Hi. Andy pointed me to this discussion a while back and I've been sort of following it. I see two issues:
1) The OP question: "if O2 sensor voltage @ WOT can be used to tune to a target AFR", and
2) Andy's discussions concerning using scatterplots for tuning and the associated software solutions (i.e. Megalogviewer AKA: MLV).
Before I give my take on these issues, a little about me. I am a masters level statistician, so I know a little about data. In my younger days, I worked in a university campus R&D/machine shop for about 6 years, so I know a bit about fabrication, troubleshooting, and problem solving. I have been tuning engines on and off for 20 years (not Harleys, although my sister-in-law works for them). Most of my tuning has been with carburetors, both fixed and variable venturi. In the past 4 years, I've been using megasquirt to tune an ITB turbocharged nissan straight six in my 1977 280Z.
For issue 1: The OP question: "if O2 sensor voltage @ WOT can be used to tune to a target AFR".
I'm not saying it is impossible, but there are lot of obstacles to using a NB sensor to tune to anything except AFRs near stoichiometric. My first question is, if this is possible, are any OEMs using this technique? I personally am unaware of any. OEMs will spend tens of thousands of dollars to save a few cents per vehicle. I'm sure they would love to have a way to read specific AFR values - even far from stoichiometric - for the cost of a NB sensor. Rather, more and more OEMs (automobiles) going to wideband solutions (granted, this is mainly for emissions reasons).
The technological challenges come from the voltage curve produced by a NB sensor. It is so flat away from stoichiometric (where you have the opposite problem of it being nearly vertical) that a very small voltage change corresponds to a relatively large change in AFR. The discussion seems to have centered on the accuracy of the sensor, but my biggest problem would be the ability to accurately read the voltage in the electrically noisy environment of a vehicle. I would also be concerned with the accuracy of the voltage measuring device, since a few millivolts can make a difference.
Now, back when it was all I had, I tuned carburetors and Bosch L-jetronic EFI pretty well with just a NB. But, I wasn't shooting for a specific AFR. I was just looking at rich vs lean and giving the engine what it wanted based on power output (using mostly my butt-dyno). I eventually built a DIY-Wideband (this cost over $200 in 2001 - and you had to assemble it yourself- of course there was no data-logging at the time, at least I couldn't afford it). Even then, the advice given was to tune to what the engine wanted and use the AFR information to monitor what was going on. I still follow this philosophy. The specific AFR value really doesn't matter, at least IMO.
The problem I see is that people want to get a number. And, they want to believe in that number. The thing is, without expensive and regularly calibrated equipment (which is out of the reach of most of us), you cannot know the actual AFR. You can get a pretty good indication of what it is (really "was"), but you cannot know.
For issue 2: Andy's discussions concerning using scatterplots for tuning and the associated software solutions (i.e. Megalogviewer).
Andy and I met through the megasquirt community. We have had many long discussions on tuning. Sometimes we agree on specific points and sometimes we don't. I do have to say that when we don't, in my experience, he is always willing to come around (as I hope I am as well) when I can demonstrate specifically why something is the way it is or should be done a different way.
As for our relation with EFIAnalytics, the producer of MLV. Andy and I use MLV because that is the tuning software that works with megasquirt. Neither of us hold a financial interest in the software or the EFIAnalytics. The software is very inexpensive and very powerful. However, all of the patterns Andy has discussed can be seen in excel with a bit more work.
Now, MAP*RPM vs Duty Cycle. Honestly, I'm still trying to wrap my head around this one. I do know that there is something to it. When you have a properly tuned and well behaving engine, there is a clear linear correlation between MAP*RPM and Duty Cycle. I suspect that this is based on airflow, because logged MAF output correlates very well with MAP*RPM. What Andy has proposed is to review a tune through this lense to see if that correlation is there. If there are areas where it deviates review the AFR error and if off, attempt to address with the tune. If the AFR is correct, you may be looking at an artifact of engine design such as runner length and configuration, throttle body placement, etc...
Hope I haven't stepped on anyone's toes,
Sam
Author
Topic: O2 Sensor Output (mv) compared to measured AFR (Read 153440 times)