Investigating a full linear PSU for the XXHE PC and NOS1

[acg]

Super stuff Anthony !

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 ?

Peter, I think that it would be best to directly measure the cpu socket with the i7-3930 installed because this is where the cpu draws all of its power (I think all).  You would have some sort of lab psu there...maybe fork out for a 4 pin or 8 pin extension cable or even just buy a 8 pin male plug, some EPS pins and make your own extension cable.  It is a bit fiddly but this is essential information.

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.

Very interesting stuff Peter...looks like there is more to be gained here.  It certainly looks like the size of the 12V rail for the CPU will be determined by the shutdown current draw which is at least 5A on my system and probably more on yours.

Regards,

Anthony

[PeterSt]

Peter, I think that it would be best to directly measure the cpu socket with the i7-3930 installed

Of course. And you did make this nice layout of it all for a reason obviously. But before I am the one to do it ... we are all quite busy with completely different things, and I don't see this going in between the jobs.

So just the question : Is someone going to do this for us ?

Peter

[acg]

P24 ATX Socket Pins Required by Motherboard

This afternoon I played around with the P24 socket which is the main socket on the mobo that carries the current to run the board, fans, USB ports, PCIE ports et cetera.  Because all motherboards seem to be selective in which parts of the ATX specification they choose to conform with I decided that I would try to figure out some rough rules for the power rails (+3.3V, +5V, +12V, -12V) and associated ATX pins.  What I did was get a P24 extension lead (like this one) and basically push out one pin at a time, startup (or attempt to startup) the XXHE PC and play some music.  The result is summarised in the graphic attached to this post.

Basically, the Asrock X79 Extreme4-M motherboard does not use the -12V rail, the 'Power Good' pin or the +3.3v sense pin.  Each of the +3.3V/+5V/+12V rails are represented by multiple pins in this socket, but the motherboard needs at least one pin (any pin) representing each rail to be powered.  It does not matter which pin this is just so long as at least one pin from each rail is powered.

What does this mean?  It tells me that even though each pin for a specific voltage is represented in this socket a number of times, they all come into one on the motherboard so there is probably little to be gained by even trying to power each pin from a separate rail.  Therefore I propose to measure three separate rails (+3.3V, +5V, +12V) into this socket by joining the wires of each of these rails into one and then running them from my lab psu.  

The '3.3V Sense' pin (#13) is used to allow the power supply to account for excessive cable drops on the +3.3v rail.  I assume that this means voltage loss due to cable resistance and the associated +/- 5% DC voltage regulation for the rail.  If this is the case I think that this can be overcome by making sure we have 3.3v at the P24 ATX socket on the motherboard and not just at the power supply.  Someone please correct me if I am wrong.

I also hope to be able to find an easy way to convince the motherboard to boot by using the +5v Standby pin and the 'Power On' pin.  The +5v Standby pin needs to be active whenever AC power is present and provides a power source for circuits that must remain operational when the computer is turned off (enables the power button on the computer case to work among other things).  The 'Power On' pin could be the tricky one, but I have a few idea on how to get it functional which I will try out when it comes to that.

Anthony

[PeterSt]

Anthony,

I really don't know anything about MoBo layouts (but have a hunch that starts to happen from of now), but those several same Voltage pins possibly allow to spread the current over different traces on the PCB. So, if you'd now use one connection only, something may get hot somewhere ?

What would be allowed though, is using the one single source for the e.g. 3 connections. Of course that one source now must be rated higher (A) then 3 singles sources. And yes, you have one lab PSU only, so for you it is not so obvious maybe. But look at what comes from your PC PSU and how the 3x etc. is organised inside (I am almost sure that at least the 12V is spread over to separate sources - and not just a Y cable split).

Again, I don't know, but take it into consideration please.
Peter

[PeterSt]

Hey Anthony,

Be careful not losing a possible new post because I am going to move this topic to a new board somewhere.

