Audio Asylum Thread Printer
Get a view of an entire thread on one page
|For Sale Ads|
Here is a power supply design for the low power PCs I've been talking about. It will also run a squeezebox very well.
The design can be built for either 5V or 9V. The original fitPC takes 5V and the new "slim" version takes 9V. Also the squeezebox SB3 (now classic) takes 5V and the new duet takes 9V. So build it which ever way you want.
Most of the parts are the same, just a different power transformer and different regulator for different voltages. For 5V there is a 5V fixed regulator, for the 9V you have to get the adjustable version and use two resistors to set the voltage.
All the parts except for the choke can be bought at DigiKey. The chokes are a little harder to find, Angela instruments and Radio Daze have them.
As with any AC supply you need to add your favorite input connector and fuse. For 120V lands use a 1/2 A fast blow, for 240 lands use a 1/4 A.
The transformers specified have dual primaries, wired in parallel (as shown in the schematic) that are good for 120V areas. For 240V areas wire in series. Check the transformer spec sheets that come with the transformer for exactly which pins are which. Not all of them will match the pinout in the schematic
Use the parts specified, the design has been tuned for these parts, if you substitute other parts it may not work as well. Don't skip the choke, it is what makes this design work. The resistors can be 1/4W or 1/2W carbon or metal film. Cheap ones work fine here, they don't have to be super expensive audiophile types. C1 should be a film type, I use the Sprauge orange drop, but other will work as well.
It is extremely important to heatsink the regulator. Without it it will burn up, not good! The one I speced is what I use, but there are many others that will work, as long as they can handle 3-4 Watts. The case of the regulator is NOT isolated! So make sure you use an insulator between the tab and the heatsink. These days I use the synthetic pads rather than mica insulators since you don't need to use the messy grease with them. Digikey has quite a selection to choose from. Pretty much any of the TO-220 pads will work.
The design is a bit unusual, it uses a choke in a low voltage design. This has some interesting properties. Primary among them is the use of a fairly small first cap which radically cuts down on the peak currents through the diodes. With low peak currents and Schottky diodes there is essentially no RF noise in this design. And on top of that I added a secondary snubber (C1 and R1) which prevents the transformer from resonating as well. The upshot is that this is one of the cleanest power supplies you will ever find, it has no RF and has one of the lowest impacts on the main AC line as its possible to get with a PS.
Since this is a power supply that plugs into the AC power line its probably not a good idea to build one of these if you have never touched a soldering iron before.
Oh yeah, you are on your own to get the right connector to fit whatever you want this to power. Radio Shack has a line of adaptaplugs which might come in handy for this.
Here is a power supply design for the low power PCs
OK, better late than never. It's been several weeks now since I finally got round to finishing a 9-volt version of the above design to drive a Fit-PC2 running cMP^2.
Even though it replaced a conventional linear supply rather than an SMPS 'brick', results have been very fine with improvements that are clearly audible - if not 'night and day' then not far off and certainly 'early dusk and late dawn'.
In passing, there seem to be a couple of minor typos in the drawing - the value for the Lower Arm Resistor is too low (calculate as per datasheet) and, at least in my case, wiring the transformer secondaries in series as shown made for a pretty high voltage at the regulator's i/p. Parallel OTOH was fine.
Whatever, it's a super circuit and I'd like to take this opportunity of saying thanks for making the design generally available.
John: Re-reading your notes, it's apparent that using it to provide one to 1.5 amps at 12 volts (for a DDDAC) would probably take the design to its limits. Is it feasible to modify it to do this without intruding (well, not too much) on your time?
I will order the parts shortly to build a PSU for fit-pc2. Here is the parts list I identified.
Triad VPS16-2700 (237-1261-ND XFRMR PWR 16.0VCT 2.7A QC .187)
4 x 20 A DIODE SCHOTTKY 120V TO-220AB (STM STPS20120D)
Hammond 63105 159ZJ (10mH 0.16) – choke open
R1 330 ohms 1/4W or 1/2W carbon or metal film
R2 300 ohms 1/4W or 1/2W carbon or metal film
R3 47 ohms 1/4W or 1/2W carbon or metal film
C1 0.022uF Sprague orange drop
C2 120uF 35V
C3 10,000uF 35V
C5,6 2x470uF 35V
Heatsink Avid 531202B02500G
BER182-ND HEATPAD TO-220 W/ADH .009" SP900
LT1084CT#PBF-ND IC LDO REG 5A ADJ TO-220-3
1/2 A fast blow fuse
I couldn't figure out what "KK1" and "SK104" shown next to heatsink refers to.
Did I miss anything?
thank you John, did you do any measurment (ripple etc.), are you able to provide that kind of info as well?
