I would like to couple a constant-current source with a high-side H-bridge and it seems to be way more difficult than expected.

In the end, it is a constant current source from 0-10 mA with a voltage of 100 V, where the current can be applied in both directions of the load.

I tried to simulate it with LTspice, but I always get high-current pulses when switching and the current has a larger offset through the leakage of the resistors of the PMOS.

To activate the current pulses, I use the shut-down channel of the op-amp together with the trigger pulses, otherwise it integrated into saturation when the high side is open-circuit.

Schematic: enter image description here

Load current through R10: enter image description here

The MOSFETs here are rated at only 100 V. I know these are the wrong ones and it is just used for the simulation and gives the same output also with a lower supply voltage.

Any ideas for another approach or improvements? I already tested with optocouplers (too slow) and directly with NMOS and a charge-pumped driver (works pretty well, but seems to be a more difficult approach).

Edit: My Goal is a bidirectional constant current source, adjustable in between -10mA and 10mA, providing up to 100V, so that it has a high load impedance range. I found several topologies that might do the job, like the howland current pump, but all of them I found need a split supply, which in the end needs +-100V. I think it should be possible to achieve this with only a single-ended 100V supply and a High Side Switch topology.

Edit2: Indeed the howland current pump seems like a good idea for a current source up to around 100mA and around 100V split supply, but what if one want lets say 500V at 10mA? OPAMP prices in this area explode and MOSFETs are very cheap, where also gate drivers with charge pump are really affordable in comparison with the OPAMPs. So, a high side switch topology seems still to be a good idea from my point of view. Any ideas what other possibilities there would be for very high voltage and low but constant current sources?

  • \$\begingroup\$ Build a -100 V inverting boost or flyback from your 100 V rail and be done with it? \$\endgroup\$
    – winny
    Commented Mar 4, 2023 at 9:08
  • \$\begingroup\$ That would indeed be possible, but makes the whole setup way more difficult. HV Opamps, isolated DC/DC converter to generate the negative line of -100V, offset adjustment to be able to connect with single-ended devices like uC and so on. \$\endgroup\$ Commented Mar 4, 2023 at 11:00
  • \$\begingroup\$ HV opamps aren’t needed. The rest is easier than an H-bridge unless you can accept a dead zone Z close to zero volt. \$\endgroup\$
    – winny
    Commented Mar 4, 2023 at 16:54
  • \$\begingroup\$ Ok, can you elaborate more what kind of circuit you are thinking about? Bootstrapping the OPAMPs? \$\endgroup\$ Commented Mar 4, 2023 at 17:32

1 Answer 1


Consider a current-steering scheme.

These waveforms look roughly like a TENS application. At least, from my admittedly vague understanding of that.

I once breadboarded a similar circuit. I don't recall it doing very much though; probably a side effect of that vague understanding.

To be clear, do not hook up any high voltage circuit to your body. Use only approved equipment, etc.

TENS generator

(Disclosure: this image is linked from my website.)

Transistors were salvaged; these were medium-voltage types common in CRT video output drivers. Besides MPSAxx, MMBT5551/MMBT5401 would also be reasonably available analogs.

The H-bridge is formed between the bottom pair (a differential pair) of KTC3206, and the upper pair of KTA1275. (Alas, younger me was not a fan of putting designators on components...) The PNPs pick which path to activate, and the differential pair steers current into the complementary side. The pair is supplied by a current sink, as is traditional. It's kind of just a balanced audio preamplifier (differential in/out), but made for higher output voltages.

For battery-operated application, a lower current level-shifting scheme might be chosen (instead of the NPN-PNP drivers), or for higher voltage, optoisolators used, even; and for greater precision, an op-amp controlled current sink can be used.

To improve settling time, and current consumption in battery-operated application, the current sink could further be driven from a setpoint, i.e. using a transconductance amplifier (the input voltage will be keyed by P1 | P2), the amp of which can be disabled/powered off between pulses. The exact timing of the input/setpoint (or, if keying is helpful at all), and power up/down, depends on how the amp responds; that seems a bit out of scope here, and a fast-settling low-idle current sink might be a good choice for its own question.

To be clear, your question here will need to address both settling of the current source/sink itself, and drive current(s) of the switch(es). To that end, the choice of BJTs here is advantageous as they have fairly small capacitances and the hFE is large (current error due to base current is small). MOSFETs can be used the same way, but very small (low capacitance) MOSFETs aren't as abundant as BJTs. Both types become more challenging when higher power dissipation is required (say for a 500V 100mA operating range), as such devices will have higher capacitance (simply, it takes a bigger junction to handle more voltage and current, and dissipate more power).

  • \$\begingroup\$ Thanks a lot, that looks really promising, I will try to simulate it. \$\endgroup\$ Commented Mar 4, 2023 at 18:56

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.