Alright, scoping it seems to have offered some clues. It is a storage scope so I could have captured the waveform, but it is chilly in the garage and just wanted to get a look with my father.
So, what I saw has made me draw the following conclusion. Basically, the MAF and converter are too damn fast and "sensitive!" They are so fast that they seem to be resolving individual intake pulses! I have no idea what speed the Motronic samples air flow voltage at, but I would guess that it is slower than the converter and might is probably taking them at pulse peaks and troughs, confusing the hell out of it. For those that know what "beats" are, the kind produced by the mixing of two or more periodic waveforms, you get the idea. This could explain the fluctuating idle...the Motronic's sampling rate at these lower engine speeds is very close to the intake pulse rate where the MAF is most prone to the induction pulses. Since it is definitely less than twice the intake pulse frequency it will have aliasing problems and resolve the wrong overall waveform (read about Nyquist frequency/sampling). Anyway I am getting waaaay off on a tangent.
Basically, this is a problem only with larger throttle openings as there is essentially very little damping of these pulses. With the throttle closed all the air is sucked through the ICV which can easily act as a damper. Even with small openings the pulses might not be getting through the cracked throttle opening without being damped out a just enough. Open them a large amount and the pulses have very little interference...and run right out to the MAF. The MAF is fast enough to resolve these.
As for the issue at the higher engine speed, this could just be due to it being a secondary or tertiary harmonic of the original problem. The cutout seems to happen around 1200RPM. The next one looks to be around 2400RPM...a factor of 2, conveniently. Why not at 3600RPM? Perhaps enough turbulence is generated in the induction system so as to damp this nonsense out.
This pressure pulse-measurement system is interesting. With the AFM, it was definitely well-damped...the flapper would not be disturbed by the individual pulses due to its relatively huge mass in comparison to the moving air's momentum, and being on a spring with friction in the moving parts. The MAF is essentially a very very underdamped system...it has no restrictions on response time other than the internal electronics and the speed of the passing air. This means I need to implement some form of damping, either in software or hardware in the converter. This means more delay (lol, like it is even noticeable by a human driver). Now, it does not matter a damn that the converter response time is presently about 970 micro-seconds...the Motronic is pobably slower since it has SO much more to do. The clock on the converter is 4x faster than the Motronic, and there is 99% less code to execute in the converter...so I think it reasonable to assume the Motronic samplng rate is a lot slower. The converter can sample at 4x the maximum intake pulse rate at 7200RPM, so aliasing anywhere in there is no problem.
I COULD look into using an RC or RLC filter to help with this, but I would prefer not to. Software is free lol, and a lot more predictable. If I were to do this in software, I would play it safe and set the averaging time to be able to deal with a 600RPM WOT condition, it would leave a 50ms delay. I do not like that, although as far as throttle response goes it would not be very noticeable...it just pisses me off that I know I could have the system run with a delay of 150 microseconds if I wanted lol.
LOL and I always hoped I could find an application for that Systems & Control class I took. I guess I did! Sorry about the world's longest post. This is how I like to stew the old brain juices...thinking "out loud." More to come. At least I now know the issue is not in the calibrations! I need to research signal filtering a little more!
