Does Improving PC PSU and Reducing PC EMI/RFI Improve SQ?

[earflappin]

Roy, thanks for the reminder on the mobo.  The GIGABYTE GA-EG45M-UD2H is a micro ATX mobo just like my current Asus - it's the same form factor and equivalent functionality. 

My Zalman PC is hardwired connected to a wireless Apple Airport which sits about 10 feet away (it was too difficult to run a hardwire from my router).  I use the Toshiba netbook wireless to connect it.  It works flawlessly with no drop-outs.  As an aside, with WinXP Pro, I would get periodic drop outs in playback.  Since I switched to Win7 for both the Zalman and Toshiba I have not had any drop-outs. 

BTW, you're up a bit late (or early I should say) aren't you?   

[PeterSt]

He's late, but that compensates for me being early a bit.

But ... don't forget to try without the wireless one day. If you don't hear a difference it's okay. But I think you will ...

[earflappin]

Peter, just so I'm clear....hehe...you're saying there would be SQ improvement if I hardwire the network connection from my router to the Zalman versus going wireless from my router to an outboard (i.e. not inside the Zalman) wireless receiver (Apple Airport Express) and then hardwire from the Airport Express to the Zalman?  Are you also saying there's a further improvement if my netbook is hardwired as well?

BTW, I am continuing to love the Quad Arc Prediction mode.  I've read where you cite Quad Arc will sound its best if the DAC doesn't over sample the input.  Do I understand that ideally the DAC should not do any filtering of its own as well?  So basically, just take the unaltered incoming bit stream and do a D-A conversion?

[PeterSt]

Hi David - About the latter, yes, that is what it is made for (and how I use it myself).
More ideal could (!) be that the DAC provides (only !) the normal analogue filtering, but this is just theory for "unknown DACs" because it should adapt to the Arc Prediction, which obviously can't happen. Then *still* the analogue filtering may destroy, depending on the implementation and used parts etc. Notice that this analogue filtering has one reason to be there only : filter out the high frequencies (above 176.4/192) your amps may not be able to cope with (linearity). In the very end it is my estimate (!) that this is all a bit of blahblah, because those who can't bear NOS/Filterless are hard to find, if there at all (and this filters NOTHING and measures the worst).
On this matter, the Phasure NOS1 will not contain any filtering whatsoever - is NOS/Filterless for those who think that bad measurements sound good thus use no upsampling - but is the best out of everything when fed with Arc Prediction which measures as good as anything else but sounds better because nothing is destroyed.


About the wire(les)s ...
The Zalman shouldn't contain any wireless device in the first place, but I think you know that.
Digital sound shouldn't travel over wireless (but that is for you to try).
When the Netbook is wirelessly connected to the router, and is doing RDP to control the Zalman, no problem.

I now think (or see) that you use the latter, so no problem. Sorry for the before stirring (if only the Zalman produces the sound, and not the Netbook).
A small but though : the sole fact that the Zalman is providing the RDP connection is not the best for SQ again. I must say honestly I never tested this (for differences), but I sure stay away from using it, even while it would be by far the most convenient control. The difference would be (theoretically) similar to the switched off Services versus not.

But then I won't talk you in the "remote troubles" we actually all have in here.

I'm just thinking ... the very best might be to let XXHighEnd shut down the RDP connection during playing (doable) while getting it up again quickly when needed (dunno).

Peter

[earflappin]

Thanks for the insights Peter. 

I'm currently getting the best SQ from my system using Quad Arc Prediction (QARC) to my Berkeley Alpha DAC (BADA)with a digital fitler setting of 2.  This is the same filter that the Pacific Microsonics ADC/DAC Model One and Model Two use (these products are used by some the largest, best known recording studios to produce the digital recordings we listen to today).  I'm trying to see if the BADA can be set with NO digital filter through a firmare backdoor or a new firmware release.

When I A/B listening to XXHE with no upsampling and using the BADA filter 1 setting (this is the filter they recommend for best SQ) VS. QARC and the filter 2 setting the latter yields a more "fleshed out" sound with more detail, texture and dimensionality AND with less glare/ringing.  In this way it sounds much more "analog".

Regarding RDC I hear you!  There may be some negative impact on SQ to using it, but I just can't be bothered with a direct connected monitor, keyboard and mouse.  I'm going to A/B the SQ just for kicks.  If you could turn off the network and RDC during unattended mode that would be great.  I'm not sure if this can be done without having to reinitiate the RDC session.  I do think if RDC could be used without SQ degradation it is a no brainer.

