I have an SMPS which puts out 35V at 5A, and it would be nice to have this adjustable. I have a couple of LM338s in my bit box, so I thought that I could just attach these to the SMPS to provide a box which has an adjustable voltage output from a couple of volts up to 30-odd, and also has a current-limiting function such that it can put out lower voltages at fractions of an amp.

There are numerous sample circuits available on the web - though nothing seemed to be quite what I wanted. This usually means that it's either not possible, or that it CAN be done, but this is not really the way to do it! I understand that running at lowered volts and amps means that I have to lose a lot of heat - luckily I have some quite big heatsinks in my box as well.... so I put together a schematic.

There may well be other problems that I am completely ignorant about. I would be grateful if someone could cast an eye over the proposal an tell me whether they think it will work or not before I put soldering iron to pcb...

thanks...

 

If you are going to go this track, I would think the current limit stage should be placed before the voltage regulator stage.
And the 200R pot will need to be a high wattage type.
It may be better to use switched resistors, all in series with the switch shorting out sections to adjust the current.

 

One problem with this is that 200Ω pot has to be able to handle 5A and that's not likely available.

 

And the regs will need a big heat sink!

 

With a resistor in parallel with the pot, the resulting output-voltage-change-per-rotation-increment will not be linear. You will much better adjustability at low output voltages compared to high. Nothing wrong with that, just something to be aware of.

The meter probably will not like a power source with an output impedance of over 600 ohms. Better to measure the meter's operating current using a real 12 V supply or battery, and then use a resistor and zener diode for the meter's power source. Because the bulk supply output voltage is so high, a 7812-type regulator won't work, but something small like an LM317L would do it.

The LM338 feedback point for regulation is its output pin. In your schematic, there are two unknown impedances between that and the external load. This will have a very real effect on the overall output regulation, making it worse. Better to put the current regulator and amp meter before the voltage regulator, so the vreg's output pin is connected directly to the load. To do this, you'll have to split the combo meter into two separate devices.

With this approach, the current going through the vreg to make it function will appear as part of the amp meter's reading. Unless you want to measure load currents of 10 mA or less, this shouldn't be a problem.



ak

 

If you are going to go this track, I would think the current limit stage should be placed before the voltage regulator stage....

What would be the advantage of that? I can certainly see the advantage of switchable resistors - a lot easier to obtain - but I am not too sure of the values - there seems to be no readily available calculation for this application of the LM338...

Thanks to all for the responses!

 

What would be the advantage of that?

Anything you place in series with the output between the voltage regulator and the load upsets the voltage regulation.
As @AnalogKid mentions in post #5.

 

but I am not too sure of the values - there seems to be no readily available calculation for this application of the LM338...

The current is equal to Vref/R or appx 1.25/R. So a 1 ohm resistor would limit the current to 1.25 amps

 

All this rather surprises me - I had simply taken these circuits off the datasheets and the web, and was expecting that there would be major issues, sufficient to stop the project in its tracks!

But from what you say, it looks as if much of the idea might work...

I will redo the schematic tomorrow. One thing I can't do, however, is separate the V/A meter into two - it's one of these:

https://i.ebayimg.com/images/g/U8cAAOSwzMtdCmnh/s-l1600.jpg

Not sure if this matters...?

 

165W(!) of heat needs a lot of LM338 regulators on an enormous heatsink with a fan. Or you can let a few regulators overheat then shut down but for more than a few times might cause thermal fatigue and failure in old regulators that soon will be obsolete and not available. The metal LM338 costs $84US today and only a few are available.

The resistance of the current meter reduces or eliminates proper voltage regulation.

Adjust current limiting with a switch and a few large resistors.

 

One thing I can't do, however, is separate the V/A meter into two - it's one of these:
https://i.ebayimg.com/images/g/U8cAAOSwzMtdCmnh/s-l1600.jpg
Not sure if this matters...?

Yes, it matters.

The interaction between the amp meter ground or return and the volt meter ground or return is curcial. For example, the photo shows the amp meter in series with the return leg of the load, while you show it in series with the source. This almost certainly is a problem. Please post a link to the product page.

Note that decent current measuring *always* is an issue. There a lots of techniques, but no single one is best for all applications. For 3-digit accuracy or better, there has to be a shunt and some voltage drop sonewhere; non-contact Hall Effect sensors are not accurate enough. So what are the trade-offs of the possible shunt locations... ?

ak

 

For example, the photo shows the amp meter in series with the return leg of the load,

If the shunt needs to be in the return leg, then to prevent the shunt drop from affecting the output voltage, the shunt should be put in the return before (to the left of) the voltage control pot connection.

For best regulation the pot (and 12k resistor) common connection should be connected directly to the supply output terminal, and the 220 ohm resistor (please label the components) should be connected directly to the LM338 output pin.
It may seem counterintuitive to connect the 220 ohm to the LM338 pin, but it can be shown that any resistance between the output pin and the sense resistor in a 3-terminal adjustable regulator seriously degrades the voltage regulation.

 

More to thinks about!

I redrew the schematic before reading these comments, so it does not take them all into account. But I expect to mount the LM338s on a big heatsink, and note the requirement to get the 220ohm connected to the LM338 pin. Incidentally, metal LM338s seem to be readily available at much less than $84 - am I looking at the wrong component?

I have posted a link to the advertised V/A meter. but this gives no data about the innards of the beast. It is marked as a KWS-DC121 (V2.03), and it seems distinctly odd. However, I did find a video describing its use - it's at

 

Rejigging the A/V meter position in the light of that video on connecting it...

(sorry - edited it because I forgot to change the position of the A/V meter...)

 

Thought about the above - what will the first LM338 (current limiting) do if the rotary switch is break before make?

 

metal LM338s seem to be readily available at much less than $84 - am I looking at the wrong component?

No.
The TO-3 metal case devices are likely near obsolescence, so are expensive.
You may have to use the TO-220 version.

what will the first LM338 (current limiting) do if the rotary switch is break before make?

Since there is no connection during the break, the output current will drop (possibly to zero depending upon the load current).
You can minimize that drop by increasing the value of the 0.1μF cap at the input to the right LM338.

What would be better is to have a switch that sequentially connects the resistors in parallel.
That would eliminate the break problem and require fewer resistors.
I've seen rotary switches that can do that but they are likely custom designed, so such a switch may be hard to find.

(Note you do not have the switch connected in series with the LM338 output pin.
Also it would be good to show that the bottom of the 10k pot and the 12K resistor are connected directly to the output pin.).

 

I have corrected the switch error, and moved the pot/12k closer to the output pin - is that what you meant? I have also increased the 0.1 cap to 1μF. I assume that all the caps are electrolytic?

The idea was to work with what I had in my bit box, so the cap values are a bit random. Do I really need a huge 4700 μF across the output pins? Will a 100 μF (at a nice fat 400WV) do? I have the LM338s in TO-3 - see below:

 

You can probably buy cheap fake LM338 ICs on ebay or buy real ones at a crazy high price at digikey:

 

Hello,

If you want a more modern regulator, have a look at the LT1084.

Bertus

 

Can't remember where these came from - but I'm sure that I would remember if I'd paid $80 or so each! We shall see how well they survive...