Hunting for noise

[Nick]

I want to close out comments on the Anadco fibre USB extension system that I made here.

http://www.phasure.com/index.php?topic=2640.msg28094#msg28094

At the time of the post tweaks to the Anadco were improving its performance significantly. However, I now don't think that the Anadco can be made to equal or exceed the sound quality performance of a direct USB connection  (assuming there are no major electrical / noise problems with the direct USB connection). 

With an unmodified PC USB interface and NOS1 the Anadco could not be made to better the direct connections performance. With the clock upgrade to my PCs USB card and the upgrades made to my NOS1 the Anadco cannot get near the performance of the direct USB PC to DAC connection.

So the conclusion here is again the same that Paul and I reached when we tried the Anadco system in both of our systems a year and a half ago. I guess that the benefits of galvanic isolation of the USB link between the PC and DAC are outweighed by the complexity / jitter introduced in to the USB link by the Anadco.

Worth a try bit the Anadco system goes back on the shelf 

Nick.

[PeterSt]

But that is sad news Nick ! And you were so enthusiastic about it (for the second round) ...

But did you use a more decent power supply for it now ? I assume yes of course ...

Peter

[Nick]

But that is sad news Nick ! And you were so enthusiastic about it (for the second round) ...

But did you use a more decent power supply for it now ? I assume yes of course ...

Peter

Peter hi,

Yes I tried a variety of PSUs and a shorter 1m fibre cable but just could not get it to be as effective as direct USB. In the end though its been a useful part of the journey, adding to the understanding that reduction of jitter in the PC to DAC USB connection is more important to sound quality than galvanic isolation of the USB connection.

On another subject I spent half a day building a transformer isolated I2S lead last weekend to try in my NOS. I was not really expecting it to be successful given the link speeds but though it would be fun to try. At 16x up-sampling I managed to get reasonably good electrical wave forms across the lead but I think the rise times of the transformers put the data streams out of sync. I did get some music playing (which was very surprising  ) but under a massive layer of static.

I might spend a bit more time with it to see if I can pin down specifically what is happening. Also it might me fun to try at different upsampling rates just to see what happens.

Here's a couple pics of the "creation" 

Regards,

Nick.

[PeterSt]

Haha, I can show you similar pics Nick; can't work. All these means (and the digital ones for sure) only add jitter if you can get a good signal from analogue to begin with. But really, an anlogue transformer is too slow. This is not related to the sampling speed. You will always get "rounded" "waves" while it should be square.

It can be done though, but the how is a big .

Regards,
Peter

[Nick]

It can be done though, but the how is a big .

Very nice  could that be on a piggy back board or built into a lead  haha 

Nick.

[Nick]

I mentioned above that the NOS1's USB interface 24mhz clock has been on the list of items for attention.

I bought a smd rework hot air station to help install a clock upgrade last week and got stuck in this weekend. I built and tested a 24mhz replacement clock from a VCXO with LiFePo4 battery power and transformer decoupled clock output yesterday. The standard NOS1 crystal clock and other components were removed and the new clock installed today. As expected the clock really dictates sound quality.

How does it sound ? With a few hours on it, its a HUGE change. I think the impact is proberbly magnified by the 6 modifications made to my NOS and PC's USB card but having said that this clock change has bigger positive effect on sound than ALL of the other changes put together and by a LARGE margin. Music has such a live quality, bass is so extended and tuneful there is such a sense of performance it's a massive step.

Today Iv been playing track after track of the harshest fastest most complex most difficult music I own. The stuff you put on to really test your system. In my case this stuff usually delivers a nugget of diapointment as the system falls apart here, there and everywhere. This afternoon has been very different. Nothing so far has tripped the system, playing this seat of your pants complex music is now simply riveting listening. And the clock is getting beter by the hour as it runs in. I knew that clock was going to be important but I'm shocked by the size of this change.

Next up is to decide how the final PC USB and NOS1 USB clocks are going to be setup. And turning the system again, the change is very large and I think PC Bios and XX Settings will need tuning to suit.

Paul is due to visit my place soon so it will be good to get his opinion.

A very happy Nick.

[acg]

Hi Nick,

WOW!!.  Please provide some more details and photos.  I am interested by your battery solution.  Are you considering a super low noise linear psu as a more permanent solution or do you think you will persist with the battery?

I have access to a SMT oven and a hot air gun here and could probably duplicate your work and provide a second opinion if you like.

