[From Bill Powers (2006.11.25.1106 MST)]
Martin Taylor 2006.11.25.12.37 –
Rick:
Then how do you explain Fred’s
last point: “if the light doesn’t interfere with my ability to do
that, then it’s not a disturbance”?
Martin:
He is correct. It isn’t, if it causes a change only in an uncontrolled
perception. That’s why his example is a good one. In his scenario, if
switching on the bright light doesn’t cause his “seeing it
well” control system’s output to change, then it clearly isn’t a
disturbance to that control system.
One problem is still an ambiguity in the meanings of words. It doesn’t
help if you switch from “disturb” to “interfere” if
it’s still not clear whether you’re talking about the interfering cause
or the interference it produces. A bright light can interfere with seeing
something else, without interfering with seeing the something else. That
is, the bright light tends to reduce the visibility of the
something else, but the squinting mostly counteracts that tendency so you
can still see the something else. The interference doesn’t interfere with
seeing the something else. The action of the control system reduces the
amount of interference caused by the interfering variable to the point
where you are not prevented from seeing the something else.
Another problem is that of using an approximation to convert a
quantitative statement into a qualitatative one. The effect of the action
of a control system is [approximately] equal and opposite to the effect
of the disturbance on the input quantity. If you leave out the word
[approximately], then you can give the qualitative impression that the
disturbance has no effect at all on the controlled quantity, which cannot
ever be true. Even integrators are imperfect in a real system, so
[approximately] integrating output functions have finite gain and the
error can never go exactly to zero. Disturbing variables always have an
effect on the input quantity. It’s small, but it exists. If it didn’t,
there would be no control.
Rick:
If the bright light were a
disturbance in the PCT sense of “disturbing variable” then it’s
still a disturbance even when it doesn’t interfere with the ability to
see the object.
It’s still a physical variable, to be sure, but if it doesn’t cause a
[small] change in some controlled variable, it’s not a disturbing
variable.
All you guys have to be more aware of alternate ways of reading what you
write. You know what you mean because you start with the meaning you want
in your mind and think you are putting it into words because you read
your own words with the desired meaning in mind. But if someone else
finds an obvious different way to read what you write, you haven’t
succeeded. You can’t say what you mean until you eliminate the obvious
alternate meanings.
Martin:
If something is to qualify as a
disturbing variable, a change in it must be associated with a countering
change in the control system’s output.
A “countering” change? How about “a change that almost
completely counters the effect of the disturbing variable?” That
“almost” is important.
If the system is
controlling for “ability to see the object” and an external
event (switching on the bright light) doesn’t interfere with the ability
to see the object, then it isn’t a disturbance to that control system
(though it may well disturb other control systems).
This is not true. The bright light might not cause interference with
seeing the object for the simple reason that the action of the control
system (squinting) reduces the amount of interference nearly to
zero.
Which end of the arrow are you talking about?
(magnitude
of
(net effect on
disturbing
variable)
input quantity)
Disturbance ----------------------> disturbance [<------ from
qo]
Interference
interference if controlled
(cause)
(effect)
Best,
Bill P.
