Building a linear ATX PSU for an audio PC. Need help & advice

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
Now regarding the currents.

I agree that those currents are high, but I want the "headroom", AND from what I have read, I've seen that the further a PSU runs from its upper limit, the quieter it is. SO there's that. I did, however, appreciate the quoted rails and current draws a couple pages back. Helped me put this in perspective more. Maybe I can drop those current ratings a bit?

THEN there's potentially this now. Using this PSU to power 2 computers...

I'm looking at putting this LPSU to use in a Dual PC Jplay setup. So I'll be needing to power 2 computers. If you're not familiar with the setup, there's a ControlPC(processing) and AudioPC(USB output) tethered together via ethernet cable. I'm shooting for the best sound I can, so if this supply really does turn out to be any better than my current Seasonic supply, I'd like to be supplying both computers with it, not just one. If that's the case, those higher current ratings would be put to use.
- if possible, I'd like like to refrain from making 2 of these LPSUs (1 for each PC) if I don't have to.

Anyway, I'm looking for guidance here, so none of this is set in stone, and just as Julf had mentioned, I'm definitely coming in misinformed. But as I'm a total amateur to going this deep, I can't tell what isn't right. That's where I need your guys' help :-D

Then besides that, I am looking to have a bit of fun AND I'm looking to test out and find out things for myself too. You can't just read everything and think you really know, some times you got to give it a shot. Maybe it works, maybe it doesn't, but it's fun to figure it out and see what you learn along the way!
 
- if possible, I'd like like to refrain from making 2 of these LPSUs (1 for each PC) if I don't have to.

The by far easiest and most logical and most economical way is using just one computer for audio playback. The dual jplay setup is a nonsense, especially when running full-blown trial/error cut-down windows just to feed incoming samples to the sound card.

If you need the second device physically close to your audio gear, get a small embedded energy-efficient linux computer reading data from some form of NAS, controlled via phone/tablet, powered by a small linear PSU (the motherboard will have SMPSs anyways just like any motherboard does) and use separately-powered asynchronous USB DAC. The NAS can provide many other handy services (e.g. backup service for other home PCs), can be placed in cellar/storage room.

For software either MPD or the squeezebox server/squeezlite combination.

But this is something you did not want to hear, I know...
 
I'm another who thinks JPlay/dual PC setups are voodoo. Far better to spend the extra money on room acoustic treatment.

I use a dedicated computer sitting in my equipment rack along with all my other equipment. The motherboard is a ASRock Z87E-ITX with a Haswell 4570T processor. Most of the time the processor is barely doing any work and it under clocks itself. At the moment, I have a single linear 12V rail feeding a pico PSU and another 5V DC regulator powering a pair of internally mounted SSDs. Windows 7 is cut down to remove a lot of stuff that's not needed and JRiver is the playback software. The OS sits on a little mSata card on the mobo. The audio server is permanently on. I listen to stereo via an ESI Juli@ sound card with upgraded clocks powered from the 12V rail. For multichannel content I use the motherboard's HDMI port. My DAC is a Theta Digital Casablanca IV with 2 Xtreme DACs.

I have measured current on the 12V rail while playing back various content. It peaks at 2.5A for playback of streamed multichannel DVD ISO using JRiver's advanced Red October settings. For multichannel audio, e.g.SACD ISO, it uses just below 1.5A. That is, not much current need. You only need a lot of current if you have whopping great big video card and moving mechanical parts.

I am in the process of designing/building a multi-rail (12V, 5V, 3.3V and 5VSB) linear PSU to replace the PicoPSU. The main power rails will have 5A capability although in my present setup little of this will be used.
 
Last edited:
The by far easiest and most logical and most economical way is using just one computer for audio playback. The dual jplay setup is a nonsense

...

But this is something you did not want to hear, I know...

I'm another who thinks JPlay/dual PC setups are voodoo. Far better to spend the extra money on room acoustic treatment

I'm personally on the fence about it. I definitely agree that it's a lot of money that could potentially be put to better use on something like room treatment. I more so want to have a look and see if what so many other users have said is true. In the end if I don't see an improvement from the 2 PC setup then I'll just repurpose the 2nd computer to an HTPC, of which is something I also really want to build. So no real loss there.

I do definitely agree that room treatment is key, but WAF is an issue for now. It's something I'll be looking into, but for now it's further down the list. And I know, I know, it shouldn't be, but I'm just letting you know it is.
 
I use a dedicated computer sitting in my equipment rack along with all my other equipment. The motherboard is a ASRock Z87E-ITX with a Haswell 4570T processor. Most of the time the processor is barely doing any work and it under clocks itself.

...

I have measured current on the 12V rail while playing back various content. It peaks at 2.5A for playback of streamed multichannel DVD ISO using JRiver's advanced Red October settings. For multichannel audio, e.g.SACD ISO, it uses just below 1.5A.
Thanks for also detailing your experience on how much current your audio PC uses, SGK. After I'm done reading into your links and seeing what else I learn I'll probably be adjusting my desired outputs. Thanks for the 2nd link too, you're being a great guy by helping me along with all this.


I am in the process of designing/building a multi-rail (12V, 5V, 3.3V and 5VSB) linear PSU to replace the PicoPSU. The main power rails will have 5A capability although in my present setup little of this will be used.
I'd really love for you to take pictures along the way and make a build thread or at least detail your process once you're done. That's something I'd love to read and maybe follow, or at least get some ideas from. I look forward to reading about it when the time comes. Do you have any idea of when it might come together?
 
a well regulated linear powersupply for computers is not a new idea.
first off, was used by overclockers, i did see my self too.
it does indeed has advantage. stability wise.
not sure of audio quality, but i guess in this case that is not the point at all, just the fun of doing it.
 
What way stability wise?
I having nothing against linear supplies as such, but I would like to see some empirical data showing improvements...or at least the differences. I have seen linear supplies used that were detrimental to noise immunity. Also the noise floor does not decrease that much due to how a lot of noise is generated in computers and all high speed switching digital circuits.....
a well regulated linear powersupply for computers is not a new idea.
Nope, I remember serving computer supplies in the early 80s, large heat generating and not very efficient, hence why SMPS's took off. But links to something relevant would be nice.
 
Without proper measurements of the PCs noise spectrum we cannot really understand what is going on, just guess....
As I have said many times the noise in a PC is not just from the SMPS, a lot of noise is generated by the circuitry operating, this is called simultaneous switching noise. The best way to get a handle on this is to read up on the differences between DDR2 and DDR3, specifically how they have spread the switching between banks of RAM to limit the noise, this technique is also employed in SMPS, call spread spectrum switching.
Personally I would not recommend playing with the RAMs power supply, unless you fully understand the implications. Now you said it gave further benefits to the sound, this could be because you are actually increasing the noise floor not reducing it adding texture to the sound further down the line, again only very careful measurements with high GHz scopes etc will give a true picture of what is going on

Thanks for the response Marce. That's an interesting take and something I hadn't considered...kinda like Valves I guess. I would be very interested to stick a decent scope across various bits of my MB and you would be more than welcome, if you ever get down to the SouthWest, to try to get some empirical evidence out of my setup. By the way, I'm also experimenting injecting better power (lifepo cells) to the motherboard clocks with good results...but that's for another thread.

I agree. The fear of SMPSs is mostly totally irrational, based on the dirt cheap wall wart ones. A well designed SMPS can be quieter than a linear supply, and superior in all other respects too.

I agree, there is too much dis-information around and they've been tarred with the cheap/crap brush.

The dual jplay setup is a nonsense, especially when running full-blown trial/error cut-down windows just to feed incoming samples to the sound card.

If you need the second device physically close to your audio gear, get a small embedded energy-efficient linux computer reading data from some form of NAS, controlled via phone/tablet, powered by a small linear PSU (the motherboard will have SMPSs anyways just like any motherboard does) and use separately-powered asynchronous USB DAC. The NAS can provide many other handy services (e.g. backup service for other home PCs), can be placed in cellar/storage room.

For software either MPD or the squeezebox server/squeezlite combination.

But this is something you did not want to hear, I know...

Have you ever heard a dual motherboard JPLAY set up running on optimised (manually or through paid for optimisation script) Windows server 2012? I agree that on the surface it does sound like a nonsense on many levels. For example, why introduce all manner of potential problems by introducing a network connection into the 'audio chain'? However, it does sound very good and in every test I carried out (albeit in 2012) it beat MPD and every variety of linux I cared to throw at it.

Having said all that if you are interested in doing a simple comparison with a single PC then do a quick google for MQn. It's free, well worth a go and you may be surprised at what you hear.
 
Member
Joined 2003
Paid Member
<SNIP>
Have you ever heard a dual motherboard JPLAY set up running on optimised (manually or through paid for optimisation script) Windows server 2012? I agree that on the surface it does sound like a nonsense on many levels. For example, why introduce all manner of potential problems by introducing a network connection into the 'audio chain'? However, it does sound very good and in every test I carried out (albeit in 2012) it beat MPD and every variety of linux I cared to throw at it.
<SNIP>

Crom,

Off-topic... can you point to a webpage or posts describing the 'dual motherboard JPLAY setup running on optimized... Windows server 2012' and comparisons with that against various Linux setups and hopefully also MQn setups?

I'm not doubting your experience, but I've been following some of the later HW setups using the small, low-powered embedded boards such as the RPi, BBB, and Wandboard and wondering if these will really be competitive against the JPlay and MQn setups you mentioned.

THANKS!

Greg in Mississippi

P.S. On-topic, I am definitely not an anti-SMPS bigot, both of my main types of amps use SMPS & the next I am planning to build will also. But I implemented a home-built multi-rail ATX-compatible linear supply a few years back and found it a very sizeable upgrade over either a well-regarded computer supply, that same supply with additional filtering, and a 'linear-hybrid' supply using two 12v linear supplies with one driving the motherboard's P4 connection and the other a PicoPSU supplying the ATX connector. Only the last provided a similar sense of musicality, ease, clarity, and purity to the full linear supply, but still was outclassed.
 
Have you ever heard a dual motherboard JPLAY set up running on optimised (manually or through paid for optimisation script) Windows server 2012? I agree that on the surface it does sound like a nonsense on many levels. For example, why introduce all manner of potential problems by introducing a network connection into the 'audio chain'? However, it does sound very good and in every test I carried out (albeit in 2012) it beat MPD and every variety of linux I cared to throw at it.

There is a single question - were you able to distinguish the two setups on the same HW in a blind comparison test? If in all your listening sessions you knew what you were listening to, sorry, no argument for me...



Having said all that if you are interested in doing a simple comparison with a single PC then do a quick google for MQn. It's free, well worth a go and you may be surprised at what you hear.

I followed that software development. The guy always refused to make a blind listening test while claiming pretty unlikely improvements. Again, sorry...
 
P.S. On-topic, I am definitely not an anti-SMPS bigot, both of my main types of amps use SMPS & the next I am planning to build will also. But I implemented a home-built multi-rail ATX-compatible linear supply a few years back and found it a very sizeable upgrade over either a well-regarded computer supply, that same supply with additional filtering, and a 'linear-hybrid' supply using two 12v linear supplies with one driving the motherboard's P4 connection and the other a PicoPSU supplying the ATX connector. Only the last provided a similar sense of musicality, ease, clarity, and purity to the full linear supply, but still was outclassed.

The interesting thing is that a lot of people think of SMPSs as "noisy", but in most cases linear supplies are much noisier electrically over a very broad frequency spectrum (starting with harmonics of the mains frequency) while the noise signature of SMPSs is restricted to much higher frequencies where it is much easier to manage.

There is a pretty good illustration here: EMC Journal: Design Techniques for EMC

An externally hosted image should be here but it was not working when we last tested it.

An externally hosted image should be here but it was not working when we last tested it.
 
The interesting thing is that a lot of people think of SMPSs as "noisy", but in most cases linear supplies are much noisier electrically over a very broad frequency spectrum (starting with harmonics of the mains frequency) while the noise signature of SMPSs is restricted to much higher frequencies where it is much easier to manage.

There is a pretty good illustration here: EMC Journal: Design Techniques for EMC

An externally hosted image should be here but it was not working when we last tested it.

An externally hosted image should be here but it was not working when we last tested it.

You guys are smarter than me but surely this chart assumes no regulation. A very simple regulator can achieve >100dB attenuation of noise across the audible band.
 
It's great to try and keep up with the posts here. Learning, learning.

I've never given MQn a shot, but is it dead in the water now or something? Over on the Jplay forums there were a few converts from Dual-PC Jplay to single PC MQn, but like I said, personally I have no idea.

I've also been in contact with Teradak, a budget'y linear power supply company. They have a linear ATX PSU called the 800W(600W usable) Linear ATX Supply, and it goes for about $1700 shipped to North America. It's pricey, but significantly lower than the ~$5k that a 4 rail ATX Paul Hynes goes for. And with 800W I was looking at powering both my PCs with it in the Dual-PC Jplay setup I'm putting together. So that'd cut the cost in half, which is much more manageable.

Unfortunately I didn't use my brain all that well and didn't consider the power-on sequence being separate for both computers, so using a single 800W ATX Supply for 2 computers it out the window now.
*yes, I'm aware that some mobos aren't picky with their start up power sequencing, but I'm not looking to bet on that right now. I'm also using server motherboards, which I bet aren't going to be as forgiving as a mini-ITX that is very likely to be employed in some DIY scenario.

I was given another solution by Teredak, however, which is with their smaller 210W(usable) ATX Supply. I can buy 2 of these guys for ~$500 a pop, and I'm in China, so shipping is negligible. That's then $700 less than the original, and it's got the correct power-on sequence that I need! And at $500, I definitely can't DIY one for that price. Not with the kind of chunky aluminum chassis this comes with.

Here're some shots I was sent. Tell me what you guys think of it, and well, with the new route.
https://onedrive.live.com/?cid=1EB7027489224A7D&id=1EB7027489224A7D!887&authkey=!AiCFCEdhmY-Jl40
 
Have you ever heard a dual motherboard JPLAY set up running on optimised (manually or through paid for optimisation script) Windows server 2012?

None of my Jplay experiments ended up in success. Yes, the dual pc is much better that the single pc. Yes, the optimised, command prompt win2012 sounded much, much better that the standard install. And yet all of the above sound dark and bassy compared to a simple win7+jriver. The superior soundstage provided by jriver was simply not enough...

One would imagine that a hardware box serving as a replacement for the "client" pc in the two pc setup should not be too hard to design for someone knowing fpga. Jplay could really think about offering a hardware solution.
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.