I am just getting to modifying the psu on my pc (Gigabyte GA-G31M-S2L for cMP2) and so far I have followed Dave's suggestion of adding capacitance to the P4 connectors - this made a huge difference. What I would like to do is have a separate psu purely for this supply - it would seem that 12v @ 2A (max) would be sufficient.
I have tried to work out what the changes are to your circuit / parts to allow for this - failed as i am not much good with circuits.
So I was wondering if you would be so kind as to supply me with the changes needed - I am in the UK so its 240volts mains (they say 230 but it never is!).
> > ”I am just getting to modifying the psu on my pc (Gigabyte GA-G31M-S2L for cMP2) and so far I have followed Dave's suggestion of adding capacitance to the P4 connectors - this made a huge difference. What I would like to do is have a separate psu purely for this supply - it would seem that 12v @ 2A (max) would be sufficient.”
I’d better own up as sondale's idea of using a linear supply for the P4 12 volt line might have followed from off-list exchanges with me. I noted that:
“I repeated [on Sunday night] the measurements I did of the current drawn by the E7200 chip [on the G31 motherboard]. On power up. it draws about 1.5 amps. As ‘underclocked’ BIOS settings kick in during POST, this falls to about 0.6 amps. Once the OS is loaded and while playing music (with a NOS DAC), the chip draws between 0.4 and 0.5 amps or between five and six watts.
”Curious is the fact that the PSU I'm using allows one to trim the voltage. I can run the chip perfectly on 11 volts instead of 12. However, despite the reduced voltage, current draw seems to go up, not down. I do not know what is going on.
“What the measurements do suggest is that the 2-amp capacity of JS's design is meaty enough for the likes of the E7200 but could well be too small for older or more powerful chips. Moral: measure first! ”
I’m grateful to sondale for confirming that adding smoothing on the P4 12 volt line improved the sound of his system.
However, I would like to stress that using a linear PSU on the same line is a significant step up from that. It is not a "nuance" but arguably the biggest hardware-derived change to sound quality that I’ve made in the 15 or so months I’ve been building dedicated audio computers.
The device I used was a 5-amp US-made unit built like a tank and using the venerable ua723 regulator, a chip design now pushing 40 years old but still doing well. The next step is to try JS’s design set for 12 volts.
I would encourage users of cMP2 or similar high-resolution systems running on suitable kit to consider giving it a try though with the following caveats:
1. I have tested a separate PSU on the E7200 and on an eight-year-old Athlon XP chip only though I dimly remember other list members reporting doing the same on other chips;
2. the PSU should really be powerful enough to run the chip even if default speed settings are restored;
3. There is a chance, even with modern devices, that running with no load on the P4 12 volt line could make some PSUs unstable - it shouldn’t but it might. It needs testing;
Can anyone explain to this baffled tinkerer why the current drawn by the chip stays the same even or even rises if the supply voltage is dropped?
Do you see any reason why your power supply wouldn't work equally well with other small computers like Excito's Bubba 2 (see link, below), and what would it take to accomplish the same things you've done, minus the waste heat?
It looks like Bubba2 takes 12V at 4A, that takes a different design, this on is only good up to about 2A. Conceptually its the same but has different part values. I'd have to run the numbers to come up with a good design for that.
There will always be heat dissipated in a linear regulator. The whole concept is that they are turning the excess voltage between the raw supply and the output into heat. How much heat depends on several factors: the current the load pulls, and the difference between the average voltage of the raw supply and the output voltage. That "average" in the above is critical.
What comes out of the raw supply is DC with some ripple. Each regulator design has what is called the drop out voltage, the raw supply must never go below the droput voltage. In the case of the LT1084, the drop out voltage is 1-1.5V depending on the current. So for a 12V reg the lowest point of the raw supply must be above 13.5V. So even if you had zero ripple coming out of the raw supply you would still have 1.5VX4A = 6watts to dissipate from the reg. Of course you always have ripple on the raw supply. So the raw supply has to be high enough up that the lowest point is high enough, but the power dissipation comes from the average voltage. So the lower the ripple the closer the average is to the lowest point. I've designed these supplies so the ripple is pretty low (.5V or so) so the average is only slightly greater than the low point. So with the raw supply as low as possible you will have about 8watts.
You also have to worry about main voltage and transformer specs etc. If a supply is designed for 115V and your wall voltage is 120 what comes out of the transformer will be higher, so more power will be dissipated by the regulator. Also the transformers don't come in infinately small steps in voltage rating so you have to choose one that is high enough, but it might wind up being a volt or two higher than necessary.
