PSU with discrete components. Any tips for CC control.

[Avelino Sampaio]

Hi

I decided to challenge myself and build a PSU for my own use, using only discrete components.

Vout = 0 ~10v
Imax = 2A
CV and CC control

The CV part is already well advanced, where I'm having good results in terms of stability and low noise. It remains to perform the transient test and thus be able to verify the phase margin.

My other challenge is with the CC control, which I want to place on the high side. I see many projects using dedicated Opamps and CI. Already using only transistors, the information is kind of scarce. I received a tip that I could use a current mirror, which would then be unbalanced by the sensor resistance. Does anyone have a tip or example circuit, on how I could design this CC control, only using discrete components? Going from 50mA to 2A

[magic]

This input stage has phase reversal. When Q3 base is at ground, the emitter is one diode above ground and the collector too. Q3 saturates and can't feed current into Q7 base, even more so if a couple mA coming from Q2 needs to pass through R13, so Q7 emitter is at +100mV or so. Then the output goes to the positive rail and there is smoke. This is exactly the same problem that TL072 is infamous for, also present in RC4558 and friends.

LM358 solves it by adding PNP emitter followers in front of the input pair. Others solve it by replacing the NPN mirror with NPN folded cascodes and adding a PNP mirror above them. Or similar tricks.

[magic]

As for current limiting, perhaps the most sane approach is to have a voltage reference and a second opamp which floats with the output voltage.

The simplest way to get the above is using the base-emitter voltage of the output darlington to power the CC circuit, but then it takes some clever low voltage tricks to get it all done at 1.2V supply. In particular, when it comes to voltage references, you are limited to a crappy diode with its 1% per 3°C thermal drift. A bandgap reference could be attempted, but it's not clear if it would work much better.

A different and maybe saner approach is to use a small transformer (or USB charger ) to provide a floating supply. Then also the CV circuit can be powered from it, and they may even share a common and decent voltage reference, like a 5.1V zener.
See here: https://www.eevblog.com/forum/beginners/shitty-chinky-psu-kit/

[Avelino Sampaio]

As for current limiting, perhaps the most sane approach is to have a voltage reference and a second opamp which floats with the output voltage.

The simplest way to get the above is using the base-emitter voltage of the output darlington to power the CC circuit, but then it takes some clever low voltage tricks to get it all done at 1.2V supply. In particular, when it comes to voltage references, you are limited to a crappy diode with its 1% per 3°C thermal drift. A bandgap reference could be attempted, but it's not clear if it would work much better.

Magic, my challenge will be to build the CC control on the high side and without appealing to opamps. I just found an article that mentions the tip I had received, which is to use a current mirror to read the sensor resistor. I'm going to replace the opamp in the example with a differential and see what happens.

https://sound-au.com/articles/bench-supply.htm

[Avelino Sampaio]

This input stage has phase reversal. When Q3 base is at ground, the emitter is one diode above ground and the collector too. Q3 saturates and can't feed current into Q7 base, even more so if a couple mA coming from Q2 needs to pass through R13, so Q7 emitter is at +100mV or so. Then the output goes to the positive rail and there is smoke. This is exactly the same problem that TL072 is infamous for, also present in RC4558 and friends.

LM358 solves it by adding PNP emitter followers in front of the input pair. Others solve it by replacing the NPN mirror with NPN folded cascodes and adding a PNP mirror above them. Or similar tricks.

I have the circuit assembled here on the breadboard. Tomorrow I will do fault simulations, to try to better understand what you are reporting to me.

[Avelino Sampaio]

Hi

I have finished devising a sketch of the CC control. I will first test with my electronic load.

[magic]

Magic, my challenge will be to build the CC control on the high side and without appealing to opamps. I just found an article that mentions the tip I had received, which is to use a current mirror to read the sensor resistor. I'm going to replace the opamp in the example with a differential and see what happens.

https://sound-au.com/articles/bench-supply.htm

A discrete opamp is still an opamp

The current mirror thing only transfers shunt voltage to ground, an opamp is still needed to compare it against the preset.
Alternatively, the potentiometer could be connected to the positive rail and an opamp could compare it directly to the shunt.

