microsoftwarecontrol wrote:
> there should be  much faster math solution, with enough accuracy, for
> this equation

I just realized I copied the equation incorrectly in the previous post.  The
part inside the square root has to be divided by Vo, not multiplied by it.
The correct equation is:

  T1 = SQRT(K * (Vo - Vin) / Vo)

The subtract, multiply, and divide aren't too bad.  A dsPIC can do the
subtract and multiply in single cycles and the divide in 18 cycles.  The
square root takes a software loop, and therefore dominates.  Unrolling the
loop should yield a little gain.  Square roots don't lend themselves well to
series expansion of the type a dsPIC can compute faster than it can to the
loops.

I had this running on a very early 30F2010 that could only go 1/3 of its
design speed and still had a switching frequency of 25KHz.  That means with
a real 30F2010, 75KHz switching speed should be attainable with the same
code.

I was trying to prove the concept, and 25KHz was fine for that.  Some
tweaking is warranted in a real application.  What optimizations
specifically do you suggest?


********************************************************************
Embed Inc, Littleton Massachusetts, http://www.embedinc.com/products
(978) 742-9014.  Gold level PIC consultants since 2000.
-- 
http://www.piclist.com PIC/SX FAQ & list archive
View/change your membership options at
http://mailman.mit.edu/mailman/listinfo/piclist