Investigating a full linear PSU for the XXHE PC and NOS1

[acg]

Hi All,

Over the coming weeks I intend investigating the effect of providing linear power to various parts of the XXHE PC and any associated changes in sound quality.

I plan for this experimentation to be methodical rather than haphazard but I am forced to approach this from a beginners perspective, rather than the perspective of someone that has electronic experience or even power supply experience.  So I propose is to break up the investigation into a number of parts:

• Define points of interest i.e. SATA, P24, P8, Fans, Peripherals
• Measure amperage requirements of points of interest.  Best to do this while the XXHE PC is in use
• Feed these individual points of interest with clean linear power and attempt to quantify any resulting changes to sound quality
• Determine which points of interest are worth providing linear power
• If feasible, design and build a high quality purpose built linear supply for the XXHE PC

So, first we figure out what to look at in regards of linear power and then we measure amperage draw, supply cheap linear power, make a judgement call, and then design a suitable supply if deemed appropriate.

To do this, I have invested in a relatively cheap, but decent quality 3 rail linear psu.  I have heard this psu in action in another system where the owner claims a very worthwhile 10% improvement in sound quality (whatever that really means), but of course how this translates to an XXHE/NOS1 arrangement is anybody's guess.  For what it is worth, ItemAudio in the UK also recommend this unit as their pick of the lab psu's that they have tried.

In the coming days and weeks I plan to measure the draw of all the points of interest that I can and publish the results in this thread.  It would be ideal if I could convince others to duplicate this approach (or another approach that can be agreed upon).  Hopefully this can become a group effort with corroborated results in different systems and different countries with an aim to really figuring out if a linear psu for the XXHE PC is a viable or worthwhile endeavour.

For the initial measuring tasks, all that is needed is a multimeter, a screwdriver, some extension cables for whatever we are to measure the amperage draw, and some time to spare.  I would guess that different computers will have different power draws at different points of interest, so the more people that spend some time measuring this stuff the better.

Within my own XXHE PC I have identified four points of interest: the P8 EPS 12 V plug; the P24 ATX plug; the Pang USB Card Molex Connector; the SATA power cable for the Samsung 840 EVO SSD.  The most complex and difficult of these is going to be the P24 ATX plug, but if we approach this in a methodical way I am sure that we can achieve the best outcome.  I have prepared a pinout of these four points of interest in my XXHE PC...
 
Another point of interest will be powering the RAM directly, which is a much more difficult task.  People such as I that use a RAMdisk for playback may get good benefit from clean linear power at this location, but I really think we should leave this until last because it involves modifications to the motherboard.

Other potential points of interest for linear power include multiple internal SSD and/or HDD and even external drives.  I do not use these things myself, so someone else is going to have to measure them and determine if they should be powered individually, as a group, or at all.

So please put up your hands and even if you cannot do these measurements yourself, feel free to make suggestions.  When my gear turns up in the coming days I will post some pictures of how to measure these things for yourself, it is quite easy.

Cheers,

Anthony

[pedal]

Wow, Anthony, this is a GREAT project by you. And I think your efforts will be rewarded. Power supply is "everything" in Hi-Fi and the general consensus is that cheap switching PSUs are bad in most (any?) stereo component.

I will follow this thread with great interest!

[PeterSt]

Great stuff Anthony.

One remark in advance of things : I would never save myself a 200 on a supply which may have a 10x worse ripple etc. then the SPSU in there right now, if that would make a difference in the first place (which I don't know).
So be careful please that you may be able to get all running all right, but that next the quality is not on par at all, and you thus won't know what you are listening to. This, while I think you are doing this for the better, right ?
But of course, if you are only doing this as a cheap trial just to see whether you can get it running, then it's okay !

Peter

[acg]

Thanks Peter,

The ripple on that supply is 1mV rms, so it is pretty good, but I don't know how well designed the supply is and how much hash it feeds back into the mains.  Yes, the $200 psu is a cheap trial just to see if I can notice differences with it.  The aim is to figure out what, if anything, works and then use what we have measured and heard to possibly design and build a specialised unit for the XXHE PC.

Is it an easy endeavour to figure out how much hash this linear psu feeds back into the mains?  Are you able to give me instructions?  I probably should get that out of the way first.

Pedal, thanks also for the encouragement.  I hope we find a way to further improve the XXHE/NOS1 sound!