[Avelino Sampaio]

I was able to implement CC control with a good result.

A quiescent current of 30mA flows through R1. With the adjustment of P2, this value was compensated in the current mirror.

I managed to set the value of R27 so that P3 was in the range of 30mA to 2A.

[magic]

I think the exact value of current limit changes with supply voltage.
And if the supply is unregulated (trasformer + rectifier, for example), the output current in CC mode will ripple.

[Doctorandus_P]

Why use two shunt resistors (R1 and R15)

I can understand wanting an extra gross current overload protection (Especially while experimenting), but I do not see a need to do this with two shunt resistors.

If you want to keep the high side measurement, then increase R1 and remove R15. Move Q9 to R1 and add an extra transistor so you can shunt the base current of your power resistor to GND to squeeze it off on a gross overload.

[Avelino Sampaio]

Why use two shunt resistors (R1 and R15)

I can understand wanting an extra gross current overload protection (Especially while experimenting), but I do not see a need to do this with two shunt resistors.

If you want to keep the high side measurement, then increase R1 and remove R15. Move Q9 to R1 and add an extra transistor so you can shunt the base current of your power resistor to GND to squeeze it off on a gross overload.

I've had previous experiences using MAX4080 and Opamp and they weren't fast enough. In my CC project I am using an integrator capacitor to decelerate and this can cause me the same problem. So I will keep R15.

[Avelino Sampaio]

I think the exact value of current limit changes with supply voltage.
And if the supply is unregulated (trasformer + rectifier, for example), the output current in CC mode will ripple.

The transformer I ordered has arrived (12 - 0 - 12 / 3A) and today I'm going to assemble the power supply to carry out more realistic tests. I intend to design the font with filtering at the minimum threshold. So I will have considerable ripple voltage at play when the current approaches 2A. I will return later today with the results, on how much this will affect CC.

[Avelino Sampaio]

Hi

Using my Electronic Load, I conclude that 3000uf will be enough to meet the project's demand.

1 - Power supply
2 - In the load of 2A, the ripple voltage in the limit to have a maximum regulated output of 10v.
3 - Ripple value at 1A load.

Now it's time to insert this source into the project and test the behavior of the CC module, with all this ripple voltage.

[Avelino Sampaio]

I think the exact value of current limit changes with supply voltage.
And if the supply is unregulated (trasformer + rectifier, for example), the output current in CC mode will ripple.

The transformer I ordered has arrived (12 - 0 - 12 / 3A) and today I'm going to assemble the power supply to carry out more realistic tests. I intend to design the font with filtering at the minimum threshold. So I will have considerable ripple voltage at play when the current approaches 2A. I will return later today with the results, on how much this will affect CC.

Magic, I made two changes to the CC module. I added a current source, using the LM334 and moved the module actuation, which was under the VAS, to the reference voltage CV.

To observe possible variations in CC mode, I monitored the voltage across resistor R15 (0.33R). And the results were satisfactory.

1- Updated schema.
2 - 1A current, without actuation of the CC Module. Oscilloscope reading on R15.
3 - Actuation of the CC module. Oscilloscope reading on R15, without oscillations.
4 - Minimum value I managed to regulate was 17.7mA

[oz2cpu]

here is another all transistor PSU
this one also provide selectable foldback or CC mode

start at 6:30
https://youtu.be/Ut5VLmpI54o

I am sorry i dont have the schematic, but it looks a lot more simple, compared to the schematics posted in this thread.
by the way: tomorrow i release a new video about one more psu, made with the same transistor only design,
i go thru its design a bit more deeply, so please come back :-)

[Avelino Sampaio]

Hi

Continuing, I performed the transient test.

Output at 5v, maximum pulse of 1A with duration of 200Us

1- Updated schema
2- result with capacitor C3 = 1nf
3- result with capacitor C3 = 2nf
4- result with capacitor C3= 3nf

Another good behavior of this project..

