Regulated Bipolar PSU

To feed modular system, you need bipolar power supply giving voltages from ±9v to ±18v depending from used system. Additional power lines might be needed. Eurorack standard suggests symmetrical ±12v line and additional +5v line which is useful for MCUs and some another digital ICs. It is possible to use computer power supply for modular systems, but PWM supplies have quite unstable voltage in fluctuating environment. Because of that, tune and volume of the synthesizer powered by PWM supply may have undesirable modulations. In some cases it can be used as artistic effect but to get completely predictable sound you need traditional power supply that uses a transformer.

Here is the one supply of this kind — Regulated Bipolar PSU. It gives enough power for almost any system.

Voltage of the regulated power supply can be precisely set to obtain ideal symmetry for the oscillators, stable tune and low noise floor. This power supply consists of bridge rectifier, filter and three identical parts of voltage refulation, protection and indication circuitry. The positive line works as following: Bridge (D1, D3) rectifies AC, then it being filtered with two 3300uF capacitors (C1, C3) and 100n ceramic disk capacitor (C5) which are connected in parallel. Then goes LM317 regulator working as Adjustable regulator. (IC2, R2, R5, D8, C8) The load resistor (R8) and LED provide both indication and minimal load required for adjustment. The resistor value should be between 2K—10K, so it’s better to find LED which is not very bright.

The negative line works identical to positive with exception of LM337 used as regulator.

This circuit is very similar to Ken Stone’s solution. I believe, he was inspired by LM317 datasheet too.


Version 1.2

Component view
PCB Layout (mirrored)


Part Quantity Value Package
C1,C2,C3,C4 4 3300u 35v E7,5-18
C5,C6 2 100n 35v Ceramic C050-024X044
C7,C8,C9 3 10u 16v E2,5-5
D1-D13 13 1n4004 / 1n4002 DO41-10
IC1,IC2 2 LM317 317TS
IC3 1 LM337 337TS
J1,J2,J3,J4 4 Terminal block 2pos 5MM

(or 2pcs 4pos)

JP1 1 Header 2×8 pins 2X08
LED1 1 3mm Red LED LED3MM
LED2 1 3mm Yellow LED LED3MM
LED3 1 3mm Blue LED LED3MM
R1,R2,R3 3 10K Trimmer Bourns W3296 S64W
R4,R5,R6 3 1K 1/4W 1% 0207/10
R7 1 4K7 1/4W 1% 0207/10
R8,R9 2 10K 1/4W 1% 0207/10

Assembled unit (old version) looks like this:
Bipolar PSU: фото

Creative Commons license
This circuit is published under Creative Commons «Attribution-NonCommercial-ShareAlike» 3.0 license.
Any use beyond this license must be discussed with author.

11 thoughts on “Regulated Bipolar PSU

    1. rpocc Post author

      Это зависит на самом деле от объёма и качества радиаторов на регуляторах. Я использовал небольшие радиаторы, и этот PSU обслуживает коробку 6U примерно с 16 модулями – все модуляционные и утилити типа LFO, источников шума, Phonogene, Ringmod, S&H и т.п., самопальный предусилитель и ещё что-то.

  1. Tim Bos

    Hey thanks a lot for posting this great psu, I made the psu on a veroboard but i left the 5v part out.
    Happy that it works, but was not so easy for me :)

    One question, what is the 100n 35v Ceramic actually doing? I am missing them beacuse i blew them up.



    1. rpocc Post author

      Hi. The .1uF ceramic at the reg inputs filter fast voltage ripples at the very moments of diode switching for which large bulk capacitors are too slow to handle. This is very common schematic and these capacitors may be omitted in some situations, but in other situations they may help to keep ripples at minimum.
      The .1 caps at the reg outputs do similar thing: filtering some hi-freq noise from the LM317 itself.
      Actually, any capacitor do not act like a low-pass filter, but more like a band reject filter having center resonance frequency on which it is most efficient. Not only at the lower, but also on the higher frequencies the reject factor is significantly lower, so for good AC filtering it’s a good practice to use a set of several paralleled capacitors, each responsible for it’s own frequency.

      The voltage rating usually should be twice higher then expected DC voltage (or even more) to keep capacitors cool and stable, and the capacitors themselves must be physically placed at some distance from the heat sinks.

      By the way, avoid using veroboard with high currents. It can have insufficient conductance and parasitic capacitance that will degrade your PSU specificatons. Just imagine, that some AC will flow across parasitic capacitance from reg input to output degrading ripple rejection, and also imagine a voltage divider consisting of lead resistance from reg output to load, and impedance of the load itself. Even at 0.5 amps the load impedance will be about 24 ohm at 12v, and even 0.5 ohm lead resistance will give you unwanted voltage drop of 12-(12/24.5*24) volts that are 245 mV, so the best practice is to use solid thick leads and solid wires up to very IDC power connector.

      Hint: adjust the desired voltage under load, it may significantly differ from the initial voltage. If it’s dropping to low, you may try to use a 2x15v transformer. Also you may directly replace an LM317 with an LT1084, which has lower dropout voltage.

        1. rpocc Post author

          Да схема стандартная, на сайте Doepfer можно посмотреть. Просто разъёмы в параллель, больше там нет ничего. Некоторые делают активные платы с +5В, дополительными кондёрами у каждого разъёма, но это уже кому как нравится. В любом редакторе плат такая фигня рисуется 10 минут по топологии “Шина”, а навороченные варианты – это уже коммерция.

  2. Tim Bos

    Hi, thanks again for the clear explanation. The caps as band rejection filter makes sense as well. I used a 18v 30vac during testing. I guess i need to get a 15v one. Will look into the LT1084. Cheers Tim

    1. rpocc Post author

      Я по-моему брал ампера на полтора. От кейса зависит. По-хорошему, догонять такую схему до максимума не стоит, лучше ограничится примерно на 0.6-1 А, поэтому сорокаваттника (по сумме плечей) должно хватить с лихвой.

  3. Your Are Logged In As Default Loser

    Hi, I don’t seem to have the range on the trimmers to get to +/-15v, if I reduce the value of the 1k resistors will that do the trick? Thanks Ryan

Leave a Reply