Regards,

Anthony

EDIT:  Peter, I had considered going the super high-quality route from the start and using Hynes regs and good quality transformers, but the cost is prohibitive once the rails add up and I really don't know how many rails I will need in the end.  It could be one, or zero, or six.  Who knows?  At $200 a pop for each Hynes reg that is a lot of money to spend on a hunch.
 

[PeterSt]

Assuming you have a scope, just hook that up to your 230V, let it show the sine which should be there and compare with PC On and Off.

A bit careful (with the Voltage too ) because you maybe see so much rubbish in the first place that comparing is difficult. On the other hand, when this is so (much rubbish to start off with) it isn't going to help you much from that (mains) side either. But *that* does not count for everyone.
The other way around just the same : when yours is clean you may achieve results (because the SPSU is not clean at all), while I will not because I have the rubbish in the first place.

Isn't there then anything which is easy ?

Peter

[acg]

Thanks Peter.  I don't have a scope but I know someone who probably does.  When the LPSU arrives I will take it and my XXHE pc around to him and see how they compare.  

Anthony

[PeterSt]

Anthony,

Uhm ... You can do that of course, but maybe it is better to bring the scope to your place and in the mean time see how your mains behaves ...
Later you will be confronted with that anyway.

Peter

[AlainGr]

It is probable that many have seen this, but anyway...

http://www.teradak.com/en/product_view.asp?ID=214

contact: teradak@teradak.com

Seen on CA: http://www.computeraudiophile.com/f10-music-servers/teradak-computer-linear-power-supply-17689/#post260016

Alain

[juanpmar]

Alain, thanks for the link, it looks interesting. Have you contacted with TeraDak?

Br
Juan

[AlainGr]

Hi Juan,

Not really... I only conveyed the information that was on CA...

Regards,

Alain 

[acg]

It is probable that many have seen this, but anyway...

http://www.teradak.com/en/product_view.asp?ID=214

contact: teradak@teradak.com

Seen on CA: http://www.computeraudiophile.com/f10-music-servers/teradak-computer-linear-power-supply-17689/#post260016

Alain

Yes, I had seen that Alain and it is interesting.  The current on rails 1 - 3 seem very excessive and rails 4 and 5 will likely be no use (apparently -12v is seldom used these days).   To me the product looks like something that may not have received a lot of testing.

On another point, my three rail linear lab psu has arrived this morning and I am working on getting a suitable oscilloscope as well.  My plan is to measure and get a good idea of what the current situation is with my mains power and XXHE PC (ripple, noise etc.) and then measure the differences when using linear power.

Anthony

[acg]

Current draw for the P8 CPU Mobo Socket

Today I had a few minutes and decided to measure the current draw from the P8 CPU Socket on the Asrock X79 Extreme4-M motherboard and the Xeon E5-2648L 8 Core cpu.  This is not the standard cpu for the XXHE PC: it is a lower powered (1.80gHz) server cpu with more cores (8 physical) and more cache (20M).  As per Peters earlier advice I measured the current draw in both minimised o/s and non-minimised o/s modes; the results are below.

o/s Mode	Computer Startup	System Idle	Playing Music	Alt-X	Computer Shutdown
NOT-Minimised	3.78A	1.15A	3.3A	--	4.85A
Minimised	3.78A	0.9A	1.2A	1.64A	2.62A

I find these results very interesting.  The CPU draws much less current at shutdown and when playing music if the o/s is not minimised, and a little more at idle when the o/s is not minimised.  Startup current draw seemed to be the same whether the o/s was minimised or not.

What was super interesting was how the current draw of the cpu was impacted by the minimised o/s.  Before minimisation, the cpu voltage was a constant rollercoaster ride that had ran as low as about 1.2A but cycled rhythmically to 2.5A and even 3.3A occasionally, but there was definitely a cycle happening where the current would ebb and flow.  In minimised o/s mode, the cpu was much, much more stable and operated in an amperage range of 1.05A to 1.2A with no real rhythm apparent.  That is a swing range of 0.15A compared to 2.1A which in the context of things is a very big difference.

Is there anybody out there that can duplicate this testing with the Asrock X79 Xtreme4-M mobo and the i7-3930K cpu?  Not only would the difference between processors be interesting, but the information is valuable should a linear psu prove worthwhile.  I will write another post shortly to summarize how to measure the power draw at this socket.

Anthony

[acg]

#1 of 2 ways to measure the current draw at the P8 CPU Socket on the Mobo

If you have a lab supply that displays the voltage and current draw, this is very easy.

