0-30V, 0-3A PSU - Audioguru's version

[Kleinstein]

The OPs for the current regulation and reference do not need to be powered from a high supply - the modified versions often include the series zener or a 7824 regulator. So it is only the OP for the voltage control, that needs to have a high voltage swing (e.g. TLE2141,MC34081,OPA604,...).
It is funny to have this old kit still around, with so many points that should be fixed or made better with not that much effort.


Using a separate floating supply for the regulator changes the circuit completely, going from a classical regulator with a low output impedance power stage to a low drop type regulator with a current setting output stage. The floating regulators has advantages (less voltage lost, usually better current control and flexible in the voltage range), but it also has a more complicated compensation for the voltage control and still often slower, unless tuned very well.

[Zero999]

If you have seen a way for PSU adjustable current-limiting without a negative rail or dedicated floating rail, let me know.

On the chinese kit, I hate the -ve rail and the need for the V ctrl op-amp to run at the top rail while the I ctrl op-amp runs on the low-side. I can find away around it all, but haven't built it to prove it works.

A negative supply isn't needed, if single supply op-amps are used. Here's a proof of concept, based on a Howland current source, with voltage limiting. I haven't simulated the transient response, which is probably horrible. U1 is shorted by Q3 and current limits, when in voltage mode, which some may not like, but that can be avoided if necessary. The Darlington pair could be driven by a current mirror based source, with each op-amp diode AND'd pulling it down.

I don't see the problem with the negative rail. I'd probably use the OP07 run off regulated +34V and -3V, derived from a 30V transformer.

[iMo]

Update: The stability analysis..

[iMo]

And some minor resistor's values changes and notes..
Update: Added V/A(out) as a function of R_Load.

[Audioguru]

I first saw the original Greek kit 15 years ago at Eletronicslab.com where its problems were discussed in the forum there. Many parts were overloaded and failed.

Paul's version will not produce 30V at 3A because his 24VAC transformer voltage is as low as the original Greek kit:
1) 24VAC has a peak of +34VDC which is reduced to +32V at 3A by the bridge rectifier.
2) Ripple from the small filter capacitors drops 2V to +30V.
3) The voltage amplifier TLE2141 opamp maximum output drops 1.6V to +28.4V.
4) R15 that is not needed drops 2V to +26.4V.
5) The driver transistor drops 1V to +25.4V.
6) The output transistors drop 0.5V plus 0.3V for their emitter resistors to +24.6V.
7)At 3A the 0.47 ohms current sense resistor drops 1.4V to +23.2V.
If the opamps and transistors have maximum spec's then the maximum regulated output might be +26V.

The negative supply is not needed for the voltage amplifying opamp since its output can go down to 0.5V above its negative supply that can be the circuit's 0V.

[iMo]

And finally the stability with a capacitive load.
15deg phase margin - it needs to play with compensation..

[Kleinstein]

One could reasonably simple modify the kit circuit to work essentially without a negative supply. The main point they use the negative supply is for the current control to pull down the set voltage alle the way to zero. So if one shifts the zero point for the set voltage to some +2 V there is no more need for the negative supply. So the feedback divider would go towards this auxiliary 2 V instead of GND and the set voltage too.

There is a slight downside that the output would not drop all the way to zero without a minimum load - depending on the minimum load circuit there could be some 50 mV or so as a minimum. To really go to zero the minimum load would need the negative supply. So the need is back, but the regulator could still work without it, just not below some 100 mV with no load.

The main limitation of the circuit is that the output OP needs to make the full voltage swing and a little more voltage lost. The good point is the the voltage compensation can be simple.

[Zero999]

And finally the stability with a capacitive load.
15deg phase margin - it needs to play with compensation..

The first iteration had a gain of 10, but I changed it to 12V, because 2.5V is a common voltage reference: there was a reason for that seemingly odd number.

You'll need a higher input voltage than 30V to get 30V out, probably at least 34V.

Try simulating the transient response. Change the load current from 0A to 1A and 1A to 3A and back to zero again. If you use a current sink, then don't forget to click "This is an active load" in the parasitic properties section, otherwise it will try to force the output to sink current, when the output transistors are off.

Another option is to use an if function with the load resistor.

