Investigating a full linear PSU for the XXHE PC and NOS1

[CoenP]

Thanks Coen.  I looked these up but their noise was rather high at about 260uV and no PSRR figures.  I am reading the specs differently to you...but that 25 page spec. sheet very helpful.

Anthony

That noise figure may turn out to be important or not, but I see other problem areas that need much more attention and are more likely to correlate with "better" sound.
Nevertheless the LT1083 also looks impressive, they have a price for a reason. There is also a 15xx series that raises my interest.

Warning/Teaser: If you converge on your regulation setup the next challenge will be a healthy supply and groundingplan. This will make or break all the hard regulation work. Next to theory and opinions it will reqiure practical experiments to find out whats best.

Regards, Coen

[Arjan]

Hi Coen,

I saw a picoPsu on the same site were the audiograde usb3 comes from.
It is Max. 160 - 168 watt. Why are these picoPsu's so limited?

Yes, a one box solution is the best. The idea of splitting the pin24 and CPU power was just my 'out of the box' thinking. But as stated before I am not an expert. So I totally missed the power management issue.

Regards, Arjan

[acg]

Warning/Teaser: If you converge on your regulation setup the next challenge will be a healthy supply and groundingplan. This will make or break all the hard regulation work. Next to theory and opinions it will reqiure practical experiments to find out whats best.

Yes...the grounding...has me a little worried as well.  When I put four rails into the motherboard and 'tricked' the XXHE PC into starting up without the ATX procedure I may or may not have been lucky with the grounding of the separate pins on the P24 socket (the P8 socket may not be so bad because it has a single 12V rail and its own dedicated grounding wires).  The P24 socket has a number of ground wires but there is no mention of grounding in the ATX power specifications, so I have no idea which ground wire/s needs to be paired with which positive wire/s let alone which rails.  I suspect that it may not matter so much because of one simple observation that I have made...that each of the three main power rails into the P24 socket (3.3V/5V/12V) has multiple wires in the connector, but the computer will post/boot/play music with ANY single one of them powered, which indicates to me not only that the multiple wires for each rail are connected on the mobo side of the P24 socket, but that the grounds are connected also, because if the grounds were not connected current would not flow and the computer would not run.

The grounding issue then becomes less a 'within the individual wires of each rail' problem and more a 'between the separate rails' concern which thankfully may be much easier to test/tweak/solve.  The fact that I randomly chose which ground wires within the P24 socket to pair with each rail of power fed into that socket and I got it to work first go suggests that it is not necessarily a major issue in getting the motherboard to function, but I agree that some noise measurements would be useful to determine if changing the grounding pins to each rail affects performance.

Anthony

[acg]

Hi Coen,

I saw a picoPsu on the same site were the audiograde usb3 comes from.
It is Max. 160 - 168 watt. Why are these picoPsu's so limited?

Yes, a one box solution is the best. The idea of splitting the pin24 and CPU power was just my 'out of the box' thinking. But as stated before I am not an expert. So I totally missed the power management issue.

Regards, Arjan

Hi Arjan,

Yes Paul Pang does offer a modified wide input picoPSU type of dc-dc converter (not sure that it is the picoPSU brand) to which he has grafted higher quality wire.

Your thinking was not far off what several people at the moment do with the P24 socket which is something I suggested at the JPlay forum: power up the pc with a picoPSU and then use a couple of relays to switch to linear on those rails when the pc has booted.  There are plenty of problems with that approach but it does work in a fashion and the guys doing it do report positive results.  Up until the last week I had no idea how to implement the ATX startup logic in a bespoke psu...but now that I do know how to do this we can get rid of the picoPSU and the ATX SMPS.

Cheers,

Anthony

[acg]

For example, I see that we are talking about support for SATA drives and such. But what if no single drive would be in there ? at least that is what I personally am aming for. Not so easy, but I will get it done.

Peter, do you care to elaborate how you intend to get away without a single drive in the XXHE PC?  Are you trying to run the o/s from an external drive?  Maybe load the o/s onto a ram disk?  If that can even be done.

Anthony

[acg]

NUMBER OF RAILS

Another thing to think about is the number of rails that the supply will require to operate the pc.

A special consideration is the 5V Standby pin on the ATX power supply.  The ATX power scheduling requires the 5V Standby line to be powered whenever the AC is turned on in the power supply.  This rail is operated by the little switch that you find on every ATX SMPS.  Turn it on and the 5V Standby rail becomes active and some of the circuits in the motherboard are powered.  The Asrock X79 Extreme4-M motherboard that is standard (and highly recommended) in the XXHE PC will draw no more than 1A from the 5V Standby rail, but the ATX power specifications suggest that 2A be made available to suit future motherboard features (at present this powers the front power button on your pc case, wake-on-lan, wake-on-usb etc).  Do we provide 2A to this rail or just 1A?  A 1A rail is easier and cheaper to build.  This rail must be powered BEFORE any of the other rails.

