I have a project where I am trying to find a means of powering a sensor with the following requirements: 24-30 V and a regulated constant current of 2 mA (about 60 mW). The latter requirement I will handle with a constant-current diode.

The source power will come from a USB port (5V up to 200 mA).

My problem is that I need a low-cost method of DC-DC up conversion with very low EMI characteristics and was looking for something more novel than a traditional switched-mode converter. Also, low heat generation and light weight would be desired.

I am not an electrical engineer and wondered what my options were.

I have some crazy ideas such as electric-light-electric conversion (such as LEDs wired in parallel sandwiched with photovoltaic cells in series), but was hoping there might be a conventional way of doing this.

  • 3
    \$\begingroup\$ I think you're going to have to quantify "very low EMI" in order to get a useful answer. Are you concerned about electrical (voltage) noise or magnetic field noise? Presumably, the output of this sensor is going to be digitized for the USB interface, so describing the sample rate and resolution of your ADC will give us some idea of how much noise you can tolerate, in terms of both level and bandwidth. \$\endgroup\$
    – Dave Tweed
    Sep 6, 2012 at 20:24
  • \$\begingroup\$ I will get a bit more information and revise the question. In terms of EMI concerns, it is magnetic field noise that may be the most problematic. I do not know if I could get specific numbers, but only mentioned it because I was warned by the vendor about close proximity to DC-DC converters - not even sure what is meant by 'close' there. \$\endgroup\$
    – nicholas
    Sep 6, 2012 at 20:56
  • \$\begingroup\$ If it's mostly magnetic fields you want to stay away from, then maybe a charge pump might do it. A few stages of Cockroft-Walton would get you to 25V, and 2 mA is doable. No magnetic fields need be abused. The only magnetic fields will be from current moving, but that's not much because there is no coiling of conductors to amplify the effect. \$\endgroup\$ Sep 7, 2012 at 0:14

1 Answer 1


A switching power supply is the conventional way. LEDs shining on lots of photocells in series would produce a higher voltage, but would also be horribly inefficient. You have 1 W available, so you would need 6% efficiency to get your 60 mW out. I haven't run the numbers, but that sounds considerably optimistic to me. You also say low heat generation is desirable, which means you need good efficiency. LEDs to photocells isn't going to meet that spec most likely (I say "likely" only because you haven't provide a real spec).

You are also confusing requirements with implementation methods. Apparently what you really want is 25 V or so with very little noise. Don't assume you need a particular method to get that. You also haven't provided any spec. "Very low" is no spec at all, so really any method is acceptable.

I would start with a normal boost switcher, filter that, then run it thru a linear regulator. Two stages of ferrite chip inductors in series with ceramic cap to ground in front of the linear regulator should get rid of the frequencies the regulator can't handle. Put a good cap on the output of the regulator too, and you should have a nice quiet supply.

  • \$\begingroup\$ +1, well said. I would add that the physical size of LED's on photo/solar-cells could be physically large and could make a good antenna (both receiving and transmitting the USB noise). This solution could have more noise than an equivalent switcher+linear. But the OP wasn't very specific about what he meant by "very low EMI", so I can't say for certain. \$\endgroup\$
    – user3624
    Sep 6, 2012 at 20:41
  • \$\begingroup\$ Well I did say I felt optical amplification was 'crazy'. This was partly a question about what alternative DC-DC conversion methods were possible (if any). I will look into the suggested answer and try to revise my question if I can get more specifics. \$\endgroup\$
    – nicholas
    Sep 6, 2012 at 21:01

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.