Beginner's Attempt at a Bench Power Supply

[acsparks]

I wanted to design my own bench power supply to try and improve my electronics knowledge.

What I came up with was a fairly effective linear regulated power supply capable of 1.25-20V and up to ~1A.

It uses 4* LM317 regulators (hence the name) and a UA741CP opamp (its a bad opamp for this implementation but I managed to force it to work in the way I wanted).

The power supply has two 10-turn pots which allow the user to dial in the voltage and maximum current they wish to use.

Pros:

• Its simple for a beginner to get their head around
• Its cheap to make
• It produces a clean output with good control over both voltage and current

Cons:

• The linear regulators waste a lot of energy as heat (switch mode power supply is probably a much better idea and will be in the next project)
• The LM317 can only supply up to 1A and will supply less than this as the difference between input and output voltage increases (i.e. as the output voltage gets lower, the lower the maximum current which can be supplied)
• When the supply is limiting the current the output signal gets choppy.  A smoothing cap is required if you want to use the power supply as a current source in this way.

I've gained a lot from completing this project and would suggest any beginner should give it a go.

Any suggestions on how to improve the design?  I've got a few ideas I may try to implement later.

[ez24]

How about a transformer with multiple output winding's and a switch to select them to reduce the voltages to the LDOs.   Like 1.25 to 10v  and 10v to 20v.  This will reduce the heat.

[bitseeker]

You can check out the Power Designs 6050A and 3650-S for a couple of different ways to minimize linear loss. Some supplies use a hybrid approach where a switching supply provides pre-regulation to optimize the voltage going to the linear regulators.

[tautech]

Couple of comments.
IMO C1 could need be larger, 2200 or 3300 uF for lower ripple at full load.....check with a scope.
There's no protection diodes on the 317's, check the datasheet for correct implementation.

[acsparks]

Couple of comments.
IMO C1 could need be larger, 2200 or 3300 uF for lower ripple at full load.....check with a scope.
There's no protection diodes on the 317's, check the datasheet for correct implementation.

It’s interesting, I was using ST microelectronics LM317s in this project and in their datasheet they recommend D1 and D2 to protect against input short and capacitor discharge during output short respectively.

I made the decision not to use any diodes after reading a Texas Instruments datasheet for their LM317 where they suggest, at low capacitor values and outputs less than 25V, you don’t require protection diodes.  This may be because the Texas LM317 is more ‘hardy’ than the ST equivalent.

In any case it would probably be a sensible idea to include the diodes in the design.  But this got me thinking why exactly do you need D2 in the shorted output + adjust pin capacitor case?  Is it simply to stop the stored energy in the C3 from finding a current path between the adjust and output pins through the device (perhaps if the value for R1 is small enough the diode could be omitted)?

I suppose D1 can also serve to protect against any inductive spikes on the output as well.

[tautech]

Read the TI protection diodes comments again and define external capacitors.

IMO without the protection of a forward biased diode on the PSU output external capacitance need be protected against.
As it is normal to add local bulk capacitance to a PCB of any project it is very likely it will be of a higher value than the TI datasheet suggests can be ignored.

[dannyf]

"The linear regulators waste a lot of energy as heat (switch mode power supply is probably a much better idea and will be in the next project)"

You can utikize a smps as a tracking regulator in front of the linear regulators to reduce the dissipation.

This will allow the linear regulator to work at a constant voltage drop.

[Dbldutch]

I suggest you use a slightly different method to set the current limiting.
If at all possible, you dont want to send the full current through a potmeter.
Just Google for 317 circuits and you will find examples.

Basically, you use the series (shunt) resistor to create a voltage that should produce just over 1.25V at max current. On the downside of that resistor, you use a potmeter with one end to the supply and with the wiper to the adjust pin. The other end of the potmeter can go with a resistor to ground. To limit the voltage span and make the adjustment more precise, use two 1N4148 or equivalent in series accross the potmeter (cathodes pointing to ground) limiting the voltage accross the potmeter to about 1.4V.

In general, the 317 can only go as low in voltage or current as much as the reference voltage, 1.25V. If you want to adjust below that, you have to tie the adjust pins to a -1.25V supply. If you use a small current sink in the adjustment line, by using a jfet, for example, you don't need the current sink on the output terminal. You can then use the free 317 to create a tracking pre-regulator, limiting the dissipation on your voltage setting 317. Again, Google is your friend.

Enjoy!

[David Hess]

You can get rid of your 10mA dynamic load by lowering the resistance of the R4 and R5 divider.  If R4 is 125 ohms, then a constant 10mA will be generated into R5.

When you mentioned using the 317 with a 741, I thought you might be using the 317 as a voltage, current, and heat protected pass element with the 741 controlling the output voltage using a separate reference.

Check out how Tektronix implemented the voltage and current control loops using 741s in the PS501 and PS503 power supplies.  I might do something like that with a 317 replacing the output transistors and a single supply operational amplifier like the 358/324 replacing the 741s.

[Messtechniker]

I am missing some slow blow fuses on the transformer primary and secondary sides.
And a mains filter to keep the mush out. And a double pole mains switch to be on the safe side.

Yours Messtechniker