Now, it may just be the case that running the maximum possible number of rails into the XXHE PC produces the best results.  But it may also be the case that we can combine some of these rails to make the final product that Peter will produce (I think that is what will happen) less expensive and better bang for the buck.

Maximum number of rails that can be fed into the XXHE PC:
#1 5V Standby line
#2 P8 12V 6A
#3 SSD 5V 2A
#4 PPAstudio USB Card - ultra high quality 5v 0.2A feed
#5 P24 3.3V 3A
#6 P24 5V 5A
#7 P24 12V 2A
#8 Fans 5v 0.5A

Minimum number of rails that can be run into the XXHE PC:
#1 5V 9A for Standby line, SSD, PPAstudio USB card, P24, fans
#2 12V 8A for P8 and P24
#3 3.3V 3A for P24

I do not think that 3 rails can work well.  Firstly, the PPAstudio USB card will be quite responsive to an ultra high quality power supply and we are just not going to get that with 9A on that rail.  Secondly, the 5V Standby will generate a lot of heat (up to 45W) and will use plenty of power if the switch on the back of the linear power supply is not turned off when the PC is not in use.  Thirdly, we may find it beneficial to run the fans on a separate rail to get rid of the electrical noise messing with other 5V components.  Fourthly, I would be a little surprised if powering the cpu separately did not provide some advantage to the entire computer, considering much of the electronic noise probably comes from there already.

So I guess that I am saying the XXHE PC LPSU will probably end up with somewhere between 6 and 8 separate rails. 

I suppose a further contentious point is the inclusion of the PPAstudio USB card in the XXHE PC build.  A couple of us have tried it now and it seems to work well: maybe we should get others to try it also and give their opinion.  The one to convince though is Peter.  Are you tempted to try one of those cards Peter?

Cheers,

Anthony

[PeterSt]

Warning/Teaser: If you converge on your regulation setup the next challenge will be a healthy supply and groundingplan. This will make or break all the hard regulation work. Next to theory and opinions it will reqiure practical experiments to find out whats best.

Yes...the grounding...has me a little worried as well.  When I put four rails into the motherboard and 'tricked' the XXHE PC into starting up without the ATX procedure I may or may not have been lucky with the grounding of the separate pins on the P24 socket (the P8 socket may not be so bad because it has a single 12V rail and its own dedicated grounding wires).  

Anthony,

This is not (so much) about properly grounding some wires and certainly not those carrying 220mV of noise to begin with. It is about the maybe achievable 220uV of noise from these regulators and how this needs the most careful groundplanes (in layered boards) in order to not make it 220mV again right from the start. And to be honest, I don't think this can be done. Oh, there can be this well designed PCB all right, but then what ? it/they will sit at a mile distance of where the voltage is needed and any regulator board neatly coughing up 220uV will show 1mV at the end where the voltage is needed.

The 1mV I make up for a reason and actually I have no idea. The 220uV was also made up, and as far as I have seen so far it will merely be in the 300uV range for 7A or 9A regulators. Also they must stay very cool in order to meet their power requierments and I am not so sure whether this "very cool" can be achieved in there.

The reason for coming up with 1mV is because that can be achieved fairly easy with a LPSU. So, what to do difficult for behind that PSU when it won't get better anyway. Or worse.

Also ... Then we entered the mother board. All sorts of sh*t happens there and although all will be made to spec (to run well), nobody is going to tell me that anything good will be happening there when the base for all MoBo designers is 220mV of noise to begin with. And oh, on-board regulators may make that better but the point is that this is out of our control anyway. So for example, who tells that there is no 12V regulator on the already 12V rail just to eliminate the expected (sure high) 220mV ? (easy to check at the consumer side where it should show less than 12V in that case). But what I mean is that such a regulator will have an inherent (say) 10mV of noise and there you go.

Or what about other regulators spitting on the same ground (assumed no low noise design for that on the MoBo) ?

I (now) say it is all completely useless to try to improve upon ~1mV of noise coming from the PSU;
This regulator subject sort of started its own life, while (I think) I proposed it as that other means to provide better (less-noise) power. But on its own useless again because it merely should be about (the SPSU) not spitting into the mains.

And hey, before we think all is arranged for regarding that spitting, we better connect our monitors to such a LPSU as well.

But correct me where you can please !
Peter

[acg]

author=PeterSt link=topic=2726.msg28752#msg28752

And hey, before we think all is arranged for regarding that spitting, we better connect our monitors to such a LPSU as well.

My monitors are run on a separate circuit to the audio gear (which includes the XXHE PC) which has its own dedicated circuit.

We would need a big LPSU for a monitor...mine draws about 70w in general use, but I've never measured it at startup.

How vibrant would the colours be!!!!  haha

[BertD]

And hey, before we think all is arranged for regarding that spitting, we better connect our monitors to such a LPSU as well.

Isolating transformer? Connect all noisy spitting stuff at the isolated end?

