An apology about all the feedback postings

<[Bill Leach (960729.0055 EDT)]

Upon receiving most of my "catch up" postings, I realize now that most of my
remarks on feedback were "off the mark" when it comes to what feedback
actually is in a control system.

I do not particularly like Bill Power's "feedback is the effect of the
controlled variable upon itself" because (in my opinion) that statement does
not aid one in achieving an "intuitive" understanding of feedback. I also
realize now that I don't even particularly like the definitions for feedback
in B:CP.

I don't believe that my thinking and expression of same on this matter was
very clear and I apologize for further muddying the waters of an already
confusing topic. Not so much an excuse but as a partial explanation, I
believe that among other things, the very long time interval involved in the
sort of interactions that were used as examples helps to lead to the sort of
"foggy thinking" that I engaged in for many of my comments. An additional
factor was doubtless almost natural tendancy to think of feedback in the
more narrow general electronics sense as opposed to the control engineering
sense of the term (both fields however regularly have to use the term in
both senses).

It seems to me now that for the Control System technical term feedback to
apply at all, the perception involved MUST be a currently controlled
perception at the time that the feedback (common definition) is "received".
Based upon these ideas:

1. The /controlled variable/ is specific aspect or property of the environment
    that is intended to be controlled by the control system. This is yet
    another one of those terms that everyone usually accepts as being
    "understood" when that is often not the case. We often say "controlled
    variable" when we really mean /controlled quantity/. The /controlled
    quantity/ is a perception of the controlled variable and does not have to
    be an exact match to the specific signal that is actually the object of
    control. The exception being that when we are talking about the subject
    control system and use the term /controlled quantity/ in a quantitative
    manner we recognize that both our own perception and the subject's
    cognitive perceptions concerning the /controlled quantity/ are both likely
    not a match to each other nor to the actual condition represented by the
    /controlled variable/. That is, except for the models we don't know the
    sensor input transfer functions nor the perceptual input transfer functions
    and thus can not in an exact fashion relate the /controlled quantity/ to
    the /controlled variable/.

    Both the theory and experience with the models however make it clear that
    this problem is a practical matter and not a theoritical one (at least to
    anything even remotely close to the levels of resolution that would
    interest us in our current work).

2. Feedback in the Control System sense of the term is any change in the
    /controlled variable/ from any cause (which by definition of the rest
    of the control system terms means the only possible causes are a
    disturbance or an action of the control system itself). To be consistant
    with engineered control system practice however I take the definition to
    that if one is considering /control system/ feedback "signal" at a "point"
    in the system then that "point" is specifically and always the input to the
    sensors.

3. The /feedback path/ is a bit more problematical. Feedback path is not, I
    believe, specifically a control system term. In the electronics sense the
    term is always applied to the situation where some amount of a circuit
    output is applied to the input of the circuit. That is not too bad but
    circuit in this sense is at least as much of the overall circuit/system as
    is necessary to complete the loop and may or may not include enough of the
    entire circuit to understand how the feedback affects the ability of the
    system to function. That is, very extensive analysis is often presented
    for a particularly complex feedback path with any reference to the overall
    purpose of the system itself.

    This is most definately not the sort of feedback path analysis that we are
    typically interested in for our analysis. We view THE feedback as
    consisting of two paths: One from at least the system output to at least
    the system input (and may include the full circular path, ie: starting and
    ending point are the same point). The other from at least the disturbance
    effect upon the environmental variable to the input (which again can
    include much more of the loop even including the disturbance entity, its'
    transform function to the controlled environmental quantity as well as the
    "reverse" transform functions back to the disturbance initiating entity
    where the required dynamics knowledge of the disturbing entity are known).
    It should be obvious that a complete closed path must be considered (as a
    continuously functioning loop) to quantitatively analyze feedback.

4. The term /disturbance/ (not otherwise qualified) means anything that
    affects a controlled environmental variable that is not itself
    _controlled_ by the control system. This definition intentionally does
    not make any "value" judgements about the specific effects of
    disturbances. The fact that another control system might just happen
    to be independently controlling that same environmental variable such
    that little or no effort is required on the part of the subject system
    DOES NOT alter that with respect to this definition that other control
    system's effects upon the controlled environmental variable are still
    /disturbances/ to the subject system.

Now then from those points I conclude:

1. If there is no control active for a perception affected then there is
    no feedback present in the control system technical sense at all.

2. Feedback (common definition) is ALWAYS a disturbance if it is
    indeed also feedback in the technical sense because feedback
    (common definition) that we have been talking about always is being
    "delivered" by another control system and is therefore NOT controlled
    by the subject system.

All aspects of the feedback (common meaning) discussed in the various
threads all are disturbance in so far as they affect a controlled
perception. The technical definition of feedback from the control system
perspective always includes the effects of disturbances upon controlled
perceptions. To me that means that any attempt to analyze feedback without
knowing such things and the reference(s), the perceived status of the
environmental variable and how the disturbance affects the perception, about
any reference to feedback is meaningless.

bill leach
b.leach@worldnet.att.net
ars KB7LX