It works like the Excel.
if(condition, value_if_true, value_if_false)

Suppose you set the value of RL to:
R=if(time>10m&time<20m,1,10)

It would set the value of the load resistance to 10Ohm and change it to 1R between 10ms and 20ms.

See example attached. Note the current and voltage spikes. I imagine the OP07 will be worse, then the ideal op-amp model I used with  GBW=10Meg Slew=10Meg. You're right, adding a parallel capacitor would reduce the voltage spikes but make the current spikes worse. A series inductor would do the reverse: better current regulation at the expense of voltage. Care needs to be taken to avoid oscillation.

I've noticed the current regulation is poor for lots of bench supplies, which are optimised for voltage. I've toyed with the idea of building my own, with a switch to select between better voltage or current regulation.

[Kleinstein]

Current regulation is often not that good for different reasons:
One is having a relatively large capacitor at the output. This limits voltage overshoot, but can add significant transient extra current.
Another point is that a good shunt is expensive - so they tend to use a relatively low value shunt to limit the power dissipation and self heating. Still heating can be a problem in cheap supplies. Without range switching it get's tricky to measure small currents with a shunt made for 3 A.
Also current regulation is often considered less important - more like a way to protect the supply and maybe circuit.

Many commercial supplies use the floating regulators, that are in principle quite good in current regulation, except for the output capacitor.
So there is not much option to switch the supply to a mode better for current regulation, as the output capacitor is required to get stability.
In principle one could use a faster, tighter tuned controlled loop for the voltage control, so that less (and with well defined ESR) capacitance at the output is sufficient. So it would be more like improving voltage control to alow a smaller capacitor and this way faster current regulation. Adding range switching would be another step.  I don't see much need to really switch the regulator to a separate mode - maybe add an optional addition capacitor to the output if needed to support the voltage regulation.
The other point is a delay before the current liming sets in - this sometimes is desirable, but sometimes not. So this could be a switchable point.

The circuit as shown, with an emitter follower at the output tends to produce current spikes from the regulator, as it takes some time for the CC regulator to kick in and the OP has to slew down - which essentially limits the rate on how fast the voltage can drop. It can work with less cap at the output, but there is a kind of simulated capacitance.

[Zero999]

Current regulation is often not that good for different reasons:
One is having a relatively large capacitor at the output. This limits voltage overshoot, but can add significant transient extra current.
Another point is that a good shunt is expensive - so they tend to use a relatively low value shunt to limit the power dissipation and self heating. Still heating can be a problem in cheap supplies. Without range switching it get's tricky to measure small currents with a shunt made for 3 A.
Also current regulation is often considered less important - more like a way to protect the supply and maybe circuit.

Many commercial supplies use the floating regulators, that are in principle quite good in current regulation, except for the output capacitor.
So there is not much option to switch the supply to a mode better for current regulation, as the output capacitor is required to get stability.
In principle one could use a faster, tighter tuned controlled loop for the voltage control, so that less (and with well defined ESR) capacitance at the output is sufficient. So it would be more like improving voltage control to alow a smaller capacitor and this way faster current regulation. Adding range switching would be another step.  I don't see much need to really switch the regulator to a separate mode - maybe add an optional addition capacitor to the output if needed to support the voltage regulation.
The other point is a delay before the current liming sets in - this sometimes is desirable, but sometimes not. So this could be a switchable point.

The circuit as shown, with an emitter follower at the output tends to produce current spikes from the regulator, as it takes some time for the CC regulator to kick in and the OP has to slew down - which essentially limits the rate on how fast the voltage can drop. It can work with less cap at the output, but there is a kind of simulated capacitance.

Yes a capacitor on the output doesn't help with current mode, but I think a glitch when switching back and fourth constant current and voltage is inevitable. When in CV mode, the CC amplifier will be trying to force it to limit the current and vice versa. Whether there's a voltage or current spike, depends on the design. Even if the rapid change on the output of either error amplifier can be greatly reduced, there's still the issue of the output having to rapidly change from a very low, to high impedance and back again.

Yes, whether allowing current surges or not depends on the application. Recently I've being using a bench supply to power some LEDs and have had issues with huge current spikes at power up. Fortunately it didn't damage the LEDs, but it isn't good. It isn't the output capacitor because it should be discharged when the power is applied. I will try connecting a short circuit, in parallel with the LEDs, during power up.

