how to design fast bench supply with CC and CV?

[temperance]

I wonder why people want to try to build a fast high current high voltage power supply without output capacitors. ...Just to be able to not burn out a small 25mA max LED's when you forgot to turn down the voltage?

Think about what will happen when there is no output cap and you connect some inductance or capacitance to the supply. How are you going to compensate for this? You can try to implement two feedback paths. A fast lane taken somewhere from a pre-driver and a DC lane taken from the output. Or just over compensate. But that's not what you really wanted to do.

Simply put, different applications require different power supplies.

If you happen to work a lot on low power applications, you will need a 15V precision 250mA power supply with a multi turn pot to adjust the output voltage. The output current limiting is not that critical, just to not damage anything.
 
For general work, 1A 25V supply is more then you'll need for every day work. The days of power hungry devices are long gone.

And for some things you might need a 40V 20A supply. In my case, I will just buy what I need because I don't have time to build one. Of course this beast will not be used to test 25mA LED's.

[exe]

I plan on building building PSU with four channels: 2x2A and 2x125mA or thereabout. I want to use the same architecture, may be even the same pcb, the difference only in output stage, shunt resistor, and compensation network.

[temperance]

A simple op amp with an emitter follower will outperform a more complicated architectures with ease when it comes to speed. Try one in LT spice and you will see what I mean.

The often copied circuit with the floating op amps is actually a fancy common emitter circuit which requires a lot of compensation. This thing will for sure oscillate with some self inductance at the output with no output capacitor. A common emitter circuit with some self inductance in it's collector is a well known oscillator circuit.

[David Hess]

I wonder why people want to try to build a fast high current high voltage power supply without output capacitors. ...Just to be able to not burn out a small 25mA max LED's when you forgot to turn down the voltage?

Low output capacitance yields better constant current performance which is important in some applications.  But the most common reason to have a low output capacitance is for fast response in automated testing applications where time is money.

[temperance]

I wonder why people want to try to build a fast high current high voltage power supply without output capacitors. ...Just to be able to not burn out a small 25mA max LED's when you forgot to turn down the voltage?

Low output capacitance yields better constant current performance which is important in some applications.  But the most common reason to have a low output capacitance is for fast response in automated testing applications where time is money.

You didn't read what I wrote. I'm referring to what people want to have in their bench power supplies and how the no output cap idea conflicts with their requirements and the very often copied and misunderstood common emitter power supply design.

They are not building an in circuit tester able to measure 100 resistors in a second.

[exe]

My friends, just in case, the thread is started exactly for this reason: share my ideas, get other's ideas and get feedback for them. So, it's totally fine to provide feedback and question requirements and assumptions.

I'm discovering what is possible, and how to achieve that. If a few cheap parts can let me achieve "100 per second", then why not? Esp. if I can make a flexible design with multiple modes of operation (or build two versions: with slow response and fast response).

[David Hess]

You didn't read what I wrote. I'm referring to what people want to have in their bench power supplies and how the no output cap idea conflicts with their requirements and the very often copied and misunderstood common emitter power supply design.

I just did not bother to address it.  It is a small step from an emitter follower to a class-AB output stage which has stability advantages whether the output capacitance is high or low.

If a class-AB output stage is not used, then it is often beneficial to add an active pull-down (or pull-up) so when adjusting the output voltage with no load, the output voltage does not linger.

Common emitter?  Earlier you wrote:

A simple op amp with an emitter follower will outperform a more complicated architectures with ease when it comes to speed.

An emitter follower is common collector and completely different from common emitter.  Common emitter output stages have a wider variation in output impedance which complicates frequency compensation.  But either can be used successfully.

They are not building an in circuit tester able to measure 100 resistors in a second.

But they might want a fast response current limit unhindered by high output capacitance.  Or some tests require high slew rate.

Where things really fall apart is poor power supply designs which have 1000s of microfarads of output capacitance which implies frequency compensation problems or a "more is better" attitude.  A good "high capacitance" design can get by with 22 to 100 microfarads per amp.

[Kleinstein]

The OP and emitter follower variant can be relatively easy for lower voltages up to about 20 V, maybe 25 V. It needs to OP to to cover the whole output voltage range + a little reserve.  Here the voltage regulation tends to be easy (at least for moderate speed), but the current regulation can turn out a little tricky. Even though there is no physical cap at the output, the limited slew rate of the OP can also cause a large current spike in case of a short. One kind of gets a simulated capacitance.

The common emitter type circuit with a current setting output stage is often used as a floating regulator. It needs some capacitance at the output, though not very much. It need a more complicated compensation even in the basic version.  If one wants fast regulation from the emitter-follower version one often ends up with a quite similar compensation. So at the higher performance level the difference is not that relevant anymore. The floating regulator is not so much limited in the voltage range and the same board (just different values, transistors) could be use for low voltages (e.g. 5 V) and high current or high voltages (e.g. 300 V) and lower current.

