123PSU - The flexible linear power supply

[ElectronSurf]

I wanted to share a project that I have done a while back as a hobbyist, it's based on old linear power supplies design that had op amps supply "floating". user blackdog also made a decent power supply several years ago based on the same principles, I did used some of his tricks in designing 123PSU. You may call it a copy of blackdog design but then you have to call all designs copy of a copy of a copy...

There are differences, compromises, added and removed features etc but there's also many similarities between two designs, my goal was to build myself something decent and stable and I think that I almost reached that goal.

The design is THT only except for the shunt resistor that I have to use metal film resistor instead of wire wound resistor. I spent DAYS designing a single layer THT only board. (pat on the back!)

The design is separated in two boards, analog and digital.

Using ATMEGA8A 10bit ADC and a lot of sampling I could get pretty stable and accurate reading of voltage and current, from left to right buttons do; voltage range switch from 20V to 5V, current range switch from 1A to 100mA, Shows current set value and last button "mutes" the output.

Shoot your questions at me if you had any but know that I'm just a hobbyist so manage your expectations...

 

[Kleinstein]

The diodes in the rectifier are rather low voltage rating. With a bridge rectifier the diodes see up to the peak to peak voltage. So the 40 V diode rating would be reached with 15 V AC.
I would more go for  more normal diodes like 1N5401 or a ready made diode bridge.

The NE5532 OP-amps for the regulators is a bit odd fast - I would expect that a RC4558 should be fast enough.

The way the voltage adjustment is done the adjustmen range would in clude negative voltages to -2.5 V, but limited to -0.3 V by D6. One may want another resistor in series to R21 to limit the adjustment range.

The extra 3rd power supply via J4 is a bit strande. AFAIK one could get the minimum load from the same power as the regulator, from the +8 V supply there.

For only 1 A the 100 m shunt is a bit on the small side. This makes the current adjustment to low values a bit tricky.  A know that finding a good shunt can be be tricky and a larger resistor would come with more heating, but it would help with settting small currents like 100 mA and below. So ideally I would go more for 300-700 mOhm instead.
Limiting the current to 1 A is a good idea, to make sure the SD1047 can handle the heat. With transformer tap switching it could even handle a little more, like 1.5 A or maybe 2 A with a good heat sink.

[blackdog]

Hi ElectronSurf

Herewith some comments from me, nice that it works!
Regarding the NE5532, if possible take the NE5532A version, you have more assurance regarding the noise and offset/bias values that they are favorable.

An RC4558 as Kleinstein suggests is not my preference.
One of the few advantages is the lower bias current, otherwise the NE5532A is almost better in everything else.
What I chose it for in one of my designs is its speed and low noise number.

When using the RC4558, the loop compensation will also have to be adjusted again so that the two control loops are again both properly stable.

I agree with Kleinstein, that you could choose a slightly higher value for the current measurement sensor.
So that the Bias currents and offset voltages of the opamp's have less influence.
Think of about 0.33 to 0.47Ohm.
Keep in mind, that by adjusting the current measurement sensor you may also have to adjust the loop offsets maybe a bit again.
That includes C19 and C21.

Also, the chosen rectifier diodes as Kleinstein indicates is on the edge, at voltage spikes on the mains, so can die too easily.

Maybe I pointed it out before, your reference voltage at 2.49V is a bit low. I would as in my versions of the schematic, take 5 or 10V if you wanted to get above 30V with this schematic.
You are amplifying the noise from the reference at higher output voltages quite a lot now.
That makes the noise and hum behavior of the power supply worse than necessary.

Also, I would at least include an additional RC filter for the reference diode,
So before R1 another resistor and an electrolytic capacitor so that the noise signal at that point goes down further.
The Resistor R1 of 1K along with the Ri of the LM336 may not be enough to adequately suppress the interfering signal from the 7808 regulator.

I certainly don't mean that the setup is that bad, it's already much better than most regulators I see coming by. :-)

Kind regards,
Bram

Translated with www.DeepL.com/Translator (free version)

[ElectronSurf]

Hi ElectronSurf

Herewith some comments from me, nice that it works!
Regarding the NE5532, if possible take the NE5532A version, you have more assurance regarding the noise and offset/bias values that they are favorable.

An RC4558 as Kleinstein suggests is not my preference.
One of the few advantages is the lower bias current, otherwise the NE5532A is almost better in everything else.
What I chose it for in one of my designs is its speed and low noise number.

When using the RC4558, the loop compensation will also have to be adjusted again so that the two control loops are again both properly stable.

I agree with Kleinstein, that you could choose a slightly higher value for the current measurement sensor.
So that the Bias currents and offset voltages of the opamp's have less influence.
Think of about 0.33 to 0.47Ohm.
Keep in mind, that by adjusting the current measurement sensor you may also have to adjust the loop offsets maybe a bit again.
That includes C19 and C21.

Also, the chosen rectifier diodes as Kleinstein indicates is on the edge, at voltage spikes on the mains, so can die too easily.

