AW: [Re: thermal tips re expo converters:]

[jhaible]

> The analogy to a regulated power supply was just to point out you can
> achieve a stable control loop with a highly non-symetrical
charge/discharge
> rate, possibly equivilent to a fast heating, slow cooling thermal control
> loop. Of course, if I am proven wrong, then I disavow everything I just
> wrote.

It's an excellent analogy, as it shows both that it is possible to get a
stable
system, and that this stability has margins that are set by the specific
design.

Again, it's all about having only one dominant pole in the system, in any 
working condition.

Most if not all voltage regulators need a minimum capacitive load at the
output.
This is to slow down the regulation, making the output resistance plus load
capacitance the dominant pole. For high frequencies, the regulator has a
high
impedance output - but it's the very same external capacitor that will make
it
low impedance again, as the load sees them both in parallel.

So the good thing is that the same component that makes the regulation loop
slow (the capacitance), also provides the "inertia" that makes it
relatively
unsensitive to fast load variations. (Sorry for the mechanical analogy
(;->) )

There are also some voltage regulators that need a minimum static load.
So there you have the "bleed rate" in the thermal analogy !
If memory serves, some 3-pin variable voltage regulators need low impedance
resistor dividers for that reason. And recently I found that a simple 7912 
also needs a minimul load as soon as you use it with input voltages higher
than 18V.

So I think it is possible to find stable solutions for systems with highly 
asymmetrical time constants, but there are some constraints about speed,
and "increasing bleed rate" sometimes is also a means chosen by voltage
regulator designers.

A related, and also very interesting problem are vactrol based limiters.
These vactrols are really bad guys when it comes to find a fast and stable
regulation. Highly asymetrical time constants here as well, but the slow
(release) time is highly dependent on several parameters, to make things
worse. All is fine as long as the vactrol is the dominant pole in the
system.
But if you try to further increase release time with an RC filter, and
*not*
further increasing the attack time of the system, that's where the trouble
starts !

JH.