Author Topic: Problems with stability of power supply circuit  (Read 2180 times)

0 Members and 1 Guest are viewing this topic.

Offline mtimmermansTopic starter

  • Newbie
  • Posts: 8
  • Country: nl
Problems with stability of power supply circuit
« on: August 04, 2019, 12:32:24 pm »
Hi,

After watching the power supply design videos of Dave, I got inspired to build my own. For my power supply however, I would like to replace the LT3080 with an op amp plus series pass transistor. The circuit worked fine when I used a resistive load, and I was able to change the output voltage by changing V3. But as Dave warned for in his video, I run into stability issues when adding a capacitive load. Of course I could just use the LT3080 to fix my problem, but for educational purposes I would like to fix this problem without one. I've added my circuit as an attachment to this post.

I am taught the basics of control systems / barkhousen stability criteria at university. So I did the analysis, derived the transfer function and plotted the transfer function in Matlab. This showed a phase margin of 36 degrees, which I guess is not enough and causes the oscillations. What they did not teach in our classes however, is how to actually solve this problem. I could of course decrease the output capacitance, but if I ever use a load with a capacitance greater than 20 uF, the circuit starts oscillating. Does anyone have a suggestion what I could do to solve this problem?
« Last Edit: August 04, 2019, 12:34:28 pm by mtimmermans »
 

Offline exe

  • Supporter
  • ****
  • Posts: 2647
  • Country: nl
  • self-educated hobbyist
Re: Problems with stability of power supply circuit
« Reply #1 on: August 04, 2019, 08:04:33 pm »
The simplest way is to add a capacitor (try 1n..10n) between inverting input and output. That would be the simplest type of so-called frequency compensation. Basically, find minimum capacitance that makes it stable and double it (this type of compensation is dominant pole compensation). Be sure to test it well with step load and test under several operating points to be sure it really is stable. The topic is quite complex, if someone can suggest a good practical guide would be great.

Anyway, the best starting point I know is https://www.allaboutcircuits.com/technical-articles/negative-feedback-part-4-introduction-to-stability/ . I suggest read the whole series of articles.
 
The following users thanked this post: ledtester

Offline David Hess

  • Super Contributor
  • ***
  • Posts: 17428
  • Country: us
  • DavidH
Re: Problems with stability of power supply circuit
« Reply #2 on: August 05, 2019, 01:02:24 am »
A real output capacitor has ESR which adds phase lead to the frequency compensation and this is why many regulators are unstable when a large low ESR ceramic or film capacitor is used.  So the first step is to include the ESR of the output capacitor in your simulation if it is an electrolytic capacitor

After that and depending on the circumstances, Exe's suggestion of external frequency compensation may be needed.  Usually this takes the form of a capacitive feedback network from the output of the operational amplifier to its inverting input.  But the LM358 is so slow that this should not be necessary.  Note that the LM358 is not too slow to make a good error amplifier; it will work fine.

Another common problem is driving the large input capacitance of a power MOSFET directly from an operational amplifier.  Doing so reduces phase margin and may result in outright oscillation.  A separate driver for the MOSFET gate would help.

A more advanced technique is to add some resistance in series with the source/emitter of the output transistor and take some AC feedback from before this point.  This allows using a low (or no) ESR output capacitor for zero AC output impedance while maintaining stability.  Integrated regulators may do this with special output transistor structures.
 

Offline T3sl4co1l

  • Super Contributor
  • ***
  • Posts: 22436
  • Country: us
  • Expert, Analog Electronics, PCB Layout, EMC
    • Seven Transistor Labs
Re: Problems with stability of power supply circuit
« Reply #3 on: August 05, 2019, 01:33:23 am »
The pictured waveform seems more like simulation noise than ordinary oscillation.  Try building a more physically representative circuit: add series resistance to the gate, capacitor, inductance to the drain, etc.  Adjust simulation parameters (the various TOLs; try GEAR 2 integration; etc.).

Tim
Seven Transistor Labs, LLC
Electronic design, from concept to prototype.
Bringing a project to life?  Send me a message!
 

