# Can you connect two LM317s either in series or parallel to help spread out heat?

I am designing an adjustable voltage regulator.

The regulator will have be able to switch between maximum Vout of 9V, 12V, 15V and 18V and will have a variable resistor that can adjust the voltage from 1.5V to its selected maximum voltage.

The problem is the amount of heat dissipated.

The input will be 21V, to compensate for voltage drop, and the max current that needs to be handled is 450mA.

Using the power dissipated formula P=(Vin-Vout)*Iout.

The power dissipated will be P=(21-9)*.45=5.4 watts.

The maximum heat that a LM317 can safely dissipate is 15 watts with a heatsink. 5.4 watts is well below that which is good but still I'd like to err on the side of caution and spread the heat across two LM317s if possible.

A couple of suggestions.

• Forget about the LM317. Consider a DC-DC converter instead. There are a lot of choices that can work.

• Arrange your resistor selection to be in series. Your approach could set a higher voltage than you intend if more than one switch is on.

If you are concerned about DC-DC ripple, you can post-regulate with an LDO (not the 317 - it has a high overhead), which is also where you would do a fine adjust.

• In the end design I'll be using different switches so that only one branch is active. But I do like your idea wish I thought of that. Do DC-DC converters not have heat dissipation problems though?
– DRC
Mar 10, 2021 at 1:49
• @drc He means specifically a switching converter which uses a rapidly switched impedance instead of a resistance to drop or even increase voltage and are capable of much higher efficiency and therefore lower watt loss/heating concerns. The LM317 is an obsolete "Low drop out linear converter". A linear converter drops voltage by wasting additional voltage as heat, and an obsolete one like the LM317 has a high "dropout voltage" which is basically the minimum wasted overhead for it to work. Linear converters are only more efficient when output voltage is very close to input voltage.
– K H
Mar 10, 2021 at 2:07
• DC-DC converters do shed some heat, but much less than a linear. They're about 80-95% efficient at max load typically. So your 450mA at 15V would have losses of about 0.6W (10%), much of that in the inductor. This part would do the job: diodes.com/part/view/AP5100 Mar 10, 2021 at 2:19
• Sorry if this is a stupid question, I don't have much power knowledge, can a switching converter still be powered by a linear source, that is a transformer with a full bridge rectifier? My power output needs to be low noise. I'm thinking of having a transformer connected to a lm317, since I already have them. That then branches off to 4 branches connected to 4 different regulators, 9 12 15 and 18V, that then connects to a lm317 for fine adjustment, like @hacktastical suggested.
– DRC
Mar 10, 2021 at 2:36
• Yes, the switcher requires DC - it does not care how it is sourced. You mention the requirement for low noise. Unfortunately switching regulators are noisy in comparison with linear regs. A common technique is to use a switching reg as the first stage or pre-regulator, then a linear reg as the final stage. This way you minimise loss in the linear reg whilst maintaining the low noise benefits. Mar 10, 2021 at 3:16

If you really want to use linear regulators, this circuit from the LM338 datasheet shows how:

The opamp regulates the upper LM338 so that the current through the two 0.1 Ω resistors is the same.

(The LM107 is obsolete; you can use any high-voltage opamp whose input common-mode range includes the positive rail.)

Instead of another LM317, one could also use a MOSFET or NPN in series as well. This can even be done with a zener to make sure only a small voltage will be across the input/output of the LM317. A minimum of around 3V is advisable to ensure proper regulation. In fact, it's even possible to use an external transistor for bigger currents as well.

Depending on the use case, but as suggested before, a DC-DC converter can also be an option as well, with much higher efficiency (well above 90% these days). Is there a particular reason for using a linear regulator?

If you want the low noise of a linear regulator, and low heat dissipation, the standard (and very nice) approach is to use a low dropout linear reg, and then put a switching regulator before it, with voltage set to about 0.5-1V above the minimum input voltage. Now the voltage drop across the linear regulator is very small, so heat dissipation is minimised.

By the way, be aware that the design you propose is likely to have horrendous spikes on the output if you operate the switches with power on, because almost all DIP switches are break before make. This could damage lots of circuitry downstream!