On Wed, 17 Mar 2004 16:16:02 +0800, Innis Cunningham [EMAIL PROTECTED]
wrote:
Hi Guys
After spending a couple of days playing with this thing I have
a question or two.
1 After you get the mode oscilating how do you get it to stop
do you have to continue to adjust all three or do you do one at
a time.
You should adjust one at a time: Adjust - observe results - adjust -
observe - etc. untill you are happy. ;-)
2 How do you know if you are heading in the right direction
what will make the oscilations slow.It seems to me that it goes
at the same rate all the time with the same amplification regardless
of settings.Is there
one property that effects the rate and one the amplification or
once again is it a combination of all three.
Generally: reduce proportional gain (Kp), reduce derivative time (Td) and
increase integrator time (Ti).
There is no one property that effects the rate or frequency of the
oscilations. We usually want to get the amplitude of the oscilations as
small as possible, the frequency is more or less tied to the
resonance-frequency of the process that is controlled.
Any help you can give or something you should read(other than
the ones in FG.I have read them but still am stumped)would be
greatly appreciated.
You could try searching for PID controller tuning in your favourite
search engine.
P.S Clarification.Where it says in the setup that Ti should be set
to infinity does mean writing inf were a number would normally
be.
No, it means setting the value as high as possible. I'm, not sure if inf
would be recogniced as infinity.
In the latest CVS version this has been addressed. You can set the
integrator time to 0 (zero). Now if you look at the PID algorithm you'll
see that this does not make sense because it would lead to a division by
zero. On the other hand setting Ti to the highest possible value might be
a bit awkward, and the highest possible value for a double might be
platform dependant. So I've decided to use zero to completely eliminate
the intergator action from the controller. The code recognices setting Ti
to zero and then removes the integrator action.
Short answer: when it says that Ti should be set to infinity, set it to
0.0 (zero).
The same can be said for derivative time (Td). Setting Td to zero would in
the previous version lead to a division by zero. The new version
recognices this and skips that particular piece of code (where the error
is filtered).
Summary: Set Ti to zero to completely eliminate integrator action. Set Td
to zero to completely eliminate derivative action.
--
Roy Vegard Ovesen
___
Flightgear-devel mailing list
[EMAIL PROTECTED]
http://mail.flightgear.org/mailman/listinfo/flightgear-devel
