Current regulation for buck converters?

[steph44]

Hi everyone,

First post on this forum (!)

I am simulating a distributed power generation system for which I need a current regulation on a buck converter (for load sharing between sources)...The model will be turned into a hardware solution so I need a practical solution, preferably analogic.

I have only been able to find ressources (articles, books) about voltage regulation of converters.

Could I simply turn a current-mode controller (for voltage regulation) into a current controller, bringing the compensator into the current loop, and getting rid of the voltage control loop ?

Anyone has some experience with this?
Is there a good ressource on current regulation ?

Thanks!

 

[berkeman]

Hi everyone,

First post on this forum (!)

I am simulating a distributed power generation system for which I need a current regulation on a buck converter (for load sharing between sources)...The model will be turned into a hardware solution so I need a practical solution, preferably analogic.

I have only been able to find ressources (articles, books) about voltage regulation of converters.

Could I simply turn a current-mode controller (for voltage regulation) into a current controller, bringing the compensator into the current loop, and getting rid of the voltage control loop ?

Anyone has some experience with this?
Is there a good ressource on current regulation ?

Thanks!

I have done buck converters with current regulation for LED lighting before. You use a low-side sense resistor to provide feedback to the buck control circuit (so it's still voltage feedback off of the low-side sense resistor). That might work for you if you can stand low-side sensing resistors in your application.

However, the "shared power sources" aspect of your post makes it more difficult. Can you say some more about the application, and what constraints and specifications you are shooting for? Can the multiple sources communicate with each other in some way other than just sensing the shared load voltage and their own current contribution? If they can communicate separately from the actual power circuit, that would make the overall system design much easier and more reliable, IMO.

 

[cabraham]

If you're driving LEDs, which is likely the case, then I recommend a dedicated IC for LED current drive. You can configure a dc-dc converter for constant current output instead of constant voltage, as berkeman pointed out, but there are reasons to use a dedicated LED driver.

For a buck converter LED driver, the power MOSFET can be placed in the low side. This beneficial twofold. The low side FET will be n-channel which is a better material since n-type Si has lower resistance than p-type. Also, being low side means that no additional circuitry needed for boot-strapping, level shifting, and/or charge pumping. The current sense resistor is located high side. The driver IC likely has a pair of terminals for this function.

A dc-dc voltage regulation IC in the buck topology uses a high side FET. This is disadvantageous because you must either use a p-channel FET, or a bootstrapped n-channel FET. P-FET is higher resistance than N-FET. If N-FET is used, extra components are needed for bootstrapping.

I'd recommend dedicated constant current IC parts from Diodes Inc., TI, Natl Semi (part of TI), Supertex, etc. These OEM also offer app notes to help you utilize these drivers. Again, a common dc-dc SMPS IC, like a 3842, or a "Simple Switcher" can be configured for constant current operation. Just be sure to read the app notes on how to do this.

Claude

 

[es1]

Pretty much what Cabraham mentioned. You can also use the article below as it includes schematics but imagine your load is where the LED is. The buck's inductor is basically a current source driving into your load.

Note: The reference for your load is the top of the current sense resistor. This might make a difference if the load needs to signal to something else on the buck's ground.

http://www.eetimes.com/design/power...atter-of-light-Part-2--Buck-whenever-possible

 

[steph44]

Thank you all for your answers. Unfortunately, my application is quite different.

It is a small mobile hybrid power system (fuel cell + battery/supercapacitors).The fuel cell will be interfaced with a DC bus through the converter. The other sources, connected to the bus, are present to improve the transient response of the whole power system and provide peak power. The current regulation would noteably limit the rate of change of fuel cell current (hereby preventing damage).

Hence, I was aiming at following a documented control strategy where I would sense load current, filtering it adequately, and combining it with a battery current reference (charge or discharge) to make a suitable current reference to be tracked by the fuel cell/converter assembly. It is therefore NOT constant current control, rather a current reference tracking.

