Stability and behaviour of current sources

[nihial]

Hi,


I'm trying to do a current source for "high" voltage (around 200 volts).

I'm trying to confirm my circuit will have a chance to be stable or if I'm missing something can someone confirm I'm on the right track ?

I used a note from analog device to evaluate the stability, I placed an AC source between the FB node and the - input of the op amp. Then I look the ratio of V(fb)/V(inm) and the phase should stay well above 0 until the gain drops.

I also have a few questions:

 - Is there a opamp technology that's better suited for this task ?
 - What is the impact of the bandwidth of the op amp ? Intuitively an opamp with a very high bandwidth might try to correct noise or whatever and may be worse than a lower bandwidth one but I'm not sure if this is the case or not and how to address that if it's the case.

Don't mind the thermal aspect for now i'm aware that the transistor will suffer

[temperance]

This type of question appears on this forum almost every week.

This might help:
https://www.eevblog.com/forum/projects/just-another-dc-load/

[mtwieg]

Basically yes that circuit is a great starting point (assuming the load doesn't need to be connected to GND).

As for analyzing stability, yes injecting stimulus in that branch should work. Personally I would put V5 in series with R3, but I don't think it really changes the results (just more intuitive IMO).

As for opamp selection, at some point the overall closed loop bandwidth is probably going to be limited by the power components, not the opamp. I would decide roughly on a target loop bandwidth (i.e. crossover frequency), and pick an opamp with a GBW 10-100x that. And simulate its stability using an accurate SPICE model, at multiple bias conditions, if possible.

As for "opamp technology", not sure exactly what you're referring to. I would definitely use a single supply opamp whose inputs and output work all the way to GND. Beyond that, hard to give advice without knowing more about the application.

[mikerj]

Note that stability will be greatly influenced by your load, making it work nicely is easy with a purely resistive load but add inductance and it could turn into an oscillator.

[nihial]

This type of question appears on this forum almost every week.

This might help:
https://www.eevblog.com/forum/projects/just-another-dc-load/

This is a nice compilation of information for my personnal knowledge I'll definitly read it entirely not sure every question is answered there but definitely worth the read

[nihial]

Basically yes that circuit is a great starting point (assuming the load doesn't need to be connected to GND).

As for analyzing stability, yes injecting stimulus in that branch should work. Personally I would put V5 in series with R3, but I don't think it really changes the results (just more intuitive IMO).

As for opamp selection, at some point the overall closed loop bandwidth is probably going to be limited by the power components, not the opamp. I would decide roughly on a target loop bandwidth (i.e. crossover frequency), and pick an opamp with a GBW 10-100x that. And simulate its stability using an accurate SPICE model, at multiple bias conditions, if possible.

As for "opamp technology", not sure exactly what you're referring to. I would definitely use a single supply opamp whose inputs and output work all the way to GND. Beyond that, hard to give advice without knowing more about the application.

I guess when you say intuitively you mean that you visually add the perturbation inside the feedback and observe the result on the output. I see it differently (injecting a signal before the A B blocks and seeing what's coming back) but I look at it with a very abstract view. If it works it's good enough for me

I know that putting it where I put it has a disadvantage, apparently to do it perfectly, I should add the input capacitance of the opamp in the feedback loop. By putting where you suggest I don't know with certainty what the implications are.

The multiple bias point seems to be a good idea.

By technology I mean, jfet input, bjt opamps,... this kind of things.
Same question could apply about using a mosfet or a jfet I don't know what are the implications of that

Application is a load controlled by a function generator for 200V power supplies with a modest current output, around 150mA

[nihial]

Note that stability will be greatly influenced by your load, making it work nicely is easy with a purely resistive load but add inductance and it could turn into an oscillator.

Valid remark, I should simulate modest series inductance and find a solution to be resilient to that

[temperance]

As for opamp selection, at some point the overall closed loop bandwidth is probably going to be limited by the power components, not the opamp. I would decide roughly on a target loop bandwidth (i.e. crossover frequency), and pick an opamp with a GBW 10-100x that.

The op amp bandwidth must be much smaller than the bandwidth of the power stage. Otherwise the phase shift introduced by the power stage is amplified by the op amp = good oscillator.

[temperance]

Note that stability will be greatly influenced by your load, making it work nicely is easy with a purely resistive load but add inductance and it could turn into an oscillator.

Valid remark, I should simulate modest series inductance and find a solution to be resilient to that

See the tread I linked to and the discussion about the RC network across the power MOSFET to compensate for the wiring inductance. The simulations are all there.

[Terry Bites]

A small series resistor in the drain can help keep oscillation at bay.

Idss and body diode leakage in the MOSFET can easily reach into the 10s of uA (and more) and the bigger the MOSFET the worse that gets. Badder with temperature too.

