DIY Digital Bench Power Supply

[Guido Socher] built himself a great little bench power supply that’s able to put out 30 Volts at 2 Amps.

Instead of taking the easy way out by putting a few taps on an ATX power supply, this project was built around a generic 24 Volt laptop power brick. An ATmega8 generates a PWM signal that is sent though a low-pass filter, allowing everything to be very precisely controlled. This DC signal is then sent through a BD245 power transistor to bring everything up to the desired output. [Guido Socher] included a USB port for computer control of everything, and the final project is something we’d be happy to have on our bench.

We’ve seen a few computer power supplies converted into a bench power source, but we’re impressed with [Guido Socher]’s build log. It’s not often we see a hack that goes over the theory of operation, and the end product is very nice (and functional) too.

30 thoughts on “DIY Digital Bench Power Supply

  1. Without any specs on the noise that this produces, I’m hesitant, at the very least, to say “good job”. Also no idea on accuracy, load recovery times, temperature compensation, etc…

  2. It’s a very good design. It uses regular components and has quite a lot of features.
    It is probably not as fast acting or regulated as an analog design, but it is enough for hobby use.

    Gotta get to work and finish my digital supply design using XMEGA.

  3. @Slanderer: I can’t speak to most of those things but, as for accuracy, I can speak from experience as I work in the calibration/support group at a testing laboratory. Here at work we, intentionally, do not calibrate power supplies even if they have displays. The idea is that the technicians/engineers using the power supplies are expected to use their DMMs/O-scopes to confirm the power output values for any test they are performing. This cuts down on the calibrations costs/labor somewhat by relying on the calibration done on the meter/scope rather than calibrating an additional piece of equipment (the power supply).

  4. Neat exercise but, for me, power supplies fall into the category of “just buy it so you can get on with your other projects”. Other things in this category are soldering irons, oscilloscopes and socks.

  5. Have my doubts about the power ratings. A single BD245 can handle 60W only in theory (that’s rating when one can keep it’s case at 25C, can’t be done with a simple cooler).

    If you are planning to really use it across the stated U/I range (especially low output voltage with high current), have no doubts that you’ll need to parallel more transistors (as stated in his write up), but also to have forced cooling of the enclosure itself (60W of dissipated power will VERY quickly warm up the interior over 25C, so even with the ideal coolers on transistors you want be able to get them to required temp)

  6. No good job @ all! He is using an anlog design voltage control with a digital backend? Wtf? Either he builds an analog supply with the known disadvantages (it just wastes the voltage that it doesn’t put out) or a digital one. But combining the drawbacks of both designs- power wasting with poor control and slow reaction, that is not the way to go.

  7. @steve you are confusing a linear supply with an analog one and a switching with a digital.
    The fact is, you can have an analog or a digital linear regulator(supply) or an analog or digital switching regulator.

    What nobody else talks about is how he makes the DAC… it’s a very ingenious idea with PWM and R2R.

  8. I saw this about 6 months ago. He sells kits, complete with all needed parts, PCB …, but the schematics and code were on his website too, so you can build your own as well.

  9. I am confused whether he’s using the BJTs in linear-mode or switch-mode. It looks to me (with the very-limited documentation) that he’s operating in linear mode.

    This is a pretty crappy design with pretty incomplete documentation — does not belong on Hackaday.

  10. Saw this quite a while ago now, I was going to copy his design, but started thinking about response time of the digital control system and how easily it could destroy my projects.

    I am now considering redesigning it with an analog backend and digital control. But as a real project with some effort it probably will never happen.

    He holds the schematics for it back, but he published one for an older version on a mailing list, so you can get an idea of the backend setup at least. I think the DAC was the biggest change.

    http://cgi.linuxfocus.org/common/src2/article379/powersupply.png

  11. The new schematic must have some sort of summing with different coefficients of the two voltages given by the two “DACs”, e.g. the R2R gives you 7 bits and then you need to add that voltage with 1/256 of the voltage given by the PWM DAC to get the 12bits resolution.

  12. “What nobody else talks about is how he makes the DAC… it’s a very ingenious idea with PWM and R2R.”

    It’d be ingenious if all you had was an atmega8 and the resistors and caps. But why wouldn’t you just start off with a microcontroller that already has a DAC and an ADC? Or just tack on a $1 I2C DAC?

    I mean, the difference between an atmega8a and an atmega16m1 is like 50 cents.

  13. I built the linear regulator part of this supply a while back, and as mentioned above, the power dissipation is scary. It (understandably) gets incredibly hot with low-voltage, high-current output. I shelved it until I can be bothered to design a switched-mode back-end.
    My thinking was you could use the switched-mode to step the input down to a value closer to your desired output, then use the linear regulator to give you a (reasonably) stable output. That way you never need to dissipate that much power in the output transistor

  14. I bought one of these kits in 2007 and I use it all the time. I haven’t noticed any output noise or regulation issues, but I generally feed it into devices which have their own regulation.

  15. The slow digital voltage control loop is a bad idea. The microcontroller can generate the reference voltage, but an op-amp should be controlling the loop. I’d even trust a slow (but inexpensive) LM324 before this.

    Low-side current sensing is also a bad idea when ground is connected to earth. If current can find another path back to earth (say via another power supply) it bypasses your sensor, rendering it useless.

    Plus a switching preregulator before the linear regulator stage, as mentioned by others, would be a very nice addition; and not very hard.

    For an extra $8 in parts, this could be a really top-notch design. But as it is, it’s certainly nothing worth the author holding back schematics for (except to hide its deficiencies).

  16. Lots of complaining about deficiencies, and yet, no-one seems to be offering a better alternative (Expressing an opinion on how and why it should be improved is not the same as actually doing it).

  17. I built a psu based on the authors design and added some extra’s as a rotary encoder knob and thermal control. The PSU fails when working with “noisy” circuits (pulses, HF, switched circuits). The atmel seems to lockup from time to time. When properly calibrated with a lab quality dmm and using high precision resistors, it’s quite accurate and dependable.

    The idea is great and costs are low. What else do you expect.

Leave a Reply to NocCancel reply

Please be kind and respectful to help make the comments section excellent. (Comment Policy)

This site uses Akismet to reduce spam. Learn how your comment data is processed.