This is great ! I'm just scanning though and already some some ideas that are simple modes and could help performance are coming to mind.

This made me laugh 

"Note 1:
An external modular oscillator cannot be used instead of a crystal, due to aggressive clock management in reduced power states."

Guess we will have to avoid reduced power states then haha.

Nick.

George,

Thanks for responding to the ask for help with these data sheets further up in this post and for your pm.  I'm been trying to get my hands on a copy for ages !

Also Peter thanks for linking them.

Now for some reading 

Cheers,

Nick.

Paul,

I think Ros deserves a medal for being so patient and I have to say I think she has "first class hearing" as she spotted the extra "good" loudness of the dexa setup   

Really though, very interesting and I hope you are both enjoying the sound.

Kind regards,

Nick. 

I should perhaps have updated on this point. With the Dexas fitted I did get that the same perceived loudness was produced with greater XX Vol attenuation settings. I know this in the past has generally been a sign that all is not well. But in this case is sounds like a richer fuller presentation whilst maintaining detail.

By way of update, it seems that the effect is to do with clock speed at the ends of the USB link. The extra loud perceived sound can be tuned in and out of the presentation by trimming clock speed.

I guess Paul might not have had his clocks trimmed in that way.

I'm waiting for my frequency counter to arrive so that I can start relating relative and absolute clock speed to perceived sound. What most interests me is understanding how to tune for the sweet spot reliably.

Cheers Nick.

Thanks for listing the chips Peter. I had not been aware that there are two versions running at different voltages until Coen mentioned it. Iv been trying to get my hands on the NEC USB3 chipset datasheet for a long time, they don't make it easy to get a copy. If anyone has it would be really great to have a copy.

Cheers,

Nick.

Peter thanks,

I will read your posts again with this in mind.

My initial toughts are that three ingreedients could be on the Silverstone card by happy coincidence.

- Improved PSU to card and clock via the molex.
- A better quality XTal / clock circuit performance
- The USB Clock speed happens to be at the sweet point frequency relative to the NOS clock speed (this seems to be slightly faster and makes quite a difference)

A while ago, I tried the same USB card type that Pauls has had excelant results from in the past when I have been to listen to his system in my system and overall it did not work so well for me.

I think the clock journey is showing that "matching" of the USB cards to the rest of the system could help. So far this has been a happy coincidence if it happens but now clocks are being looked at I think it is becoming possible understand how to tune the USB card to the NOS.

Best,

Nick.

Edit,

Actually I should also have mentioned the two versions of the NEC chip that Coen pointed out in an earler post. Could one sound better that the other ?

Peter hi,

Great to see such excitement and focus on improving USB clocks.

I am really keen to respond to some of the comments because at various points over the last few months and with different clock configurations I think I recognise so much of what your describing. What has become apparent is that the details of the clock implementation totally drive the magnitude of the changes in sound and just as importantly the residual "problems" that remain. Because the magnitude of "improvements" any remaining problem really stands with upgraded clock in place (a nice problem to have  :-) )

Anyway this is a long way of saying I would really like to offer comments but its not really possible without understanding the config you have at the moment :- number of clocks, clock types, PSUs, clock coupling (transformer / capacitive / direct), clock speed. All make such big differences to sound.

Can you let us know what config is generating the thoughts above ?


I think we very likely to be listening to similar changes. Remember this from recent PMs.


"To be honest, I have really underplayed the effect of the clock upgrades on line, off line i'v never heard such a change in sound. The NOS is simply absolutely stella set up in this way. I think you might even want the guys at 6 moons to re-review, I just don't know what they would make of it though, or what comes after a statement level recommendation for that matter."

I still feel a common reference would be extreamly useful and now we are back to the North Sea again haha.

Anyway, fantastic to hear about so much progress.

Regards,

Nick.

Coen hi,

I am not aware of the spec. of the tranformer used by the Paul pang card so difficult to comment there but concentrating for a moment on my DIY clocks build and the Dexa clocks. The main application of the output transformer used in both of these is for ethernet network transceivers signal isolation, bandwidth is rated >100Mhz. I guess to preserve the signal waveform the transformers effective maximum pass band will be much higher than this.

What has struck me with the experiments on USB clocks so far is how much more that jitter reduction seems to impact on sound quality than clock RF impacts sound as HF components of square clock waveforms increase.