[Kleinstein]

The power up phase is a different issue from the normal regulation. In better supplies there is some extra circuit to keep the output off before the supply to the control is fully there and when the supply breaks down (turn off - which can be the more difficult case to detect). In the simple form this could be a way to disable the output if the raw voltage is below a certain level, needed to operate the OPs.

The simple circuits have usually quite some glitch when switching between CC and CV mode, because the non active regulator will often run all the way to it's limits. In normal control theory this is call windup and in better regulators handled separately. This is not standard for lab supply circuits, but it is possible to add this to the circuit.  With some extra circuitry the change over can be quite smooth.

The circuit from the start of this thread uses a slightly unusual form, as the CC mode reduces the set voltage and thus works with both regulators in series. So there would be no (or only a small) spike when going from CC to CV mode as the voltage regulation is always active.

[Zero999]

The power up phase is a different issue from the normal regulation. In better supplies there is some extra circuit to keep the output off before the supply to the control is fully there and when the supply breaks down (turn off - which can be the more difficult case to detect). In the simple form this could be a way to disable the output if the raw voltage is below a certain level, needed to operate the OPs.

The simple circuits have usually quite some glitch when switching between CC and CV mode, because the non active regulator will often run all the way to it's limits. In normal control theory this is call windup and in better regulators handled separately. This is not standard for lab supply circuits, but it is possible to add this to the circuit.  With some extra circuitry the change over can be quite smooth.

The circuit from the start of this thread uses a slightly unusual form, as the CC mode reduces the set voltage and thus works with both regulators in series. So there would be no (or only a small) spike when going from CC to CV mode as the voltage regulation is always active.

I'm interested to see how the seamless switching between modes works.

The original draft of my Howland current source based design did a similar thing to the circuit at the start of the thread, except it was the CV amplifier reducing the CC amplifier's set current. I only changed it to challenge the myth that a negative power supply is required, otherwise I probably wouldn't have posted it, since it was just a quick sketch, rather than a complete design.

[not1xor1]

If you have seen a way for PSU adjustable current-limiting without a negative rail or dedicated floating rail, let me know.

On the chinese kit, I hate the -ve rail and the need for the V ctrl op-amp to run at the top rail while the I ctrl op-amp runs on the low-side. I can find away around it all, but haven't built it to prove it works.

A negative supply isn't needed, if single supply op-amps are used. Here's a proof of concept, based on a Howland current source, with voltage limiting. I haven't simulated the transient response, which is probably horrible. U1 is shorted by Q3 and current limits, when in voltage mode, which some may not like, but that can be avoided if necessary. The Darlington pair could be driven by a current mirror based source, with each op-amp diode AND'd pulling it down.

I don't see the problem with the negative rail. I'd probably use the OP07 run off regulated +34V and -3V, derived from a 30V transformer.

I ran a few tests on variations of this circuit with LTspice.
The circuit might work without negative supply with a LT1013. It has a better phase margin with low loads if you use 100k/470k resistors.
The main drawback is the resistor tolerance. Even just 1% of umbalance would cause about 15% of current variation vs. load value (or output voltage if you prefer). 0.1% tolerance resistors would provide about 1.7% current variation which might be acceptable in a cheap DIY PSU provided that aging and temperature drift do not make it much worse.

BTW the output current can be calculated like:
I = (Vref * Rhot) / (Rsense *Rgnd)
where Rhot is the resistor connected to the emitter/output Rhot are the 2 resistors connected betwen the BJT emitter/output and the opamp inputs, Rgnd are the resistors which connect the opamp inputs to ground... and Rsense... guess it  .

[soldar]

I've noticed the current regulation is poor for lots of bench supplies, which are optimised for voltage. I've toyed with the idea of building my own, with a switch to select between better voltage or current regulation.

I think this is natural for starter, basic power supplies because they are used mainly in voltage mode and the current limiting is just a rough setting for protection and is not as critical as the voltage setting.

[Tek Tech]

Probably one could increase the available power by using an optional tracking pre-regulator feeding the power stage. The tracking pre-regulator will increase the output noise somewhat, but one can get higher output current with less heat.

That's a brilliant suggestion! Do you have any idea how to make it relatively simply?

[Zero999]

I've noticed the current regulation is poor for lots of bench supplies, which are optimised for voltage. I've toyed with the idea of building my own, with a switch to select between better voltage or current regulation.

