"=C1lvaro Deibe D=EDaz" wrote:
> You only need a basic hardware PWM to generate your 4 servo outputs.
> Five resistors and a little ISR will do. Easy to get 250ns from a 8MHz
> clock from the hardware PWM, without glitches, although deadbands of
> 2-4us are usual in the middle-priced servos, so usually there is no
> need to go beyond 1us of resolution in the PWM's.

I agree there are various ways the hardware could be minimized, especially
with more responsibility pushed onto the firmware.  Those would actually
matter if this were a production product.  However, it's a one-off design,
and largely a research test bed at that.

I am finishing up another side project now, and haven't really thought about
the electronics for the flying sheet yet.  I want to use a dsPIC because I
have no idea how compute intensive the control algorithm may end up.  Most
likely a 18F could do it, but I want to have room just in case I want to try
something complicated.  With a dsPIC running at 30 MIPS, I'm pretty sure CPU
power isn't going to be a problem.

Since quite a few dsPICs have lots of PWM modules, dedicating one to each
servo is the immediately obvious thing to do.  Another simple approach is to
externally multiplex the output of a single PWM module and run the servo
pulses in sequentially.  If that reduces overall weight I may do it.

> Split ailerons (split "ailevators", in this case) (or whatever the right
> name
> is in english) are usual to control fin-less flying wings (Northrop
> YB-49, i.e.).

If I remember right, the Northrop flying wing and the B-2 split the
"ailevators" such that they have two surfaces, one directly above the other.
That means you can use them both together to create drag without any net up
or down force.

Having two control surfaces above each other would be too mechanically
complicated for what I'm doing, so I have them next to each other.  The
drawback is that when you activate both in opposite directions you get a net
torque applied in the roll axis.  This then needs to be cancelled by the
surfaces on the other side of the wing, but with less drag else you haven't
achieved control over yaw.  This part I have thought about in my head and I
believe it is possible to do, just that the math gets a little more
complicated.  In other words, I'm deliberately making this mechanically
simple and making up the difference in the math.  I really think it's
doable, but either way, it will be fun to try.

> If you keep the numbers in the integer field, I think it
> will comfortably fit inside a PIC18.

It probably would, and that would be a appropriate thing to look at if this
were a real product.  But it isn't, so I won't.  I'll most likely use all
floating point on a dsPIC.  I'm interested in the control algorithms and
don't want CPU power or numeric problems getting in the way, even for trying
things lots more complicated than may ultimately be necessary.

> (Perhaps) interesting links:

No!  I know flying wings exist, but the fun part is breaking new ground,
even if it's just new ground for me.  I don't know if others have solved
exactly the same problem (same arrangement of control surfaces) as I'm
trying to solve, but if they have I don't know about it and certainly don't
know of their solutions.  And I want to keep it that way until I've come up
with my own solution.

This is for two reasons.  First, inventing the solution is the fun part.
Otherwise I'd be just a dumb monkey building a model, and there's not fun in
that, at least for me.  Second, it's a lot easier to think outside the box
if you've never looked inside.  Right now I have the advantage of not
knowing the "right" way to do this.  All too often conventional wisdom gets
stuck in a rut about how something should be done.  If I don't know what
that rut is, I can't get stuck in it.

After I've got my flying sheet working, then it will be interesting to see
how others did it and how my approach differed.  When I've got mine working,
I'll obviously have found a right way too.  Maybe I'll converge on their
right way after a lot of trial and error, but maybe I'll come up with
something a little different too.


********************************************************************
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