Given this my thinking is that bandwidth limiting the clock waveform might improve general electrical noise but is likely to also impact phase performance which in turn could have a big impact on sound.

Im still very keen on clock transformer decoupling as it is appears to take ground loops on the clock connection out of the picture which can otherwise be a big problem for sound quality. But generally it think high bandwidth passive components on the out put of the clock may be the best way to go.

Kind regards,

Nick.

Coen hi


My DIY clock used capacitor decoupling on both output of the clock module in serise with the transformer primary coil and also on the secondary of the transformer on its output in serise before the signal connects to either the NOS PCB or the PCIE USB card.

I tried direct coupling of the clock module to the transformer primary and the transformer secondary to the PCB clock inputs but this resulted in a loss of signal amplitude and would not drive the USB PCIe card and NOS PCB IIRC.

Hope this helps,

Regards,

Nick.

Paul,

Bingo my thought too !

 This is one of the reasons I have been saying that I think clock speed faster than 24mz may be needed. My scope will not measure to 8 digits but something that will is on the way. Im really wanting to understand the variables,  I think the understanding could really help tune the usb link .

Cheers,

Nick.

Paul,

Fantastic !   normal service resumed.

Interesting to here that the trimmers do have an effect now. It took a while to work out what they were doing because most settings seem to have qualities that are appealing in some way, so I agree that it is not so clear cut.

The sound I'm homing in on has a nice balance of dynamic, tone sound stage and emotion. It seems to be just moving from the slight sharp setting to towards the smooth setting on the trimmer, just so that the edge disappears.

It might be worth having a chat about this, as mentioned above I think absolute speed and relative speed of the clocks seems to have an effect so it would be good to bounce a few ideas about and check and compare results.

Speak soon,

Nick.

Paul hi,

That seems like a better way to do it I guess on Paul Pang's card.

Sorry I realised that I said that I was adding to your point but actually my comment above were about the molex on "vanilla" USB cards, sorry if I did not make that clear. As you know I have only seen your Paul Pang card for a limited amount of time so I'm not really able to say anything about the supply of the card.

Best,

Nick.

Peter hi,

Just to add to Pauls points.

Most cards will use the 3.2v pcie bus voltage by default. In this situation this bus voltage is used to power the nec chip and via a dc to dc converter the the 3.2v is stepped up to 5 volts to power the aux usb supply (5volts carried via the usb cable)

With the molex I think only the 5 volt supply is used. I know it will be used to supply the 5v aux supply (the enable pin on the dc to dc converter above is used to switch off the 5v generated from the 3.2v pcie bus). I am not sure if the molex 5v supply is used to power the nec chip, I suspect not.

One thought about the card that will not work. Is the clock live signal connected to the right xtal pcb pad. There will be two traces from where the old xtal is mounted to the nec chip. The correct connection usually has a series resistor on its way to the nec chip. If in doubt it should be ok to just try to solder to each pad in turn (this is what I did on the nos pcb to work out which xtal pad to inject the signal. If the nec chip has no clock signal the pc will not see the device.

Apologies if you already know some of the above. Hope it might help and best of luck.

Regards,

Nick.

Ps I lost another post yesterday.
Im reasonably sure now that sound quality is dependant on both absolute clock speed AND relative clock  speed. There are some other effects in play as well which ill come back on later but tuning the usb link clocks appears to have more variables that need to be understood. The good news is that this is about getting the last 20 to 30% of performance on top of the really significant improvement that just putting the clocks in provides. 
 

Paul,

Good to give it a try anyway but a distinctly "average trimmer". I hope it might go better when dexa no 2 is back in place.   My guess would also be that the gap in speed is too big to "match" well. Thanks for coming back with the results, it all adds to the picture of what might be going on 

Nick.

Hi Peter,

Here I have 3 dexas, two 24mhz for usb and one 24.5mhz for audio stream. No Paul Pang card used here, prior to the dexas two vcxo 24mhz diy built clocks were used to test the approach using transformer coupling before fitting the two 24mhz dexas when I was sure that a reference grade clocks were warranted.

Here trimming of the usb clocks is carried out on one of the 24mhz  dexas which happens to be at the pcie card end of the link. It can just as well be at the nos end dexa as the result in the same.

I'm not sure why Paul is not experiencing change but my initial guess would be, as you both mention, that the trim available on Pauls dexa is not sufficient to match pauls one daxa with his Paul's Paul Pang card.

Regards,
Nick.