[earflappin]

Regarding PSU impact on SQ I thought I'd share my experience with power cords and power conditioning units (PCU) - i.e. those class of components which take the mains power and attempt to improve the quality and regulation of power to those components plugged into it.

Over the years I've experimented with alot of different PCU's ranging from the early PS Audio PowerPlants, Shunyata Hydra's, Sound Application, Walker Audio Velocitor, Running Springs Audio, Environmental Potentials (these EP devices install in the electrical panel itself and are used in noisy industrial plants to feed power to CNC and robotic machines), and, most recently the PurePower APS.  Similarly, I tried a range of audiophile PC's and my own DIY cables.  I've also owned an array of amplifiers and pre-amps that used various types of PSU's including battery and tube.

Regarding PCU's, I inevitably found that while these devices reduced noise they imparted their own sonic signature which included a dulling of dynamics (likely due to current limitations).  Ultimately, I found the EP devices to be most effective with the least sonic impact (an added bonus is that the EP devices provide surge protection for the entire house). 

About 6 months ago a good friend and serious audiophile told me I needed to try the PurePower APS claiming it was the biggest breakthrough in SQ he had ever heard from a single component.  Yeah right I thought.  Anyway, I got one on trial - it took about 30 seconds to hear the transformation.  Significantly improved SQ - similar to what QARC does.  The PurePower APS is a regenerative power supply, but unlike the PS Audio design, the PurePower regenerates from a battery so the unit is "off the power grid".  And if there is any current limitations I can't hear them even with power hungry amps. 

Regarding power cords I've had a similar experience.  Generally I've found that they all improved some aspect of SQ while trading off others.  When you look at the design of these PC's alot of them have filtering networks.  With my PurePower APS I get the best SQ by using a simple DIY power cord comprised of 1 16ga solid core copper wire for hot, neutral and ground.  If I use audiophile PC's I hear a slight smearing or slowing of the sound.  I think this is due to their filtering networks.  Since the PurePower removes most/all of the electrical artifacts on the power grid there is simply no need for any filtering.

Regarding the PC, I have never liked switch mode power supplies and I remain anxious to learn about Peter's experience with trying battery power for the PCI card.  If I wanted to try battery power for my entire PC can somone recommend a commerical unit I could buy? 

Thanks!

[Fidelio]

If I wanted to try battery power for my entire PC can somone recommend a commerical unit I could buy? 

But the PC PSU would in this case still be switch-mode, only with a more stable 230V going into it.

For pure battery power, you would have to have multiple battery pack, one for each PSU rail you need. That would mean 3.3V, 5V and 12V in various configurations. However, even the most expensive Li-Ion or NiCad batteries are not linear regarding voltage, so the mainboard should have onboard power regulation similar to that on laptops, built especially for being powered from batteries. I think powering a desktop PC from batteries is too much of a hassle. If one is very adept at electronics, one could of course build a non-switch mode PSU, but that is another question.

[earflappin]

Thanks Fidelio.  That's the conclusion I was coming to as well.  For now I think I will leave it with powering my PC from my PurePower APS. 

I think the idea of trying battery power just for the PCI card remains the best first experiment and will eagerly await Peter's findings in this area.

[earflappin]

Peter, dang...I hate this obsession...err...I mean hobby....   

Anyway, I did play with RDC and there seems to be a slight, but audible, improvement in SQ when I shut down RDC.  I get most of the improvement if I select a playlist, run XXHE in unattended mode, and the log off my RDC session.  The issue right now is I can't restart RDC without a reboot of the music server once I've run XXHE (see my post under problems).  I'm reinstalling Win7 64 bit to see if my tweaks to the op system are the issue.

[PeterSt]

Peter, dang...I hate this obsession...err...I mean hobby....   

Well, sorry about that !

I too have been experiencing a lot with power regeneration, and at some time I ended up with a normal computer on-line UPS which I used for over 10 years.
Later I got myself a crazily inexpensive off line UPS (1800 euro or so, generating real sines), but again later it appeared to create problems with grounding, plus I never got excited from the sound.
Next I got another kind of problem with high frequency noise on the means, in the end coming from the loud speakers as well, and I got myself a couple of transformerless regenerators meant for 30w only. I think these were the only ones I ever experienced not harming the sound at all (but doing a great job on the mains noise and everything) ... until one broke, another one broke and a next one got fire. I won't name the brand, because although my main amps are rated under 30W and the small regenerators they should cope, they did not, and I guess a 30W amp is different from a 30W DAC. However ...