[Avelino Sampaio]

This input stage has phase reversal. When Q3 base is at ground, the emitter is one diode above ground and the collector too. Q3 saturates and can't feed current into Q7 base, even more so if a couple mA coming from Q2 needs to pass through R13, so Q7 emitter is at +100mV or so. Then the output goes to the positive rail and there is smoke. This is exactly the same problem that TL072 is infamous for, also present in RC4558 and friends.

LM358 solves it by adding PNP emitter followers in front of the input pair. Others solve it by replacing the NPN mirror with NPN folded cascodes and adding a PNP mirror above them. Or similar tricks.

Magic, I forgot to mention that I performed tests and in fact the output goes to the positive rail. However, in my perception, this failure condition is of very low risk, given that c8 or q3 will have to fail for it to occur. Or is there some other situation that leads to the inversion, and I'm not noticing it?

[magic]

What happens when you simply try to set output voltage to zero, by turning P1 all the way down?
I'm referring to the latest (final) schematic.

[Avelino Sampaio]

What happens when you simply try to set output voltage to zero, by turning P1 all the way down?
I'm referring to the latest (final) schematic.

2.8mv reading as you can see in the image. Same value if I hit the mute switch.

[Avelino Sampaio]

What happens when you simply try to set output voltage to zero, by turning P1 all the way down?
I'm referring to the latest (final) schematic.

2.8mv reading as you can see in the image. Same value if I hit the mute switch.

Magic, this value of 2.8mv comes from the base of Q4, as transistor Q8(tip120) is off. This is a situation I don't want, so I added P4 to set the minimum output voltage value to 0.5v.

I noticed that the CC control didn't work well at low voltage, so I returned it to the VAS. I also changed the position of the MUTE switch.

[Psi]

If a super simple CC with minimal components is desirable. You can use a P transistor with emitter and base across a high side current sense resistor and set the value of the resistor so the transistor starts to turn on when the current through the resistor is higher than your CC limit.

Then you connect the collector through a resistor to your voltage feedback line.  When the current is above a threshold it will start to reduce the output voltage to keep the current under that point.

Being so crude it has some disadvantages.
- It's not very accurate as you a relying on the transistors exact turn on voltage/current.
- The resistor value needs to be quite large to get enough voltage drop, which means you have to burn off some power in the resistor.
- It's only suited to a fixed current limit, rather than something adjustable

[magic]

What happens when you simply try to set output voltage to zero, by turning P1 all the way down?
I'm referring to the latest (final) schematic.

2.8mv reading as you can see in the image. Same value if I hit the mute switch.

Thank you for checking, I didn't expect this to work correctly.

R9 must be helping by limiting how much current can be pulled out of the base of Q3 and preventing the base from really going all the way down to 0V. Perhaps a useful trick to remember...

[Avelino Sampaio]

Checking the behavior of the output voltage and the stabilization time, when energizing the PSU.

1- Output voltage at 5v and no load - time of 57ms and no overshoot.
2- Output voltage at 5v and with a load of 0.6A - time of 62ms and no overshoot.
3 - Output voltage at 5v and no load - 1s time to discharge the 100uF capacitor.
4 - Chinese power supply (30v/10A), with 5v output voltage and 0.6A load - my project is twice as fast as my switching power supply.

[Avelino Sampaio]

If a super simple CC with minimal components is desirable. You can use a P transistor with emitter and base across a high side current sense resistor and set the value of the resistor so the transistor starts to turn on when the current through the resistor is higher than your CC limit.

Then you connect the collector through a resistor to your voltage feedback line.  When the current is above a threshold it will start to reduce the output voltage to keep the current under that point.

Being so crude it has some disadvantages.
- It's not very accurate as you a relying on the transistors exact turn on voltage/current.
- The resistor value needs to be quite large to get enough voltage drop, which means you have to burn off some power in the resistor.
- It's only suited to a fixed current limit, rather than something adjustable

Thanks for suggesting this option to me, however the CC module presented is working very well!

[Avelino Sampaio]

New schematic update, adding the cooling system. I also added the Digital Meter.