Predictive (analog) Feedback Element....Musings on Linear Analog op amp

[RJSV]

Hey!
   I'll post a diagram in a minute, but this is one of those 'musings' that hit every once in a while...(I heard Edison did that).

   Thinking goes, a conventional OP-AMP circuit often has gain controlled by a little resistor network, such as having a 1 to 2 ratio in the two resistors, for having GAIN controlled actively.  That example feeds back at 1/2 which causes the circuit to 'control' stabilize at GAIN of 2X.   All conventional.  But I've started thinking, that feedback signal is in the PAST, from a time-line point of view, even if it's a short delay.
What is possible, if your feedback term is up to date, meaning no (effective) delay ?
   Of course, any circuitry will have a propagation delay, even the wires having a smaller delay contribution.
Imagine your 'Predictive' block has a digital processor, and you are passing a (predictable) sine wave, and thus that processor could adjust the output phase.   I'm talking about a rather capable little component, often much more complex than the analog OP-AMP and passive components, in bulk.
   Containing moderate, or even larger RAM resources, and perhaps with an AI slant to it !
The sine wave example was just for illustrating the set-up, and I don't know a huge amount regarding predicting noisy signals, but some 'Enterprise' related aspects would be such things as:
   Example:  Delivery truck comes every Tuesday, approx. 9 am, causing some circuitry activity.  The AI feedback component knows this, and can predict certain functions and sequences very likely.   An activated motor would perhaps run to a STOP switch soon, which this 'feedback element' predicts.  Of course the enterprise activity is on a way higher scale, than immediate OP-AMP component signals.

   That's a casual musing.

[Kleinstein]

With control theory such a forward looking element is called feed forward part. It can be quite nice, especially with relatively slow systems, like a thermal system that should follow a temperature profile.

For just an amplifier, there may not be that much need for a feed forward part as the electrical feedback can be very fast.

[RJSV]

   It's hard to conceptualize this type of feedback, and using an example, of warehouse truck arriving, say, causing a slight sag in the power, as various electric actuators are opening doors, etc.
That voltage sag might not relate exactly, to an analog OP-AMP circuit, but here is one concept:
   Let's say (a delivery truck) goes through one gate.  The processor would, using AI type process, know that one mechanical switch always predictably opens back up, after 2.3 seconds of motor power applied, and processor can 'anticipate' by way of starting an intense monitor, of switch, right just before expected to open.  I think that starts to be a good example, but that particular action doesn't usually need some super accurate timing.  So it's a partial example, but the concept is for whatever process NEEDs that kind of accurate anticipation, of analog feedback to be applied.

   Whew, it's difficult to describe the full application, but would be a special, (higher cost) and higher performance result.

[Infraviolet]

I think any digital processing to do this, especially the ADC and DAC parts, would always inevitably be slower than an op amp which keeps signals in the analog domain.

[RJSV]

Forgot to mention,  I'm the formerly named ID RJHayward, an E.E. near Berkeley, Ca.
   I got to the point, where the blue-collar mix and near-poverty doesn't always get a good welcome, and so brought up a nearly- politically neutral ID dropping the Hayward reference.   (Not that I'm not a proud poverty stricken fellow, but involving 'Hayward' likely stirrs up visions of street weirdos.
   Anyway it's now, me, as 'RJSV'.  meaning rick-jack Silicon Valley.
Thanks

[thermistor-guy]

With control theory such a forward looking element is called feed forward part. It can be quite nice, especially with relatively slow systems, like a thermal system that should follow a temperature profile.

For just an amplifier, there may not be that much need for a feed forward part as the electrical feedback can be very fast.

Harold S. Black invented feedforward before he invented feedback.

Feedforward is used in some RF amplifiers, because it avoids feedback instability issues.

I've used feedforward error correction in a 3rd-order digital PLL. It's a useful technique to have in your conceptual toolkit.

Oral history from the man himself: https://ethw.org/Oral-History:Harold_S._Black

[RJSV]

Thanks, but even a (likely) impractical suggestion gets provided motivation, especially when I'm rather weak in analog circuits.
   I'm actually reminded of stock market watchers (gamblers).
Predict such and such.  Heck, 5 years ago, I wouldn't have thought to toy around with 'smart' feeds, to a 5 cent analog stage !

[langwadt]

With control theory such a forward looking element is called feed forward part. It can be quite nice, especially with relatively slow systems, like a thermal system that should follow a temperature profile.

I'd say that it usually makes sense to have feedforward with a thermal system. If you know it take approx X watt to maintain a set temperature it make sense to start with X watt and then let a pid loop handle only the errors

[thermistor-guy]

...
I'd say that it usually makes sense to have feedforward with a thermal system. If you know it take approx X watt to maintain a set temperature it make sense to start with X watt and then let a pid loop handle only the errors

It's common for commercial thermal controllers to be self-tuning. They have adaptive feedback, where the control errors are used to adapt the model of the plant.

The practical advantage is you can connect the controller to the plant, and the controller learns what to do. So you don't have to be a controller expert to use it.

[Kleinstein]

When speed is not critical the I part of a PID regulator can compensate for the typical output needed. So no advantage of feed forward for a static PID system.  One may get away without the I part and use learned Feed forward instead - still this gives slightly less accurate regulation, though possibly faster on some transients.

The feedforward part is especially helpfull for a dynamic system. So when you want the temperature to follow a given profile and have a good estimate for the added power needed to get a rate of change.