Continue to Site

Welcome to

Welcome to our site! is an international Electronics Discussion Forum focused on EDA software, circuits, schematics, books, theory, papers, asic, pld, 8051, DSP, Network, RF, Analog Design, PCB, Service Manuals... and a whole lot more! To participate you need to register. Registration is free. Click here to register now.

Unstable output voltage with LM317 regulator

Not open for further replies.


Jul 3, 2022
Reaction score
Trophy points
Activity points
I'm building a variable power supply that converts 120 V AC from the mains to DC voltage adjustable between 1.25 V and 30 V depending on the value of a potentiometer.

Here's the schematic

30v dc psu schematic.jpg

I simulated the circuit with Multisim simulator and it works perfectly fine.

However when building the circuit on real life there's a problem, the voltage at the output of the power supply is not stable, it decreases over time, for example if I try to set the output at say, 20 V, my multimeter reads 20.06 V, ok that's fine. But after 5 minutes the voltage decreases to 20.04 V, after 10 minutes is 20.03 V, after 20 minutes is 20.01 V and after 34 minutes is 20.00 V.
The same thing happens with lower voltages, if I set the output at 5 V, multimeter reads 5.01 V but after 15 minutes is 5.089

And this is with no load connected at the output other than the multimeter, I guess the voltage will decrease even further with a load connected. On the simulation this doesn't happen

I tried adding a small aluminium heatsink on the potentiometer in case it was thermal drift but it didn't help.

What could be causing this? Is this variation expected due to things not being perfect like a simulation?

Heating seems to be the main cause as you assumed by improving heatsinking, but given the extended time pattern it sounds not semiconductor related, but rather due to transformer core heating, and this is expected. BTW, I assume the correct value in the measurement was 5.100 and not 5.010v, since this 0.2% reduction would be consistent with the 0.3% reduction of the input voltage from 20.06v to 20.00v

There is a min load current spec to stay in regulation.

You do have a load, the large output cap charging initially. And of course the
R bias divider network.

But I think what you are seeing is just a small rise in die temp, the effects of
quiescent current, cap charging, and thermal mass producing the long time
constant drift. There is a loose thermal stability spec in datasheet.

If you are looking for high stability, low drift, an LM317 is NOT the solution.

Regards, Dana.

operation seems normal as various components heat up after switch on, if you want 0.1% tracking accuracy over time, you need 0.1% resistors and a much better regulator than the 317

There can be many things including your DMM.
Every electronic component has a Value vs. Temperature characteristic and this drift is normal.
Hint : Resistors are heated and their values also drift. Potentiometer too.
Internal Reference of the IC has also a Drift coefficient. Everything.
Simulators don't know this mystery ... ;)

operation seems normal as various components heat up after switch on, if you want 0.1% tracking accuracy over time, you need 0.1% resistors and a much better regulator than the 317
Ok, I added small heatsinks on the regulator and the potentiometer but didn't help, perhaps a fan blowing air would be better?

Also, what regulator could I use that is more precise and accurate than the LM317?

At the heart of the regulator is a bandgap reference
and every bandgap reference has a curvature to the
tempco. The factory trims to a minimum overall deviation
but there is always a residue (worse on these older
simple non-curvature-corrected bipolar parts with
diffused resistors; thin film resistor improves curvature
(not common on cheapo linears), curvature correction
circuitry can be more trouble / variation than it's worth).

Check your heatsink and case temp. Specifically don't
bother yourself about voltage moving, if temperature is.

Those numbers don't seem unreasonable to me, you're
talking something like 0.1 - 0.3% on a couple-% tolerance
regulator. Check the datasheet tempco chart and see
whether you are really seeing anything unexpected.

R2 should be no more than 120Ω to provide the minimum load current required by the LM317.

Those output voltage variations seem normal.
Do you need a regulator with tighter regulation?

Here there many precision adjustable voltage regulators. If you carefully design one of them, you'll have a really stable power supply.


Long term drift of a couple 10mV is expectable. If you want better, then specify it... as value with unit.

What are your goals for -

1) Startup time to a stable V ?
2) Temp range design must operate in ?
3) Max load current ?
4) Max change output V due to load changes ?
5) What is model # of your multimeter ?

Of course noise, line changes, all have respective specs and influences on output V.

Regards, Dana.

Not open for further replies.

Part and Inventory Search

Welcome to