I will do this 15 minutes from now ...
Peter

[acg]

Anthony,

I really don't know anything about MoBo layouts (but have a hunch that starts to happen from of now), but those several same Voltage pins possibly allow to spread the current over different traces on the PCB. So, if you'd now use one connection only, something may get hot somewhere ?

Yes, that was what I was thinking.  The fact that any pin on a rail is enough to power everything indicates that they have been split for a reason, heat is probably it.

What would be allowed though, is using the one single source for the e.g. 3 connections. Of course that one source now must be rated higher (A) then 3 singles sources. And yes, you have one lab PSU only, so for you it is not so obvious maybe. But look at what comes from your PC PSU and how the 3x etc. is organised inside (I am almost sure that at least the 12V is spread over to separate sources - and not just a Y cable split).

I will look into this.

Thanks,

Anthony

[Nick]

Anthony hi,

Another consideration is the sequencing of the power rails during turning on. There mobo will expect a "power good" signal to be provided after the rails are sequenced on.

I had a look at options for building a linear supply a while ago but did not get further than reading arround what the supply would need to do.

Take a look here the "power good" information is half way down the page.

Cheers Nick.

http://www.tomshardware.com/reviews/power-supply-specifications-atx-reference,3061-2.html

[acg]

Hi Nick,

Yes, the 'Power On' may be a difficulty, but see my next post for what I have measured with this rail.

Cheers,

Anthony

[acg]

P24 ATX SOCKET CURRENT DRAW MEASUREMENTS

So last night I got out my trusty multimeter and hacked into the P24 Extension cable that goes between the motherboard and the P24 plug of the BeQuiet SMPS.  My aim was to figure out the current draw of the various rails and ancillary pins.


12V Rail

There are two 12V pins on the P24 socket each of which will successfully power up the mobo without the other being connected.  I cut both wires, twisted the mobo end of the cut together, removed one supply wire from the supply side so that only one 12V wire was coming from the SMPS and made the measurements below...

o/s State	Startup	Music	Shut-Down
Maximised (no cpu fan)	0.83A	0.48A	0.56A
Maximised (cpu fan plugged in)	0.93A	0.59A	0.85A
Minimised (no cpu fan)	0.81A	0.57A	0.81A
Minimised (cpu fan plugged in)	0.92A	0.57A	0.81A

The 12V rail powers the cpu fan in the XXHE PC so I though it prudent to measure with the fan connected as well as disconnected.  The amperage varied when I pressed a key or moved the mouse, so perhaps the PS2 ports are run from this rail also.  As you can see the fan does not make much of a difference to the power draw on this rail, neither does minimising the o/s.



5V Rail

There are 5 5V pins on the P24 socket and each pin alone will power up the mobo if all the other 5V pins are not powered.  I cut all the 5v wires, twisted the mobo end of the cuts together into one joint, remove all but one cut 5V wire from the SMPS side of the cuts and made the measurements below...

o/s State	Startup	Idle	Music	Shutdown
Maximised	4.0A	2.2A	2.2A to start, 2.0A stable	2.3A
Minimised	4.1A	2.2A	start 2.2A, stable 2.0A	2.2A

Like the 12V rail, the 5V rail does not seem to be impacted by the sate of the o/s.  I am not confident that 4.0A peak during startup has been recorded accurately because it is displayed on the multimeter as a rapidly changing amperage which may mean that the peak has been missed.



3.3V Rail

There are three 3V pins in the P24 socket each of which will successfully run the mobo in the absence of power to all of the remaining pins.  I cut the 3.3v wires, twisted the mobo side of the cuts wires together into a single joint, removed  all but one of the cut wires from the SMPS and made the measurements below...

o/s State	Startup	Idle	Music	Shutdown
Maximised	1.43A	1.42A	start 1.43A, stable 1.35A	1.42A
Minimised	1.43A	1.38A	start 1.4A, stable 1.34A	1.4A

Like the 12V and 5V rails, the 3.3V rail does not seem to be impacted by the state of the operating system.



