# LM317 voltage regulator varying voltage output depending on the load

I have designed a circuit using a LM317 to provide a 13 V voltage rail. The Vin is 24 V. I have voltage voltage divider networks which supply different voltages to a logic comparator which is connected to the Vin. From this voltage regulator I am simulating a Hall sensor and hence have added a potentiometer. The signal is sent to the same comparator mentioned above. (The comparator is comparing voltage levels from the voltage divider network with the Hall sensor.) The circuit also has a 5 V voltage regulator which is supplying its voltage rail to a few more components.

I noticed that the voltage output from the LM317 fluctuates by about 30 mV depending on the load I have set using the potentiometer. Because of this, the output from the voltage divider network is also fluctuating. This causes the voltage levels to change and hence get a undesired output from the comparator.

My question is:

1. Is it normal for the LM317 voltage output to fluctuate by 30 mV?
2. If this is not normal, how can I fix this issue? I read that if the load output from the LM317 is not much, then there is a chance that the output voltage may fluctuate (this maybe due to the Iadj current). My guess is it might because of this issue, and connecting the 5 V voltage regulator to the LM317 might solve this issue, but I am not sure.

Edit: I haven't updated the schematics. I changed the resistor values from 5k6, 44k2 and 750R → 220R, 1960R and 100R, respectively.

• LM317 works poorly with too high voltage divider resistance. The behavior will be just as described: voltage will vary depending on load. Mar 10 at 11:53
• Additionally, LM317 might melt through the floor if you go from 24V to 12V. It's essentially a radiator with the side effect of also producing an output voltage. There's very few reasons to use it for such voltage gaps. There's TO 220 drop-in replacement switch regulators that can replace it, for example. Mar 10 at 11:56
• A SO-8 version is even worse at dealing with the heat. Don't expect to drive more than 100mA or so from it. Mar 10 at 12:42
• Back in the days you'd use one of these regulators like LM317 from 24V to somewhere around 15V and another like 7805 to take it from 15V to 5V, only to spread the heat between several parts. In modern design you'd rather use a buck regulator to go from 24V to 5V and then maybe from there a linear regulator from 5V to 3.3V, in case you need to power RF electronics or similar sensitive parts. Mar 10 at 13:37
• If 30mV variation is unacceptable, LM317 long-term-stability spec of 1% might be a red-flag, especially if those comparator threshold voltages are critical. But do stay with a linear regulator (versus a switcher). Mar 10 at 14:37

1. All regulators have varying output voltage due to varying output current. It is hard to say how much in your circuit the current varies, and for the LM317L the change could be 1.5% over full temperature range with load currents between 5 and 100 mA.

2. It may be due to adjustment pin current. The adjustment voltage comes from a voltage divider that has far higher impedance than typically recommended in data sheets. The change could be max 5 microamps.

5 microamp change through a 44k resistor is already 220 millivolts.

The fluctuation could be much smaller if you change the 5k6 resistor from the adjustment pin to be 10x less, 560 ohms. Generally a common value would be 240 ohms, for two reasons. It would provide the minimum current draw of 5mA without any other load, and it would make the adjustment pin current to have much smaller effect on output voltage. It is not guaranteed that by reducing the resistor by about 20x you get 20x better performance, as the LM317 is a rather generic regulator, not a precision reference.

• I did not update the schematics, but yes I noticed a 30mV change when i reduced the resistor values i mentioned above, as compared to 50mV+ change i was measuring before with the values shown in the schematic. Mar 10 at 9:59
• In my case then, I would need to look for a precision voltage regulator that can provide a stable output? are there such regulator available? I thought they are only available for standard values such as 3V3, 5V, 12V 15V etc. Mar 10 at 10:08
• Precision voltage regulators don't exist. Precision voltage references exist, but the can't provide much current. There are many ways to go forward here, like, it might be useful to know why you even need 30mV precision from a regulator, and do you even know what precision you need, will 10mV do or 1mV or is it just that the circuit could be made differently so even a 100mV change does not matter? What is the circuit doing? Mar 10 at 10:23
• That was my first question, if the 30mV is expected/lower/higher. if it is expected, then that solved my question. If it high and can be reduced, I would like to try, But as for a limit for 'precision', I don't have one, I can theoretically be .1V above/below the 13V. but having it closer to 13V reduces the my error %. The circuit would be used for testing steering columns. Mar 10 at 10:51