Some designs (even some commercial lab supplies) use a large capacitance in the 1000 µF range  as a brute force way to avoid overshoot in the CC to CV transition. However there are more intelligent ways to avoid this and avoid the overshoot without a large capacitor.
A large capacitance at the output does not really help the regulator. It is more making things more difficult and the regulation slower.
A large capacitor at the output can also make a fast current measurement more tricky.

[Zero999]

I wonder why people want to try to build a fast high current high voltage power supply without output capacitors. ...Just to be able to not burn out a small 25mA max LED's when you forgot to turn down the voltage?

Low output capacitance yields better constant current performance which is important in some applications.  But the most common reason to have a low output capacitance is for fast response in automated testing applications where time is money.

You didn't read what I wrote. I'm referring to what people want to have in their bench power supplies and how the no output cap idea conflicts with their requirements and the very often copied and misunderstood common emitter power supply design.

They are not building an in circuit tester able to measure 100 resistors in a second.

You clearly didn't read the original post or much of this thread. No one is talking about building a high current, high current power supply. The specification in the original post is 15V, 3A maximum.

I don't see how not having a built-in output capacitor can cause any problems. As mentioned earlier, any circuit sensitive to voltage troughs and spikes, should already have good enough supply decoupling, that it shouldn't matter. Even if the power supply has a very low output impedance, up to hundreds of MHz, this will no longer be the case for a circuit connected to the output, via a 1m length of cable, which will have an impedance of the order of 100R, higher if it's two separate flying leads, at RF.

A solution could be to have ground-referenced voltage, and somehow with some level-shifting set current on a high side. Like on the included schematic. I discarded that idea. One of problem is that it didn't want to start, so I added this startup resistor across pass transistor. The second is I cannot use a clamping diode with this design (this is the only type of clamping I mastered so far, but I'm studying other proposed methods).

The main show stopper there is the minimum compliance voltage of the current sink will limit the minimum output voltage. It's better to go for an inverter output, like Blackdog's circuit or high side current sensing, such as the Howland topology, in the circuits I've posted. Low side current sensing is also a possibility, but that makes controlling the voltage reference more tricky, as it will float above the output's negative rail.

[temperance]

Picture on the left: emitter follower
Picture middle: common emitter circuit. The usual floating op amp circuit.
Picture on the right: an oscillator. Especially with C2 installed

[exe]

My friends, where do you get the idea that I want to build a power supply with no output cap? I think mentioned "minimum" capacitance*. Even if I said so, all requirements are flexible.

Nonetheless, it's a very nice remark and a reminder that some output capacitance is needed for stability. Esp. if it's a high-speed design.

*I never defined what minimum is. As a benchmark, my current power supply has 4.7uF and outputs 1A. The output cap is mlcc, so it can discharge quite fast (and this is probably not a good thing).

[Zero999]

Picture on the left: emitter follower
Picture middle: common emitter circuit. The usual floating op amp circuit.
Picture on the right: an oscillator. Especially with C2 installed

Sorry, I don't know what you're on about. All of the circuits posted in this thread have an emitter follower output stage. The schematics you've posted there are nothing like any others in this thread. Only the left one makes any sense and bares some resemblance to a power supply and a basic one at that: just an emitter follower and voltage reference. The middle one has no 0V reference and the right one has no DC connection to 0V. 

My friends, where do you get the idea that I want to build a power supply with no output cap? I think mentioned "minimum" capacitance*. Even if I said so, all requirements are flexible.

Nonetheless, it's a very nice remark and a reminder that some output capacitance is needed for stability. Esp. if it's a high-speed design.

*I never defined what minimum is. As a benchmark, my current power supply has 4.7uF and outputs 1A. The output cap is mlcc, so it can discharge quite fast (and this is probably not a good thing).

It should be possible to make a voltage regulator which is stable, without any output capacitance. It just won't have so good transient response, in voltage mode. Lots of voltage regulator ICs, such as the LM317 will work without an output capacitor, with minimal oscillation. Try a simple LM317 circuit, set to say 5V, with on output capacitor. It might undershoot and overshoot a bit, when a load is applied and removed, but it won't go into full blown oscillation.

Whether there's an output capacitor or not, depends on how important the voltage regulation is, compared to the current regulation. Adding a capacitor will improve the former, at the cost of the latter, which ideally needs an inductor. A simple DPDT switch can be used to select between either a capacitor or inductor on the output, do change from either better voltage or current regulation.

