DIY Lab PSU (or bench power supply): what to choose?

Question

I apologize if I will write a lot, I invite you to read only if you have a lot of free time...

For some time now I have needed a versatile bench power supply, which would be useful for several general tests. I understand that the word "general" in electronics does not mean anything, because an instrument suitable for any situation is only a utopia, but from my experience I can say that it has hardly happened to me that I need voltages higher than 24V and currents higher than 3A, for small applications.

I would therefore like to build a power supply with two separate channels (possibly usable in series or in parallel), for example with maximum 2x15V.

I would like to have a very stable power supply with very low noise, so I would like it to be linear, not switching.

I have available (in addition to a nice metal case with "generous" dimensions) a 2x12V 6.25A (150VA) toroidal transformer. These two outputs, after being rectified, should give me 12 x 1.41 = 16.92V. After subtracting a dropout voltage, typically 1.5V, I should get my 15V per channel.

I also have two LT1083 linear regulators, which hold up to 7.5A continuous. They are very beautiful and have a TO-3P package easily mountable on a heatsink. But unfortunately it is an obsolete model, not replaced by any equivalent, so I would prefer not to use them, to make a future replacement easy, should one break.

I have found many solutions with linear regulators and a current of many amps, both with many regulators (like LM317 and equivalent) in parallel, and with bridge-connected mosfets.

Then I would have to make other choices:

1) Start from 1.2V of minimum voltage or adopt a more complex circuit to start from 0V

2) Adjust only the voltage or also the current (not of primary importance to me)

3) Use a simple analogue control with multiturn precision potentiometers and display to measure voltages and currents, or insert a microcontroller to control the settings with PWM, with a digital interface (in the future possibly also controlled by software)

But the crucial point is another: the maximum current and its dissipation.

If I want each channel to go from 0 to 15V, for low voltages I would have a maximum dropout of about 15V, which at 6A would require a dissipation of 90W !!!

Normally I would not reach this condition, because when could I ever need a 1V to 6A output? For example, it would be more likely to use 12V at 6A ((17-12) * 6 = 30W of dissipation). But I should still be in the worst conditions and I think that dissipating 90W of heat is really too much for a bench power supply, even with forced ventilation.

What do you think? What would you do for me? Settle for a much lower maximum current, not taking full advantage of my toroid? Make sure to limit the current as a function of the voltage, with a digital control, so as to keep the total dissipation below a set threshold?

After all, a good compromise could be having a precise linear variable power supply, with a low current, and a second switching power supply in case I need a higher current.

But even if I limit myself to 3A per channel, dissipating 45W of heat may not be that simple.

Any advice and opinions will be welcome...thank you in advice!

dougw · Answer

As you mention, you will likely need to keep the current to limits implied by practical heat dissipation methods. You could apply any excess current capacity to extra channels. For example some standard fixed voltages can be pretty handy - 3.3V , 5V , 12V.

shabaz · Answer

Hi Simone, I feel for your desired voltage range, a good achievable aim for a linear PSU is 1A per channel. Once you've got that built you'll have something always usable, and can consider a higher-current supply later. I have higher current power supplies, but I still use my home-made dual 0-15V linear supply which barely achieves 1A. The transformer is too large however. I think it's better to buy an approx 50VA transformer, with 0-15V windings for slightly more headroom in case that is needed. Personally I'd use decent potentiometers from Farnell/Newark, and an off-the-shelf low-cost digital panel meter (e.g. from aliexpress). There's some nice examples of home-made power supplies if you examine jw0752 blog posts.

