Hi Sean,

On Thu, Apr 18, 2013 at 3:27 AM, Sean Breheny <shb7@cornell.edu> wrote:
> Hi Manu,
>
> I am not familiar with the PWM peripheral on the pic you are using but
> usually cycle-by-cycle current limiting means very simply this: constantl=
y
> checking to see if the current is outside of some window (say, for exampl=
e,
> +/- 10 Amps) and when it goes outside this window, terminate the remainde=
r
> of the present PWM cycle (i.e., turn off all FETs and let the current go
> through the body diodes back into the bus) and then recheck the current a=
t
> the beginning of the next PWM cycle. If it is now inside the window,
> re-enable PWM as normal. If not, keep the FETs off for another whole PWM
> cycle and check the current again at the beginning of the next PWM cycle,
> etc.
>
> "cycle-by-cycle" meaning that the exit from current limit is synchronous
> with the beginning of a new cycle, to avoid increasing switching losses b=
y
> allowing the system to introduce additional on/off edges beyond what woul=
d
> happen in normal PWM.
>
> I don't think this is what you want to prevent flux walking because, if y=
ou
> used such a scheme, and the current tended to increase in one particular
> direction, then it would just keep hitting current limit in that directio=
n.
> You want some scheme which prevents the average flux from growing beyond
> some tiny value in the first place, and this value would be much less tha=
n
> the peak flux achieved during each ON cycle.
>
> I've worked this out before in an SMPS design and I think, if I remember
> correctly, that I got around it by simply increasing the PWM dead-time
> (time from turning off one set of FETs, or in your case, one FET, to
> turning on the next FETs) a little bit so that the current would always
> decay to zero within each half PWM cycle. I'll have to look back at my
> notes again.

This was what I was thinking. I did simply play with the Dead time a bit
here and there. It did help to some extend, but it still didn't fix the pro=
blem
as expected.

In short, there are 2 problems that were occurring, quite inversely:

#1, Flux walking problem: this reduced a bit by increasing dead time
#2, As pulse width reduced, current increased per pulse, causing
switching loss again.

It looked like a bit too hard to reach the in-between stage empirically.

It would be great, if you could have a look at your notes again,

Thanks,

Manu
--=20
http://www.piclist.com/techref/piclist PIC/SX FAQ & list archive
View/change your membership options at
http://mailman.mit.edu/mailman/listinfo/piclist
.