[exe]

Try a simple LM317 circuit, set to say 5V, with on output capacitor. It might undershoot and overshoot a bit, when a load is applied and removed, but it won't go into full blown oscillation.

How did they achieve that? I always thought it's because they are very slow (overcompensated) and because they use emitter follower output stage. It also has an internal shunt which, I guess, prevents fast oscillation of the output stage.

[temperance]

It's correct, you never said something like that. Under 6 you mention: minimum output capacitance. Others have taken this to the extreme in post 14.

The circuit in Reply 57 is a common emitter circuit which you can not simulate properly in spice without mortifying the voltage source.

The LM317 doesn't oscillate without output caps because the common emitter output stage is fed by a current source rolling of the gain of the output stage early on.

@ Zero999
The circuits as drawn don't need a reference. They only demonstrate that the circuit in post 57 is a common emitter amplifier. You can put R4 in the second schematic in the collector if you want. It's still the same circuit.

[Zero999]

There was a very good post here and it's just vanished. One of the points made was a emitter follower is better for voltage regulation, as it has a very low output impedance and a common emitter is better at current regulation, as it has a very high output impedance. I did design a simple CV/CC PSU circuit using three output transistors, so when in CV mode the output stage was a Darlington pair and when it CC mode, it switched to Sziklai pair configuration. I'll have to dig it out and finish it off.

It's correct, you never said something like that. Under 6 you mention: minimum output capacitance. Others have taken this to the extreme in post 14.

The circuit in Reply 57 is a common emitter circuit which you can not simulate properly in spice without mortifying the voltage source.

The LM317 doesn't oscillate without output caps because the common emitter output stage is fed by a current source rolling of the gain of the output stage early on.

@ Zero999
The circuits as drawn don't need a reference. They only demonstrate that the circuit in post 57 is a common emitter amplifier. You can put R4 in the second schematic in the collector if you want. It's still the same circuit.

What's your point? All of the circuits here have an emitter follower output stage. It isn't obvious to me what the circuits you've posted there are supposed to represent. Of course a power supply circuit needs a reference. 

[udok]

I deleted my post because i don't want to get into too long discussions.

But one main point was, that a voltage source and current source do not mix up well.

It starts with the output stage design, a voltage source needs a low impedance node.
Therefore choose an emitter follower and not a collector output.
Global feedback will lower the output impedance by the feedback factor,
but it is better to start with a low value if you want a voltage source.
If you want a current source, start with a collector output which has high impedance
right from the start.
The output impedance of an emitter follower is less than 1 Ohm right from the start,
and the second pole which comes into play with a capacitive loads is often at a much higher
frequency than the unity gain frequency of the regulator loop.
This is the case with the LM317, which often does not need a capacitor for stability.

With "fast" regulator loops you have to cope with the second pole introduced by the
open loop output resistance and the load capacitor.
Two poles together have a -12 db/octave Bode plot which is instable if the
unity gain frequency is in this region.
In this case you can use tricks like the -3 dB/octave Bode plot mentioned by David Hess,
or you could place a zero in the loop, or you can specify requirements for the output
capacitor and the ESR.  The ESR together with the output cap introduces a zero,
and the -12 dB/octave becomes a stable -6 dB/octave Bode plot.
Fast designs use almost always sense wires.  This creates more problems,
as the resistance and inductance of the power wiring does not isolate
the capacitor anymore from the regulator loop.

The problems with mixing up current and voltage sources does not end here.
The very efficient Harrison circuit is not a good current source because of the
bias and leakage currents which flow through the current measuring shunt.
Then the current shunt itself should be higher for a good current source
which means more wasted power.
Almost any modern power supply uses a preregulator to reduce power waste
and costs for cooling.
These designs have only one or two Volts of headroom.  This makes
them useless as a current source with dynamic loads.
 

[Kleinstein]

What's your point? All of the circuits here have an emitter follower output stage.

Not all the circuits shown here have the emitter follower output stage. Especially the floating regulators are usually not. They may look like an emitter follower, but the control is relative to the positive side / emitter. So the control is actually common emitter and thus a current setting output stage.

Switching between the 2 types of regulator can be quite confusing. Both have there pros and cons and one kind of has to decide which way to go. In such a forum it gets very confusing of different circuit types are mixed in one thread.  Quite often the emitter follower type is easier to design and can be a little simpler, at least at low speed. However it is also usually limited to low output voltage, like 20-25V. Already 30 V may need special OPs that can work with a little more supply.

[temperance]

Thank you Kleinstein for confirming my input in here about common emitter stages.

@ Zero999 thank you for those multiple 

Some advise: learn the basics before you give other people advice and spice up power supplies without understanding what you're doing. Also, keep the  to yourself even if you disagree. It's not very polite and you look stupid if you turn out to be wrong. What does the person who opened this topic has to think about this?


