Feedback Stability versus Noise Reduction: A trade off?

Discussion in 'General Electronics Chat' started by MasterSnow, Apr 10, 2009.

  1. MasterSnow

    Thread Starter Member

    Jan 18, 2009
    22
    0
    Hi all,

    Imagine that we have a closed loop control system that goes like this:

    A/D -> DSP -> Switching Power Amplifier -> Fine Motion Actuator -> Feedback -> A/D (thus completing the loop)

    I have a dilemma in understanding here. Assuming that this system was trying to achieve the most precise control of the actuator possible. Obviously, we'd want sensitive feedback, a high resolution A/D, and a well damped system. However, we'd also want lowest noise possible as I see it. However, this can only be achieved, best I know, by providing extensive filtering of unwanted bands (which is A LOT for a slow system like this, I'd wager) and selecting low noise components, low drift components. Not to mention that we'd need to limit bandwidth to avoid A/D aliasing.

    But here's a problem that I can't seem to find anyone talking about. Anytime you try to filter (especially if there is a high gain analog input stage from the feedback to the A\D), you get wild open loop phase shifts in the analog circuitry...and often well above the critical 1dB cutoff point. SO...close that control loop and tell me what you think will happen? :(

    Thus the issue. How can one filter out unwanted bands and get away with not having total system oscillations? I'd assume that digital techniques for the DSP would introduce phase shifts as well, so it seems like a serious limitation to any kind of precision control. I would LOVE to be told that I'm wrong about this by someone much wiser than I, or that another way exists. :)

    -Logan Snow
     
  2. beenthere

    Retired Moderator

    Apr 20, 2004
    15,815
    282
    There are too many unknowns for a substantive response. Quantifying -
    - would help.

    From a certain amount of work in servo systems, I can say there is a difference between "well damped" and critically damped. Well damped is going to require patience for the actuator to crawl into the final position. Critically damped, the actuator will get to the desired position with minimum over/undershoot.

    That said, the whole system -
    - is a black box. The inputs of both those A to D converters is pretty mysterious. Or, for that matter, why and what kind of signal processing goes on.
     
  3. MasterSnow

    Thread Starter Member

    Jan 18, 2009
    22
    0
    Thank you for the reply, beenthere.

    I'm afraid that this might be too specific of a question that I am trying to turn into a general question. It is difficult to fully quantify, as I have no current system parameters other than the basics.

    My real question is: is it ever ok, under any known circumstances, to utilize multi-pole filtering techniques in a closed loop control system. I need no specific info about the system in mention, it was just a hard example. Because it seems to me that people who design closed loop controls must filter unwanted bands, but I would assume that that would tend to make the system far less stable. I mean...in theory I could frequency compensate the entire system maybe, just like with an operational amplifier?

    Ideally there'd be a sharp noise cutoff and almost no overshooting. However, I know that while it may be possible to tweak the parameters carefully, I'm just looking for a third party to say whether or not it is "practical" to assume I can use filtering in the open loop gain of any feedback system. It might be possible, but I just want to know how difficult.

    Thanks,

    Logan Snow
     
  4. italo

    New Member

    Nov 20, 2005
    205
    1
    Know this the higher the gain and more feedback is the stability factor in electronics. there are other things that can make a system unstable but it basicaly a phisical dilemma like a motor mechanical instability to seek a null.
     
  5. nigelwright7557

    Senior Member

    May 10, 2008
    487
    71

    Look up critical damping.

    Too much gain will oscilate (overshoot)
    Too little gain will undershoot.
     
Loading...