@Cabraham: Merci Claude for the detailed info but I wld need a "current tracker" controller rather than constant current. Any commercial hardware comes to mind?


@Berkeman : So I was not thinking of making them communicate with each other directly, only in the sense that the auxiliary sources' state of charge and the power circuit demand make up the fuel cell current reference.
What were you thinking about when you said that it wld make the design easier if they could communicate?

@es1: Nice link



I am still at the simulation stage, therefore the control strategy/system topology isn't fully fixed yet. As I am still quite inexperienced with power electronics, any thoughts or perspectives on this are welcome.

For the controller design, I suppose I can slightly modify the standard control-to-output transfer function for current (or voltage) mode control of output voltage, so as to target the sensing resistor voltage, rather than output voltage. Hence I can track the output current indirectly.


Ultimately, for info, the load will be interfaced with the DC bus through an inverter.

Can I just use a sensing resistor on the inductor branch of the buck ?
Nonetheless, wouldn't a Hall effect sensor be more apropriate here ?


Here attached is one of the journal articles that I am using for the control strategy (see fig. 7). My problem lies in the design of the "current controller", which is said to be in "average current mode control". But to the contrary of standard control methods, there is no outter voltage loop.

I am kind of worried about stability issues, as it seems they are numerous with power converters, especially if an inverter is added between the load and sources (source load interaction/"input filter problem" ?)

I am surprised I can't find detailed info on current control for this type of application...

(Sorry for the long reply)

 

[jim hardy]

i'd be surprised if this hasn't been adddressed in appnotes from IC manufacturers.
I for some years maintained a reference tracking current control system built in 1960's from discrete transistors and phase control SCR's...

is this appnote the right flavor? Perhaps it'll have some good terms to serach on...

http://www.ti.com/lit/an/slua079/slua079.pdf

 

[steph44]

Yes, it is the right flavor. But I had already downloaded it !

It is about "average current mode control" (instead of "peak"), but still, the ultimate goal is to regulate voltage.
Going through an inner current loop is beneficial to the controller design (pole/zero structure of TF,stability issues etc...).

I had found some other interesting appnotes of this kind ("sluaXXX.pdf" and smthg like "slupXXX.pdf") from TI / Unitrode (they seem to provide the same appnotes...are they part of the same company?)...

But I could not find any appnote regarding actual current regulation. Any other source for manufacturer "appnotes"?

I suppose I can find my 'inspiration' in such references for "average current mode control", and doing without the outter voltage control loop. But Id rather work on a proven and documented technique.

But I am as surprised as you are for not being able to find what I want. It does not seem to be a peculiar application !

So, Jim Hardy, was the system you worked on doing voltage regulation through current mode control, or was it actually tracking a current reference..?

 

[jim hardy]

But I could not find any appnote regarding actual current regulation. Any other source for manufacturer "appnotes"?

hmmm i'll think for a day. It's been literally a couple decades since i looked. I know there'll be some in motor controls... TI has a knowledge base that's overwhelming, accessible from their site. I used to have a book "DC Motor and Servo Controls"

Yes TI bought Unitrode some time ago.

...

was the system you worked on doing voltage regulation through current mode control, or was it actually tracking a current reference..?

It was a dirt simple system designed in late 1960's. All discrete semiconductors, a mix of analog and logic circuits..
It tracked a voltage proportional to desired DC current.
Measured current was compared to desired and a DC error voltage developed.
It used (edit) SCR phase control, simple half wave (edit) rectified three phase to control DC current through electromagnets which in turn moved the rods in a nuclear plant.
It was i believe adapted from motor controllers used in steel mills.
As i said above it accepted a voltage proportional to desired current, compared measured current to desired current resulting in error voltage.
Error voltage advanced or retarded firing angle in proportion to error, to push current in desired direction - higher or lower .
Since load was electromagnets their inductance made current response well behaved. You get a natural integration with inductance.