Because I just *had* to solve my mains-noise problems, I found a grounding scheme which is a 1000 times better (think dBs here) than what a regenerator ever can achieve. This is part of the DAC and proprietary and noise becomes unmeasurable because of it (unless the -160dB my measuring equipment shows is to be seen as reliable ). Notice on this matter, it is the DAC that creates the "mains problems", and not anything else, as per my own findings.

Currently I use nothing for mains protection etc., except shunt regelated PSUs, and I think it is the best.

To my findings, batteries are not the best at all, and those with the proper math can point that out (someone in here can do that, but I forgot who it is). This is speed related.
So ... to be clear on things, the "PSU's" I am going to let loose on the PCI bus will not be batteries of any kind. They (I've quite some to test) will be non-switching though, *if* it needs that in the first place ! Anyway, think more in the direction of a 100000uF "backup". I must say honestly though, that right at this moment I don't even know what or how to exactly test it (with electrical noise appearing at -160dB), apart from listening of course. One thing : measuring tells all (there was a time I was not of that leage) if you only know what (and how) to measure. And to keep in mind : I can measure the analogue out of the DAC software wise in details no measuring equipment can do. So I guess I am going to do that.

Btw, somewhere next week I'll be having the stuff here.
Peter

[AUDIODIDAKT]

I hope this will work,

To keep the soundcard out of the mobo, and give it its own psu.

Have you thought about integrate this in the DAC, or would you rather keep it outside. (boxed)
And maybe the integrate of word-clock (on both sides) would become easier this way. (as in bring PCI slot with soundcard into DAC)

If this all works for the better, anyway.
Really great someone capable of doing these kinds of measurings!

[PeterSt]

I hope this will work

What to say, if it already works now ? But if you mean teachnically (and not sonically), yes, that is to find out (but it just should !).

Have you thought about

I am twisting my mind on these things for over two years now, and right from the beginning (ok, after a couple of hundred pages of writing) I kind of knew what actually *is* the best solution. Like the matrix (I think) I talked about earlier. However, not everything is feaseable for normal money, and and getting it to you all ("all" is those with these kind of needs of course) is the big sport. You (Roy) have seen some of my half creations because they were just visible, but there is more. And there's also the most obvious like Firewire, which just is *not* going to work. Not in practice with all those half baked chip morons around it (by now I hope they read this). And in the end ? well, even that is not the best solution.

You will see ...

[PeterSt]

This is just for fun and

With my ever planned 32/384 DAC I planned a special connection to the PC which has never been done before. That is, not that I know of, and there isn't any reason for it also with other manufacturers;
Then, in this topic I expressed about wanting a general I2S connection. This I2S connection should work for PC's just the same. And as you may know, it is (AFAIK) the most jitter free connection;
I tried this with Firewire as the first carrier, but this is a dead end. I just can't find any Firewire boards with manufacturers that want to sell something, or otherwise the drivers s*ck and access to the SDKs etc. is (too) difficult.
So now I combined my old plan with this new DAC ...

I guess I again came back to this, because it is

- fun
- the best
- fresh
- the most obvious.

haha

[earflappin]

Peter, thanks for sharing your experience.  Glad you didn't burn yourself up experimenting....   

Eager to hear the results of the outboard PCI card experiment.  BTW, I tried using some EMI/RFI shielding inside my Zalman, including on my Lynx card, and couldn't hear any improvements in SQ. 

When will you be shipping your DAC and what will one need to buy for the PC side of the i2S interface.  Is it possible to use the standard PC HDMI connector with special driver code?

[PeterSt]

[most vague post ever which is a first brainstorm about why hires material is not right]

I still wanted to respond on this one :

I'm currently getting the best SQ from my system using Quad Arc Prediction (QARC) to my Berkeley Alpha DAC (BADA)with a digital fitler setting of 2.  This is the same filter that the Pacific Microsonics ADC/DAC Model One and Model Two use (these products are used by some the largest, best known recording studios to produce the digital recordings we listen to today).  I'm trying to see if the BADA can be set with NO digital filter through a firmare backdoor or a new firmware release.

