Need voltage regulator 20v at 3A - ideas?

[andyjevans]

I have an application (tube DC filament supply for 10Y DHT in filament bias) requiring 20 volts adjustable +/- 2v at 1.25A steady current. I'm planning on at least 3 amps turn on surge. Currently using a LM1084 which though rated 5 amps is only rated 15v, and it cuts out on turn on.

So what's a simple circuit for achieving a comfortable 20v at 3A with plenty of margin? Supply voltage is choke input and can be anything from 24v to 29v depending if I put a small cap in front of the choke (280mH).

Need to get this built, so all ideas, links and schematics very welcome!

andy

[db!]

Have you looked into the LT1083?

[abraxalito]

Is a (switching) buck regulator out of the question? It could handle a wide range of input voltages so the choke in the pre-reg might not be necessary.

[birdyman]

You can use 15 v regulator with 5v zener attached to its ground pin making "ground pin" of 15v regulator 5 v higher from common ground. now you have 15 + 5 = 20v regulator.

[andyjevans]

Many thanks for your contributions! First the LM108x series - the LT1084 may work better than the LM version - possibly it's better built to a higher spec, though generically it should be pretty similar. The LT1083 does look as if it would do the job - but ouch! Expensive at over £12 from Farnell. There is a LM1083 version if I can find a supplier in the UK.

I don't want to eliminate the choke - with filament bias the signal goes through the filament supply so the quality of the filament supply is absolutely critical - it doesn't just have to do the job it has to be sonically as perfect as possible!

I'll look at the zener option - that might help. I was discussing this with the London Audiocircle in the pub last night and suggestions were:
- a Darlington pair instead of a voltage reg
- a thermister input so the voltage would come up more slowly and maybe not trigger the cutout on switch on, which is the current problem.
- a 1 ohm resistor in series with the positive leg before the reg, again to try and avoid triggering the cutout.
- using the LM1084 as a CCS

More suggestions?

andy

[kavermei]

A 317 with one or two 2N3055 helpers maybe?

An NTC is a good idea to prevent triggering the protection.

K

[AndrewT]

Hi,
if the circuit needs 1.25A then I cannot see any reason for supplying a constant (adjustable) voltage.

I would use a Constant Current Source (CCS) set to 1.25A to feed the circuit.

A 317 can meet this current requirement but if the voltage drop across the 317 increases then the total dissipation of the 317 also increases.
Might be better to use a pair of 317 in parallel each set to ~630mA

The big advantage of the CCS feed is that current surge into the cold filament does not exist.

[andyjevans]

Yes - a CCS needs trying out for sure. But in filament bias there is a 10 ohm resistor in series with the filament, which is also directly heated so comes up at once. So the device doesn't actually see a short circuit at switch on. But the choke input seems to send a surge.

andy

[AndrewT]

Originally Posted by andyjevans

So the device doesn't actually see a short circuit at switch on.

what short? Post circuit.
The cold filament does not have zero resistance.

[RJM1]

I recently built one of these to power a laptop computer(3A). With the 20 ohm pot it will give you 17.9 to 21.7V. You can add an output cap if you want to improve HF rejection but this will give you about 68db power supply rejection at 120hz without any capacitors.

M2 needs a good heatsink.

Needs about 3.5V Vin to Vout.