Chronology Current Month Current Thread Current Date
[Year List] [Month List (current year)] [Date Index] [Thread Index] [Thread Prev] [Thread Next] [Date Prev] [Date Next]

Re: [Phys-L] feedback control for zero output



On 08/07/2013 01:55 PM, Carl Mungan wrote:
I have a device I'm trying to construct in which I want to null all the
light coming out one port and have it all go out another port. However, the
light is very intense, so I don't want to sample the "on" port. I'd rather
have a detector monitoring the "off" port to use as the feedback signal.
But the problem is "off" isn't a very good condition to monitor. (It's very
subject to noise.)

Is there a way to build a feedback control using a null as input? I tried
googling various terms and couldn't find a good starting point to get me
thinking about this stuff productively.

There is a treeemendous amount known about feedback
loops. There are fat books on the subject.

To make progress, it would help to know more about the
application.
-- I assume we're talking about an interferometer, right?
-- Is the intent to do something with the light itself,
or is the intent to use the thing as a sensor for the
optical phase, perhaps as a way of measuring the relative
length of the arms of the interferometer?

Normal feedback/control theory applies to linear systems. Part
of the problem is that ordinary optical detectors measure the
intensity, which goes like the /square/ of the electromagnetic
field strength. This makes the null (and the peak!) grossly
insensitive to the optical phase. Nevermind the noise, this
creates infinitely bad "leverage".

Any serious suggestions would have to wait for more information,
but here's a hint of the sort of things that might go into the
mix: Suppose the goal is to measure the arm length. Then you
could /modulate/ half of the light with some convenient IF
(intermediate frequency) signal. You then use a lockin to
make a measurement that is phase-locked to the IF frequency.
This linearizes things *and* gives a tremendous reduction in
noise bandwidth.

I wouldn't be surprised to hear that's not feasible, but it
is important to know why it's not feasible ... i.e. to know
more about what constraints there are, and conversely what
sort of design flexibility there is.