If its really important to get the voltage reg to have as low a power dissipation as possible you can run the transformer off a variac and look at the waveform going into the reg with a scope and adjust the variac until the lowest part of the waveform is just above the dropout voltage. But then if the mains voltage drops due to the neighbor's air conditioner trurning on you go out of regulation. Thats why I always build some margin into my designs and live with the heat.
john, i jumped the gun and ordered the parts for the power supply from digikey, and scored the choke from mouser (they still had a few in stock).
i was going to build one up as a straight 12v for the sim phono preamp, which draws very low current, and figured the other one could be devoted to the sound card (and give me a good excuse to buy one, as soon as someone has them in stock!).
what sort of changes are required, other than the trafo, to accomodate either higher or lower current delivery? also, i have a custom wound electraprint trafo, with electrostatic shield, which is rated at, iirc, 12v and 1 amp. would that work as a drop-in replacement for the triad, in the straight 12v design?
thanks for hadling lots of questions,
whether that 12V 1A is going to work depends on the current draw. At 1A, no way. At much lower current it might, let me know what current you need and I'll see if it will work or not.
Lower currents will work fine with the design as is, its just that the peak currents through the diodes can be made a little lower with a design optimized for the low current. It would be something like (these are off the top of my head not real numbers) at 1A you have 1.5A peak. At 0.5A you have 1A peak but you could get 0.7A peak with an optimized design.
Higher currents are a different story. Much above 2A is going to stress the chokes badly, they will get hot and probably will start buzzing. For higher currents it will definately take a different choke (the one that is one slot down in the 159 series. Its like 4mH I think) It will also take different cap values as well.
the word on the sound card (which i ordered today) is that it draws a max of about 20 watts on the 12 and 5v lines combines, but "typically much less". this from a 3rd party source who is reviewing the card, and probably got the info from a legit source.
i bought the big 12v tranny that you mentioned 36-2200, and what i think are some bigger heatsinks than you speced (to use for the 5 v line. i may try it as is, and see how it sounds just connected to the molex from the antec psu that came with my fusion case, so in case it sounds horrid, i won't waste the effort on the psu.
most of the parts are pretty cheap, and look like worthwwhile parts-bin inventory anyway, so if you;re game to do some tweeking, so am i. the card may be here in time for the weekend, so we shall see what happens.
sounds like i may be able to get away with building the supply pretty much as-designed, but using two chokes and maybe even two of everything except the tranny for the two different voltages. we shall see......
Unfortunately that might mean 3A on the 5V and .5A on the 12, 2A on the 5 and 1 on the 12 would be fine.
Send me an email with the specifics of what parts you have and I'll see what I can come up with.
Just your average linear power supply.
Any good electronics textbook would have the same circuit along with the theory you need to understand how to customize it to fit your needs.
> > "Any good electronics textbook would have the same circuit along with the theory you need to understand how to customize it to fit your needs."
I'm sure that textbooks give informed readers the basic theory but not all of us on this list are competent to use it. Besides, the difference between a textbook circuit that works and a practical one that works well needs theory informed by experience and an eye for detail.
I do not understand why folk feel the need to be ungracious about a useful contribution to the list. If they think the design lacks merit, they should of course say why but it is wrong to suggest that John has merely rehashed a design note.
I've looked at dozens of power supply designs over the years, many of them very good. I've seen several broadly similar to John's (as you'd expect) but none identical and certainly not in the data sheets or in this context.
I find his argument about its merits convincing and it looks easy to build. It seeks to address a significant issue with PC audio and adds support to the notion of "satellite" players that he has been plugging.
I'd like to pursue the idea of adapting cMP2 to use such a device as (a) I don't know Linux but am reluctant to make the effort to learn it right now only to find I can't get it to work with my kit and (b) I am very satisifed with cMP2 and would need some convincing that Linux alternatives are better. (Never say never and all that but also a step at a time.)
Nevertheless, the core argument for using the likes of Fit-PC is strong and John's proffered PSU design adds to it. Many thanks for the effort.
Not really. This design has been carefully optimized to give very low current spikes through the diodes and transformer secondary. Most modern linear designs will just use a large cap after the diode bridge. For a 1A load that will produce somewhere between 12 to 20A current spikes depending on what capacitors you used. Even if you go with a standard PI filter out of the textbook and design it the standard way you will still get current spikes in the 3-4A range, alot better than 12-20A, but this design gives you current spikes in the 1.5A range.
I never stated that this was a new unique never seen before power supply design, just that it was a carefully optimized design targeting parameters I have found to be important for audio purposes.
1) Choke in low power supply. No, this isn't the first one to do it -- Twisted Pear Audio uses small chokes in their low power supplies, for example, and DHT filament supplies have contained chokes for many years, but it isn't exactly "average".
2) TUNED secondary snubber network which as John has mentioned in previous posts can make a measurable difference with power transformer resonances.
4) Lower value first C after diodes.
5) Extremely low DCR
6) Tested with specific components as a SYSTEM to be low noise, etc. This is a valuable and underrated aspect, IMO. Most of us don't even have the appropriate test equipment (or free time) to check things like transformer resonances.
Thanks for the design. I plan to build some of these.
Question: Does the secondary RC snubber use the same value for the 5 volt version?
