I just read through the tread on "black magic" and thought I should clarify something (and provide a clean opening for the discussion).
When controlling systems in closed loop there are different schemes of control when implementing the PID system.
1.) position based
2.) velocity based

It has come to my attention during my normal work that many electrical control people "EE"'s tend to assume that PID is always based on the fact that a point in the control system is to be reached, ie: a position control. I say this because I've had to argue the point a few times.
In a "position" based control the "P" gain is used to increase the output to reach the point desired by the command, "I" gain or "reset" is kept to a minimum, and "D" is used to control overshoot or allow for some "Mush" (techno term).
The output will ramp up and try to get to the desired setpoint and then ramp down due to decreasing error when getting close to avoid overshoot and then finally stop at "zero output" in the commanded position.

For systems attempting to maintain a setting durring a moving type of operation "speed control" the set up must operate differently because the actual setpoint (if using position control) must never be reached, Therfore a different formula is used and the setup operates differently.
The Velocity type control is setup so that the "I" gain is the primary driving force in the output by resetting the baseline output more often to maintain the output required to obtain constant speed. In this operation the "P" gain is kept to a minimum and gives the advantage of faster response when a large change is required but should not be attempted to use for the continuous control or occilation will occur due to overshooting in the calculations. "D" is not even used, it just causes problems of inducing lags into the system.

I wrote this from personal experience in tuning fast response electrohydraulic control systems, my terminology is not the best and I'm NOT an EE.
I just wanted to be sure that my usage of the closed loop terminology is better understood before I go into my response to the actual question.

Finally, The question/statement:
If in PE mode and the ECM is in "Closed Loop" (bit set), Learning disabled (per the old thread) and
THE INTEGRATOR is reset
(saw that flash by in the last thread and sparked this whole response)

If the integrator is constantly reset then that means closed loop SHOULD be functioning but is not able to make any changes.
The BLM adder/subtraction may/may not be used (I don't know)
I would think it applies but,

Does the actual output stay locked at the previous setting when PE was engaged then adjust from there to obtain the WOT AFR?
or
Is there a complete change in the output based on the AFR and INJ constants? (among other things)

just a note,
this control acts more like a rolling correction register with a lookup table, not really PID. I think the IAC has more of PID control than this but I have to learn that next.

 

This is all good info...... I'll have to read it again when my brain isn't so fried...(long week).

To really confuse this, from what I've seen, GM didn't do this the same for every mask! I'm sure that will keep the arguments going for years, because I'm sure that not everbody will see the different ways they do things in different masks!

 

I was pondering the possibility that the mode was left in closed loop to provide for "stepless" changeover from so called open loop WOT to closed loop operation.
It kind of makes sense, otherwise there may be some overcorrecting going on upon return to O2 fuel control.
JP

 

JP86SS,

I have to admit you lost me in your second (velocity?) example of PID. In either example, it should all come down to the "error" you're trying to reduce and inputting to the PID logic. Whether that error is "velocity" or "position", the behavior of the P, I and D terms on the chosen error feedback will show the same trend on any (linear) system. Overgaining any of the terms will cause unstable behavior regardless of the error feedback chosen. Again, I admit I didn't follow your 2nd example very well.....

Anyway..... It would seem that the BLM/INT terms within the controller both act as simple integrators. I don't believe (could be wrong) that there is any proportional or derivative correction applied. Any proportional correction shouldn't have any memory of previous events. In other words, the correction would only be based on the current measured error rather than being summed/incremented/decremented with the last measured error. I think your last statement is right on the mark in that it's more of a "rolling correction register".

In the logic I've reviewed, the INT is reset (i.e. no adjustment based on this value) and the BLM is used only if it was adding fuel during PE. I understand the application of existing BLM or resetting BLM varies by mask. Not updating INT makes sense since our stock feedback signal can't be used (assuming NB O2 sensor) at AFR != ~14.7 due to unknown or variable voltage-to-AFR transfer function. Letting the BLM add fuel in some masks was probably done as a CYA. We wouldn't want the engine to run lean when transitioning to open-loop with a clogged injector. Now that I think of it, I know it was used in the SyTy code. Maybe this was done in masks for engines that might have been more sensitive to fuel delivery than others...? I have to admit I've sometimes wondered why the software didn't look for an O2 voltage above a nominal "rich" reading when commanded AFR was less than 14.7. (Maybe they have done this in some mask I haven't seen??)

Rather than just looking at the "IAC", you'll want to consider the whole idle algorithm to see an example of P-I control. I'm just getting into this myself recently. But, I believe we could consider the spark stabilizer/correction due to RPM error as a proportional control. The integral control would be with the IAC steps/Idle airflow.

 

This too is my thought. While in open loop after a startup there are timed delays, coolant threshold(s) and the checking for an active O2 sensor before switching to closed loop.

The transition to PE is sudden, same as the transition out of PE mode. If the ECM was in closed loop before entering PE, it immediately reverts back to closed loop (O2 feedback controlled) mode upon exiting PE mode. The delays and checking for an active O2 are bypassed.

While in PE mode the ECM basically just locks the PID parameters. The INT & BLM may ot may not be used, and may or may not be locked at 128. Proportional gains and timers get held in reset.

It is difficult to specify exactly what the ECM(s) do with the PID parameters while in PE mode. Why? Because there are variences between masks. On some masks if the INT is greater then 128 at the time of PE entry, it will lock the INT at that value (a value greater then 128).

It is possible to specify what happens if we also specify which maskID we are referring to.

One interesting aspect of the fueling PID routine is that the proportional gains are set high. So high that it causes the loop to oscillate. This is what keeps the cat-con happy. And forces the O2 to swing about the desired O2 values, which keeps the INT happy.

Over on the diy-efi site there is a paper that covers a bit of the '7747 closed loop fueling. A little dated but still a good read.

Notice that I didn't use the term 'learn mode' at all. Learn is something different then closed loop.

RBob.