So, I revisited the data that I took this summer and put together another plot. This time I displayed the air flow rates and AFR's from all 4 runs, with RPM as the X-axis. The most obvious thing is that the first run from each sensor is MUCH richer than the second run. Remember that I unplugged/reset the Motronic before swapping sensors, so it was re-learning all of the short- and long-term fuel trim settings. There was ~5 minutes of run-time after reconnecting the Motronic and doing the first pull, and then another 5 minutes after that until the second pull. Since the second pull is leaner, a big factor must be in allowing the Motronic to adjust based on run-time data from the O2 sensor.
OK. Here's the plan. I reinstalled the datalogger hardware this evening. I am to put at least 100 miles on this thing so that the Motronic will (hopefully) be fully adjusted, and then I will log some more 2nd gear pulls with the MAF (which is on there now). I will probably also log the idle and some part-throttle driving just for the sake of completeness. Thereafter, the VAM will get stuffed back in, the Motronic reset and after 100+ miles I will do the same measurements with it. Does anyone here know how many hours of operation it generally takes for the Motronic to establish its long-term fuel trim values?
Here is the plot of data that I mentioned at the start of this post.
The green and orange lines are the first runs after resetting the Motronic with the engine at full operating temperature (green is MAF, orange is VAM). Similarly, the blue and yellow lines are the second runs, at which point the Motornic has had 10-15 minutes of operation at full operating temperature (blue is MAF, yellow is VAM).
You will notice that the first runs are much richer than the second ones. My main interest here is the blue and yellow lines below 3500RPM. On these second runs, the MAF runs a lot leaner than the VAM (AFR of 15.0 vs 13.5). This corresponds exactly to the big "bump" in air flow signal coming from the VAM in that range. Someone correct me if I am wrong, but an AFR of 15.0 at WOT is not a recipe for power, and I think that this must be why I "feel the bog" in that range.
I have no reason to believe that the VAM+stock air box is magically flowing that much more air in that range, and as noted in a prior post, I think that it is a resonance effect interacting with the VAM's door which causes an artificially high output voltage from the VAM sensor, which just about any M42 tune will have factored in. Since the MAF accurately reports air flow here, it ends up passing through the tuning tables and leads to leaner running since the table tuned with the assumption that air flow is over-reported in this range. I do seem to recall hearing unverified information over the years that indicates that the Motronic runs off of the 2D maps below 4000RPM regardless of throttle position, which means that it is using the air flow voltage as a primary load signal, and then above 4000RPM if the throttle is open more than 80% it switches to 1D maps for faster response (in which case the air flow signal is largely ignored). From looking at how well the AFR's match above 4000RPM, this seems like it could be the case.
The initial implications here are that there may very well be no way that the M42 can ever have a "perfect" plug-and-play MAF conversion. You would either need a chip with a tune that is not based on a big bump in voltage from 2500-3500RPM, or you would need a MAF converter that taps into the crank position sensor and throttle potentiometer...in which case it really can't be called plug-and-play anymore since you are cutting wires. Maybe I could work an FFT algorithm in to the converter that basically allows the RPM to be read from the pulse frequency of the raw MAF signal, and then a correction curve is applied which mimics the VAM's goofy signal bump from 2500-3500RPM.
Phew. That is a lot of text. At this point, I know one thing for certain. It is time to find a tuner that can work with M1.7 and get a custom tune for this engine with the MAF installed.
For a little bit of extra fun, this link has a ~11,000 pixel wide plot. This one has the raw MAF output signal on it, and it is super duper wide which allows you to see how ugly the signal is. Every single intake valve opening is clearly visible! Remember, in this plot the horizontal axis is in milliseconds, not RPM.
http://www.e30tuner.com/assist/b21secondgear_superwideintakepulses.png