I'm out of here.

[Zero999]

Thank you Kleinstein for confirming my input in here about common emitter stages.

@ Zero999 thank you for those multiple 

Some advise: learn the basics before you give other people advice and spice up power supplies without understanding what you're doing. Also, keep the  to yourself even if you disagree. It's not very polite and you look stupid if you turn out to be wrong. What does the person who opened this topic has to think about this?


I'm out of here.

There's no need to throw your toys out of the pram. A simple, succinct explanation of the confusing schematics you posted would have done. Post crap, incoherent, garbage and you will get plenty of emojis. I get that they weren't supposed to be complete designs, but come on, at least show where the load, voltage reference and 0V are supposed to be. I know this is the beginners section, but I thought you were supposed to be advising the original poster how to do it properly.

What's your point? All of the circuits here have an emitter follower output stage.

Not all the circuits shown here have the emitter follower output stage. Especially the floating regulators are usually not. They may look like an emitter follower, but the control is relative to the positive side / emitter. So the control is actually common emitter and thus a current setting output stage.

Switching between the 2 types of regulator can be quite confusing. Both have there pros and cons and one kind of has to decide which way to go. In such a forum it gets very confusing of different circuit types are mixed in one thread.  Quite often the emitter follower type is easier to design and can be a little simpler, at least at low speed. However it is also usually limited to low output voltage, like 20-25V. Already 30 V may need special OPs that can work with a little more supply.

Which circuits, posted in this thread don't have an emitter follower output stage? Could temperance be confusing this thread with another one, or am I mistaken? If it's the latter, please link to the ones which don't have an emitter follower output stage.

I don't see the issue with using an emitter follower at higher voltages. There are plenty of ways of boosting the output voltage from an op-amp to drive an emitter follower at higher voltages, allowing a low voltage op-amp to be used. Granted, adding another stage of amplification will increase the risk of oscillation, but I don't see how it's as bad as having a common emitter output stage who's gain is heavily dependant on the load.

[xavier60]

Technically, the position of the CS resistor in the diagram in  Reply #57 makes the Darligton an Emitter follower.
The function of the Darlington in the design as a whole is of a common Emitter driven current source.

[exe]

Hello again.

I tried my idea. There is no problem with setting voltage, but I have problem reading it back (see the picture). Any current drawn by the floating circuit goes through the shunt (because floating ground is connected to the output which is happen to be shunt). So, even if I buffer output voltage (in this case it will be virtual ground), the current drawn from the sensing circuit will go through the shunt (the return path).

There were several proposals using differential amplifier. My reasons to avoid howland/differential circuits were:
1) they have low input impedance
2) CMRR is poor and I need to match resistors
3) limited output swing (afaik)

Now I'm thinking that it might be not so bad idea, esp. because common mode shouldn't be too big across the shunt. So, I'll try circuits posted earlier.

There are ready instrumentation amplifiers, but they are quite slow, I'm worried this may affect performance. They are also mostly low-voltage (3-7V range typically, afaik).

PS emitter follower vs common emitter: it's an interesting detail that I missed originally. This explains why moving reference voltage from positive output to ground make the circuit unstable, especially when switching CC/CV mode. I spent a lot of time stabilizing it, and mostly failed. So, good to know.

PPS I also tried doing some analog switching with jfets and low-level mosfets, but this didn't get me far.

[Zero999]

Yes, the Howland circuits need well-matched resistors to work well, otherwise the output impedance will be low, when in CC mode. I don't see how having a relatively low input impedance is an issue, just add a unity gain stage. If an instrumentation amplifier IC, with a sense, as well as a reference input, can be used to make a Howland current source. Some examples include AMP02, INA103,  INA125.

I'm not sure what you're trying to achieve. The current is only significant because you chose very low resistor values.

 no_cc_floating_ref_sensing problem.asc (3.12 kB - downloaded 69 times.)

[exe]

The current is only significant because you chose very low resistor values.

I didn't use high-value resistors to avoid the noise. Perhaps, this is not so important for readback as I can heavily filter it, I only need may be a few kHz of bandwidth. In simulations I used small values (way smaller than I'd use) to exaggerate the effect to ease the analysis.

EDIT: also offset currents may come into play. So, I only use high-value resistors if I have to.

[exe]

Ah, silly me, I think I can buffer the output on the "floating" side. Stay tuned

[exe]

The idea didn't work. In this arrangement, in CC mode the reference voltage V7 goes below ground. The clamping diode would clamp it forever. So, time to try other solutions from the thread. I have thoughts to give up on this idea, but I still would like to figure out how other solutions work.