[time-nuts] D term (was no subject)

[Didier Juges]

Maybe we are getting a little off-topic here, but a very long time ago I
was dealing with industrial ovens used to braze ceramics used to make
microwave tubes.
It was very difficult to maintain the precise temperature ramp up and down,
particularly as the oven was not always loaded the same way.

In order to automatically compensate for different oven loading (and
ambient conditions), the controller injected a very low level "random"
noise over the temperature setting and by analyzing how that noise was
filtered by going through the oven, was able to determine the response of
the oven itself and from that optimize the PID terms in real time as a
function of the load. This was in the early 80's. It was pretty hot stuff
then, even for an oven :)

Didier KO4BB


On Sun, Jan 25, 2015 at 8:12 PM, Jim Lux <jimlux at earthlink.net> wrote:

> On 1/25/15 1:30 PM, WarrenS via time-nuts wrote:
>
>>
>> I second  Poul-Henning Kamp's comments concerning D-terms,
>> (mostly) as done in the TBolt and likely other GPSDOs.
>>
>>
>>
>
> Bear in mind that a PID loop is basically a fairly simple control loop
> that is easily susceptible to linear analysis.
> They're simple to implement with analog controls, they're simple to
> analyze, and for a whole lot of applications, they'll work just fine.
>
> And there's decades, if not centuries, of experience with P, PI and PID
> controllers in a practical sense.  A lot of people know how to *tune* the
> parameters based on observed system dynamics.
>
>
> But for a lot of systems:ones where there are significant nonlinearities
> and/or time delays and/or saturation/limiting effects a PID loop might not
> be a good choice.  (for instance, if you're doing a closed loop position
> control with a stepper motor as the actuator, with a small motor and a big
> heavy load, with low friction...)
>
> For myself, I am seduced by the idea of a control system that builds and
> adjusts a model of the system being controlled, and then derives the needed
> control inputs from inverting that model (whether arithmetically, or by
> some clever algorithmic means).
>
> For instance, a PID controller doing temperature control doesn't have a
> *good* way of incorporating side information like the outside temperature.
> There's all kinds of schemes for doing this (double loops, extra terms,
> etc.)
>
>
>
> _______________________________________________
> time-nuts mailing list -- time-nuts at febo.com
> To unsubscribe, go to https://www.febo.com/cgi-bin/
> mailman/listinfo/time-nuts
> and follow the instructions there.
>