STEP 1 - Make a loom for the P8 socket.  I purchased a 8 pin EPS extension cable (such as this) and cut one end off it (the end that does NOT plug into the mobo), twisted all the positive wires together and fed them into a banana plug, then twisted all the negative wires together and fed them into a different banana plug.  If you already had the 8 pin EPS plug and pins you could make up a cable yourself in about 5 minutes.  Apparently you can also feed the 8 pin P8 socket with a 4 pin P4 connector: you simply install it in the half of the 8 pin socket that is nearest the edge of the mobo.

STEP 2 - Turn on the lab psu and set the rail you are going to use to 12V and plenty of amps (minimum 5A, preferable 8A+)

STEP 3 - Remove the existing psu plug from the P8 socket on the motherboard and replace it with the new cable.  Plug the new cable into the 12V rail on the lab PSU.

STEP 4 - Turn on the lab psu.  You are now using two psu's for your motherboard.  For the computer to start the P8 cpu socket must be receiving its 12V feed before you press the start button on the computer.

STEP 5 - Press the start button on the computer.  The computer should start normally.

STEP 6 - Take note of the maximum amperage draw displayed on the lab psu at startup (i.e. from the time you press the power button to the time XXHE is loaded), at idle (i.e. when the computer is sitting there with XXHE open, but is not playing music or doing any other tasks), playing music, pressing Alt-X to come out of unattended mode, and shutting down the computer.  This should be done multiple times with both minimised o/s and not minimised o/s

STEP 7 - Post the results in this thread

 

[acg]

#2 of 2 ways to measure the current draw at the P8 CPU Socket on the Mobo

If you DON'T have a lab supply that displays the voltage and current draw, but you do have a multimeter this is still very easy.

STEP 1 - Unplug your XXHE computer from the wall.  Remove the 8 pin plug from the P8 CPU socket on the motherboard.  Plug in the 8 pin EPS extension cable (such as this) and cut all the POSITIVE wires in half (the yellow ones, not the black ones).  With the yellow wires that are connected to the compute PSU, twist them together and screw them into a terminal strip so that all four wires go into the same socket of the terminal strip.

STEP 2 - On the opposite side of the terminal strip screw in the positive probe of the multimeter.

STEP 3 - Screw the negative probe of the multimeter into a different socket of the terminal strip.

STEP 4 - Twist the remaining four yellow wires (that you cut earlier) into one group and screw them into the terminal strip opposite the negative probe of the multimeter.

STEP 5 - Plug the free end of the extension cable, with the multimeter wired into the middle, into the P8 socket of the motherboard.

STEP 6 - Make sure you don't have any potential shorts anywhere and then plug the computer power supply back into the wall and press the start button.  The computer should start normally.

STEP 7 - Take note of the maximum amperage draw displayed on the multimeter at startup (i.e. from the time you press the power button to the time XXHE is loaded), at idle (i.e. when the computer is sitting there with XXHE open, but is not playing music or doing any other tasks), playing music, pressing Alt-X to come out of unattended mode, and shutting down the computer.  This should be done multiple times with both minimised o/s and not minimised o/s

STEP 8 - Post the results in this thread


Note that in the photos below I have also directed the black wires (ground) through the terminal strip.  This is not necessary, they can be left entire and uncut.

[PeterSt]

Super stuff Anthony !

Btw, when you have all still connected, you might consider Alt-x in Normal OS Mode just the same. Just do not tick all the "shut down" buttons and use Unattended as usual (oh, you can use the "shut down" stuff just the same but I'm not sure which comparison would be the most fair).
And of course, is using Alt-x in Normal OS Mode in the game anyway ? maybe not.

FYI : I can tell (by guarantee) that your measured 1.2A during playback can be pulled down to 0.5A. This is a bit unsafe though, so count 0.72A (this is processor frequency related which can be brought down more). Notice that I talk about super-stable current draw and not about your measurend 0.15A "swing range". For that matter, my 0.5A is related to the upper side of your swing, while my "super stable" setting would use the under side (so, roughly 0.49A now - didn't do the math).
I am not sure whether this can count for booting, although I do know that a certain BIOS setting would let it stick to 1.2A (not my 0.49A !). But this setting is not the best for SQ, so actually we can't use that.

I wonder, can't we make a comparison with the 3930 by means of "deduction" or how to name it ? So, you have some (one only) type of hdd in there for the OS (preferrably a 2.5" because I have that) and now you measure what happens at booting for the total draw (like you did previously actually). I do the same, and it should be higher. The difference should be caused by the CPU, so we know what the 3930 does. Same for the other measurements.
Or would that be too rough ?

Thanks mate,
Peter