC 5v USB line which might present as common mode noise on all three USB lines

That. Or at least that is a good idea. But also outside of common mode noise it is a good thing to have one type of noise only, even if that's there two times.
So either way, I think this is a good hunch of you ...

Peter

[Nick]

Haha Anthony, you don't want to know. But most of these things can just be measured. And otherwise much is (my) theory only. Point is though : all these combinations (and different values !) are hardly feasable to test. So I really don't do that, but apply the theories.

Peter

Peter,

I really agree with you here, I dare not imagine how much time went in to tuning the DAC board ! Im looking at cap types but would not touch values its very obvious that there is design that's beyond me, so i'm not about to meddle haha.

The only change I did try  to a cap values on the board did not work (ref post above) !  .

Nick.

Edit - I selected the wrong post of Peters to "quote", it now reads as intended.

[Nick]

Hey Nick,

it might be that the d+ d- lines may be carrying similar noise spectrum to the PC 5v USB line which might present as common mode noise on all three USB lines

That. Or at least that is a good idea. But also outside of common mode noise it is a good thing to have one type of noise only, even if that's there two times.
So either way, I think this is a good hunch of you ...

Peter

Thanks Peter, I see what you mean. Im not sure about overall positive effects of the additional cap yet. I think some more listening and maybe changes to size are needed to decide.

Its interesting thought that at least in my system pc supply sounds better and that the attempts to "clean" the supply introduce such changes to the sound.

Im not sure it will lead to anywhere but fun to try out 

Cheers,

Nick.