Home Computer Audio Asylum

Music servers and other computer based digital audio technologies.

Another Linear Supply sees the light of day!




Dear cMP-philes,

Following is a description of a project that I embarked upon concerning the building of a bespoke Linear Power Supply (LPS) to power my G31M+E7200 cMP2 rig.

Back in the good old days of cMP, the recommendation was to power the motherboard (mobo) by a quality ATX Switch Mode Power Supply (SMPS) and the HDDs by a cheaper (Granite) SMPS. This seemed to work quite well. However, it was not long before some cMP-philes began to experiment with other strategies to improve performance. Some of these strategies are documented on the cMP web site (Appendix B – Advanced Optimisations). Anyway, my configuration prior to this project was: (i) An ATX SMPS powering the P20-24 on the motherboard; (ii) A cheap 12V Linear Power Supply (LPS) powering the CPU (P4); (iii) A Granite SMPS powering the HDDs.

Some time ago there was a post on the Asylum by John Swenson who proposed a 9V LPS incorporating a Hammond Choke. John reported that this improved the sound quality of his fit-PC. This, I believe, encouraged some cMP-philes to think about developing a LPS for a detuned cMP. I don’t know who amongst the fraternity was the first but I recall seeing a post and photos of Mihaylov’s LPS. By my reckoning, Mihaylov had gone to great effort to build a LPS that handled the three voltages necessary to power the mobo (12V, 5V, 3.3V). I remember looking at those photos and thinking that it was an “all out” attempt to build a quality computer power supply for audio. Yes, a great attempt but too large a step for me.

I began to think about the idea of up-scaling John Swenson’s design to power my mobo through a PICO PS. I thought the advantages of this solution were: i) The PICO could handle the drop-down voltages of the 5V and 3.3V lines, and ii) The PICO could handle the start-up “power-on” sequence on the P20-24 required to bring a mobo to “life”. I thought a 12V LPS+PICO would be a step in the right direction and would cost considerably less than a LPS similar to Mihaylov’s. In summary, I figured the alternatives were:

1. Build three LPSs; one for each of the required voltages (12V, 5V, 3.3V)

2. Build a 12V LPS+PICO (hybrid solution)

3. Build something else..?

Option 1 seemed to me to be too extravagant – as one who is trying to keep this hobby within bounds. Option 2 seemed a good one for reason of its more modest cost and relative simplicity but perhaps a compromise for the reason that the PICO used switch mode regulators to achieve the drop-down 5V & 3.3V required by the P20-24.

Option 3 was considerably open in possibility: Given that I am not a fan of up sampling – preferring instead to playback at the native rate; Given that I was not using a soundcard – instead outputting via USB to DAC (see below System for details). I began, at first, to consider the approach championed by Ryelands to use a “headless” fit-PC2. I think what Ryelands has done is very neat and elegant indeed: the fit-PC2 has modest power requirements and its on-board regulation meant that a single good quality 9V LPS would suffice. In the end, however, I settled upon the idea of building a “dual” 12V and 5V+3.3V LPS for my existing board with two transformers and two chokes – in an effort to keep the cost down. As I was not able to attempt the skilled “grunt work” myself, I contacted a friend (by the name of Elson) who was adept. Soon after, I took delivery of the “Elson” LPS. Total cost was ~US$450. (incl. parts & labour. And, before anyone asks: Elson will not build another!) I connected the Elson to the P24 of my G31M mobo using a schema suggested by GStew and powered it up according to his instructions. My rig came to life on the very first attempt. Thank you GStew!

Description:

The design is based on John Swenson’s offering but with some modification. (See John’s post here.) It uses two toroidal transformers and two Hammond chokes (159ZJ) – one for the 12V supply and another for the 5&3.3V supply. Caps: Jamicon (15000uF 25V). Resistors: Takman metal film. Diodes: Motorola bridge rectifier type. Max average current: 2A for 12V line, 4A for 5V line, 1A for 3.3V line. Dimensions: 9”/225mm(w) 3”/75mm(h) 11”/275mm(d).

(Please don’t ask me for a schematic of the Elson. The “source” was John Swenson’s design and Elson took it from there – keeping it as close as possible to John’s spec but also taking into consideration parts cost and availability.)