Others have pointed out that your values for R102 and R103 are far too large. One of the purposes of R102, besides setting the adjustment pin potential, is to draw current from the LM317's output, even under no load conditions.

The datasheet says on page 6, that "minimum load current to maintain regulation" is typically 3.5mA, or at worst, 10mA. Example schematics in the datasheet all use a value of 240Ω, for R102, resulting in 5mA through it, even with no other load.

Even if you fixed that, I think your expectation of achieving an output voltage variation of less than 30mV, under various load conditions and input voltages, is optimistic. If 30mV change is a problem, then you should really consider using a precision reference. The go-to jellybean device is the TL431, and it absolutely rocks.

It's a shunt regulator, like a zener diode, and you use it in the same way, but it uses two external resistors (R1 and R2 below) to obtain whatever voltage you desire.

It seems, from your schematic, that you need references of 12.7V, 11.3V, 8.8V and 7.7V. All these can be derived like this:

simulate this circuit – Schematic created using CircuitLab

12.7V is a bit too close to the 13V output of the LM317. If you power this circuit from +13V, R3 would have to be very small to pass sufficient current for the TL431, and all the dividers. Then, even small deviations from 13V would result in large changes to current through the TL431, that would diminish the otherwise excellent regulation of this device.

As shown, you can expect way better than 1% regulation, perhaps even approaching 0.1% of variation in the 24V supply.

• Thank you for the detailed explanation. I wish I had posted this question earlier. I looked briefly through the data and your excellent explanation of the working, This circuit would work when there is a constant stable Vin. If for example I was to change the Vin from 24V to 16V, then I would need to recalculate R3? Mar 15 at 8:46
• @EmeraldMonk My value for R3 was "guessed" based solely on how much current I want "available" for all the dividers, plus some for the TL431 itself. I settled on 10mA-ish total, 5mA for the dividers and 5mA for the TL431. R3=1kΩ got me in that ballpark. If that ballpark is good (it is), you should aim to choose R3 that will pass 10mA, too, so $R=\frac{V_{IN}-12.7V}{10mA}$. With a 16V supply, 300Ω-ish is good, it's not that critical. Mar 15 at 10:09

As pointed in the other answer (and now done), reducing R102 to the recommended 220Ω (and R103+R104 in proportion) can improve things. Other ideas:

• Make sure the wiring resistance from VOUT to load is low, and the ground connection for R104 is to the ground point where you want a regulated voltage; otherwise said, make sure the wiring resistance of the load does not interfere.
• The recommended output capacitor is 1μF, C101+C102 is only 20% of that, it could cause output ripple/oscillation, which variation under load could be part of the DC variation you observe.
• Try a small (100nF) capacitor between ADJ of the regulator and ground. This can reduce the output ripple.
• I replaced the Cap C101+ C102 with a 1µF cap. It I noticed a little less fluctuation. Now its in the order of about 20mV which I think is not too bad. I can deal with that much variance. Im yet to try out a cap between the ADJ and GND, I will most likely add that cap in when a redesign of the circuit is being designed. What value cap do you recommend? Mar 15 at 8:29
• @EmeraldMonk: In the answer I recommended C=100nF. It essentially forms an RC circuit with R102. With R102=220Ω, that's τ=RC=22μs (a few times that adds to the turn-on time), and it should tend to reduce ripple above roughly fc=1/(2πRC)≈7kHz. You can increase that C manyfold (increasing τ and reducing fc in proportion), the limit is the increase in turn-on time. Mar 15 at 13:02