Cheers,

Anthony

[PeterSt]

Hi Nick,

Replacing that 24MHz with something better does not really make sense to me. Or not yet ... 

You shouldn't have replaced these two components at the same time with the suggestion that the "clock" makes it for the better.
Small problem of course : how to make that original crystal run from a battery ...

Anyway, without real further investigation I say that it is the supply noise you got rid of. Nothing about the clock itself.

For others : It may not be fully clear to you, but this is not about the clock(s) to the DAC; this is only about the USB interface.


In the mean time I am "able" to theoretically (!) see how even this USB transfer oscillator(s - PC end as well ... although ... maybe not that one) can contribute to jitter. So, not talking about noise incurring for this jitter, but just the oscillators and their own jitter spectrum. Well, *that* would be something new.

Nick, how easy is it for you to use this new oscillator but now grab a normal nearby 3.3V from the USB board (hence don't use the battery) ?
Judging the result could be difficult because the low noise battery supply will be a first thing what helps. Still (and contrary to my first sentence in this post) I have the hunch that it is the clock itself you are hearing. So what if the result stays as good with the normal 3.3V supply ?


Next step : feed the whole USB interface with that battery. It's only that you shouldn't plug in the USB cable too often while that will suck a few hours of normal operation because of doing that. Switching on (also the PC) same story.

Peter

[Nick]

Pater hi,

Jut a quick response for now, ill have to come back in more detail later today.

There is only one change to the NOS USB board, the extra components I mentioned that were removed are the resonator capacitors which are part of the clock circuit so I'm commenting on only one recent "change" which is the clock replacement.

I can try using the NOS USB interface 3.3v supply easily so I will give this a go. I am very sure that this is not about power however, so I'm not anticipating a significant change. I think this is a jitter improvment. Ill post back my thoughts about what is happening later.

Best,

Nick.

[PeterSt]

I am very sure that this is not about power however, so I'm not anticipating a significant change.

Nick, maybe I should emphasize that I think the same (my post started out a bit confusing). At least I hope you won't find a significant change ...

Thanks !
Peter

[acg]

Peter, do you know off-hand the voltage and amperage draw of the USB board?  This is the board that the 3A supply feeds is it not?

Anthony

[PeterSt]

Anthony,

I especially forgot the figure needed for plugging in the cable, and all I know is that the 3A PSU is in there because it was used in the predecessor of the NOS1-USB which needed 1A as far as I recall. But 1A it won't be because that this beyond USB specs (what it that ? 500mA ?).

The continuous usage while playing fits your observation of something like 20mA (see XXHighEnd PC topic). But the real figure will be shown by the USB port (concerned) properties. So what's listed in there was my own observation back at the time (and the reported figure in the USB port's properties are official; notice though that this is not real time judgement of the OS - it is the figure put in there by myself ...).

Peter

[Nick]

Peter hi,

I only had time for a partial reply earlier so to expand a little...

For others : It may not be fully clear to you, but this is not about the clock(s) to the DAC; this is only about the USB interface.

Agreed, this is the 24Mhz USB 3 receiver clock on the NOS1 USB interface (not the audio 22Mhz and 24Mhz audio clocks on the NOS USB board already work great).

You shouldn't have replaced these two components at the same time with the suggestion that the "clock" makes it for the better.
Small problem of course : how to make that original crystal run from a battery ...

I’m not sure exactly here about the thought that two thing have changed, The only components replaced on the USB interface PCB are the standard 24Mhz clock crystal and the resonator capacitors for the clock crystal.  I think the (battery) power supply change is not very important but I will try powering the new 24Mhz clock from the 3.3v power supply used for the NOS USB interface PCB just to see.

The existing clock is a crystal driven by the USB chip not a active clock module a so it not possible to power the existing clock arrangement with an "external" supply (battery etc).

Replacing that 24MHz with something better does not really make sense to me. Or not yet ... 

Anyway, without real further investigation I say that it is the supply noise you got rid of. Nothing about the clock itself.

In the mean time I am "able" to theoretically (!) see how even this USB transfer oscillator(s - PC end as well ... although ... maybe not that one) can contribute to jitter. So, not talking about noise incurring for this jitter, but just the oscillators and their own jitter spectrum. Well, *that* would be something new.

My thought is that this does come down to Jitter in the end but in the digital processing domain which must ultimately somehow result in jitter in the audio data stream. I think we are all resonably comfortable that you can influance jitter in the software domain from within XX so it did not seem a big step to assume that data hardware might be able to as well.

