Hi,

My Adnaco system recently returned from a holiday in Holland (thanks for looking after it Peter ).

I intended to use it only as a diagnostic tool to use whilst understanding electrical noise within the PC DAC and AMP. This was because when I first bought the system more than a year ago to try out electrical isolation of the PC from DAC, the system showed promise but ultimatly could not exceed the sound quality of the normal USB connection.

(The findings from a year ago are posted here
http://www.phasure.com/index.php?topic=2089.msg21556#msg21556)

So fast forwards to working with Anadco now. First thing is that in the intervening time frame the PC OS, XXHE software and my PC have all very significantly changed. On reconnecting the Anadco SB3 system the slightly darkened (though very pure sound and totally without "hash") sound was present again. However, in my current system the sound using the Adnaco is much better than I remembered and so I decided to spend a little time understanding the system to see if there were any settings or electrical changes that would make it perform. Well it turned out that this can be done and by a couple of relatively simple changes to the Anadco USB board. PC Bios, and XX Settings have been all over the last few days but its coming together now.

This will need other ears to verify what I am hearing and there are still a couple of points to sort out but I am really thinking that with the modification Anadco is making a significant positive contribution to sound quality.

Paul hi,  I think we have yet another configuration to test next time we meet !

Regards,

Nick.

PS Alan I think you may have bought an Anadco system at the same time I did, was yours an S3B (USB) system ?

Paul hi,

Comparing the PCs again would be fun. I still believe your X58 Asus setup to be a real sweet spot setup. I am really pleased with sound quality here at last though and I hope that our PCs will be much more comparable than they were when we tried this with my Asus X79 board !


Best,

Nick.

Brian hi,

There is always a desire to share experiences but I'm always worried by doing so it may encouage others may invest and the outcome is not so nice. So I'm really pleased and more than a little relieved that you have found the ASRock board to be so good. After a month or so of using the board I'm still really enjoying it as well 

Regarding settings, I don't understand quite why but when I came back of holiday I spent more time on the BIOS settings and there are slightly different now, so I'm not sure that those posted may give the best possible results. The settings that do seem to help are clock ratio at 24, memory frequency at 800mhz and memory voltage up a little to 1.6 to 1.65v. I'll post more if I get to a definite set of values.

Anyway really please again to hear that it's gone so well.


Mani hi,


On my ASUS I tried mobo and PCIe USB ports and the settings that you have above re ps/2 connections and USB turned off. The settings also helped to give the best sound from the ASUS board here as well. Even with these settings applied to the ASUS the ASRock was I very big step forwards at least in my system. It should be fun to swap out the board this week and take a listen. Fingers crossed here that you get a positive result.

Regards,

Nick.

Mani hi,

I think there is some real progress here in getting to an understanding PC generated noise. The debate is really helping to draw together snippets of information and the wider picture is coming into view.

Thanks for posting your measurements of Case 1, 2, and 3 tests.

Was your higher value of 22mv for case 2 measured on you Asus mobo PC system by any chance ?

Interesting about the Entopias cable having a break in the shield ground. I had also tried lifting the USB shield connection at the NOS end some time ago when it became clear that there is coupling within the NOS of PE and signal grounds. I was in mail discussion with Peter and Paul about it, but we all seemed to get slightly different results so I did not post at the time. Now I tend to move back and fourth between lifted and connected shield at the NOS but generally I also experience a cleaner top end with less hash when the shield is lifted which seems consistent with your findings.

The light that has gone on for me is that the USB transmitted noise is not a DC offset or even audio frequency. For these noise types we could expect connections to PE to drain the noise resonably effectively. The noise is however in the MHz range touching the 100s MHz range which means that although it can be attenuated to some extent by grounding at the NOS (about 3db improvment in my case, see our case 2 and case 3 measurements), the impedance of the ground is too high at these frequencies to reduce the noise level with any great effectiveness. I agree with you about your connections 2 and 3, and your measurements really show their effect very well. At these noise frequencies even these low resistance grounds can only help reduce and not eliminate the noise, but there is good correlation between managing the noise down and improved sound !

