Working with adjustable voltage regulators

Hi guys, I'm building a bench power supply out of an old PC PSU.

I want to be able to regulate +12v so I can have access to roughly 0-12v.

I was looking at the LD1084 (Datasheet: http://www.st.com/web/en/resource/technical/document/datasheet/CD00003086.pdf)

Looks like it can provide up to 5A, which should cover anything I'll be working with.

I've never worked with adjustable voltage regulators before, what supporting circuitry will I need to make this work?

Thanks!

That regulator requires 1.5V minimum differential across the regulator, so you can only get to about 10.5V if feeding it 12V. The minimum output is 1.3V.

Do you -really- need 0V output? Everyone says they want a sig gen that does 0-30MHz, but 0Hz is DC.

I'm in the process of modifying a PC power supply so that it puts out about 18V from the 12V output. My plan is to rewire it so that the 3.3V secondary windings go to the 5V line, and a voltage divider from there makes it look like the 3.3V output is still there. Then the power supply should boost the PWM so that the 3.3V output will be putting out 5V, about 1.5 times more. So the 12V output should put out 18V.

OId PC PS used to just regulate 5V and trust the other outputs to be within spec. Newer power supplies regulate 3.3V, but check all the other voltages. So I'll need to spoof the other outputs back to the chip.

I'm not sure yet if it is really checking each one, or just feeding them all into a voltage divider so if any one output goes dead, it shuts down.

I am going to use the same linear regulator, and use 10 turn potentiometers to adjust voltage and maximum current. I really, really hate having a "coarse" and "fine" control, that is the mark of a cheap POS power supply.

polymorph: That regulator requires 1.5V minimum differential across the regulator, so you can only get to about 10.5V if feeding it 12V. The minimum output is 1.3V.

Do you -really- need 0V output? Everyone says they want a sig gen that does 0-30MHz, but 0Hz is DC.

I could make due with a minimum of 1.3v.

So how do I control it? A potentiometer I assume, but how do I spec/size one?

Oh, and a huge heat sink. Worst case, 5A @ 1.3V output and 12V in means 5A x 10.7V = 54W. Quite a lot of heat.

See page 6 of the datasheet. Vo is the maximum, R2 will be the potentiometer.

So if you do modify the PC PS to put out at least 14V so that you can get a max of 12V:

Vo = Vref(1+R2/R1) Vo/Vref - 1 = R2/R1 R1(Vo/Vref - 1) = R2 120(12/1.3 - 1) = 987 ohms

So a 1k ohm 10 turn front panel control would work great to make a 1.3 to 12V adjustable power supply.

Keep in mind, an unmodified PC power supply regulates the 3.3V line, and without a load the other voltages will be all over the place. Usually too low. So put a load of about 100mA to 500mA on the 3.3V output, and the 12V line should be OK.

polymorph: Oh, and a huge heat sink. Worst case, 5A @ 1.3V output and 12V in means 5A x 10.7V = 54W. Quite a lot of heat.

Yeah... My rule-of-thumb is, "For more than 1 Amp, use a switching regulator". This is especially true with adjustable regulators/power supplies.

The power (heat) dissipated in a linear regulator is the current x the voltage dropped across the regulator. So at lower output-voltages there is more voltage dropped across the regulator and there is more heat. I've used a lot of 78xx regulators and they get very-hot at around 1 Amp, even with not much voltage dropped across them. The good news is, the LD1084 is thermally protected so it will just "shut down" if it overheats.... You won't burn-up the part.

A switching regulator requires more parts (including an inductor) and they are more complicated to design, but they are nearly 100% efficient so heat is MUCH less of an issue.

And it gets worse if you mod the PC supply to put out, for instance in my case 18V in order to get at least 15V at the highest point: P = (18V - 1.3V)5A P = 84W

That is comparable to a computer CPU.

For mine, I'm going to do some tests with a few heat sinks, with a small fan for forced air cooling. This is for myself and will be done as a club project at OlyMEGA ( Olympia, WA, USA).

Back in 1982, I built a “very linear” based bench power supply using a junkbox 30V transformer from my stash and a handful of parts. The only expensive item was a 5K ohm 10 turn potentiometer for voltage adjustment. The design was straight from the manufacturers datasheet. I guess my point here is that 1) it still works 2) seldom if ever gets warm and certainly not hot 3) Proves that I have skills from longer ago than I can comfortably remember 4) over engineering something means that yes, you paid a little more, but it lasts longer. 5) use heatsinks.

Note: I seldom make it provide more than 500mA The regulator uses a TO-3 package that is designed for a heatsink. I use a small finned sink and the metal case for cooling.

So what does all this mean. Build the reference design from the datasheet (or sadly, since the LD1084-adj does not really have a nice reference design like the LM317, use the datasheet as a guideline.) It actually appears that the design of the LM1084-ADJ is close enough to the method employed by the LM317 that you could put a potentioemeter between ADJ pin and GND and a output feedback resistor to Adj pin and be done.

Note: Make sure you have the LD1084-ADJ not one of the fixed voltage ones.

It actually appears that the design of the LM1084-ADJ is close enough to the method employed by the LM317 that you could put a potentioemeter between ADJ pin and GND and a output feedback resistor to Adj pin and be done.

Not only -could- you, you are -meant- to use it in this way.

I expect the reason the datasheet does not have more examples is because there are so many reference designs and examples out there for the LM317, and they all equally apply to the LD1084V.

Looks like ST makes the LD1084V and Linear makes the LT1084 in only one version, the adjustable one, while TI makes the LM1084- series in several versions, so good advice to make sure you get the adjustable version.

BTW, Linear makes the LT1083, which is rated for 7.5A. Edit: oops, mistakenly said LT1085 the first time...

Also: in my calculations above, I should have said 1.25V to be more accurate. In which case, 110 ohms and 1k ohms will allow adjustment from 1.25 to a little more than 12V.

Oh, yeah... a reverse biased diode connected from the regulator's input to output is not a bad idea.

This prevents a situation happening where the output pin has a greater voltage on it than the input pin. Not sure about this series of chip, but the LM317 and LM78xx regulators can be damaged by this. Possible if you have an outside extra capacitance connected to the power supply, via some external circuit or you are experimenting with supercaps, and you turn the power supply off.

Thanks for the advice guys.

How would I go about boosting the available voltage? It would be cool to be able to get 18-24v out of it.

You will need to determine which control IC is on the board and let us know.

If it is a TL494, it is quite a bit simpler as that chip has been around for a very long time and is well documented.

If it is labeled "3528", it gets more complex. I have a half-dozen PC power supplies with that chip I'm figuring out how to mod, but haven't had time to dig any deeper. Try googling for "FSP 3528" without the quotes.

This is giving me the impetus to work on this project again.

This forum seems to indicate that the Fairchild KA3511 is the publically available version of the FSP 3528: http://www.badcaps.net/forum/showthread.php?t=22171

And here is the app note mentioned, AN4003: http://www.alldatasheet.com/datasheet-pdf/pdf/50691/FAIRCHILD/AN4003.html

Looks like we could just rewire it a bit, take out a few resistors and replace one or two, and thereby boost the 12V output. Also, move the higher current shottky rectifiers to the 12V output.

Modding the psu is a good idea.

But if you can't, leave the 12v rail as it is power directly from it for 12v applications.

For any voltage lower, use an lm317 or a switching reg, you could use a pass transistor and pass control of the regulated output to a transistor (unless it's a switching reg)

Already covered. The 1084V or 1084-ADJ is an LM317 on steroids, up to 5A and low dropout.

Making a switching regulator to cover that wide an output range could be interesting. And noisy, potentially.