Background Story:

my application involves in sensing low light, as low as intensity as 1mW/m2 so i have picked an APD than a normal PIN

usually after reading some APD science i have came to know if you want to increase the gain and operate at maximum, one has to change the bias to very high voltage as specified by factory specified spec for that APD, that too when operating at higher voltage one has to very careful with temperature and the bias voltage should be changed according to the temperature to maintain the gain

to be brief observe the effect

enter image description here

importantly an APD of same design(discrete APDs) may have different breakdown voltages, if you see the datasheet of APD i have shared the very first table shows you that the breakdown voltage of APD may varry from 315 to 490, which clearly means every APD will have different breakdown voltage and different operating voltage for same gain, one observation is the

dV = Vbreakdown - Voperating

this dV is same for same design of APDs


now my application needs 10 APDs to be used, i was initially thinking i can happily gang them to a single power source as in case of PINs


simulate this circuit – Schematic created using CircuitLab

but after knowing some APD science, its clear that it would be a flaw in design to gang APDs, so should i have to use 10 power supplies for 10 APDs ? :/ , makes design bulky, noisy and costly

An Idea :

i want the circuit to be something like this


simulate this circuit

the voltage at P1 would be temperature compensated maximum voltage(i feel all APDs would have same temperature coefficient - correct me), the digitally configurable block should be able to take 450V maximum and based on requirement should be able to provide a voltage between 350-450V, this kind of approach would really help in achieving my task of using a single power supply and biasing them,

here i am clueless about what can be this configurable block, i didn't see any buck converters or regulators with such a high drop voltage and really skeptical about this concept, this is just a idea


i have changed my mind, its even OK for primary design to go for non programmable version of P1

a new thought came in my mind is why don't i use a simple potentiometer ?

the maximum current the dc-dc voltage source can give is 1mA so even though the voltage applied is 450V, if i need 400V at output i would put a potentiometer in place of P1

but will this bring a performance degradation ? because i am trying to process 100ns pulses falling on APD diode repeating at a rate of 20us, so a 50K resistor after power supply is going to add up capacitance ?

  • \$\begingroup\$ First, you want to apply positive bias to the cathode of the APD, not the anode. \$\endgroup\$
    – The Photon
    Jul 16, 2016 at 6:06
  • \$\begingroup\$ sorry for the blunder, i was in a haste, edited \$\endgroup\$
    – kakeh
    Jul 16, 2016 at 7:59
  • 1
    \$\begingroup\$ "based on requirement" what requirement? i.e. how do you want to configure the voltage block? Is manual switching OK or do you need MCU control of the adjustment? What maximum current do you anticipate in each APD? Do you need continuous adjustment or would 10V steps be close enough? You could easily generate 12 10V steps (340 to 450V) and use a rotary switch per APD for example. \$\endgroup\$
    – user16324
    Jul 16, 2016 at 8:56
  • \$\begingroup\$ @Brian Drummond as the TIA i am using has maximum saturation current of 2.5mA(gain 1K swing 2.5V), so a maximum current of 2.5mA and bias voltage of maximum 500V would require me use a 200K of limiting resistor, which i feel enough and would not effect my bandwidth of 10MHz, as shown in fig i want the control to be digital via an MCU, it can be in steps of 10V, i dont think high precision is required \$\endgroup\$
    – kakeh
    Jul 16, 2016 at 9:42
  • \$\begingroup\$ What would this a better question is a clear design spec for Input and Output in point form like any datasheet to include all requirements. Performance challenges can then be included. \$\endgroup\$ Oct 4, 2016 at 13:11

1 Answer 1


I've seen a cute trick performed inside a 1KV PMT supply regulator. Rather than using an expensive 1500V transistor plus a fancy high-side driver circuit, they used a bunch of transistor-output opto isolators in a series chain. They were wired as a HV shunt-regulator.

Choose devices with sufficient collector voltage (divided by N.) Fairchild H11D3 can take 300VDC, with 5KV isolation. Apply your LED current (also a series chain,) and you set the transistor operating point (current.)

The LED input is isolated, so the trick should also work for high-side pass-regulators as you show above.

Note that 2.5mA and 450VDC is over one watt. Your 200K resistors must not be tiny or close-spaced. And, if using an opto-iso chain, employ enough series devices to spread out the half-watt, keeping each DIP package milliwatts low enough.

I find this:

Jim Williams, "Feeding and Reading the APD:" PDF (Lin Tech Inc.)

  • \$\begingroup\$ isnt a potentiometer solution, much cheaper and simple ? how can i tune voltage if opto isolation is used, i need to provide the tuned HV to photodiodes, APDs work at very HV \$\endgroup\$
    – kakeh
    Sep 28, 2016 at 16:11
  • \$\begingroup\$ Shunt regulation was my instinctive reaction to this problem. Allows you to keep the control circuit at the cooler side. Run the numbers on typical component values, and dissipation - see if it works. \$\endgroup\$ Oct 3, 2016 at 9:22
  • \$\begingroup\$ @kakeh > potentiometer solution ...yep, but didn't you want microprocessor control? Replace your pot with a transistor, but then you need a fancy 500V floating base drive. To control the HV biase voltage, just vary the LED current while measuring the voltage. \$\endgroup\$
    – wbeaty
    Oct 3, 2016 at 23:24

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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