Of course the SQ changes could still be due to better ground referencing at lower frequencies (audio down to DC) as you mention, so this must be considered. Looking at the noise in the scope trace though, you can really see how jitter in the DAC and resends of USB data could be caused, this would defiantly leading to the hash and hardness in SQ.

Seeing the the problem in this way is leading to all sorts of thoughs on what to test next and how it might be managed which is great. I'v got stuff in mind to look at over the weekend and will post back but I'm wondering now if this could really lead to something.

The USB shield is a supper kick off point. Understanding what is going off on the USB d+ d- and signal earth is next to help understand the overall connection (we should be able to forget the +5 volt USB wire as its not needed by the NOS).

Best,

Nick.

Mani hi,

I just read this again and your question about why there is 6mv over a 1m shield and would this meant that you get 36mv from your basement PC with longer USB lead.

It just clicked looking at the trace I posted above. This is not a steady state offset we are looking at across the cable (or even audio frequency). If this were just DC offset then the resistance (proportional to the length of the shield) would determine the voltage drop. What we are seeing is a ver high frequency noise profile (a wave length of 1 horizontal division in the trace is a 2mhz frequency and as you see the noise is much higher frequency than that). What our volt meters are measuring is the rms value of the noise which in my case is 10mv with out the USB shield connection between the pc and nos1 and 6mv with the USB connection. The USB connection dose allow the NOS1 to effectively "pull down" the amplitude of the noise by connecting it to PE but the impedance of this route to PE is not low enough (being meters long) to completely drain the noise away to the actual protective earth. However your additional connection 3 probably lowers the impedance to PE and helps to reduce the noise amplitude. If you want to see what causes that hard top end sound, I think that scope trace could be the culprit.

If I am right your other PC in the cellar will have a similar USB shield voltage (provided that the ATx psu and mobo produce the same offset (as the PC you have measured) above chassis ground when the PC is booted.

I think things are starting to make more sense.

Nick.

EDIT - a little more thinking and I think this now links into Paul's experience using a star PE setup. Im thinking that just as your additional connection "3" may reduce impedance to ground for this noise, Paul's star earth is possibly doing the same thing at the PC and generally in the NOS1 ?

What would happen if a choke was placed in line with the shield connection between the PC and NOS could the noise be filtered ??? Of course this may help with the noise transmitted over the USB shield but since the PC's signal earth, is linked to the chassis earth, this noise is likely also to be in the USB signal ground wire and possibly the d+ and d- data USB wires. Still it should be easy enough to check these other connections for noise.

Hi,

A few notes from measuring round NOS1 and PC chassis ground voltages measured at the USB shield connection points of the PC and NOS1. I currently have standard PE arangemnts (eg no star PE eathing of NOS PC AMPs etc).

Measurements are taken at:
a) USB shield connection at the PC.
b) USB shield connection at the NOS.

Both of these points have low resistance connection to the PC PE connection and the NOS PE connection respectively.

Case 1
With PC turned off, NOS1 turned off, NO USB lead connection between PC and NOS1.
Voltage a) to b) = 0mv.

As expected 0v as both chassis grounds are connected to PE via the mains leads to the components.

Case 2
PC turned on and booted, NOS1 turned off, NO USB lead connection between PC and NOS1
Voltage a) to b) = 10.6mv (eg PC is is now 10.6mv higher than NOS1 PE).

Almost all PCs that I have owned elevate their chassis ground potential with respect to mains PE when the PC is booted.

The oscilloscope trace below is taken between point a) and b) under Case 2 test conditions. This gives some idea of the level of noise that is set up between the PC and NOS1 PE / chassis grounds when the PC is booted.

Mani mentioned being able to hear mouse movement through his speakers. I found that moving the mouse did visibly excite the noise shown on the scope trace, particualy the spikes above the main noise level (in real time these spikes appear relatively random and frequent).