Maybe I pointed it out before, your reference voltage at 2.49V is a bit low. I would as in my versions of the schematic, take 5 or 10V if you wanted to get above 30V with this schematic.
You are amplifying the noise from the reference at higher output voltages quite a lot now.
That makes the noise and hum behavior of the power supply worse than necessary.

Also, I would at least include an additional RC filter for the reference diode,
So before R1 another resistor and an electrolytic capacitor so that the noise signal at that point goes down further.
The Resistor R1 of 1K along with the Ri of the LM336 may not be enough to adequately suppress the interfering signal from the 7808 regulator.

I certainly don't mean that the setup is that bad, it's already much better than most regulators I see coming by. :-)

Kind regards,
Bram

I've finished this project about 6 month ago, it's working and I'm not in the mood to do any changes, except for those rectifier diodes.

Built it just for the sake of building it, the extra details about noise is not that important to me but thanks for pointing it out.

I shared it so others can make another versions of this with different features and specs, hence the "flexible" is in the title.

[dobsonr741]

Is that right the “Emitter” current sense net is direct to U3B via R32? Or perhaps some mislabeling? Current limit would not work as drawn IMO.

[ElectronSurf]

Is that right the “Emitter” current sense net is direct to U3B via R32? Or perhaps some mislabeling? Current limit would not work as drawn IMO.

Why not? the shunt voltage drop is at the emitter. maybe the floating voltage confused you, the op amp supply ground is referred to the positive output.

[dobsonr741]

Got it. Yes, GNDREF was my misunderstanding. Little unusual, that’s all.

Why did you make this design decision? Could have measured the current on the low side, thus moved ground ref to the negative terminal.

[Kleinstein]

The circuit follows the pretty standard flaoting regulator concept, as used in quite some commercial lab supplies. There is nothing wrong with this, especially with a digital display, as the 2.nd transformer can also be used for the display / µC.

The slightly odd parts are
1) the circuit around U7, as this could be done easier (current source from the +8 V to GND)
2) Q - just a resistor would be fine too, maybe a zener diode would be better and help with a possible spike on turn on.
3) The part around Q6: looks like a fix for a possible spike on turn off. Looks strange, but could work.
4) using 2 x relay tap switching at such a low power level ( 1 relay is plenty)

Using a 2.5 V reference is not that bad. It would not get much better by amplifying the reference first and than use 5 V at the regulator - this would even loose more of the pot to negative set voltages.

[ElectronSurf]

Got it. Yes, GNDREF was my misunderstanding. Little unusual, that’s all.

Why did you make this design decision? Could have measured the current on the low side, thus moved ground ref to the negative terminal.

When sensing the current at the low-side of circuit your ground is not actually ground anymore, you have raised it by some amount.

[ElectronSurf]

The circuit follows the pretty standard flaoting regulator concept, as used in quite some commercial lab supplies. There is nothing wrong with this, especially with a digital display, as the 2.nd transformer can also be used for the display / µC.

The slightly odd parts are
1) the circuit around U7, as this could be done easier (current source from the +8 V to GND)
2) Q - just a resistor would be fine too, maybe a zener diode would be better and help with a possible spike on turn on.
3) The part around Q6: looks like a fix for a possible spike on turn off. Looks strange, but could work.
4) using 2 x relay tap switching at such a low power level ( 1 relay is plenty)

Using a 2.5 V reference is not that bad. It would not get much better by amplifying the reference first and than use 5 V at the regulator - this would even loose more of the pot to negative set voltages.

1- How could it be done with the +8V rail when its ground is referred to the positive output? doesn't it short out the regulator?

2- Yes a single resistor sufficed to provide the base current needed but JFETs is used because of almost perfect low noise nature to suppress any noise coming from the regulator. not to mention stability of output current to the input voltage changes.

3- It was blackdog idea😍

4- The transformer had 4 taps, lower voltage drop over the main transistor doesn't hurt.

Thanks for the input.

[Kleinstein]

The probably simplest form for a constant biasing current source would be the follwing:
A PNP transistor with base to GNDREF ( = pos output) ,  collector to GND ( = neg output) and the emitter with a resistors (e.g. 2 K range) to the +8 V regulator supply.

For the analysis of the circuit with the floating regulator it can help to see it as a negative regulator. So take the REFGND as reference point and consider the negative output side to be moving from the collector side of the power transistor with the raw input supply as a level shift to get to the negative side.

If the transformer has 4 voltage taps, it is OK to use them, especially for learning. For tap switching with relays it also helps if the steps in the taps are not so large, as there is quite some current peak from charging the main filter capacitor. So the relays see quite some stress and possibly wear. The is likely not a problem with the low power version, but it can be for higher current (e.g. 5 A versions). To reduce very frequent switching of the realys the control part should have quite some hysteresis as done with R12 and R27. The capacitors C17 and C20 are rather large and I would normally drop them all together.