When I A/B listening to XXHE with no upsampling and using the BADA filter 1 setting (this is the filter they recommend for best SQ) VS. QARC and the filter 2 setting the latter yields a more "fleshed out" sound with more detail, texture and dimensionality AND with less glare/ringing.  In this way it sounds much more "analog".

Well, I don't think the Alpha can do without any filtering !

I am not sure what to think from this setting of 2 and the theories behind that, but I have the hunch there is "something" important in there. So, notice this is a combination of "pre-arranging" (Arc Prediction) while something else (the filtering in the DAC) does not expect that to have happened. I may have said elsewhere that the filtering as such is hardly an explicit means to get the analogue wave right, but is merely a principle of "flattening" with the effect of digital (stepping) turning into analogue (rounded stepping). There's no other theory than "flattening" BUT the mathematics behind it highly anticipate on the ADC process and what actually goes wrong there, which is nothing else than a too low sample rate to make it analogue in the file, and with this knowledge of the "band limited" registration this math can turn that back to normal analogue. Read a (27 page) paper of Dan Lavry and you'll get the hunch of "this is Walhalla" indeed.

With the danger of me myself not even being able to understand what I just wrote, I am convinced that something else is wrong. I mean, the current means of this filtering s*cks all over in the first place and the people who should know do actually not (like D.L., your setting 1, name it), but I am afraid there is something going on beyond (!!) the red book means of thinking that can't work while again the knowledgeable are convinced it does.

Ok, this latter again :
There is a big difference between red book and its 22050 boundary, and say 96KHz material with its boundary of 48000;
The 22050 boundary *must* be created or otherwise we'll starve of harmonic distortion, with the result of that boundary being 20000 and (much) less because of the roll off the filtering "needs", while the 48000 boundary is okay by itself, but is dealt with the same way as red book for filtering. Or at least that is what I think.

Now, don't try to understand the above, but try to get the hunch of this :
At "decimating" or IOW going from the high resolution recording to anything else but 44100, I think there is an error applied. This is similar to what I just babbled about and it keeps on anticipating on the 22050 boundary. My thinking springs from knowing how the normal filtering works, and that actually the same filtering is used in the ADC. Not exactly the same, but anticipating on the same error applied when doing it the other way around, in the DAC. So, it is always to keep in mind that the normal filtering applied in the DAC anticipates on the knowledgde of how the ADC worked, and actually both are a combo.

What I am all heading for the the stupid (as seems) fact, that native high resolution material as offered to us does not sound as good as my own "faked" high resolution upsampling from 44.1. In order to see this, one has to listen through an NOS/Filterless (and I mean really filterless) DAC in the first place. Look :

The principle of Arc Prediction as how it works out net, 100% anticipates on no further filtering needed. Its principle would come down a 100% to native 88200 - or better 176400 material just not needing the upsampling, because it already has that resolution in the first place.
(thus, we feed both a 176400 Arc upsampled from 44100 and a native 176400 to the completely filterless DAC, and the upsampled sounds way better)
Now, how can it be that this native 88200 / 176400 through that NOS/Filterless DAC does NOT sound as good ? This can only be because it already contains some "filtering" in the material, which filtering is NOT there in 44100 material. It almost looks like the decimation process went via 44100 and is then upsampled by the wrong filtering means.

Of course the latter is not true for the explicit means of doing it, but the same as the normal filtering means are wrong and work out so indirectly, I guess similar is happening at decimating.

At some stage I created the decimating from 352800 to 176400, and the (too stupid) way I did it implied quantisation noise. I then aksed 2L (which I kind of worked together with on this) to create me a 176400 version of their 352800 recording so I could compare. They did (and of course applied one of the existing routines in their software), and I myself ended up with some commercial product doing the same, which commercial product also can do it the other way around (upsampling). Notice that the math doing this (sincx) is reversible ...

Now, coming to an end of the most vague post ever (sorry for that), it is my idea that at decimating, only when doing that to 44100 the area under 22050 remains untouched because everything and all tells it should (meaning : everything and all avoids filtering in the audible band as good as possible). Thus, when creating a 176400 file out of a natively recorded 352800 (DXD) the *resolution* is maintained, but the filtering is the same as ever, and while in effect this rolls off after 22050, its *means* are again the "flattening" I talked about earlier. In the eyes of formal filtering theories there is no reason not to do so, while forgetting the ringing implied, which is the one thing Arc Prediction does NOT.

If I am right on this, all currently produced hires material is lost forever already.
Peter