Great PC Tweak - A Bert "Must Have"

[Nick]

Nick, I reckon you guys should pull that card out after you listen to it so you can have a good look at what has been done to it.

Keep us informed.

Anthony

Anthony hi,

I'm really keen to take a look at it, also to listen to my modified card along side it. Im guessing Paul is focused on tweaking it so some of the details may out before we get together  . I'd be interested to put my scope on the clock. It wont tell anything about phase noise etc but will give a picture of signal amplitude and rise time. The key points from messing  about iv done here appear to be oscillator performance, quality of the decoupling (the transformer and matching) of the clock circuit and the clock's power supply. No surprises about these factors but as always implementation is so important so i'm really looking forward to seeing Pauls card.

Best,

Nick.

[Arjan]

Hi Nick,

I have the audio grade usb3 card and did some test today. First I used a external battery Konig 5V-2.1A 7000 mah (normally used as backup for an ipad) via a usb-molex cord. This is no success! Maybe not a good battery to test with, but that is the outcome. Next I tested with an external linear peaktech lab power supply. Much better. But the normal internal bequiet power supply (pure power 530W L7) still beats that....

Some other results of the tests: PC without the hood is much better, more open sound. Also the harddrive outside the pc box made a nice improvement.

I have a little problem with the usb card, only the 5V leads may be connected. It seems that the card is leaking through the 12V leads. But it works fine without the 12V leads in the molex.

I checked the card it seems that the additional components are connected via some small leads soldered to the molex. So connect this to a separate battery seems possible.

kind regards 

[Scroobius]

Further tests with the USB Audio Grade PCie card. The card comprises a normal USB3 PCIe card with a piggy back card that has a low jitter clock and a silver output isolation transformer. The piggy back card picks up 5V from the incoming MOLEX connector which I disconnected I then connected a dedicated linear 5v supply to the piggy back board. The supply from the PC SMPS remained connected to the molex on the main PCIe card.

This seems to be slightly better than having both the PCIe card and its piggy back card from the same supply from the Seasonic SMPS. But it is a small improvement I would say.

The piggy back board seems to have a DC to DC converter but it has a big heat sink glued to it so not easy to see but anyway it probably produces 3.3V for the clock. It gets the heat sink warm to the touch so it is probably a regulated supply rather than "chopped" supply - it would be interesting to see if a battery supply 3.3v direct onto the clock improves things. Well maybe but it is not an easy thing to do with this board maybe something to do on a rainy day.

Cheers

Paul

[Nick]

I have the audio grade usb3 card and did some test today. First I used a external battery Konig 5V-2.1A 7000 mah (normally used as backup for an ipad) via a usb-molex cord. This is no success! Maybe not a good battery to test with, but that is the outcome. Next I tested with an external linear peaktech lab power supply. Much better. But the normal internal bequiet power supply (pure power 530W L7) still beats that....

Hi Akremer,

I'm going to get a look at the card when Paul comes over to mine soon. As I mentioned above if the molex powers either the cards USB chip or is used for the 5v USB leads power supply then using a supply other than the pcs own internal supply proberbly won't sound best. My guess is that the PCIe bus signal / USB d+ d- signal suffers greater noise if the signal source and end points are not powered by the same supply. If this is right then it's better to use the pcs supply as this is used at both ends of the PCIe bus transmitting  / and receiving the data.

The exception is the USB boards clock power supply. If this is isolated from the power to the isb chip and the 5v usb lead supply then using a good supply is likley to improve the clock and since it is transformer decoupled at its output the smoothed supply to the clock doesn't cause the problem above. I think Paul's post is saying he found the clock supply to make a small difference.

I can not measure to pinpoint these ideas so could be something else happening.

Cheers Nick.

[PeterSt]

Yo,

Although this is not completely off topic, I have the feeling it is distracting.

Peter

[Nick]

Yo,

Although this is not completely off topic, I have the feeling it is distracting.

Peter

Peter hi,

Yes definatly off topic, might it be a good idea to move the last 8 posts concering the clocked USB 3 card to this thread ?

http://www.phasure.com/index.php?topic=2749.msg28645#msg28645

Regards,

Nick.

[PeterSt]

Ok, I just did so. Thank you Niick.

Btw, split from under this post : Re: Investigating a full linear PSU for the XXHE PC and NOS1.

