Thanks Nick, this is helpfull!

The signal loss would be due to the dc component saturating the transformer core.

Next to providing galvanic isolation, which the capacitors allready can do if you would add one on the other leg, I think the benefit of the transformer is that it acts as a band pass filter. by its nature it removes the steepness of the signal. This leaves a signal with less unneccesary hf energy that would be detrimental to the sound.

Regards, Coen

I just modded my paul pang card a little and ran into an odd situation.

What i at first did was connecting the secondaries of the hf trafo to the PCB x-tal and PCB gnd at the removed capacitors respectively.
The result: no card detected.

This got me thinking: the original capacitor on the secondary could be necessary in the end to block a dc component at the NEC chip side.

Reconnecting the capacitor made the card work again. So that makes me wonder how the DEXA clocks work. They must have some capacitive decoupling on the pcb to avoid this situation. Or maybe this only holds true for the way the pp usb3 card is working.
Nick, how did you connect you previous clock experiments clock signal to the usb chip?

Anyway, the sound has more bass and naturalness than I can rember. It is also a bit too laid back and soft for my taste. This could just be the clock difference thing here.
Unfortunately i have no heavenly trimmer available.

Now lets see what the one clock thing brings. Suggestions how to get this working with a minimum of effort are welcome. (Does the NOS1 usb oscillator also have a dc component?)

Regards,  Coen

Why not forward engeneer? Linear ADC has been done for ages.  Silicon is much much more capable than ten years ago. Maybe we can do true 20 bit acurracy in a real ladder adc at 384 KHz. Caveat: you will probably have to design a soc to meet to the specs. This is very high speed ultra precision stuff. Now that still isn't well suited for experiments to say the least.

Technically the delta sigma dacs will still run circles around it, are they really that bad? Theory says that when properly bandpass filtered and decimated, there will be no difference in sampled signal. The delta sigma noise is gone, in either case the signal has to be filtered (hence ringing if done steep) to avoid aliasing.

It would be usefull to have a true ladder adc'd signal to assess the true merits of the ladder concept. Anyway it is my optionion that the ladder has the edge in practical playback, where (multibit) delta sigma the edge has in adc.

dSD/DXD is a different story though, now that would be interesting. Jitter issues deluxe, but no mathematical conversion necessary only filtering...

Lets move on!

Regards,  coen

Paul,

You are probably violating the startup requirements with a running clock on the card before power-up. This could be either a bios or a pcie card startup problem. My guess is the latter.

I would suggest you use a card that works for others. Nick?

Regards, Coen

The latest nec chip runs on a very low voltage (like 1.3V or something). Like a pc motherboard this voltage is made from probably the 3.3v by a switching regulator but I did not measure its source. The switching regultor is on the pp card on the right of the nec chip.

Pci-e has no native 5V connection, so that has to be made on the board or by an external supply. In case of the pp card this is the only function of the molex. The voltage is filtered by an rf pi-filter with a "big" rf choke on the top part of the board.

The card designer has two voltages available on the pci bus to make the 5V voltage from: 3.3 and 12V. I guess the 12V will be preferred for the 5V USB power if no external supply is available, at least that makes the most sense to me fom an efficiency and power availability perspective.

Anyway you for shure don't need the extra molex for the 12v and in case of a molex, likely the 12v is not used at all. Too bad you can't prevent it from entering the card since its also on the bus .

Regards, Coen

I recently read about a genius idea for clock distribution that I like to share.
Credit is due to the guy from ecdesigns who has developed a 44.1/16 SD card reference player.

The principle is that splitting a clock signal and distributing over a distance it will degrade the clock quality and introduce noise. The strikingly simple answer is to filter the incoming distribited signal with an chrystal filter, which has a very narrow passband letting only the original signal trough. In his wording this filter acts as a highly specific "tuning fork" requiring little energy to resonate and as such providing for a light load tot the master clock generator. One can use this principle for any clock distribution ánd/ór transport.

How about that!

Regards, Coen

Paul,

The jitter won't be the same and they have both different tempraturesensitivity (and different temperatures during the year).

But it is closer to ideal than having an unoptimised clock sync.

Regards, Coen

Hi Anthony,

Imho this is serious exitement on Peter's part.

I am the one that did 1+1=2, Peter is the genius who knows like no other what its implications are.

But I think we're catching up

Regards, Coen

P.s. Enjoy your "popcorn" !

This description makes sense. The offset increases (hf?) distortions on one end and filters the sound on the other. It makes one wonder how complex this gets when serious jitter comes into play....

Regards, Coen

Great work Nick,

Now that was a superfast practical experiment!

If you hear such differences in sound whilst only offsetting the clock a littlebit, it makes sense that we might be lucky or lost either way with standard "large-but-sufficient" tolerance clocks. This also means that our milages will vary and we could have quite different 'objective' sonic experiences with the same hardware...or different hardware....

Please elaborate on your one clock solution when you get it done. I gather you have to arrange for some kind of clock distribution that is least detrimental to the clock signal synchronicity.

Regards, Coen

Ps you beat me on this one!

 


Regards, Coen

Just a thought:

You can reduce the number of clock variables in the system by slaving one of the USB ends to another. Iow like slave the PC to the NOS clock. Wether this one clock approach works out for the better remains to be seen. This seems to work for recording studios, but then again that is a true synchronous signal. Anyway the low oscillation will not be there.

Regards, Coen

Thanks for the elaboration Peter.

The clock experiments hold a key to better understanding what makes good sound in a computer audio setup. Since it seems that many influences are somehow powersupply related it raises also the question what the major contributor is: the precision clock or the PSU (as far as they can be seperated influences).

I am pleased to read that there are healthy theories popping up that can be scrutinised on a software level. This also promises great progress still to come. We are not near the point that we Háve to resort to hardware upgrades at all. Still these enable you the enjoy today the benefits that are yet to come .
Anyway this process also relies on the bold experiments of our most commited forumfellows, so keep up the good work and enjoy your results!

Regards, Coen

Hi Paul,

Apparently there is great merit in upgrading the clock on both ends of the USB link. I feel that the XX software based compensation of clock imperfection will have its limits and that the last mile only can be addressed with hardware.

Anyway I wondered how you connected the PSU mains and where you placed the PSU. Also the Neutronstar board seems quite large. Does it come with mounting stuts?

regards, Coen

That's more like it!

I can't see an orderable 24MHz clock though....


Regards, Coen