I have built a simple LM317 regulator power supply and would like to add a pass transistor to provide more current and lower output impedance of the circuit. Here are two alternatives, one I copied and adapted from another question here, using a PNP and NPN transistors, and another which is using only an NPN transistor (adapted from a source using LM338 and multiple 2N3055s in parallel).

The supply will be used for powering an Allo.com Katana DAC and a Raspberry Pi (two similar supplies) so the requirements are 5V at 1.2A (DAC) and 5v at 3A (Rpi with extra usb devices), with very low ripple (in microvolts) and very low output impedance. I have already measured the supply with and without Cadj, and the ripple is so low (50uV) that I decided not to add a capacitance multiplier. The problem is that the audible DAC sound dynamics were much lower than even a cheap smartphone charger, and it was pointed out to me that the regulator has a relatively high output impedance that I should mitigate by adding a pass transistor and/or much more output capacitance.

So my question is which of the alternatives should provide better performance, and does using only a NPN transistor negatively affect the ripple?

Circuit #1 - R values are copied and possibly should be revised

Circuit #2 - transistor voltage drop needs to be taken in account for output voltage

  • 1
    \$\begingroup\$ Why did you choose the LM317? This seems to be an extremely odd choice, seeing that you seem to need more current or voltage drop than the LM317 can provide itself. So, explain how you came to the conclusion that sticking to the LM317 was the right choice, instead of using a voltage regulator design that is actually meant for your use case. Also, you don't define anything about your requirements (voltage in, out, current, acceptable ripple), so how on earth are we supposed to assess whether your circuits fulfill the requirements? \$\endgroup\$ – Marcus Müller Aug 5 at 12:38
  • 1
    \$\begingroup\$ (and CADJ doesn't provide "ripple rejection" in general - it provides a modified control loop behaviour. Without thinking too much about this, a 100 µF capacitor in this position basically makes the regulator not react to any fast load variations at all – leading to basically no regulation, which is the oppositive of what you'd normally use a voltage regulator for. So, really, state your requirements before choosing an architecture and parts.) \$\endgroup\$ – Marcus Müller Aug 5 at 12:41
  • \$\begingroup\$ I’m voting to close this question because questions like this are best self-served by using one of the many free simulation tools available. \$\endgroup\$ – Andy aka Aug 5 at 12:56
  • \$\begingroup\$ @MarcusMüller thanks for pointing out the lack of clarification and requirements, I edited the question. And for the choice of LM317 - I want to explore it's possibilities since I'm just learning electronics. I can use the LM338 for higher current, but that doesn't solve the output impedance issue. \$\endgroup\$ – nahero Aug 5 at 14:06
  • 1
    \$\begingroup\$ You have a high end DAC coupled with the opportunity to learn electronics. I think a rethink is in order. \$\endgroup\$ – StainlessSteelRat Aug 5 at 19:00

So, yes, the LM317 is really not a good regulator.

You can certainly extend its capabilities with external pass transistors, but you're suddenly building a control system for a voltage of your own – the LM317 in itself has been designed to be stable given some capacitance and load variation, your own system needs to be designed so that the higher current gain doesn't lead to instability.

As you can see, that makes the whole thing rather complex (11 or > 15 components!).

Using another nearly 50 years old circuit, the LM338, doesn't make this any better. (And I heavily doubt your 50 µV measurement, unless you do have an oscilloscope.)

Pick a less ancient voltage regulator, if you need better regulation. Your use case suggest you'll want one supply for the digital part (the rPi + digital side of the DAC) and one separate one for the analog part (analog side of the DAC + amplification).

The design of your katana DAC thing doesn't seem to allow for that.

So, a super low-noise low-impedance power supply doesn't matter at all. You've got the noise generated by your rapidly switching digital logic in your analog supplies, and no matter how low the impedance of your voltage source gets, you won't get that out of your system. Go for a well-designed high-frequency switch-mode power supply design; 1 MHz switching speed is so far away from your audio frequency, that unless your system starts running on a synchronous clock to that, you'll not have anything in the audible / amplified frequency range. Adding low-noise linear regulation, again, doesn't help when the source of the noise is sharing the same supply.

Regarding your DAC board: They probably designed a voltage regulator stage in there, too, because no logic actually runs at 5V these days. However, they also decided to build opamps out of discrete parts, claiming these are better. Um. That's what us engineers call audiophile hogwash. A well-designed, factory-trimmed opamp IC will inherently beat discrete opamp designs by far, simply because of thermal coupling of the transistors on the die. So, probably things aren't going to get any better than using a normal not-the-cheapest-you-could-get USB power supply, and be done with it.

| improve this answer | |
  • \$\begingroup\$ Thank you for the break down, and I agree with most of what you say. Audio is finicky, and best discussed on a different forum. That being said, I didn't get an answer to my questions - how does adding pass transistor in the two variants affect impedance and ripple? Granted, I should build both and then test what I can myself, but an experienced opinion or a point in the right direction is much appreciated. \$\endgroup\$ – nahero Aug 6 at 10:11
  • \$\begingroup\$ no, audio is not "finicky"; audiophile device suppliers are just borderline scammers. Engineers don't believe in technical magic. \$\endgroup\$ – Marcus Müller Aug 6 at 10:28
  • \$\begingroup\$ I did answer that: you're building a control loop of your own, now you have to measure and simulate. We can't tell you. \$\endgroup\$ – Marcus Müller Aug 6 at 10:28
  • \$\begingroup\$ Ok, thanks. Will return with some measurements once I build this. \$\endgroup\$ – nahero Aug 6 at 11:03
  • \$\begingroup\$ Don't build it. You're very stubbornly trying to use the wrong tool for the job. Throw out your LM317, get a good voltage regulator that actually does the job you want. \$\endgroup\$ – Marcus Müller Aug 6 at 11:06

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.