What at first seemed counterintuitive about the system is what happens when zero current is called for.
Initially , while current is still pesent, the SCR's get fired during the negative half cycle to force current lower rather than wait for natural L/R time constant.
Upon reaching zero current the firing angle shifts to center around zero crossing. This provides equal positive and negative part cycles. That makes applied voltage average zero.
It is at first hard to accept that scheme could work, for how does negative voltage pass backward through the SCR to get from source to load ?
Once one accepts that the negative voltage, by Lenz's Law comes from the load and not from the source, which is the nature of inductance, it seems an epiphany and that's when we came to appreciate the system's simple elegance.

My system controlled current through 135 electromagnets, three each for forty five control rods.
If you are curious about the system go to USPTO dot gov and look up US patent 3,796,890. It's pretty much out of the GE SCR manuals of the time. You'll need a TIFF reader to see the drawings and USPTO has links to some. I use Alterna's free one which seems to work well with IE .

old jim

 

[jim hardy]

you probably have this one, too.

http://www.fairchildsemi.com/onlineseminars/regularTopologies_implementations.pdf

 

[steph44]

@jim_hardy : I took a look at your system. Quite interesting indeed, even though I am probably not experienced enough to grasp its elegance..! Thank you for the link (no i did not have it actually), some schematics could be useful to me.

While reading this pdf, as well as a journal article, made me wonder :

Would a SEPIC converter, instead of a buck, be more suitable to my application ? Since the converter would be directly connected to a fuel cell, couldn't a SEPIC limit the current ripple from the fuel cell, hence improving its durability while not requiring any sort of input filter..? (maybe i shld start a new thread for this)

Also, I think I will use some kind of sensing resistor, as berkeman suggested. What would be the reasons for not "standing low-side sensing resistor" ? (see berkeman's post)

 

[cabraham]

Thank you all for your answers. Unfortunately, my application is quite different.

It is a small mobile hybrid power system (fuel cell + battery/supercapacitors).The fuel cell will be interfaced with a DC bus through the converter. The other sources, connected to the bus, are present to improve the transient response of the whole power system and provide peak power. The current regulation would noteably limit the rate of change of fuel cell current (hereby preventing damage).

Hence, I was aiming at following a documented control strategy where I would sense load current, filtering it adequately, and combining it with a battery current reference (charge or discharge) to make a suitable current reference to be tracked by the fuel cell/converter assembly. It is therefore NOT constant current control, rather a current reference tracking.

@Cabraham: Merci Claude for the detailed info but I wld need a "current tracker" controller rather than constant current. Any commercial hardware comes to mind?


@Berkeman : So I was not thinking of making them communicate with each other directly, only in the sense that the auxiliary sources' state of charge and the power circuit demand make up the fuel cell current reference.
What were you thinking about when you said that it wld make the design easier if they could communicate?

@es1: Nice link



I am still at the simulation stage, therefore the control strategy/system topology isn't fully fixed yet. As I am still quite inexperienced with power electronics, any thoughts or perspectives on this are welcome.

For the controller design, I suppose I can slightly modify the standard control-to-output transfer function for current (or voltage) mode control of output voltage, so as to target the sensing resistor voltage, rather than output voltage. Hence I can track the output current indirectly.


Ultimately, for info, the load will be interfaced with the DC bus through an inverter.

Can I just use a sensing resistor on the inductor branch of the buck ?
Nonetheless, wouldn't a Hall effect sensor be more apropriate here ?


Here attached is one of the journal articles that I am using for the control strategy (see fig. 7). My problem lies in the design of the "current controller", which is said to be in "average current mode control". But to the contrary of standard control methods, there is no outter voltage loop.

I am kind of worried about stability issues, as it seems they are numerous with power converters, especially if an inverter is added between the load and sources (source load interaction/"input filter problem" ?)

I am surprised I can't find detailed info on current control for this type of application...