jeff_electron · Answer

Hmm tricky, the real problem is to get the best potential out of your toroid. power dissipation can normally be greatly reduced by putting a variable switcher behind the linear regulator allowing for a relatively constant power dissipation across the voltage range for a given current. this goes against the desire to make a dead-quiet linear supply though. you could go for a small maximum current, but this unnecessarily cripples the supply for higher output voltages where it is much more likely to be wanted anyway, and you find that you end up wanting a higher minimum output voltage to get more current out. on the basis of making the most out of your toroid, avoiding a complicated variable constant current mode or digital control circuitry: I would, in your shoes choose a power dissipation target, use a few parallel pass transistors, and implement a temperature based current limit. this is a pretty simple circuit - just stick a thermistor next to your output transistors, and compare the voltage with a comparator to some reference. use the comparator output to shut off the output, and make sure there's a lot of hysteresis so it doesn't flick on and off around the critical temperature. you could go one step further and make the output latching so it requires manual reset once it trips. this will protect your transistors from the actual cause of failure - over heating rather than a proxy for it like too much current / power etc. you may or may not need to independently provide an instantaneous current limit (fast blow fuse?) or protect the transformer independently. i would say that your rectifier circuit under load is very unlikely to provide you with the full 16.9V. for a start you've got the diode drop to consider, but you also have the instant current from the transformer. because the input is a sinewave, if you want to draw full current but operate near the peak value you will have to take huge gobs of current in the very small time that the sine wave is high enough to forward bias the rectifier diodes and actually top up the reservoir caps. of course this gets vanishingly small as the output of the rectifier gets higher. then of course theres that the output of the transformer will sag under load proportional to the instantaneous current drawn. i'd say as a rule of thumb that this all works out to be about a 12V DC regulated output for an appreciable current. (i.e, Vrms ~ Vdc after rectification and regulation) one final thought if the above feels like too much effort is that you could split the ranges using relays and window comparators on the output. 10-15V @ 6A, 5-10V @ 3A, 0-5V @ 2A say gives your linear circuit maximum power of 30W, 30W, 30W respectively. more divisions = more relays but lower power dissipation per range. again, you have to be a bit careful about flip flopping around the switchover point. in this case as well as hysteresis i would use a time delay so that inrush current doesn't collapse the output voltage momentarily causing it to switch to a lower range. (its also really important to go by measured output voltage rather than voltage demand, since its the measured value which determines your power) of course, if you didn't want to avoid complicated constant current mode circuits then we could have think about that...

Andrew J · Answer

It's worth thinking about cost as well, depending upon what you hope to get out of the project. I know you already have a number of the parts, but once you have to start acquiring more to support 'features', you'll find it hard to keep the cost below the purchase price of a cheap but useful unit.

If the aim is to get your power supply and have a great deal of fun getting there, which is what I did , then crack on and keep us up to date on progress, it's always interesting to read about someone's project. If cost may be an issue, CPC often have deals on cheaper PSUs that do all you want. Check out their Bargain Corner - Test Equipment. They only have single channel supplies right now but it changes; Shabaz and others have bought one of these and found it to be very good value. CPC is part of Farnell in the UK - if you don't live here, check around local suppliers.

s1m0n3t · Answer

Thank you shabaz for your precious point of view. I agree with the fact that 1A can be enough for many applications, but perhaps it is not so rare to need a little more current. I realize, however, that hoping to get 6A for the entire voltage range, while maintaining an exclusively linear configuration, is really too much.

The idea of @dougw to get some useful fixed voltages (3.3V, 5V, 12V, for example with 1A) seems to me absolutely valid, but it would not meet the trivial needs of powering for example a 5V device with a current of 2 or 3 amps.

It is true that on the one hand I would like to fully exploit the potential of the toroid that I have, but on the other hand I would like to obtain a good flexibility on the load that I can power: it is more likely to have to supply a device with a "generous" current at a single voltage , rather than several voltages simultaneously, each with a low current.

Thank you so much @jeff_electron, you offered me so many ideas that could be developed. Despite all the controls mentioned could be done with analog stages (amplifiers, comparators, etc.), I could in fact centralize all the control logics on the firmware of a microcontroller.

I was already undecided whether to keep everything analog, with two simple potentiometers, or to insert a digital control. But considering that I would still need to buy some panel displays to measure the voltage and current of the two channels, I might as well use a nice 20x4 alphanumeric display (which I already have), a microcontroller, perhaps with an external 4-channel ADC and a 4-channel DAC (for greater precision), and one or two temperature probes.

In this way I could adjust a current limit selectable by the user (useful in some cases) and impose a limit according to a maximum tolerable temperature.

I should choose whether to insert an analog constant current control stage, or two relays that simply isolate the outputs when the threshold is exceeded.

DIY Lab PSU (or bench power supply): what to choose?

Top Replies