Power On Pin

I was unable to measure the voltage variations of this pin.  It seems as though the act of measuring this voltage with either of my multimeters is enough to drop the voltage out of spec and the mobo will not power on.  I could, however, measure the amperage draw which was a consistent 150uV which must be present whenever the computer is turned on.  If the pin was pulled when the computer was powered on it instantly performed an ungraceful shutdown.  When the computer is turned off, this pin does not draw any current.  The literature indicates that this pin must be supplied +5V whenever the computer is on.



+5V Standby Pin

This pin will supply power to the motherboard even when AC is connected and the computer is turned off.  Measurements were as follows...

o/s State	Computer Off, AC On	Startup	Idle	Music	Shutdown
Maximised	0.353A	1.0A	0.9A	0.9A	1.0A
Minimised	0.353A	1.0A	0.9A	0.9A	1.0A

Draw from this pin was identical regardless of o/s state.  The ATX Standard specifies 1A minimum for this pin, with a recommendation for 2A to be provided to meet future requirements.



Summary
At this stage, a maximum of 5 separate rails could be run into the P24 socket alone: +12V 2A min.; +5V 5A min.; +3.3V 3A min.; +5V 0.5A min. (power on); +5V 2A (Standby).  Likewise, a minimum of 3 separate rails may suffice if all the +5V rail were combined.  The real question is whether the extra two rails would be worth the extra expense.

The next step is to have a look at some ATX psu's to see whether all the pins for each voltage are generated from the same rail, or whether the load is split according to pin.

[acg]

Just looked at the specs for the BeQuiet L7-530 SMPS that Peter recommends (and I use) in the XXHE PC.  There are two +12V rails: one for P24 ATX socket and PCIe ports; the second for P8 cpu socket, molex and sata peripherals (video cards, HDD, SSD, USB cards et cetera).  As far as I can tell the +3.3V and +5V rails are shared between the motherboard and peripherals.

I wonder how much there is to be gained by separating power supply to these components to minimise interplay?

Another interesting thing is the ripple rating of the BeQuiet.  Both +12V rails have a ripple/noise (max.) of 120mV while the +5V/+5V Standby/+3.3V rails are rated at 50mV maximum.  Compare this to the supposed specs. of my lab linear psu of 1mV rms, so I should be able to hear a difference if the ripple is an issue.

It's time to organise an oscilloscope.

[PeterSt]

You are really into this eh ? Well, I wonder when (!) this is going to bring us other things ...

12V Rail

There are two 12V pins on the P24 socket each of which will successfully power up the mobo without the other being connected.  I cut both wires, twisted the mobo end of the cut together, removed one supply wire from the supply side so that only one 12V wire was coming from the SMPS and made the measurements below...

o/s State	Startup	Music	Shut-Down
Maximised (no cpu fan)	0.93A	0.59A	0.85A
Maximised (cpu fan plugged in)	0.83A	0.48A	0.56A
Minimised (no cpu fan)	0.81A	0.57A	0.81A
Minimised (cpu fan plugged in)	0.92A	0.57A	0.81A

The 12V rail powers the cpu fan in the XXHE PC so I though it prudent to measure with the fan connected as well as disconnected.  The amperage varied when I pressed a key or moved the mouse, so perhaps the PS2 ports are run from this rail also.  As you can see the fan does not make much of a difference to the power draw on this rail, neither does minimising the o/s.

First the notice that those first two measurements may have their results twisted ? If not, something (cpu) heats up causing more current.

Then, in order to test that "PS2 movement draw", go into Minimised, start playback and on that black screen move the mouse. It now totally depends on what you see for a difference with your earlier tests regarding this;
I think it can be so - at least partly - that the mouse movement shows less draw now. If that indeed is so, that "over" part you measured earlier is just CPU usage. Notice the importance of the mouse needing to be over a black screen. BUT :
Writing this, I now realize that this won't be a valid test for W8 because that black screen is a "form" as such. The test would only be valid in W7 or when you'd be able to have "nothing" there.