(Sorry for the long reply)

But what would "current control" buy you? I'm just wondering why an inner control loop for current is even needed. With an SMPS, which is what many commercial ICs are designed to control, the output being regulated is a fixed voltage. As load impedance changes, current varies, but the servo loop forces Vout to remain fixed. An inner control loop can sense the change in load current & adjust so that the output voltage remains fixed w/o relying entirely on the outer voltage control loop, resulting in better dynamics, among other things.

But with LED drivers, we are not outputting a fixed voltage, but rather a fixed current. More importantly, our "load" does not have to be ground-referenced. The LED string can be placed in series with the choke, on either high or low side. The current sense resistor translates the load current into a voltage. This voltage is actually the output state variable being controlled by the servo loop. But this voltage is just the load current times the resistor value.

So the output state variable is the load current. Unlike an SMPS, we allow the LED string total voltage drop to vary. As temperature increases, an LED forward voltage drop decreases at the same value of forward current. My question is "what is gained by adding an inner current control loop to an LED driver?" The PWM IC's outer ("voltage") control loop is controlling the load current by controlling the voltage drop across a low values sensing resistor. So the "current control" is already achieved by the outer loop. This outer loop is of course regulating a voltage, but it is the voltage across a fixed precise resistor value, so that controlling this voltage is equivalent to controlling the load current.

Maybe that is why the commercial offerings of LED driver IC controllers are the 1-loop variety instead of 2 loops. The single loop does the job of controlling current very effectively. If anyone here knows something I've missed, please share it. That is my answer. I've been driving LEDs since around 1991. I currently develop products using high power leading edge LED lamps & drivers. I use a 1-loop control method. Did I help?

Claude

 

[jim hardy]

Would a SEPIC converter, instead of a buck, be more suitable to my application ? Since the converter would be directly connected to a fuel cell, couldn't a SEPIC limit the current ripple from the fuel cell, hence improving its durability while not requiring any sort of input filter..? (maybe i shld start a new thread for this)

Well you have exceeded my knowledge base.
But it does appear the SEPIC draws current from source during both off and on cycles so your observation looks sound to me.
In a vehicle aren't you dealing with extreme currents? Might C2 in SEPIC become unwieldy? I'd think a commutaion duty capacitor might be up to it, though.

As to low vs high side sensing
the only reason i know of to prefer one over other is convenience.
It's sure nice to have one side of your load connected to circuit common, especially in a car, which dictates high side sense.
On the other hand it's sure nice to have your current measurement ground-referenced for input to contol circuity, which dictates low side sense.
Note my system was low side . It almost predated the integrated circuit.
Nowadays they use an opamp to change reference from high to low side
and there's plenty of high side sense IC's made for just that task. Some even have a large shunt resistor molded in. I stumbled across them about ten years ago, probably it was TI or IRF made them... my memory is becoming 'write only'.

I hope this helps.
I am sensitive about "talking through my hat" , and I'm almost there now.

So i'll retire to the sidelines.

Good luck with your project ! It sounds really interesting and cutting edge.

 

[steph44]

@Claude :

My application is not about driving LED as you may have seen here above. But anyway, my goal is to track a current reference. That's what I meant by "current control".

In most applications, as you mentionned, the output voltage is to be kept constant. This is achievable through "voltage mode control" (1 loop), or "current mode control" (2 loops), where the inner loop modifies the pole/zero structure of the transfer function to be compensated in a favorable manner.

I am not aiming at using 2 loops, just one.

Indeed, as you explain, I will probably use a sensing resistor and doing "voltage mode control" (see attached image), except that my voltage reference (indeed equivalent to current reference, since sensing resistor is known) is not constant. Neither is my load, nor my voltage input (fuel cell).

Do you have a good ressource (scientific journal article preferably) for the control loop of a LED driver ?

 

[steph44]

@jim_hardy :

Well thank you for your answers, it does help. Note that in my case, this not a vehicle so currents will be moderate. But it is a very cool project indeed !
I still haven't found the references for what I want, but some of your words have put me on interesting tracks, so it was helpful.

And not talking through your hat on a forum honors you ;-)