This is often called feed-forward control - you predict how much control
input your system is likely to need and then perform feedback control
around that point to correct for disturbances and uncertainties. You can
even make the system slowly auto-tune the feed-forward value based on
observation of the amount of error.

Sean



On Wed, Jul 10, 2013 at 3:26 PM, Spehro Pefhany <speff@interlog.com> wrote:

> At 01:45 PM 10/07/2013, you wrote:
>
> >The bit you are missing is that a proportional only control scheme will
> >almost always have some residual error, since it is the error that drive=
s
> >the output.  The higher the gain, the less the error, but of course too
> much
> >gain will lead to instability.
> >
> >A proportional term is used in conjunction with an integral term if
> >minimising error is important.  The P terms gives you fast response, the=
 I
> >term reduces the residual error to zero some time later.
>
> It's possible to add a variable offset to reduce the error in a strictly
> proportional controller (you could think of it as the nominal output %
> right
> at the setpoint). Usually that's something like 30-35% on a thermal syste=
m
> (IOW, the heater is running ~1/3 of the time if it's time-proportioning).
>
> It works pretty well if the conditions are fairly stable, because you can
> keep the gain high without losing stability in most well-designed* therma=
l
> systems. If a 3% change in PV results in a 100% change in output,
> a 5% line voltage variation or a 10% change in demand will only result in
> a 0.2% change in the PV. This was really common in the days of analog
> electronic controls.
>
> The main issue is cosmetic with digital displays because you can see a
> small
> persistent deviation from the SV.
>
> Best regards,
> --sp
>
> * by well-designed, I mean the system has a single dominant lag, not so
>    easy if you need hefty thermowells and such like.
>
> --
> http://www.piclist.com/techref/piclist PIC/SX FAQ & list archive
> View/change your membership options at
> http://mailman.mit.edu/mailman/listinfo/piclist
>
--=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
.