I want to make a power DC supply with voltage rating 0-200V.

I want it to be a variable power supply with voltage from 1V to 200v.

What should be my approach?

  • \$\begingroup\$ Well, first off I'd recommend writing up a better specification. For example, what do you hope will be its current compliance range? Also, do you want it to achieve an exact zero volts? Or would a higher minimum be okay with you? Do you also want a variable current limit? Do you have an accuracy requirement? Etc. It's very important to work out the details you feel are important to you, first. You don't just write "variable voltage from 0-200 V" and stop there. \$\endgroup\$
    – jonk
    Commented Jan 21, 2021 at 5:48
  • \$\begingroup\$ @jonk the load will draw current in range of micro amperes. Error margin of 1% is acceptable. Exact zero volts is not desirable. \$\endgroup\$ Commented Jan 21, 2021 at 5:56
  • 1
    \$\begingroup\$ @mkeith Yeah. But a calibrated VOM isn't cheap. And keeping it calibrated also isn't cheap. (And the compliance current would still need to be about 100 times what's needed by whatever load is applied.) \$\endgroup\$
    – jonk
    Commented Jan 21, 2021 at 6:50
  • 1
    \$\begingroup\$ @mkeith We use multiple wavelengths to get \$50\:\mu\text{K}\$ precision over \$500\:^\circ\text{C} \to 1650\:^\circ\text{C}\$, where an immersible sapphire light pipe probe with platinum sputtering at the end was used. Without the platinum tip, it becomes an area-integrating probe when used from a suitable distance (the acceptance angle of sapphire is twice that of quartz) to achieve a desired spot-size. Or a suitable camera-like lens may work. In those cases, the range is from \$500\:^\circ\text{C} \to 3000\:^\circ\text{C}\$. Calibrated accuracy is \$10\:\text{mK}\$ across the range. \$\endgroup\$
    – jonk
    Commented Jan 23, 2021 at 4:23
  • 1
    \$\begingroup\$ @mkeith Only one company achieved it. No competition. The founder who started it was a physicist from NIST. Not a cheap box. ;) Funny thing is that he got started down this path because he wanted to study internal combustion engines (ICE.) So he first developed a sapphire rod that could be placed via a drilled, angled hole through a spark plug so as to directly observe the explosions in real-time. The only buyers of that equipment were Japanese car companies. No company in the US was sufficiently interested. He was a very tough man to work for -- expectations were very very high. \$\endgroup\$
    – jonk
    Commented Jan 23, 2021 at 4:30

1 Answer 1


My approach, but I've got not experience in building measurement-grade supplies, would be the following. Please, do take this slightly "and the details are left for the reader to implement" approach with a grain of salt.

  1. You need a reference voltage and a control loop. Doing things in open loop over a large range is probably a bad idea.
  2. Your large voltage range requires the control loop to control a divided output, and that will get inaccurate for small output values, so I'd recommend doing multiple separate controllers, and just picking the right one for every output voltage. (E.g. one for 0.5V–2V, for 2V–8V, from 8V to 32V and for 32V to 200V)

You can buy voltage references (do not go too cheap on this, a few cents more here will help you not worry about that source of error later on). Say, you get one with 2.048V +- 0.1% over the temperature range you care about.

Then, you'd take the voltage output by that, and pass it through a multiplying DAC, or voltage DAC with external reference input. That DAC has more effective bits than your intended resolution needs. Using more bits than strictly necessary allows for later software calibration.

Then, you'll need a > 200V DC source, ideally filtered against noise, but I don't know what you need in spectral hygiene.

You'll want a high-side pass bijunction transistor, with a load resistors stronger than your maximum load. (Which means you want different load resistances, swapped based on your output voltage range, effectively. You can also do that with e.g. a single n-channel mosfet controlled by a relatively rough DAC to give the desired output current "minimum load".)

That pass transistor's base voltage is controlled by a control loop that compares the voltage on its output, divided by a different resistive voltage divider (0.1% resistors are relatively cheap) for each range, to the voltage generated by your precision DAC above.

Use a low-offset, low-bias-current opamp, or instrumentation amplifier circuit, to implement that feedback loop.

Sprinkle in a series R and a parallel C filtering on the opamp's output to dampen the oscillations you'd get on voltage adjustments/load changes. Put analog switch IC's on the output of that – another few ohms of resistance doesn't hurt the control loop, and neither will the parasitic capacitance; now, replicate for each voltage range, and you'll have selectable controllers for all of your output.

What I described above is basically an opamp-controlled linear regulator, one for each double-octave of output voltage. The redundancy comes from the fact that getting the same resolution over such a large voltage range using a single resistive divider ist hard, and I don't see an elegant way to get a high-accuracy multiplying DAC to work at 200V.

Now comes the annoying part: everything done, this thing will probably not be 1% accurate. You'll need to have a microcontroller that translates the desired voltage to both a selection for the active controller and a non-linearity-compensated DAC value. Depending on your resolution (1% accuracy sadly doesn't tell us what the resolution is) and thermal and time-stability needs, this might imply hacks like keeping all your hybrid/analog circuity in a temperature-controlled heated enclosure, or a temperature-compensation calibration, or bright "do not operate when temperature is outside 18°-22°C" stickers or something.


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