3
\$\begingroup\$

I'm a recent graduate EE student and this is my attempt to gain experience and strengthen my resume so I'm eager to figure this out. I'm designing a dual channel floating output bench power supply that will supply 0-32V @ 5A with CC and CV operation modes. I intend on this being a long-term solution so I'm aiming for good power efficiency with a clean output signal. The safety and protection features I plan on including are a microcontroller based emergency thermal shutdown, reverse current protection, overvoltage protection, and of course a fuse.

At first I intended on creating it using IC voltage regulators: switching pre-regulation + second stage linear regulation. My plan was that the switching output would match the linear output to minimize power loss in the linear regulator. The problems with this are wildly varying CC under load and difficulty finding regulators with the desired specifications.

Having decided to check out the few schematics for commercial bench power supplies that I could find, I noticed that they don't seem to use IC voltage regulators at all! This leads me to wonder if it would be easier to create the linear regulator portion with discrete components instead of ICs. I've been having trouble finding resources which discuss the best methods of PSU design so I wanted to come here for a second opinion.

Which method do you think is the best for me to create my desired product? Switching + linear, switching + discrete or something else?


The post here Integrated voltage regulator chip versus discrete circuit utilizing op-amps and power BJTs for DIY bench power supply poses the same question but it doesn't look like there was an answer to the actual question. I looked at it and it seems like they just discussed the safety of using a negative reference voltage.

\$\endgroup\$
  • 4
    \$\begingroup\$ @ElliotAlderson I actually saw this earlier too and I don't think so. The only "answer" is by the OP and they're just saying something that they'd overlooked in the datasheet. There's some discussion in the comments but it seems like it's just an explanation of how to use an lm317 as an LDO. \$\endgroup\$ – Dailyyoda Jun 20 at 18:26
  • \$\begingroup\$ Depends on what your requirements are: bout iout and noise and thermal water and efficiency would determine what circuit is best \$\endgroup\$ – Voltage Spike Jun 21 at 23:20
1
\$\begingroup\$

The only linear voltage regulator IC I know of that meets your specs is the LM338. The TO220 package version is a reasonable price, but with a minimum input-output differential of 3 V the power dissipation at 5 A is at least 15 W, which would require a large heat sink. To work over the full 0-32 V range at 5 A with current limiting a tracking pre-regulator is essential.

For efficiency the pre-regulator needs to be a switching type, and the output tracking needs to be tight. Therefore I would look at making a pure switching supply that meets the spec, then consider methods of reducing noise on the output. This could involve a discrete linear pass element (bipolar or FET) controlled by an op amp, whose task is simply to reduce ripple with a low voltage drop.

Switching supplies generally have automatic protection against reverse current. If the supply needs to absorb reverse current without raising the voltage (eg. for powering a motor controller or simulating a battery) use a synchronous switching regulator with 'half bridge' output.

You might also consider providing an auxiliary low noise output using a fixed pre-regulator and lower current linear regulator IC. For example the LM317 can easily supply up to 12V at 1A with current limiting when powered from 15 V. This could be useful for bench testing projects that need multiple supply voltages (often the higher voltage supply would normally come from a battery or unregulated mains supply, so low noise is not so important).

| improve this answer | |
\$\endgroup\$
  • \$\begingroup\$ I'll probably use an IC like you've suggested. I was primarily wondering if there was some crazy good benefit of discrete parts that I was overlooking. Everyone suggests finding a workaround for problems with specific IC problems rather than designing a discrete circuit, so I guess that should tell me everything I needed to know. The LM337 is really good for my purposes but I didn't find it before because I was looking at regulators with near 0V output. Now that I think about it, I can have a relay switch the output to the filtered switching regulator output at low voltages. \$\endgroup\$ – Dailyyoda Jun 21 at 17:59
  • \$\begingroup\$ With discrete components you can optimize the circuit for best performance, but the design can be tricky. Integrated regulators are easier to get going and 'bulletproof', but get in the way when you want more. In 1979 I built a variable power supply using an LM317. It is still my main bench supply today. I got the output voltage down to zero using a negative bias voltage. (circuit here:- bhabbott.net.nz/circuits/LM317%20Variable%20PSU.jpg). I also built a fully discrete linear supply that puts out 25A @12V through 4 2N3055's in parallel. It's a monster! \$\endgroup\$ – Bruce Abbott Jun 21 at 23:11
1
\$\begingroup\$

A 0-32V supply with 5A and CV and CC is a very challenging design. You'll likely have your hands full designing that, debugging it, characterizing it and understanding it no matter what approach you take. If you don't have a reasonably strong background in analog and power then adding the design of a high speed power stage is really a whole new layer on top of some already challenging layers.

Since you want to boost your experience and resume, I think you'll get the most by doing a rough design both ways - pencil out a dc-dc plus integrated linear. Select parts, look at low-Vout, mid-Vout, high-Vout conditions at light and heavy loads. Look at thermals. Pencil out the design of a high power, high speed discrete linear. What do you learn are the pros and cons?

I think that exercise will provide a lot of experience and wisdom. Once you feel like you understand the trade-offs then pick the approach that has come to appeal to you most and build a prototype and reconcile it with your calculations.

From a purely practical standpoint, a DC-DC with maximum integration followed by a linear like the LM338 or another linear with an external pass device (to boost Vin and Iout capability), seems like the most natural solution to me as a starting point I'd have to identify a critical shortcoming to dive into designing my own high speed, high power discrete linear regulator.

On the other hand, doing both constant voltage and constant current isn't what integrated linear regulators are built to do. A discrete approach might turn out to be better, especially if you aren't looking for ultimate PSRR and loop gain.

| improve this answer | |
\$\endgroup\$
  • \$\begingroup\$ Hmm, this seems like the best route to take where I will learn the most and be able to make an informed decision. I imagine that the major trade-off is between noise and specifications. When you make it using discrete parts you can tailor it to your specific needs but you also provide a lot more places that can introduce noise. While it's difficult to find specific advice about the specific trade-offs of this situation, I've come to realize that there's a reason most people always suggest using ICs. \$\endgroup\$ – Dailyyoda Jun 21 at 17:44

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