Current balancing using a PIC

[Guest]

Hi All,

I've been tasked with designing a circuit that maintains the current balance between 3 MOSFETs connected in parallel. The MOSFETs will be passing up to 200A DC, and because they are inherently not well matched, would not pass an equal share of the current without some sort of active balancing when drive by a common gate signal.

My plan is to use a hall effect current sensor (CSLA2CD) in each series with each of the MOSFETS to measure the current in each leg. I plan to independently control the current in each leg by using an op-amp buffer on each MOSFET gate with an adjustable offset voltage controlled by a digital potentiometer (AD8400). With this method I can slightly adjust each gate signal up or down as necessary relative to the original gate signal to maintain current balance.

The code would do something like this:

1. Measure current AMP1, AMP2, and AMP3
2. Calculate the abs. current difference in the 3 legs (AMP1-AMP2), (AMP2-AMP3), and (AMP1-AMP3)
3. Determine is any of the 3 differences exceed the "difference threshold" (maybe 2 or 3 amps)
4. If 'Yes', find the HIGHEST leg and knock it down a notch, find the LOWEST leg and knock it up a notch. Repeat at the top.
5. If 'No', repeat at the top.

It seems like this method would always adjust the 3 gate circuits until current balance was achieved in the 3 MOSFETs. I think it might work quite handily especially since the currents only need to be balanced within a few amps for this method to achieve the desired result.

Obviously, this is just a rough framework, but I think conceptually it is a sound plan. Does anyone have any particular comments on why this is, or is not, a good idea?

Thanks,

John

[treitmey]

200A is a LOT. What you making? an electric car?

Post your electrical idea here
http://www.edaboard.com/index.php?sid=cc3d54876664cfa3971be44e57880a57
Too have some experts look it over. and give suggestions.

[Jerson]

Hi John

This may be not directly supporting your idea of using a PIC. I will recommend you look at the Cypress PSOC and their app note for a multi cell charger. See AN2234 and AN2258.

http://www.psocdeveloper.com/docs/appnotes/an-mode/detail/an-pointer/an2344/an-sword/multi%20cell%20charger.html?tx_piapappnote_pi1%5BtypeS%5D%5Bnoteid%5D=Y&tx_piapappnote_pi1%5BtypeS%5D%5Bauthor%5D=Y&tx_piapappnote_pi1%5BtypeS%5D%5Btitle%5D=Y&tx_piapappnote_pi1%5BtypeS%5D%5Bdescription%5D=Y


Basically what the note states is to use voltage sensing of the individual cells and then use PWM to control mosfets across each cell to balance the charge on each cell.

This might be close to what you want to achieve.
_________________
Regards
Jerson Fernandes

[FvM]

I understand, that the MOSFET are connected in parallel directly, so individual PWM isn't an option. I further assume (it's not exactly said in the post), that the FETs are operated linear rather than as switches. It's the only configuration, where current balancing makes sense in my opinion.

But I wonder, why you don't use simple analog current control loops for the individual FETs. It's much simpler than using a processor.

[languer]

I think FvM is approaching this the right way. Why would you need current balancing (gate voltage adjust) if it is a switch?

Are you making a big-a** amplifier? HF, VHF, or what? - doesn't really matter.

As I read it, current is not defined; it could be up to 200A, but not necessarily 200A. This makes analog current control loops a bit more difficult. At the very least you would require to control the set point.

Your approach is probably ok. Would take great care on the sampling rate, update rate, and overall bandwidth of your loop. Since I am assuming these will be used on some sort of amplifier configuration for an AC signal, you want to make sure the current control loop does not influence the data you are passing through the amplifier. Or I maybe wayyyy off, and these are not used for amplification of an AC signal at all.

[Guest]

Hi Again,

It's not a switch, the MOSFETS will operate in linear mode, and the output will be continuously variable from 0 to 200A DC. This is the output stage of a parametric test system.

Hmmm, I'm not familiar with current control loops that incorporate three separate branches (the parallel MOSFETS). That would mean feedback from 3 separate sources (the current sensors), and three control variables (the MOSFET gates). I've done single loop current control before using analog circuits, but never multiple loops. Do you have a reference that I could look at?

Thanks,

John

[FvM]

Yes, the operation is as I expected from your post.

In my opinion, the design doesn't actually involve a multiple loop controller. You have three current sources operated in parallel, each is supplied with the same setpoint of 1/3 Iout,set. There are some implementation variants depending on how Iout,total is controlled, but this doesn't change much to the operation of the "inner" current controllers.