Ok, sorry, never mind this. I don't think knowing the result of this is much useful anyway. But what I was actually heading for is the more precise knowing what does what for possibly noise creation. So as you know I advise PS2 connections. But what when USB is used after all ?
This is more complex because in the end this also heads for shutting down USB2 completely, which in your situation can't be done unless you have such a separate PCI USB3 card (then you can shut down USB2 while your NOS1 keeps on working on the card).

And on the other hand ... I am changing the subjects as it seems ...
But no necessity to give you pointers to this regard, because I see you are "rolling" into such subjects (of noise) anyway automatically. And FYI (because I never told you), this is how the 0.5mV/Div scope comes handy later; you just might be able to measure noise differences anywhere outside of the PC (or at "outputs" like USB) because of changes. "Just" because 500uV is actually very much but with that per division (of 1cm of height) there's a bit more to see than that of course.

Hey, it seems you like this much eh ? Well, I hope you can still play some music once in a while ?

Peter

[acg]

Nice pickup on the 12V rail Peter.  I have checked my notes and edited that post so that it is now correct.

Ok, sorry, never mind this. I don't think knowing the result of this is much useful anyway. But what I was actually heading for is the more precise knowing what does what for possibly noise creation. So as you know I advise PS2 connections. But what when USB is used after all ?
This is more complex because in the end this also heads for shutting down USB2 completely, which in your situation can't be done unless you have such a separate PCI USB3 card (then you can shut down USB2 while your NOS1 keeps on working on the card).

I have shut down all my USB2 and USB3 ports because I use the PCIe USB3 card from PPAstudio.

But no necessity to give you pointers to this regard, because I see you are "rolling" into such subjects (of noise) anyway automatically. And FYI (because I never told you), this is how the 0.5mV/Div scope comes handy later; you just might be able to measure noise differences anywhere outside of the PC (or at "outputs" like USB) because of changes. "Just" because 500uV is actually very much but with that per division (of 1cm of height) there's a bit more to see than that of course.

Hey, it seems you like this much eh ? Well, I hope you can still play some music once in a while ?

Peter

Oscilloscope's!!  When you want to buy a cheap one they are all too expensive and even then they may generate too much noise to be entirely useful. 

I am enjoying tinkering with this stuff and this is a subject that I have been considering for a while now.  Don't worry though, I am listening to plenty of music.

Cheers,

Anthony

[pedal]

Anthony,

As a layman I consider these measurements to be of significant value, because they *prove* (partly) why different PCs sounds different.

Different motherboard, soundcard, OS, etc, taxes the power supply completely in different patterns.  Knowing that the PSU is a weak link in common PCs, it is clear that this behaviours influences noise and sound quality.
-PSU being the heart of all electrical Hi-Fi equipment.

As Peter points out, all your measurements are perhaps not perfectly carried out yet, there are still areas to improve/discover. However the basic conclusion is clear.

These kind of measurements also "proves" why (partly) XX highend performs better than other music player software, because functions as "minimize OS, etc" allows a more even and lesser taxing of the PSU.

(This should be a wake up call for ignorants who believes that anything beyond "Foobar on an ordinary PC" is a waste of money).

Keep up the good work! 

[acg]

Thanks for the kind words Pedal.  I am not yet convinced that better power in the XXHE pc will directly improve anything, but I think that the computers power supply may impact other components via the AC lines.  We will see!

Tomorrow I am auditioning my nos1 in a high end system and if I find some time I will pull out my lab LPS and see if a few of us hear a difference.  I will power the P8 cpu socket, my PPAstudio usb card and the ssd.

Cheers,

Anthony

[pedal]

Reducing RF from a noisy component: One trick is to take a regular kind of Schaffner filter and REVERSE the direction of the plugs. Then it will filter whatever your PC leaks back to the AC connection.

This trick can be done even on noisy refrigerators etc . But don't tell anybody you have a net filter on your refrigerator, they think you are crazy!