Peter

[CoenP]

Now I've had this card in my system for 300 hours and I am sorry to say that it doesn't deliver the goods.

The burn stabilised at about 200 hours as indicated by PP.
I've noticed a mounting issue late in the process, but fixing that didn't bring the sq forward.

The card deliveres a smooth subtle kind of sound where all sharpness has been tweaked out. The tone can be described as natural with lots of details. However, the low end is unfortunately completely gone as is the impact of the fundamentals, leaving the sound robbed of dynamics and musical momentum. Also I've noticed "strangeness" for lack of an adequate description in echoes and instrument decay. Note that most listening was with win7.

Other XX settings did not bring any relief.
Furthermore I've found that the sound was somewhat inconsistent, sounding better or worse at times even after burn in.

Installing the old card again, the sound becomes much more interesting and instruments have real body again. This is especially noticable on the sound developement (attack, decay etc) of keystrokes on the piano.

Regards,  Coen

[CoenP]

I forgot to post some ponderings about this experiment.

1) an "improved" clock at the pc side does NOT make the influence of XX settings powercords etc inaudible. Actually they have the same effect.
The PC "noise" is still able to slip through. Maybe for the better .

2) the perfect clock signal and its alledged beneficial effects may be ruined at its implementation. To me the x former is highly suspect, it is able to pick up any noise in the PC since it is unshielded.

3) I think this "better" clock needs to be at the other side of the USB cable ie in the NOS. First any good clock gets noisy again as it passess the USB link (pcb, contacts, plugs, cable, etc). Second the USB mode of operation (async) sets the timing on the receiving end.

Some food for thought for those who are experimenting with clocks. I am looking forward to your findings!

Regards,  Coen

[PeterSt]

Great stuff Coen.

I myself think that any transformer isolation will radiate as hell. But this is quite personal thinking as usual.

Best regards,
Peter

[Nick]

Coen,

Not a great result, sorry to hear the clocked card is not working out there. I have not heard the card you have but will get a chance in a couple of days, Paul will bring his PP card to my place so it will be very interesting to take a close look and have a listen.

The replacement USB clock I built started out with a similar sized SPDIF transformer but this was only temporary and it was replaced quickly with a 100mhz rated micro signal transformer / choke combination. The replacement units are targeted at ethernet isolation. Anyway sound really improved. Another thing that helped was capacitor decoupling the transformer on its input and outputs but capacitor size was important to sound. This all sounds like it would degrade the waveform but it seems not from the audiable results.

Re-reading your experiences it sounds like something is wrong. It's really the opposite  of what I would have expected. Of course is could be that you existing set could just have an excellent USB port setup at the PC and the new card is just not a good !

Anyway I'm looking forwards to hearing Paul's PP card and we will report back on what we hear.

Regards,

Nick.

[Nick]

2) the perfect clock signal and its alledged beneficial effects may be ruined at its implementation. To me the x former is highly suspect, it is able to pick up any noise in the PC since it is unshielded.

3) I think this "better" clock needs to be at the other side of the USB cable ie in the NOS. First any good clock gets noisy again as it passess the USB link (pcb, contacts, plugs, cable, etc). Second the USB mode of operation (async) sets the timing on the receiving end.

Coen hi again,

I'm defiantly with you that the transformer is very important I don't know about the PP card's trafo so could not comment specifically on that.

I was looking through USB host chip data sheets and it seems they commonly use a x 20 multiplier circuit to generate the 480mhz line speed of the USB link from the 24mhz clock. My thinking was phase noise in the 24mzh base clock that generates the 480mhz clock has to be a bad thing. With this in mind the transmitting USB device's clock phase performance could be important. I'm trying to find an explanation of how this happens but at the receiving end the USB chip will have to 'clock in' the data as it arrives. The clocks at both end of the link are nominally 24mhz (for the NOS and and NEC usb chipset at the PC) but they will not match exactly. For this reason I'm guessing that the receiving USB end my have to lock onto the transmitting (480mhz) data rate and it might be that a good clock at both ends improved to stability of the lock and reduces transmission errors.

Experience here has been that the PC end clock does help quite significantly but the largest difference by far has been changing the clock at the DAC end of the USB link.

Best regards,

Nick.

EDIT put the /quote in the wrong place

[acg]

I forgot to post some ponderings about this experiment.

1) an "improved" clock at the pc side does NOT make the influence of XX settings powercords etc inaudible. Actually they have the same effect.
The PC "noise" is still able to slip through. Maybe for the better .

2) the perfect clock signal and its alledged beneficial effects may be ruined at its implementation. To me the x former is highly suspect, it is able to pick up any noise in the PC since it is unshielded.

3) I think this "better" clock needs to be at the other side of the USB cable ie in the NOS. First any good clock gets noisy again as it passess the USB link (pcb, contacts, plugs, cable, etc). Second the USB mode of operation (async) sets the timing on the receiving end.

Some food for thought for those who are experimenting with clocks. I am looking forward to your findings!

Regards,  Coen

Interesting report Coen, and very informative in a number of ways.  One of the things that I have always wanted to do since I got into this computer audio thing is to build my own pc case.  There are a number of reasons for this but one of them is so that I can run the USB card in relative isolation to the mobo and power supply to minimise the amount of unwanted electrical action between them.  I have toyed with building a little metal shield to slot over the USB card in an attempt to isolate it, but have always come up with other projects.

I have the original V1 version of the PPA USB Card, which is a more elaborate and more difficult to build version of the V2 card that you have there, so there are some physical differences, but my listening impressions are quite different to yours Coen.

Based on your reports, and Nicks experiments with the NOS1 USB clocks, it does seem at this stage that the best gains are to be made with the NOS1 USB clock.  I really want to hear about the Nick/Paul get-together this weekend and associated investigations and impressions.

Cheers,

Anthony

[CoenP]

Hi all,

Thanks for the replies. I genuinely feel that there is very little positive to report especially now my regular card is working out splendidly.

Re-reading your experiences it sounds like something is wrong. It's really the opposite  of what I would have expected. Of course is could be that you existing set could just have an excellent USB port setup at the PC and the new card is just not a good !

Anyway I'm looking forwards to hearing Paul's PP card and we will report back on what we hear.

Its not what I expected either, but somehow this doesn't work out.
I allready have an (rather cheap) IBM OEM USB3 card with similar TI chip. It does draw its power from the PCI-bus and not from a separate Molex and it has bigger on board buffer caps. The crystal seems some generic type in a cylindrical container. Maybe its just a lucky combination in my setup.
Perfect? Well it does work particularly well, but I can identify areas that could be improved upon. I consider archieving these improvements a bonus rather than a neccessity. With the PPA card the sound was and remained unacceptable.

Anyway I am also interested in the findings of others about this card especially when used as intended so power from the psu and no batteries and stuff for the clock.
An "improved" clocking still seems like a good idea though Peter may have run lately into the limits of lowering the noise (like the phase and amplitude noise in the clock signal). At present time does not permit experimenting with clocks in the NOS1. I am not surprised sound differences are more profound there.

Wrt the working of USB2/3 likely the datasheets of the chips will state the tolerance between the masterclock and the external clock. Internal circuitry of the USB transciever will lock in if you stay within these limits. The main benefit of a better clock is a lower (phase) noisefloor IN the substrate of the chip translating into less (phase) noise on the datalines and thus in the DAC chips. Something like that, because the DAC clock signals are generated by separate clocks.

Though we can reason out a lot only the experiment will tell what it practcally means. Keep up the good work guys!

Regards, Coen

[PeterSt]

I really start to wonder ...

For a completely different project I have had made a bunch of small PCBs which go in beteen USB and the DAC board and they digitally isolate the i2s. This can not work in itself because it implies 250ps of jitter caused by the isolation chips;
I had the PCBs made to proove that isolation can't work anyway for several reasons in my mind and they were made for testing on the NOS1. So, it is my assumption that it goes wrong for other reasons than too high jitter (and notice that the NOS1 oscillators carry 200fs of RMS jitter, while this thus implies 250ps - some 1000 times more).

But what if it works after all, outside of the jitter issue ...

From another situation I already know how this amount of jitter sounds and I think it is just enough "doable" to judge whether any PC (or USB) influence can still be passed on. But who knows, maybe the net result is even better. So while I have these PCBs laying around for 3-4 months by now, maybe I should put some components on them and try it out.

In the mean time, right at this moment we're applying the 24.xxx audio oscillators on USB ... See if Nick's failure on this happens here too ...

Peter