Case 3
PC turned on and booted, NOS1 turned off, USB lead now connecting PC and NOS1.

Voltage 6.4mv (eg PC is is now 6.4mv higher than NOS1 PE).
The NOS Chassis ground connection to the PC via the USB shield is pulling down the PC chassis voltage from 10.6mv to this new value of 6/4mv.

Thoughts.

The PC sets up noise on its chassis ground when it boots and elevates its chassis ground by an average of 10mv but a detailed view shows the type noise that is present below. This must set up a ground loop with some complex noise current patterns circulating. The shield of the USB cable in actually connected to NOS signal earth via a 104kR and 4.7nf capacitor which form a parrallel connection between chassis ground and signal ground (this seems to be normal practice in USB interfaces not just for music devices). This means that the noise that is in the trace generated by the PC can in theory transfer in to the NOS signal ground (I must take some more traces to see it this is actually happening).

Mani’s connection “3” may be presenting a low impedance route for this noise (in the trace below and the associated loop currents) to be lead back to PE before it transfers into the NOS Chassis ground / NOS signal ground.
As Coen points out there is the possibility of EMI pick up in the PE wiring loops which but I could not really guess the effects of these.

Hope this helps.


Nick.

Paul hi,
Im really interested in this the star earth seems very encouraging, since you posted on this I have been working through how to put a star earth back to my systems dedicated PE spikes as well. I’m just placing the order for wire and sockets etc today and I’m hoping for good things when they go in.

I’v been poking around with my ‘scope and logic analyser for a while now. There is MUCH more happening than I expected with respect PC to DAC to AMP noise. I am still looking for the flash of inspiration that will help to understand and make sense of what I am seeing but thoughts right now are.

In the audio frequency range noise on the USB connections and measured at the output of the DAC is low. However wide bandwidth measurements (up to 200mhz) at the same test points show SMPS and general noise (often with characteristic SMPS and other frequencies) at these test points in the 10mv+ range. Given that my amp inputs generate audible sound from input signal levels of around 100uv this is stuff to worry about. Interestingly these noise measurements improved when I changed my mobo recently and sound quality also improved significantly.

What I am seeing is showing that the electrical noise from the PC is carried both via the USB connection the DAC and also via the mains and the mains PE connections to the DAC and Amps. In addition the DACs own USB interface creates is own noise signature at the DAC outputs (eg the I2S transmission noise I posted on and other and characteristic frequencies from oscillators etc on the board).

Evidence of mains and PE borne noise can be found by measuring at the DAC outputs with only the PC only switched on (eg DAC turned off). It is easily possible to see SMPS and other elements of PC noise at the DAC outputs. Even with the USB lead disconnected there is still noise on the DAC outputs which can only be transmitted via the mains wiring (or radiated) into the DAC from the PC. In fact when only the ATX PC supply is switched on and without the PC booted SMPS noise can be seen on the DAC outputs and with the PC booted disk access noise etc can be seen !

Frankly it’s a little depressing seeing what we are up against and I am struggling to decide what noise is relevant and how to manage all this high bandwidth noise out of the system (I don’t have experience filtering SMPS noise and it seems to propagate everywhere). Perhaps the best approach would be to use the PC only to transfer processed Data to the DAC interface and have an FPGA time the data out of interface memory and into the DAC chips. This would allow for total disconnection of the USB and PC whilst music is played. I guess even wireless transfer to the DAC interface of some sort might allow the PC and DAC to be electrically disconnected from one another.

Going back to the star earth and why it works so well, I am thinking that is could be to do with better referencing of the signal grounds to a common potential within the systems components but also it may be that in some way the star earth is helping manage the USB and Mains wiring transmitted noise between components.

I am looking forward to trying the earth and will try to get some comparative measurements with and without the star earth in place.

Cheers,

Nick.

Paul hi,

Absolutely top stuff ! This sounds really interesting for my system too given the similarities  I'd like to sit down and draw out what you are describing to see if its possible to work out why the noise level is improved.

