[time-nuts] Improving the stability of crystal oscillators

[Bruce Griffiths]

Neville Michie wrote:
> Hi All,
> I am not trying to start an argument, but I would like to point out  
> that PID controllers are only
> good at controlling a certain class of system.
> For a system that has a coal truck that must dump its coal down a  
> hole, the system has mass, velocity
> and distance, all the qualities to get a perfectly damped system with  
> PID control.
> For thermal control, the function is more likely to be a Bessel  
> Function, and a Z transform filter
> is more likely to find a match.
> In any case, PID controllers are often to be found in totally  
> unsuitable situations giving worse control
> than even a bang-bang controller.
> The thermal block controllers work well because of the dominant  
> integrating effect of the block,
> the time delay for a heat front to propagate through the block is the  
> only concern for instability.
> When instability is a problem I relocate the thermistor closer to the  
> heater, giving a marginal degree of under-
> control.
> Because the block is well insulated it soon becomes very close to  
> isothermal.
> cheers, Neville Michie
>
>
>
>   
Neville

If a purely proportional control loop has such great performance why
does the 10811A use a PI temperature controller and the E1938A use a
PII^2 D controller?

Surely the finite offset between the setpoint and actual temperature
achieved by a proportional controller is a source of long term
temperature instability?
If one uses resistive heating then some linearisation improves the
performance as the heat from the heating element is proportional to the
square of the voltage across the heating element.

A state space controller may give improved performance but PI(10811A),
PID and PII^2 D(E1938A) controllers seem to work well when used to
regulate crystal oscillator temperatures.

Bruce