Offline exe

  • Supporter
  • ****
  • Posts: 2647
  • Country: nl
  • self-educated hobbyist
Re: Problems with stability of power supply circuit
« Reply #4 on: August 05, 2019, 10:49:22 am »
I suggest add a compensation capacitor even if it seems to be stable, unless you can do a bode plot to verify that phase and gain margins are good-enough not to bother with compensation.
 

Offline David Hess

  • Super Contributor
  • ***
  • Posts: 17428
  • Country: us
  • DavidH
Re: Problems with stability of power supply circuit
« Reply #5 on: August 05, 2019, 12:33:59 pm »
I suggest add a compensation capacitor even if it seems to be stable, unless you can do a bode plot to verify that phase and gain margins are good-enough not to bother with compensation.

The LM358 is slow enough that external compensation should not be required unless something else is wrong.  With an electrolytic output capacitor, the only remaining problem is driving the high capacitance of the power MOSFET.

 

Offline mtimmermansTopic starter

  • Newbie
  • Posts: 8
  • Country: nl
Re: Problems with stability of power supply circuit
« Reply #6 on: August 05, 2019, 06:59:03 pm »
Thanks for all the suggestions! It seems like the oscillations were caused by a wrong simulator parameter. Now that I switched it to GEAR, the oscillations seem to be gone. To be sure, and because I am interested in learning how to do one, I would like to verify the stability with a bode plot. After googling around, the easiest option seems to be to add an enormous inductor in the feedback path to set the DC point, and inject the signal through a enormous capacitor. According to him (my source: ), you should set all independent voltage sources to zero. In my case I think this doesn't make sense because I want to linearize the circuit around my operating point, so I left my input voltage at 16V DC. Also according to the man in the video, it should be possible to add the signal source plus inductor anywhere in the loop. When I test this at two different nodes, I get different results. Attached to this post you can find the two circuits I used for testing. I applied the AC signal using V3 and V7, and measured at the voltage probes PR1 and PR2. Is this the correct way of doing it, or am I misunderstanding something?
 

Offline exe

  • Supporter
  • ****
  • Posts: 2647
  • Country: nl
  • self-educated hobbyist
Re: Problems with stability of power supply circuit
« Reply #7 on: August 05, 2019, 07:06:37 pm »
The LM358 is slow enough that external compensation should not be required unless something else is wrong.  With an electrolytic output capacitor, the only remaining problem is driving the high capacitance of the power MOSFET.

I'm not sure how if it is really slow enough. Datasheet from TI mentions this:

Quote
Capacitive loads which are applied directly to the output of the amplifier reduce the loop stability margin. Values
of 50 pF can be accommodated using the worst-case non-inverting unity gain connection. Large closed loop
gains or resistive isolation should be used if larger load capacitance must be driven by the amplifier.

I myself built this circuit and it oscillated at currents above 0.3A or so (pass element is irfz24n or irfz44n). I had to add compensation. It still caused oscillation when I connected it to my lt3080 power supply. So, I had to increase capacitor even further, to quite some big value. Not sure how big my gate resistor is.
 

Offline exe

  • Supporter
  • ****
  • Posts: 2647
  • Country: nl
  • self-educated hobbyist
Re: Problems with stability of power supply circuit
« Reply #8 on: August 05, 2019, 07:11:05 pm »
Is this the correct way of doing it, or am I misunderstanding something?

Didn't read your message as I have to go, but check this thread, it provides quite a comprehensive guide how to to this analyzis. Also provides complete schematic of an electronic load: https://www.eevblog.com/forum/projects/dynamic-electronic-load-project/
 

Offline KE5FX

  • Super Contributor
  • ***
  • Posts: 2096
  • Country: us
    • KE5FX.COM
Re: Problems with stability of power supply circuit
« Reply #9 on: August 05, 2019, 07:20:10 pm »
Note discussion of the R2C2 time constant in this thread.  As George H. suggests, this type of circuit tends to be happier with a nonzero R2.

The LM358 is slow enough that external compensation should not be required unless something else is wrong. 

Using a slow opamp is no guarantee of stability.  The opposite is more likely to be true IMO.
 

Offline David Hess

  • Super Contributor
  • ***
  • Posts: 17428
  • Country: us
  • DavidH
Re: Problems with stability of power supply circuit
« Reply #10 on: August 06, 2019, 06:26:30 am »
The LM358 is slow enough that external compensation should not be required unless something else is wrong.  With an electrolytic output capacitor, the only remaining problem is driving the high capacitance of the power MOSFET.

I'm not sure how if it is really slow enough. Datasheet from TI mentions this:

Quote
Capacitive loads which are applied directly to the output of the amplifier reduce the loop stability margin. Values
of 50 pF can be accommodated using the worst-case non-inverting unity gain connection. Large closed loop
gains or resistive isolation should be used if larger load capacitance must be driven by the amplifier.

I myself built this circuit and it oscillated at currents above 0.3A or so (pass element is irfz24n or irfz44n). I had to add compensation. It still caused oscillation when I connected it to my lt3080 power supply. So, I had to increase capacitor even further, to quite some big value. Not sure how big my gate resistor is.

That is why I mentioned the problem of driving a heavy capacitive load.  Either an external buffer to drive the gate of the power MOSFET is required or external compensation between the output and inverting input needs to be added.

The LM358 is slow enough that external compensation should not be required unless something else is wrong. 

Using a slow opamp is no guarantee of stability.  The opposite is more likely to be true IMO.

With a slower operational amplifier, each degree of phase is equal to a greater amount of time so a given amount of delay within the feedback loop is fewer degrees of phase lag.

Emitter or source follower power supplies using slow operational amplifiers like the 324/358, 741, 301A, or my favorite, the 308, almost never require extra compensation unless something is overlooked like driving the large capacitance of a power MOSFET directly which is the case here.  Note that higher gain relaxes the requirements further (attenuation within the feedback loop) and a gain of 1 (voltage follower) is the worse case.  Incidentally, these operational amplifiers are slow enough that the output transistor can be replaced with an integrated regulator although a drive transistor should be used to handle the regulator's quiescent current.

Faster parts like the OP27, LT1007, or TL071 are much more likely to require extra compensation and parts which are faster yet will always require extra compensation.  But they may be worth using despite this for lower noise.

Jim Williams discussed this at a practical level in appendix C of Linear Technology application note 47.
« Last Edit: August 06, 2019, 11:03:36 pm by David Hess »
 

Offline mtimmermansTopic starter

  • Newbie
  • Posts: 8
  • Country: nl
Re: Problems with stability of power supply circuit
« Reply #11 on: August 06, 2019, 11:11:53 am »
Is this the correct way of doing it, or am I misunderstanding something?

Didn't read your message as I have to go, but check this thread, it provides quite a comprehensive guide how to to this analyzis. Also provides complete schematic of an electronic load: https://www.eevblog.com/forum/projects/dynamic-electronic-load-project/

Thanks for the suggestion! However, also this method seems to depend on where you inject the disturbance. I added the disturbance after R3, and now the circuit looks like:


The however changed the loop gain to


Since the simulated loop gain changes when you add the disturbance at different points, how do you know the correct location to add the disturbance?
« Last Edit: August 06, 2019, 11:17:33 am by mtimmermans »
 

Offline exe

  • Supporter
  • ****
  • Posts: 2647
  • Country: nl
  • self-educated hobbyist
Re: Problems with stability of power supply circuit
« Reply #12 on: August 06, 2019, 11:25:27 am »
Since the simulated loop gain changes when you add the disturbance at different points, how do you know the correct location to add the disturbance?

You add disturbance to the input where feedback is taken (imho), and watch output changing. Disturbance should be small so it's "small signal analysis". So, I'd say before R3. There should also be a big inductor so you can measure open loop response. But I'm no expert.
 


Share me

Digg  Facebook  SlashDot  Delicious  Technorati  Twitter  Google  Yahoo
Smf