I think this is natural for starter, basic power supplies because they are used mainly in voltage mode and the current limiting is just a rough setting for protection and is not as critical as the voltage setting.

Perhaps in the past but nowadays lab PSUs are often used to power LEDs, in which case the constant current setting is more critical.

[Kleinstein]

A good current regulation is not often required and it also adds quite some costs for the shunt. To get good resolution at low currents the shunt can not be very small and with more drop the self heating of the shunt can limit the stability of the current regulation.

Another point is that quite often the capacitance at the output is quite large and this makes the dynamics of the current regulation not that good.

[Zero999]

A good current regulation is not often required and it also adds quite some costs for the shunt. To get good resolution at low currents the shunt can not be very small and with more drop the self heating of the shunt can limit the stability of the current regulation.

Another point is that quite often the capacitance at the output is quite large and this makes the dynamics of the current regulation not that good.

I would quite happily pay good money for a lab power supply with both decent voltage and current regulation. I don't demand anything fancy, just reasonable current regulation (5%) from 30mA to 3A, without high current pulses, more than double the setting.

If it's well-designed, an output capacitor shouldn't be necessary. Any circuit which needs a power supply, with a low impedance at high frequencies, should have a decent decoupling:  the transient voltage response, of a lab PSU, isn't as critical, as many believe it to be.

[Kleinstein]

The output cap is there for 4 reasons:
1) some capacitance could be needed for the stability of the CV loop  (usually not very much though), especially if made to also tolerate a larger capacity.
2) To avoid to much voltage drop on pulsed load the capacitor can reduce the drop, though making the time longer. Kind of having the cap to cover up a slow regulator and making it even slower
3) To reduce possible voltage overshoot on transients (turn on, CC to CV transition)
4) for the stability of the CC mode with inductive loads  (this usually does not need much).

Especially point 3 may need quite some capacitance.

A very fast acting current limit can also be quite confusing as it may be too fast for some applications.

[Zero999]

Oh I know why it's there. I just think it can be greatly reduced, if not eliminated.

Local decoupling (which you should have anyway) will overcome the issues with pulsing and overshoot for sensitive loads. Seriously, who connects a microcontroller up with no decoupling? Even with the capacitor, inside the PSU, there can be spikes due to lead inductance.

I think a switch to select between optimal current and voltage modes. One day I might get round to putting together a proper design.

[Kleinstein]

The usual circuits don't have much anti-windup for the regulators. This gives some extra current peak when going from CV to CC mode - this may be not that bad and even could be a good thing in some cases. However for the CC to CV transition this gives an extra charge pulse and thus possible voltage overshoot - a larger cap at the output is the easy way to limit this overshoot. Here one can not really rely on the cap at the circuit itself. In addition it does not look good if there is a 5 V overshoot without an extra cap at the output.

Just for stability a capacitor in the 1 µF range could be sufficient for many circuits.
I agree that the transients in CV mode are not that important, but a fast regulation is also the obvious way to reduce the capacitance at the output.

[wandows]

Hello everybody! I would like to count on the help of friends!!!

In the offset adjustment, my power supply has the output voltage >= 32.9V, and does not reduce to 30V.

Any tips on what can be done to solve this problem?

[antoniodv]

Hi everyone.
Thank you all for the information you have shared.
Inspired by the best electrical diagrams of linear power supplies made in the past and available online, I digitized them by adding a display microcontroller and DAC to have everything in a single PCB.
As you can see in attached pdf , for the final stage I used the K7200 scheme.
The firmware for ATtiny3216 is also ready.
I'm waiting to complete the final PCB, all tests went well. I just doubt the voltage peak at shutdown I would like to remove it. Advises?

[MrAl]

Hello there,

My question is mostly for the original circuit.

Why use a circuit so complicated?  What are the specs and do the specs warrant such a complex circuit design?
If it was 30 amps I could see it, but 3 amps, that's nothing these days.

As to the newer circuit, still a bit complex.  What are the specs.

Maybe a related question:
What specs are you looking for.

[xavier60]

Maybe the -3.3V is discharging too quickly. Im curious about the CV loop's stability?
The design that I like, doesn't need a negative control rail,
https://www.eevblog.com/forum/beginners/lm324-power-supply-with-variable-voltage-and-current/msg3582664/#msg3582664