[mtwieg]

I guess when you say intuitively you mean

Nah more like that's my gut instinct, without any specific reasoning.

that you visually add the perturbation inside the feedback and observe the result on the output. I see it differently (injecting a signal before the A B blocks and seeing what's coming back) but I look at it with a very abstract view. If it works it's good enough for me

I know that putting it where I put it has a disadvantage, apparently to do it perfectly, I should add the input capacitance of the opamp in the feedback loop. By putting where you suggest I don't know with certainty what the implications are.

It's interesting to think about. I can't really explain my preference, so I resorted to doing a simulation to see the difference (if any), see attached.

I set up three perturbations sources:
A is the place in your schematic
B is where I suggested
C is on the reference input (for getting the closed loop transfer function)

The loop gain transfer functions according to A and B are plotted on top. The closed loop transfer function is plotted on the bottom. You can see loop gain B crosses 0dB around 680kHz with a phase margin of 34 degrees, which agrees with the closed loop gain peaking at that frequency. The loop gain A crosses 0dB at around 9.5MHz with a phase of around 100 degrees, which definitely isn't right. But maybe a different way of calculating loop loop gain is needed when perturbing at A.

By technology I mean, jfet input, bjt opamps,... this kind of things.

Eh, probably not critical unless you're really trying to push bandwidth/noise. I wouldn't go for a JFET-input opamp unless you also provide it a bipolar power supply.

Same question could apply about using a mosfet or a jfet I don't know what are the implications of that

For the power transistor, MOSFETs tend to do much better for these circuits, especially at high power levels. Would only consider BJTs if you need extreme noise/bandwidth (and don't care about the error from base current).

[nihial]

Except you don't break the loop by putting it there, there's still the path from the output to input with the 10pF

[Sassy Taste]

Op amp are high gain DC source amplifiers.  The ones you're talking about are big Class D amplifiers the type they use in concerts in front of the stage.  Those connect to the speakers above or on the ground and sound really loud.  They run on 120 volt lines.  So you're trying to see if they use OP AMP for the amplifier Class D?  No.  Just for DC supply volts.

[temperance]

Op amp are high gain DC source amplifiers.  The ones you're talking about are big Class D amplifiers the type they use in concerts in front of the stage.  Those connect to the speakers above or on the ground and sound really loud.  They run on 120 volt lines.  So you're trying to see if they use OP AMP for the amplifier Class D?  No.  Just for DC supply volts.

-Op amps have high DC gain when the feedback network doesn't reduce the DC gain to 1 by AC coupling the feedback network.
-Class D amps are full of op amps.

It is rather unclear what you are trying to say and to whom.

[ledtester]

Op amp are high gain DC source amplifiers.  The ones you're talking about are big Class D amplifiers the type they use in concerts in front of the stage.  Those connect to the speakers above or on the ground and sound really loud.  They run on 120 volt lines.  So you're trying to see if they use OP AMP for the amplifier Class D?  No.  Just for DC supply volts.

-Op amps have high DC gain when the feedback network doesn't reduce the DC gain to 1 by AC coupling the feedback network.
-Class D amps are full of op amps.

It is rather unclear what you are trying to say and to whom.

Reading the user's other posts it is very clear their posts are AI generated.

[EEVblog]

Reading the user's other posts it is very clear their posts are AI generated.

Unless an AIbot has emailed me, seems human.

[coppercone2]

oh shit, its the second variety

[Analog Kid]

Just an idiot, not "AI".

[ledtester]

Reading the user's other posts it is very clear their posts are AI generated.

Unless an AIbot has emailed me, seems human.

The content of the posts still could be AI generated.

[EEVblog]

The content of the posts still could be AI generated.

Possible, but they contacted me twice asking why they had been banned, so they passed the usual human filter.

[mtwieg]

Except you don't break the loop by putting it there, there's still the path from the output to input with the 10pF

I know what you mean, and you definitely have a point. But also consider that the FET is also an amplifier just like the op amp, and from its perspective there are two feedback paths, one via the 10pF cap and one via the op amp. From that perspective, one could make the same objection as you are, that injecting at A does not fully "break the loop," while injecting at B does.

It's certainly possible to change the circuit in such a way (like by adding a delay in the opamp input, for example) that measuring loop gain by injecting stimulus at A reveals instability while injecting at B does not. I think the complementary is also true, that injecting at A may also miss something you would by injecting at B under some conditions.

My preference for injecting at B isn't that it's more "correct", but rather that it is tells you much more about what the closed loop transfer function will look like (because it actually injects at the node we're trying to control).

[coppercone2]

find my thread about high voltage bjt current mirrors, it might be enough for you and their robust as hell. It said 10mA absolute maximum i gave it like 100 and it drifted a little lol


I think their a super solid design choice if the accuracy is good enough, compared to trying to wire in some APEX amplifier (which I already had a mega problem with)