Don't forget the NOS1 internal signal ground IS connected to PE. The connection is via a 4.7nf cap and 110k ohm (EDIT just checked the resistor is 104k ohm) resistor (R and C parallel with one another) on the NOS1 USB board which connect the NOS signal ground via the USB lead's shield (I must also check the USB signal earth as well) back to PE in the PC.

You really have me thinking haha.

Nick.

Ps what gauge of wire did you use for the additional chassis earths to the NOS1 and the amp, how are they connected to the chassis and have you lifted the main plug earth now that you have the star earth arrangement ?

Peter hi,

I "think" this maybe what I may be finding as well. When xx starts time stability says something like unknown (I am travelling at the moment so cannot check exactly). Click "refresh" and it goes to stable. Could be placebo but I think sound improves after hitting refresh.

Nick.

Brian hi,

That is some praise for the sound of your system if your better half is an audiophobe. Never mind drugs, play some music 

I'm pleased the new board is working well. I also had to drop 3db or so from listening levels which I took to be a good thing. It seems like the noise floor is lower with much higher dynamics which brings the perceived loudness back up to asus board levels.

I'm in Chamonix on hols right now so this is from memory. When setting up the asock I found the main sweet spot parameters were the cpu ratio setting memory speed, memory timing and surprisingly memory and cpu voltage.

I have not totally finished playing but it would be good if you could try this and let me know what you think.

Cpu ratio at 24 (a change of +-1 here collapses or sharpens the sound)

Cpu clock frequency at 103

Take the first 5 memory timming settings off auto and and use the speeds  listed on the dimms (this will be timming for the max rated speed of the dimms but since we are setting memory frequency to 800mhz it is slower timings than with auto set. In my case I think the settings are 11, 11, 11, 11 and 30).

EDIT I should also have said that I find running only 2 (not four) memory dimms is important. You can just take two out the mobo sorts it all out automatically.

Then most importantly up the memory voltage by 15% (here that means 1.65 v)

Last put cpu voltage up by 10 to 15 %.

For me this lot gives a significant sweet spot. As I say its still a work in progress. The board took a week or two to run in but worth taking ten mins to try these settings they do really help my system.

Cheers,

Nick.

Peter,

 Thanks for confirming,  must be a placebo effect.

Cheers,

Nick

Arvind hi,

I had the same happening here a while ago. For me it was fixed by going from minimised OS to normal OS and back to minimised again. After doing this time stability and time performance index toggled fine including XX forcing a reboot.

Might be worth a try.

Cheers,

Nick

Ps I think that I need to click the XX Tweaks "refresh" button each time I play music to obtain best sound. I had meant to ask Peter if this is likley to be placebo or if there could be some logic in this.

Anthony hi,

Got it, your Xenon has 8 physical cores, 16 threads and 22mb of cache, it must have been too late for me last night when I read your first post. That's interesting, Peter would know if the extra threads might help XX directly or indirectly by servicing background OS calls  better, but i'm guessing the larger cache might be useful regardless of the effect of the number of threads.

Nick

Anthony hi,

Im guessing that the core (thread) count is not too important ( 8 v 12 ) I may try limiting the active cores here to see what happens. It seems that underlocking is a good way to go anyway so a slower processor speed may actually help 

I'v been playing with a lot of BIOS settings trying stuff out since getting the mobo. Pulling out a couple RAM DIMMs (two fitted now), slowing down RAM frequency and timings and lowering the CPU ratio is producing interesting results. Its looking like there is quite a sweet spot to be had with these settings.

Regards,

Nick.

Brian hi,

Haha i wondered how long you would last, I really don't think you will regret it  It turns out the board is a tweakers delight. It sounded good the moment it went in compared to the ASUS board but after 3 days of setting it up the sound is exceeding expectations. I'll PM you some settings to try 

Peter mentioned in another post setting up a thread for XXPC tweaks. I think that would be great idea.

Cheers,

Nick.