How about for a 12v version with a higher rated transformer?
Cappy, the RC network is described in detail by Jim Hagerman in a white paper linked on his website. The values are calculated based upon the specific circuit. However, if I recall correctly, John has experienced benefit from using this same combination of values for the RC on several of his components even though they require a range of voltages....at least the lower voltage ones.
John, i, and several others, are contemplating building linear supplies for the pci-e version of the xonar soundcards, which, i believe, take +12 and +5 via a std molex connector. would you recommend building two whole supplies, or just using two of the "regulator modules" fed from a common choke regulated front end? what voltage is available at the output of the l-c section of the ps? would the trafo you spec-ed be sufficient va rating to do this?
thanks for your contributions.
you can certainly run two regulators off the same raw supply, you just have a lot more voltage drop for the 5V reg so the heatsink will have to be bigger.
The design shown is good for a max of about 2A. So if the combination of the 12V and 5V is less than 2A you can do it.
For 12V you want to use the VPS36-1200 transformer with the secondaries in paralel. Thats one will be right on the edge for 2A, if you do need 2A I'd go with the VPS36-2200.
If you need more than 2A total you have two choices, make 2 supplies, or I can design you a higher current version. Or if the current will be a lot less let me know as well. This design was optimized for the 1 to 1.5A range. It will work well up to about 2A, and it will work as well at lower currents but it will not be optimized. For example if you know you anly need a combined total of 200mA I can design a supply optimized for that.
john, i don't have the card yet, and the info online does not specify current consumption, or even if it actually uses the 12 and 5 volt rails, or just one or the other.
i may try to contact asus and find out, or check the boards to see if anyone who has one even knows.....
'til then, i am pricing out parts, and seeing whether i would target building two whole supplies, or a dual-voltage one.
i've also got a couple other ps projects pending. i've got some custon wound electra-print tafos, with electrostatic shields. one is for a +12v low power supply for a phono pre, and i've also got a headphone amp that uses 9-12v, which have stealth bridges, and a regular r-c configuration along with the low-dr
opout regulators. i may try your supply design on one or both of those as well.
thanks again for your generous offer.
Thanks for your great work, but fit-PC Slim specifications declares power 12V single supply +/-20% (9.6 minimal), not 9V.
You can change the 300 to a 330 and you will get 10V. I would not go 12V with the speced transformer. But there is no need for 12V. The voltage just goes into a 5V regulator in the slim, all going to 12V gets you is a hotter box because the regulator has to drop more voltage.
If you really want to go with 12V use the VPS36-1200 and run the secondaries in paralel instead of in series as in the schematic. Change the resistor values to 430 and 51 that will give 11.8V. If you use 1% resistors you can tweak the voltage even closer. But again I don't think that is necessary.
Why wouldn't the 16VAC @2.7A rated Triad VPS16-2700 in your present design not be sufficient for a 12VDC regulated supply, especially with lower voltage drop Schottky diodes, low drop-out voltage 3-terminal regulator, and significantly lower current draw than the transformer's rated 2.7A for 16VAC. I don't see a need for going with an 18VAC secondary on the transformer.
When using 115V main the bottom of the ripple is going to be about 12.5V which is a little less than whats necessary for regulation. Yes the voltage will be a little higher due to less than max current draw, but not by all that much, these transformers actually have pretty good regulation characteristics. If you can guarantee to be running at 120V at all times then it would probably work fine. I'm just trying to give enough margine so that at worst case conditions (high current draw and low mains voltage) it will still be in regulation.
You can adjust the value of R2 and R3 to get 12vdc out if you like or use the fixed value version of the IC regulator: LT1085CT-12 or LT1084CT-12. (In the case of the fixed regulators, simply ground pin 1 and eliminate R2 & R3).
The power supply is pretty much text book & app note from the chip maker. It's a basic bridge rectifier followed by a pi section filter and low drop-out voltage adjustable regulator IC made by Linear Technologies. While it's more common to find a pi filter with choke in a high voltage tube power supply, nothing says you can't use one in a low voltage setup.
The link does not work; it says personal pages are not available.
I just use a linear laboratory supply with voltage and current readouts. The Chinese ones are not expensive and are ok loaded to about 1/2 rating. Otherwise they get hot. I just increase heat sinnking by clamping on additional fins.
Many thanks John!
As usual you are one of the most communicative technical experts on this site, and one of the most generous. Though your power supply design may not seem sophisticated to some, I appreciate your willingness to share the simple genius of your solution to those of us who are in constant need for cost-effective and efficacious power supply resources.
Having been a dedicated user of battery power supplies for my entire audio system for over 15 years, I can see that with a few twists on your layout, I can replace a few batteries!
All the best to one of the best.....
Post a Followup:
Post a Message!
This post is made possible by the generous support of people like you and our sponsors: