Author Topic: Current limiting stability problems  (Read 18949 times)

0 Members and 1 Guest are viewing this topic.

Offline RemondepemonTopic starter

  • Contributor
  • Posts: 16
  • Country: nl
Current limiting stability problems
« on: July 10, 2014, 08:19:57 pm »
Hi everybody,

I'm also working on power supply project.  ;D My voltage control works fine, but I have problems with my current control. I use the schematics of David's power supply  with an LT3083, but with an ASC712-5 current sensor. Everything works fine, 1V/A into the comparator and it actually limits the current. But when it limits the current, the voltage goes down and the current as well. Then the current starts to rise and so on. This results in an oscillating current control loop.

I have been searching and reading, but it seems that nobody experiences this problem. What am I doing wrong here?
Second question is the location of the current sensor. What are the pros and cons of placing the current measurement before or after the regulator or even in the negative line.

Hopefully somebody can help me out.

Thanks,
Remon
« Last Edit: July 11, 2014, 07:49:48 pm by Remondepemon »
 

Offline void_error

  • Frequent Contributor
  • **
  • Posts: 673
  • Country: ro
  • I can transistor...
Re: Current limiting stability problems
« Reply #1 on: July 10, 2014, 08:33:58 pm »
Everything works fine, 1V/A into the comparator and it actually limits the current. But when it limits the current, the voltage goes down and the current as well. Then the current starts to rise and so on. This results in an oscillating current control loop.

I have been searching and reading, but it seems that nobody experiences this problem. What am I doing wrong here?
Second question is the location of the current sensor. What are the pros and cons of placing the current measurement before or after the regulator or even in the negative line.

Hopefully somebody can help me out.

Thanks,
Remon

I actually had that oscillation happen in a simulation. One way of dealing with it is to reduce the high frequency gain of your current comparator (I suppose it's an opamp) by placing a capacitor in the negative feedback loop. Reducing the opamp gain to something less than open loop gain will also help improve stability.

The ACS712 is isolated, so it doesn't actually matter where you put it. It's series resistance is also tiny. Since it can measure currents of both polarities it's output voltage for 0 current flowing through it will be half the supply voltage. You could take a look at what I've done so far with my PSU design, maybe it helps.
Trust me, I'm NOT an engineer.
 

Offline dannyf

  • Super Contributor
  • ***
  • Posts: 8221
  • Country: 00
Re: Current limiting stability problems
« Reply #2 on: July 10, 2014, 08:34:31 pm »
Two issues:

one, for most of the designs, the constant current portion oscillates. That's by design thus unavoidable.

two, most designs get around your issues by using two diodes, and making the regulators :active low: when a faulty condition, either over voltage or over current, the circuits pulls down the output. The diodes act to OR the conditions.
================================
https://dannyelectronics.wordpress.com/
 

Offline RemondepemonTopic starter

  • Contributor
  • Posts: 16
  • Country: nl
Re: Current limiting stability problems
« Reply #3 on: July 10, 2014, 08:51:13 pm »
Thanks void, I will try this tomorrow as it is almost midnight now.  I will post my findings here of course.

Dannyf, I have seen the dual diode designs as well. This design requires a negative power supply to work. That is something I try to avoid.
« Last Edit: July 10, 2014, 08:54:10 pm by Remondepemon »
 

Offline void_error

  • Frequent Contributor
  • **
  • Posts: 673
  • Country: ro
  • I can transistor...
Re: Current limiting stability problems
« Reply #4 on: July 10, 2014, 09:27:04 pm »
Thanks void, I will try this tomorrow as it is almost midnight now.  I will post my findings here of course.

Dannyf, I have seen the dual diode designs as well. This design requires a negative power supply to work. That is something I try to avoid.

You're welcome :)

You won't need a negative rail unless you're using an LM317 or something similar. I might have found a way to avoid it https://www.eevblog.com/forum/projects/psu-design-help/?action=dlattach;attach=101510;image.

The outputs of the opamps have to swing close enough to 0V. Even a dirt cheap LM358 can do that at a low enough Isink.
Q1 & Q2 will start conducting when their added base voltages are above about 1.2V (more as Ic increases).
To put it another way, if the voltage on the base of Q2 is below 1.2V with respect the the output, none of the two will be conducting. The minimmum base voltage is V(D2) or V(D3) + V(out) of LM358. According to the datasheet, output voltage will be about 1V while sinking 1-10mA. It's pretty close but it might work (if not then I'll use Schottky diodes). I have yet to test it.
Trust me, I'm NOT an engineer.
 

Offline akis

  • Frequent Contributor
  • **
  • Posts: 981
  • Country: gb
Re: Current limiting stability problems
« Reply #5 on: July 10, 2014, 09:31:27 pm »
In all my designs (three so far) I do exactly that, 10nF capacitor on the op-amp that controls voltage or current since there is no feedback resistor.

Current limiting sense resistor behind the voltage regulator so that the output voltage is not offset by the voltage drop on the sense resistor. I have had good results with 0.05R (1%) or two 01R 1% in parallel, and an Av of about 10-15 at the diff op-amp stage. Unless you have sense wires on the load, in which case it does not matter where the sense resistor is.
 

Offline David Hess

  • Super Contributor
  • ***
  • Posts: 16545
  • Country: us
  • DavidH
Re: Current limiting stability problems
« Reply #6 on: July 11, 2014, 12:42:28 am »
Two issues:
one, for most of the designs, the constant current portion oscillates. That's by design thus unavoidable.

Most of which power supply designs?  Mikes?

Is there a specific reason the current limit would be deliberately designed to oscillate?
 

Offline David Hess

  • Super Contributor
  • ***
  • Posts: 16545
  • Country: us
  • DavidH
Re: Current limiting stability problems
« Reply #7 on: July 11, 2014, 01:00:17 am »
In all my designs (three so far) I do exactly that, 10nF capacitor on the op-amp that controls voltage or current since there is no feedback resistor.

Frequency compensation in power supplies is a common problem.  One disadvantage to using excessive frequency compensation at the error amplifier with separate voltage and current loops is the long recovery time as one or the other amplifier comes out of overload.

Quote
Current limiting sense resistor behind the voltage regulator so that the output voltage is not offset by the voltage drop on the sense resistor. I have had good results with 0.05R (1%) or two 01R 1% in parallel, and an Av of about 10-15 at the diff op-amp stage. Unless you have sense wires on the load, in which case it does not matter where the sense resistor is.

Some designs avoid using a differential amplifier by referencing the current limit to one side of the sense resistor.  This neatly avoids adding another active device to the current limit feedback loop.
 

Offline akis

  • Frequent Contributor
  • **
  • Posts: 981
  • Country: gb
Re: Current limiting stability problems
« Reply #8 on: July 11, 2014, 08:40:04 am »
I presume I could have the sense resistor on the ground (negative) terminal, and therefore measure the voltage drop directly on it without a need for the differential amp, I would only need a comparator amp. I am not sure of the implications that the voltage drop might cause eg putting two supplies in series to make a symmetrical one. Maybe it's a good idea.
 

Offline RemondepemonTopic starter

  • Contributor
  • Posts: 16
  • Country: nl
Re: Current limiting stability problems
« Reply #9 on: July 11, 2014, 09:51:13 am »
Thanks for all the replies, they have been very helpful.

This morning I took the time to draw a schematic diagram of the power supply. Let me comment on it.
At this time, the Voltage and Current are manual controlled. At the end, there will be a DAC that controls this.

First of all, there is a switching regulator that tracks the second linear regulator, the LT3083. This regulator is fed by a voltage from a amplifier (opamp). There is also a transistor that can cut off the voltage when the maximum current is reached. I have been experimenting with C's over de feedback loop and found out that 1 uF was sufficient to deal with the oscillation. I decided to measure the current after my linear regulator since the current there is more stable. My ACS712-5 delivers 2.5V at 0A, so I decided to use an opamp to substract 2.5V and multiply the voltage by 5.4. This gives me 1V/A on the output. Unfortunately I cannot go all the way down to 0V @ 0A, so I decided to set it to 1V @ 0A. This gives me enough resolution for my 10bit ADC later on.

All comments on this design are welcome!

Thanks,
Remon

 

Offline David Hess

  • Super Contributor
  • ***
  • Posts: 16545
  • Country: us
  • DavidH
Re: Current limiting stability problems
« Reply #10 on: July 11, 2014, 10:15:20 am »
I presume I could have the sense resistor on the ground (negative) terminal, and therefore measure the voltage drop directly on it without a need for the differential amp, I would only need a comparator amp. I am not sure of the implications that the voltage drop might cause eg putting two supplies in series to make a symmetrical one. Maybe it's a good idea.

It is a great idea and I have done it in the past without problems except maybe for the issue you mention.  The old Tektronix PS501 has the current sense resistor in series with the output after the pass transistor instead of the ground which makes the tracking implementation in the PS503 straightforward.  It might be fun to redraw the PS503 schematic and see what changes are necessary to move the current sense resistor to the ground side.

These power supplies avoid using a difference amplifier in the current sense loop in the way I mentioned.

I have really grown to like the PS503 design for development work.  Having a variable tracking supply with a floating ground is incredibly useful.
 

Offline blackdog

  • Frequent Contributor
  • **
  • Posts: 736
  • Country: nl
  • Please stop pushing bullshit...
Re: Current limiting stability problems
« Reply #11 on: July 11, 2014, 10:38:45 am »
Hi Remon,

Uhm... You wil NOT get this stable!
Look @ your schematic, how many poles thit you create?
1e 10K+100nF opamp A=5.41, 2ecurent set opamp 1uF+10K, 3e BC547+100-Ohm+22uF.

Your disign is NOT nice to its load...
Think about this, set the current to 25mA and your Voltage is 12V.
You want to test a LED, connect the LED now to your powersupply and its gone  ;D

Wat happend? 1e the charge from de 22uF output capacitor is dumpt in your LED and afther that your LT3083 is pomping 10A in your LED!!!
It wil take about 0.1 a.2 Sec to get stable, bey, bey LED. PIC Arduino enz.

It wil get even wors if you use the recomended output capacitor on de LT1083, ~150uF Low ESR.
Sorry, there is in my opinion no solution for kindness to the load and stability in your design.

If you wanth a good design use the HP/Harison desing from de late 50  :D
A lot of powersuply builders use this design, old but stil very capable.

Take a look @ this topic: http://www.circuitsonline.net/forum/view/110029/1
It is in Dutch, but i think you can read it...


Kind regarts,
Blackdog




Necessity is not an established fact, but an interpretation.
 

Offline akis

  • Frequent Contributor
  • **
  • Posts: 981
  • Country: gb
Re: Current limiting stability problems
« Reply #12 on: July 11, 2014, 10:41:59 am »
I agree about the floating supply. All my 4 bench supplies are floating, that allows me to power anything and then attach the Ground crocodile clip of the oscilloscope anywhere I want, pretty much.

However I am not sure why having one more stage (a differential op-amp  that references the Vsense to ground) in the current limiting circuitry is considered a problem. I believe we are in a thread where people are discussing placing two LM317s in series...  Maybe it introduces some extra latency in the limiting, maybe, I have not checked it on the scope. That would come into play if the DUT is very sensitive to ns spikes, like for example I was reading about how to power/drive laser diodes, where even a 1 ns spike will kill them.

 

Offline akis

  • Frequent Contributor
  • **
  • Posts: 981
  • Country: gb
Re: Current limiting stability problems
« Reply #13 on: July 11, 2014, 10:52:58 am »
Uhm... You wil NOT get this stable!
Blackdog

I agree about the possible latencies and in addition the circuit looks way too complicated and for what?

In my opinion what you need in a power supply is:
1) Switcher to track the output voltage so as not to waste heat on the pass transistors - you already have that, albeit in a very complicated way
1A) get rid of switcher and use large heatsinks and fans, that creates a very clean output, but be prepared to waste 150W-200W as heat (30V-35V/5A input, 0.1V/5A output).

2) Current limiting circuit
3) Voltage control circuit
4) Voltage sense on load circuit with two extra leads, and compensation when load is not connected or grounded
5) Relay to disconnect load on push button (imagine you have X number of power supplies in the same box, and you wish to switch them on/off individually and imagine you want to set the voltage/current BEFORE taking the load live...

6) Relay or circuit to discharge the transformer's filter's large caps when power switch is off.

In my opinion functionality comes first, and things like the relay to take the load offline instantly or remote voltage sense are extremely important.
 

Offline akis

  • Frequent Contributor
  • **
  • Posts: 981
  • Country: gb
Re: Current limiting stability problems
« Reply #14 on: July 11, 2014, 11:13:34 am »


Not the LED test for the current limit again. The current limit is their to protect the power supply not the load, if the load is drawing more then anticipated chances are it’s already fried.

The ADJUSTABLE current limit is precisely there to protect the load, down to a few mA. On my supplies I go down to 0mA, ie the current limit starts from negative values.

The problem with using the LM317 type regulators is that you cannot go to 0 unless you have a negative rail but why bother with those regulators and extra complexity in the first place, an op-amp driving a FET or transistor and a good voltage reference does the job very well.
 

Offline blackdog

  • Frequent Contributor
  • **
  • Posts: 736
  • Country: nl
  • Please stop pushing bullshit...
Re: Current limiting stability problems
« Reply #15 on: July 11, 2014, 01:48:38 pm »
Hi AcHmed99,

Hi AcHmed99, try to see what I'm trying to explain ...
I do NOT test LED's this way on a power supply as I suggested.
I try to let you understand, what kind of energy there are present, in a power supply, and wat the reasen can be, that your load is killed...

I "kind" LAB powersupply needs a fast/stable CC/VC circuit!
If you want energie for short pulses, than you need good decoupling in your circuit, not a slow acting powersupply.
My opinion is, that a LAB power suply is not used, to start cars, bus sensitiv electronics :-)

I use different power supply's for different jobs.

Kind regarts
Blackdog


Necessity is not an established fact, but an interpretation.
 

Offline akis

  • Frequent Contributor
  • **
  • Posts: 981
  • Country: gb
Re: Current limiting stability problems
« Reply #16 on: July 11, 2014, 02:11:54 pm »


Not the LED test for the current limit again. The current limit is their to protect the power supply not the load, if the load is drawing more then anticipated chances are it’s already fried.

The ADJUSTABLE current limit is precisely there to protect the load, down to a few mA. On my supplies I go down to 0mA, ie the current limit starts from negative values.


Why does a load draw more current then anticipated? Either inrush or something is shorted. If your current limit is hit before your voltage setpoint your current will likely have minimal overshoot  if thats what your talking about. Thats because your current limiter has control of the loop early. That assumes the Voltage loop hasn't already pumped up the the voltage to max out.

If the current loop controls the loop before the voltage loop then the current loop dominates and you have a current source, with a voltage limit.
Try your test with your supply outputting max voltage (already powered) and current limit set to 20mA then hook it up to your LED let me know what happens to your LED?

I don't think I would want a PSU that kicks in CL fast enough to save a 20mA LED, that would be similiar to what I'm experienceing with the Analog Discovery. It's a bloody nussiance, you want a supply to be able to deliver short burst of peak current. If I want to test LED's I would have a dedicated CC source output on the regulator.

You are confusing two separate issues. One is the idea of an adjustable and very low current limit to protect any kind of device you are currently testing. It is not as black and white as you have mentioned either. For example you could have a badly biased transistor running away thermally over 15 seconds, at 1A it is toast, your PSU is 5A, what do you suggest, we wait for the full 5A to kick in before the red lights come on? The answer is we adjust the current limit to the smallest we can get away with, especially when the DUT is new and we have low confidence.

Secondly there is the idea of quality and speed - how quickly does the current limit react, how much does it overshoot, or ring, what sort of energies are stored in the output capacitors that the current limit cannot control and so on. This is an entirely different matter. It is perfectly reasonable to have a current limit that reacts within, say, 100ns, and during that time it may allow the current to overshoot by X%, and this may be perfectly acceptable, or completely disastrous for some sensitive device, eg a laser LED.

But it does not mean to say that the Bench PSU should not have an adjustable current limit. It just means that your PSU is not appropriate for the type of load.
 

Offline blackdog

  • Frequent Contributor
  • **
  • Posts: 736
  • Country: nl
  • Please stop pushing bullshit...
Re: Current limiting stability problems
« Reply #17 on: July 11, 2014, 03:18:44 pm »
Hi AcHmed99,

I get the inpression that you'r talking about current foldback wat you don't like.
That kind of current control is not my favorite, the same for SCR control, thats "Switch Off" above a limit.

The power supply's i design, are not of the "Fold Back" limiting type, but CC.
And for the electronic's i design this CC is a must.

Kind regarts,
Blackdog
Necessity is not an established fact, but an interpretation.
 

Offline RemondepemonTopic starter

  • Contributor
  • Posts: 16
  • Country: nl
Re: Current limiting stability problems
« Reply #18 on: July 11, 2014, 03:43:28 pm »
Is has been busy here since I posted my 'unworkable' design, but I leared stuff and hopefully others do too. Many thanks to everybody that want to help guys like me!

I decided to borrow the current limiting implementation of Void from the "bench power supply design" blog.
Also, I realised that my original design did not have a feedback loop for the output voltage. Now it has.

I have changed the schematics and I have to say that this works a lot better. :D As Blackdog was discussing, I had a very large overshoot in current and this is not acceptable. This design works fine with the LT3083 as well since there is a current source inside. I did power the opamp only from +25V and GND, no negative rail. Now, my power supply starts around 0.9V. With a negative power rail this might go to zero. Also, there is quite some noise from the current sensor that is superimposed on the output voltage.

Here is the schematics v2, so please shoot again!

 

Offline David Hess

  • Super Contributor
  • ***
  • Posts: 16545
  • Country: us
  • DavidH
Re: Current limiting stability problems
« Reply #19 on: July 11, 2014, 04:43:28 pm »
Here is the schematics v2, so please shoot again!

I guess the output does not need any bulk decoupling because the current sense element is essentially zero ohms but this points to a problem.  The 22 microfarad bulk decoupling capacitor at the output of the LT3083 is going to screw up the frequency compensation of the current limit feedback loop.  The current sense should be between the LT3083 and the bulk decoupling capacitor.

I would tune the constant current feedback loop by adding zeros so that value of the feedback capacitor can be lowered significantly speeding up the response time.  That would entail at least a resistance in series with the feedback capacitor and another series RC in parallel with the 4.7k input resistor.

 

Offline David Hess

  • Super Contributor
  • ***
  • Posts: 16545
  • Country: us
  • DavidH
Re: Current limiting stability problems
« Reply #20 on: July 11, 2014, 05:06:22 pm »
I agree about the floating supply. All my 4 bench supplies are floating, that allows me to power anything and then attach the Ground crocodile clip of the oscilloscope anywhere I want, pretty much.

I like them because they can be used to implement a bipolar power supply at any low impedance node in a circuit.  This is handy for powering auxiliary test circuits during development without worrying about single supply operation.

Quote
However I am not sure why having one more stage (a differential op-amp  that references the Vsense to ground) in the current limiting circuitry is considered a problem. I believe we are in a thread where people are discussing placing two LM317s in series...  Maybe it introduces some extra latency in the limiting, maybe, I have not checked it on the scope. That would come into play if the DUT is very sensitive to ns spikes, like for example I was reading about how to power/drive laser diodes, where even a 1 ns spike will kill them.

It just adds another delay into the current limit feedback loop resulting in longer recovery and slower response because of the increased frequency compensation needed for stable operation.  They make fast difference amplifiers but they are not common and you still have to account for common mode rejection.

The single ended designs which I mentioned are much faster and are arguably simpler as well although probably not as intuitive without more analog design experience.  The first bench power supplies I designed for myself used difference amplifiers in the current limit feedback loops but I would not do that now.

I disagree with AcHmed99 on this.  A fast current loop has many advantages and few or no disadvantages.  It does not create startup problems except when driving a negative resistance load and even that is an advantage because it reveals something the designer may want to consider carefully.

As far as the LED test, in many cases even the minimum of bulk output decoupling will damage the LED before any current loop responds anyway.
 

Offline David Hess

  • Super Contributor
  • ***
  • Posts: 16545
  • Country: us
  • DavidH
Re: Current limiting stability problems
« Reply #21 on: July 11, 2014, 05:12:53 pm »
Take a look @ this topic: http://www.circuitsonline.net/forum/view/110029/1
It is in Dutch, but i think you can read it...
I notice that the second schematic shows type of single ended current loop without a difference amplifier that I mention.  It uses a separate floating bias supply but that is not always a requirement.
 

Offline blackdog

  • Frequent Contributor
  • **
  • Posts: 736
  • Country: nl
  • Please stop pushing bullshit...
Re: Current limiting stability problems
« Reply #22 on: July 11, 2014, 06:13:27 pm »
Hi David Hess,

Not the last version but good workable is this one...
http://www.bramcam.nl/NA/NA-01-PSU/NA-PSU-SCH-17.gif
Forget the Fet on the output, and for the transformer i have a different plan.

No power On or Off Glitch.
Fast acting U en I control.
Low output resistence, DC not measureble (depents on were to connect the sense wires)
Max output current limited outside the "i" control loop.

This is a "real" picture of the load stability, Wire = 2x0.5M to my Active load, over de active a 6,8uF Low ERS capacitor.
That why you see a little ringing, short cable's and no capacitor, no ringing!
http://www.bramcam.nl/NA/NA-01-PSU/NA-PSU-30.png

Its a big topic on that forum, a lot to read, i explane a lot why and how.
Happy reading :-)

Kind regarts,
blackdog
Necessity is not an established fact, but an interpretation.
 

Offline David Hess

  • Super Contributor
  • ***
  • Posts: 16545
  • Country: us
  • DavidH
Re: Current limiting stability problems
« Reply #23 on: July 11, 2014, 06:57:20 pm »
No power On or Off Glitch.
This is important but often overlooked until it is too late and something expensive is in ruins.  DSOs are especially handy for finding output glitches.

Another consideration is that shutting the power supply off or losing AC power should not damage the supply itself when a large output capacitor is used.  At most, it should blow a fuse.

Quote
Fast acting U en I control.
It looks like you clamped the integrators so they never saturate to prevent excessive overload recovery time.  My first adjustable current bench supply design suffered from this making the current limit very slow to respond.

Quote
Low output resistence, DC not measureble (depents on were to connect the sense wires)
This is not normally important in a bench power supply but I have had the same results.  0 to 30 amps with an output voltage change of 10 microvolts or less makes me wonder if my voltmeter is broken.

Quote
Max output current limited outside the "i" control loop.
One reason I like to use integrated regulators as pass elements is their built in fast current limit , safe operating area protection, and thermal shutdown.

Quote
This is a "real" picture of the load stability, Wire = 2x0.5M to my Active load, over de active a 6,8uF Low ERS capacitor.
That why you see a little ringing, short cable's and no capacitor, no ringing!
http://www.bramcam.nl/NA/NA-01-PSU/NA-PSU-30.png
It is difficult to argue against the results.

Quote
Its a big topic on that forum, a lot to read, i explane a lot why and how.
Happy reading :-)
Your English is better than my Dutch so I will have to be satisfied with Google Translate and the schematic which is easy to understand.  Thanks for the link.
 

Offline RemondepemonTopic starter

  • Contributor
  • Posts: 16
  • Country: nl
Re: Current limiting stability problems
« Reply #24 on: July 11, 2014, 07:47:53 pm »
Hi everybody,

Thanks to David that pointed out my output cap messing up the current loop. I moved the current measurement before the cap and lowered it from 22 to 10uF. Also, added the -5V ICL7660 to be able to regulate all the way down to 0V. I have to say that everything feels better since I added the -5V to the opamps.

There is still some noise on the output, especially when in current limiting. What can I expect here?
I measured my factory regulated power supply and that also gives quite a lot of noise when in current limiting state.

Here is the updated schematic v3.

Have a good night,
Remondepemon
 


Share me

Digg  Facebook  SlashDot  Delicious  Technorati  Twitter  Google  Yahoo
Smf