Can i use a computer power supply to power audio amplifiers?

[N-Channel]

MODERATORS: Please bump this to the Power Supplies Forum. THX!

Hi Carlos,

Yes, this is a great thread you started here!

There are several threads dealing with this subject:

ATX (computer) power supplies for power amp
Anything I can use for DIY amp from PC powersupplies?
Line-powered SMPS for Power Amp?
ATX PS Gainclone?

Using ATX PSU trans for 12V SMPS?

I like the last one here as it has some very good advice. it's not the longest one, but it hasa some really good advice.

I have a few comments I would like to share about using old AT & ATX power supplies for sudio applications:

1) Yes, they can be used for audio applications, BUT, some modifications will be necessary. The extent of modification will be determined by what type of amp you want to run. For example, if you want to run a gainclone, or any amp (discrete or chip-amp) that requires a bi-polar supply in excess of +/-12V, then you're in for at the very least, a trasnsformer rewind, major diode rectifier replacement and re-working of the output voltage sensing section.

2) In my experience, AT supplies lend themselves to modification a little better than ATX ones, unless you're going to use that +5Vsb housekeeping supply to run some things while the main converter is off. But please, Carlos, don't let this dissuade you from trying an ATX.

3) Jackinnj has posted several times on the DIY forum, the link to QEX magazine article that covers converting an old AT or ATX supply to put out +13.8V at something like 15-20A, specifically for use to power ham radio gear. Although it could just as easily power a BLT chip amp, like an NEC uPC1230H2. Perhaps he could be pursuaded to post that like here. Please Please Jack!

4) Carlos, referencing your post about the noise at 30-40kHz, this is the PWM chip's switching frequency (usually 36kHz). For most AT & ATX Power Supplies, this PWM chip is almost always a TL494. Sometimes, it is labelled "7500" or "KIA7500", but it is the same chip. Since the '494 is a dual output PWM chip, the chip's oscillator frequency is actually double the switching frequency, because it has to turn on each transistor alternately for one complete cycle. The chip's oscillator frequency can be easily be changed by substituting the timing resistor, R(t), for a different one. Be careful on changing the frequency more than +/- 15%, because the transformer windings, core material and alot of other things in the PSU are designed around this frequency. If you were going to kick it up to, say, 100kHz, or even 200kHz, you are in for a major re-design. A little trick to deal with the noise, is to replace the timing resistor with one, say, 10% lower and putting a panel-mounted potentiometer of 20% of the original timing resistor's value in series with this pot. This will give you a frequency adjustment of +/- 10%, allowing you to "slide" the noise off the band you are on, or ourt of thre audio range. If the noise re-appears, simply re-adjust the pot, now called a tuning pot, until the noise again disappears.


See this link about the '494 to get a good idea how it operates:

http://www.onsemi.com/pub/Collateral/TL494-D.PDF

Page 6 of the Datasheet gives some insight on how to set the oscillator frequency.

5) I have to agree with EVA on not using a ferrite toroid in place of the existing transforemr core. Adding the three layers of transformer tape would make it even more difficult to deal with winding. Since you already have a good core, use it.

I think I have emptied my head out of all I can think now, since I have been typin this for about an hour now, so I will stop here.



Just throwing my two cents' worth in the pile.

'73, de N8XO,

Steve

[N-Channel]

I forgot one other thing, Carlos. In your pics of the 200W AT board there, see the missing components L1, C2, C3 &C4?

These were left off to minimize the cost of making the unit. They form the AC line filter that keeps the switching noise generated by the TL494 PWM chip from getting out onto the AC line. Some manufacturers cut these necessary components out, just to shave some of the costs off, and boost their profit a bit. When you're making millions of units and you save, say, 50 cents, then the "savings" can add up quickly. however, this occurs at the expense of the supply not meeting EMC or noise requirements of the various regulatory agencies.

If you can calculate the values of these missing components, and replace the jumpers with the correct values, then you will attenuate the noise considerably.

Also, see the jumpers across where L5 &L6 are supposed to go? Same thing again, just trying to shave a few cents off the manufacturing cost. If you can replace the jumpers with a couple of inductors at the right value, you should improve the noise reduction a little bit more.

[Eva]

Some useful schematics for better understanding of simple off-line half-bridge self-starting SMPS:

Typical ATX power supply:



Typical AT power supply:



A custom made 13.8V 40A PSU using the same topology and working principles as above:



Anyone interested in PC PSU modification should study and understand these schematics first (I take that kind of things to the sofa or the bed for detailed analysis).


N-Channel:

Believe it or not, these slow-switching old designs based on bipolar transistors and positive-feedback-turn-on are really EMI-quiet in comparison with the MOSFET things full of RF ringing that are more in fashion nowadays, they may even pass conducted EMI tests without any filter.

[N-Channel]

I knew you'd chime in right away, EVA!

The 40A Supply comes from the DEC 1998 issue of QST. It was a good article, and it is good for 40A continuous because the author stacks something like 4 or 5 cores together to enable the supple to handle that power level.

The real key here is the saturation value of the driver transformer. This needs to be kept much smaller than the that of the main power transformer to enable the self-starting, since the PWM chip is on the secondary side of the supply.

[Eva]

The transformer of the 13.8V 40A power supply is strongly oversized, maybe because the author didn't calculate it properly, because he didn't got the self-starting thing right, or because he wanted to operate the cores at very small flux densities. With a proper design, the same power may be just obtained from a single E42/21/20 transformer. As you know, in my 32Khz power supply with a similar topology (full bridge really) I'm getting 1.8KW from just two E42/21/20 transformers (15V 125A output fully regulated). And my transformers are usually operated around 100mT (if I remember properly), quite far from saturation, while having a lot of free winding space.

[N-Channel]

Are you using MOSFETs or bi-polars? What is your experience with driver transformers -v- floating-driver chips like the IR2110?

[Eva]

Don't you remember? That thread covers it:
0-15V 0-120A adjustable PSU attempt with average current control.

It's a full bridge made with just four transformer driven TO-220 bipolar devices, and tricky base drive. Also, I have a few IR2110 in my parts box that I have not tried yet, I think that each one costed me as much as the four MJE13009 bipolars

[N-Channel]

Yes, Eva, now I remember it. Ouch! That's expensive! I have some '13009s laying around, but I was gonna use them to upgrade some '13007s in another AT supply I am thinking of modding.

Hey Carlos, sorry if we've strayed off subject here. Hope all this back-and-forth is helpful for you in your decision on how to best proceed........

[Eva]

We are not so off-topic: My PSU started as a modified AT PSU capable of 15V 20A output (I removed redundant stuff, added a voltage adjustment pot, replaced rectifiers and increased a bit the current limit). Then I bumped it to 36A with additional (breadboarded ) synchoronous rectification and by rewinding the transformer to a convenient 7:1 turns ratio. Then I abandoned the AT case, PCB and transfomers and jumped to a full-bridge with a E42/21/20 12:1 transformer and a masive output inductor to get 72A...

[N-Channel]

OK. What do you think of my idea in trimming the switching frequency +/-10%? Pretty cool, eh?