Converting a ATX PS to 13.8v

Glad to hear it. It's always better to start off with something designed for the job IF the price is reasonable.

PlainBill

There's a joke in here somewhere.

I didn't intend any offence, Dave - probably could have picked my words better. All I was trying to imply, is that switch mode power supplies are a bit of a black art, as you have discovered, and unless you are working with them all the time, and thus understand the subtleties of some of the circuitry and regulation techniques which they employ, then this project is quite likely to tie you up in knots. I must admit that I didn't look too closely at the application schematic to see how exactly they went about regulating the thing, but when I saw the two undedicated adjustable zeners in there, I assumed that they were probably making use of at least one of them in the primary feedback loop. It is a common technique to use an (usually external) adjustable zener to achieve main rail set voltage and regulation.

Franc is probably right in this case with his proposal of altering the resistors on the IN pin. You may well find that when you put the pot in, you actually had the desired effect of increasing or decreasing the output voltage, but then fell foul of the under / over voltage lockout feature, which I think I said that you might have to disable. If you put a digital meter on the output, and watch very closely, you might see the voltage start to alter up and down, before it trips out.

Arfa

On Wed, 05 Aug 2009 14:48:07 +0100, "Dave Plowman (News)" put finger to keyboard and composed:

You are not actually altering the IN voltage. It remains at +2.5V due to the action of the error amp. The other input to the amp is internal to the IC and is fixed at +2.5V.

Were you testing the supply without a load?

If you did not remove the +5V sense resistor, then the output may fluctuate as the load varies.

- Franc Zabkar

Right - so what is happening to alter the output voltage?

No - I was loading it with a 55 watt car headlamp bulb.

Seems pretty stable - at least for my purposes.

No offence taken - it is Usenet, after all. ;-)

Which is why I'm happy to experiment and hopefully learn.

Snag is my actual circuit differs considerably from either of the examples given in the spec sheet - including a sprinkling of external zenners. But it's beyond my patience/available time to trace it all out properly.

The network between actual outputs and sensor inputs is far from simple - the spec sheet just shows a direct connection.

Indeed.

On Thu, 06 Aug 2009 10:57:08 +0100, "Dave Plowman (News)" put finger to keyboard and composed:

I don't know much more about the black art of switchmode PSUs than the average tech, but AFAICS in this particular case all one needs to understand is that the error amp inputs must be equal when the supply is regulating correctly. If the output is too high, then the error amp reduces the drive to the chopper (by narrowing the pulse), thus bringing the output back down. Conversely, if the output is too low, then the error amp increases the drive to the chopper.

I mentioned this possibility because I had a fake 400W PSU (really only 200W) where the +12V rail moved about in response to changes in the +5V load. Also ISTR that some HDD enclosures include a dummy load resistor for load balancing.

- Franc Zabkar

Why wouldn't you soak test at worst case numbers rather than nominal?

Certainly depends on the power supply, but there needs to be enough flux in the transformer to make the outputs work. If the 5V is the regulated output...or 3.3...or whatever, and there's no load, there won't be enough flux to make the 12V work at any significant load. You're gonna need to redesign the circuit to sense the 12V output so you can get the current you need. The rest is just resistor ratios.

But wait...if there's no load on the other outputs, you run the risk of transients charging the caps till they explode.

And depending on the physical location on the circuit board you pick to connect your new components, you may make significant changes to the transient response. With high-currents, one point on a trace ain't the same as another. You can compensate the feedback loop, but load transient induced spikes can/will cause you grief if your layout doesn't consider it. Building the transient test fixture to test your design is non-trivial.

There's an interesting ratio. How much you are saving on the power supply (including costs of design/debug/labor/testing/documentation/training) divided by the cost of blowing up all the stuff connected to it.

And what do you do when it breaks...or you need another one and the original PS product is no longer available?

Few people understand the subtleties of power supply design. It's a black hole (art) that sucks in everything around it.

Go buy a real honkin' power supply and use a slew of small circuit breakers to power your individual test stations. I don't know your burdened design labor rate, but modifying PC power supplies is often false economy.

Your time might be better spent simulating the automotive power environment for your tests.

littelfuse appnote an9312 is a place to start.

Are we having fun yet? mike

Worst case for soak testing wouldn't be lower than nominal.