The thoughts below are very circumstantial but its possible to see how important the 24Mhz clock and derived frequencies are to the async USB link and moving data downstream into the FPGA.

My first reason for trying an upgraded clock is that as mentioned earlier in the thread I have re-clocked USB receivers on Audio interfaces in the past and had significant changes to sound quality. This would probably have been enough to try a new clock but the NOS USB PCB proberbly a very  expensive board and I did not want to commit to a possibly irreversible change without at least some research into what the 24Mhz clock drives and the areas that might be impacted by an improved clock.

This is what jumped out from data sheets.

• Looking at the NOS USB transceiver chip (I cannot get hold of the NEC USB Chipset datasheet), the USB transmitter multiplies the 24Mhz frequency by 20x to generate the 480Mhz USB 2 standard line transmission “clock”.  I guess that phase noise in the 24Mhz clock can really impact the X20 process and the stability of the 480Mhz derived transmission clock.
• The 24Mhz clocks at both end of the USB link (NOS and PCIe USB 3.0 card) will not be at exactly the same frequency (10s, 100s or 1000s of beats per min out, I have actually tried but failed to tweak my oscillators to match the crystal speeds  ). Could the receiver at both ends of the USB link be using somthing like a PPL to “clock in” received data. If this is the case the more stable data speed arriving the lower the chance of the PPL hunting or losing its lock ?
• To generate a 20x clock 480 Mhz transmission clock speed (or PPL lock if they are used), the phase accuracy of the 24Mhz source clock is going to very important. The standard clock is a simple sine wave oscillator with relatively poor phase performance. When compared to a low phase noise clock the standard crystals sine wave means accurate phase triggering must be hard for the USB chip and noise in the detection circuit of the USB chip can create a relatively large time shift in the detected clock edge.
• Final thought is looking down stream from the USB cards usb transceiver chip to the FPGA. I don’t know how the FPGA code is designed and therefore which source clock drives the data handling within the FPGA. I would guess that some of the FPGA could be slaved to frequencies derived by the 24Mhz USB Clock.

I could not really give hard arguments about the points above they are as likley to be wrong as right but it just felt like the 24Mhz clock was important enought to try an upgrade. Power supply and electrical noise might come into the overall equation but the change in sound is just too fundamental IMHO. It would be the biggest “noise” clean up tweak ever if it were just this alone. Whatever the cause,  the change in sound quality is defiantly for the better and IMHO is really, really large. I was rocked on my heals again this morning when I played a few more tracks.

I’m happy to send the implementation details off-line Peter if you would like them (its not too complex really but might help if you want to try it out), I don’t know if you have tried the capacitor modification to the USB board yet but the clock change is really is worth some time.

Best regards,

Nick.
Edit - language tidy up

[PeterSt]

Nick, not responding to everything ...

Sorry about the confusement about my "two components" but I really meant the oscillator as one component and the noise-free battery as the other, while knowing that you couldn't apply the battery-only to the crystal. So that's how I thought to test it the other way around, and apply the normally 3.3V supply to that better oscillator. In the mean time though I changed strategy (in my first post about this) and thought that it wouldn't be the noise making the difference but just the oscillator. So this would even justify trying the normal 3.3V supply. It would prove it is the oscillator doing it.
Of course I did not know that you were on that track too ...
So all consensus.

Meanwhile, already yesterday I thought to replace the oscillators at both ends with just the 24.xxx audio clocks. Haha. I think it will just work. What do you think ? So I wanted to suggest that to you, but you have one of those only (spare, I know ).
Next it is not so easy for me at this time because I'd need to replace the oscillator on the MoBo. Can be done, but is not the most convenient of course. So ... what USB3 interface do you suggest ? Should be normal PCI (not PCIe) or otherwise I can't get it in.

And oh, my thoughts on the software being able to influence and from there think about this (still asynchronous !) interface to tweak are again the same as yours. The "why" I am not sure about at all at this moment, but all further theories tell me that there's some work to do here. All quite new ...

So far for now !
Peter

[acg]

So ... what USB3 interface do you suggest ? Should be normal PCI (not PCIe) or otherwise I can't get it in.

Peter, the Asrock mobo has 3 x PCIe slots and only 1 x PCI slot.  I understand that the silent video card has the heatsinks and takes up a couple of slots, but can't you move it to the slot furthest from the CPU and then slip in a PCIe card behind it?

As far as cards go Peter, I would recommend you try two...a standard USB based on the NEC Chipset and the PPAstudio USB3 Card.

Cheers,

Anthony