[PeterSt]

A just-not different subject :

For a longer time I have been thinking to produce this small box with such a LPSU in it for any device we just *need* around the audio rack and assumed no separately earthed mains ring is available for that. It's probably only that I do have these separate rings why I didn't make a bunch for myself.

For example think about all the self-powered USB devices we may use. All could be powered from this one little box.

One problem I foresee at this moment : how to make general "adapting" inlets so all devices can go in there (I am not talking about the different voltages). Also, not all devices have detachable cords hence those cords must be cut and some new connector must be soldered on it. ... And next it can't be used elsewhere anymore.

Let's not proceed with this subject in this topic, but I only wanted to make clear what we are actually working on, and how that is related to more, or otherwise it is a useless endeavor.

Regards,
Peter


PS: Saw the bottom of my sig ? - yea yea, I know. There's one left - in the PC ...

[PeterSt]

And hey, before we think all is arranged for regarding that spitting, we better connect our monitors to such a LPSU as well.

Isolating transformer? Connect all noisy spitting stuff at the isolated end?

Hey Bert - Do you mean such a beast like e.g. Mani has ?

[PeterSt]

My monitors are run on a separate circuit to the audio gear (which includes the XXHE PC) which has its own dedicated circuit.

And then to think that I scratched this subject from my post (after yours Anthony).

But if you really have those rings separately earthed, measure the voltage over the two earths. Just for fun.
So this is what I do *not* do (PC is in the audio ring).

Peter

[acg]

My PC is on the dedicated audio circuit (dedicate from the mains box)...my monitors are on another circuit for the other computers.  No separate ground for audio...I'm pretty sure that's illegal here.

[BertD]

Hey Bert - Do you mean such a beast like e.g. Mani has ?

Something like that but smaller will do as well as long as you do not connect all your PC's to that one transformer... ;-)

I am using a 1000VA for both my PC's (and related switching adapters), one monitor and one Behringer.

http://www.amplimo.com/index.php/products/safety-transformers-for-medical-use

Bert

[acg]

Anthony,

This is not (so much) about properly grounding some wires and certainly not those carrying 220mV of noise to begin with. It is about the maybe achievable 220uV of noise from these regulators and how this needs the most careful groundplanes (in layered boards) in order to not make it 220mV again right from the start. And to be honest, I don't think this can be done. Oh, there can be this well designed PCB all right, but then what ? it/they will sit at a mile distance of where the voltage is needed and any regulator board neatly coughing up 220uV will show 1mV at the end where the voltage is needed.

The 1mV I make up for a reason and actually I have no idea. The 220uV was also made up, and as far as I have seen so far it will merely be in the 300uV range for 7A or 9A regulators. Also they must stay very cool in order to meet their power requierments and I am not so sure whether this "very cool" can be achieved in there.

The reason for coming up with 1mV is because that can be achieved fairly easy with a LPSU. So, what to do difficult for behind that PSU when it won't get better anyway. Or worse.

Also ... Then we entered the mother board. All sorts of sh*t happens there and although all will be made to spec (to run well), nobody is going to tell me that anything good will be happening there when the base for all MoBo designers is 220mV of noise to begin with. And oh, on-board regulators may make that better but the point is that this is out of our control anyway. So for example, who tells that there is no 12V regulator on the already 12V rail just to eliminate the expected (sure high) 220mV ? (easy to check at the consumer side where it should show less than 12V in that case). But what I mean is that such a regulator will have an inherent (say) 10mV of noise and there you go.

Or what about other regulators spitting on the same ground (assumed no low noise design for that on the MoBo) ?

I (now) say it is all completely useless to try to improve upon ~1mV of noise coming from the PSU;
This regulator subject sort of started its own life, while (I think) I proposed it as that other means to provide better (less-noise) power. But on its own useless again because it merely should be about (the SPSU) not spitting into the mains.

And hey, before we think all is arranged for regarding that spitting, we better connect our monitors to such a LPSU as well.

But correct me where you can please !
Peter

At the moment any gains/losses in sq from this linear psu project are theoretical and speculative at best.  We just don't know, and as you said early in this thread this is a VERY complicated area.  

I know of one respected audio engineer with good products (that I have owned) that was convinced by his clients to try a simple single rail linear power supply into his computer.  Eventually he gave in, put together a good one with a Hynes reg and could not figure out why power supply made such a difference in his system as it had in his clients systems.  No real explanation...it just worked above and beyond the "hash into the mains" theory.  Adding a LPSU to the USB card that I have in my computer ratchets up SQ in my system.  Why?  I don't really know but it may be something to do with grounding or noise on the USB cable or even placebo.  I've tried it a few times now just for a sanity check and I get the same result every time.

Anyway, the reason I have started this thread is because I want to know what will happen if I put high quality linear power into everything and learn which components benefit from that power.  If we can get away with two or three rails that would be awesome and relatively inexpensive, but if we don't try we won't know for sure.

Anthony