One unusual aspect of the design is the use of voltage trim-pots. By adjustment of three screws, the voltages can be varied up or down with a consequent effect on sound quality. I experimented with voltages on the 12V and 5V line. Unfortunately, the way things were implemented I was unable to lower the voltage on the 3.3V line; only increase it which didn’t seem to make for improvement – so I left it at 3.3V.

Following is an excerpt of a communication to another cMP-phile at the time I was fiddling with this stuff. It tells the story reasonably well, I think:

“I have been experimenting with Voltages in the 4.5-4.7 to 11.5-11.7 range. The first thing to say is that there appears to be a noticeable change in sound for such small changes in voltage. I was listening to three tracks in particular yesterday: Eva Cassidy’s Songbird Tk 3; & Dave Brubeck’s Time Out Tracks 2 and 3. With Cassidy, I want to hear a voice to die for (as female voice is “my thing”.) Also, Brubeck’s tracks are a good check of how things are going in the sense that I want to hear a realistic piano; sax that has a little “bite”; realistic bass strings; dramatic and percussive drums (particularly Tk3).

At the lower end of the voltage range (4.5 & 11.5) things get very clean and liquid: Cassidy’s voice is gorgeous and dynamic but I seem to notice that some of the low-level stuff melts a little too much into the background such that they become a little indistinct. Whereas, at the top end of the voltage range (4.7 & 11.7), the background stuff (strings and other instruments) begin to take over a bit to the extent that I lose the focus on that yummy voice.

…………..

After more listening, I am of the opinion that 3.3, 4.6, 11.8 is the “go” with the current mobo.. The performance "hit" is definitely on the 5V line while the 12V line only gives subtle changes in SQ. Whereas, the difference between 4.5, 4.6, 4.7, 4.8 is not subtle, IMO.”


Note that these experiments were done on my trusty G31M board. I would not necessarily expect the same results on a H55 or H67 board.

Outcome:

Does the Elson make for an improvement in sound quality? Most definitely! I have, on several occasions reverted to my old ATX Corsair SWPS to perform an AB. The Corsair sounds “grungy” by comparison. The Elson offers better clarity and is far more revealing. Blacker blacks. More micro-detail; instruments and voices have a much more natural sound.

Is the Elson as good as a (full) 12V 5V 3.3V LPS solution? I don’t know as I have not had the opportunity to compare but would say it is unlikely. I suspect a “full implementation” would yield the best performance. However, the upside of the Elson is that it would not be as expensive.

Is the Elson better than a 12V LPS+PICO solution? Again, without the opportunity to A-B, I don’t know. It may be but then again it may not. What the PICO has in its favour is simplicity and lower cost.

Is the Elson an elegant solution? At this time, I have the Elson powering the P24 on the mobo. A cheap 12V LPS powering the CPU P4 (cost US$30). A Granite SMPS powering the HDD. In all, a lot of wires going everywhere! I crave something neater.

Components

As far as a bespoke LPS is concerned, I believe there is improvement to be obtained by selecting the “right” components: regulators, caps and so on, as these are all going to make a difference to the sound. My selection of components is one of many. There are other contributors to this thread who have tried other designs and components with success. If you have a mind to attempt a similar project then I suggest you review their component list and leverage upon their trials and outcome as well as mine. Cost and availability were considerations for me.

Thanks:

Many thanks to GStew (Greg), Ryelands (Dave) and John Swenson who have been helpful in this project. I would also like to thank Mihaylov, jackwong and hfavandepas who have been so generous with their discoveries. Please forgive me if I have omitted to mention others who have also contributed.

System:

cMP2>Empirical Audio Freeway2 (USBtoS/PDIF)>Audio-GD DAC-19DSP>Supratek Sauvignon>Blue Circle BC26> Thiel CS3.6

AB.






This post is made possible by the generous support of people like you and our sponsors:
  Atma-Sphere Music Systems, Inc.  


Follow Ups Full Thread
Follow Ups

FAQ

Post a Message!

Forgot Password?
Moniker (Username):
Password (Optional):
  Remember my Moniker & Password  (What's this?)    Eat Me
E-Mail (Optional):
Subject:
Message:   (Posts are subject to Content Rules)
Optional Link URL:
Optional Link Title:
Optional Image URL:
Upload Image:
E-mail Replies:  Automagically notify you when someone responds.