Innovate wideband going full lean
 

So, this has literally nothing to do with Miatas and/or turbos, but I figure the people here know more about this stuff than many other corners of the internets.
I have an old 1990 Nissan D21 King Cab pickup truck, that has got it's ancient carbed Z24 engine replaced with a slightly lower-mileage fuel-injected KA24E. I didn't get the engine with a complete wiring harness or the stock ECU, so I'm installing a standalone in the form of a Speeduino. Most of that is slightly beside the point, but anyway.

It starts and runs and idles well, but basically nothing outside of the cells around idle are tuned yet. The issue I'm having is that the wideband O2 after a while runs off to lean numbers, eventually pegging at max lean (22.4). It's an Innovate MTX-L (combo controller and gauge) left over from another abandoned project.

When starting it from cold, the AFR values seem perfectly sane. It runs a little bit rich (hasn't been tuned perfectly, and getting a few percent WUE still), sitting in the high 12's to low 13's, but smooth and steady with no misfires or anything weird. Then, without much warning, AFR values start climbing, and over the course of about a minute they go from "normal values" to crazy lean, pegging at 22.4. Throughout this, the engine is still idling smoothly (at around 1100rpm, I haven't bothered getting it lower yet). Open-loop idle, no fuel correction from O2 sensor, so it should be getting a constant level of both fuel and air coming in. If the AFR readings are correct and it was indeed running this lean (from getting way less fuel or more air for whatever reason) I would expect it to misfire and run badly, or at least slightly change RPM, but it doesn't.
When this initially happened, I believed the numbers and started cranking up the VE table to keep AFR down. I ended up with about 3x as much fuel, the engine was sputtering and running like shit, and it was stinking of unburnt fuel out the exhaust. On later attempts, I've cranked up the fuel a little bit as this happens, and while it doesn't seem to affect the AFR readings it does by smell and sound seem to be making the engine run too rich. So, I see no real reason to believe other than that the engine is getting roughly the right amount of fuel as it is and that the readings are off. Can't be certain, but that's at least my hypothesis.
The MTX-L gauge and the ECU (as seen in TunerStudio) agree on the values to within 0.1 (about as good as the precision of these analog signals ever get, I guess). I'm mostly going by the gauge, but the ECU agrees. I have a also tried just disconnecting the signal to the ECU (as it's not doing closed-loop fuel anyway), but the gauge still says the same thing.

The problems happen when the engine is warm. No idea if that causes it, but at least there is some correlation (be it from temperature, or time running). From a cold start it happens after a few minutes at idle (just as engine is reaching 80°C or so coolant temp) and drifts off to full lean over about a minute. If I do it with a warm engine (and of course also warm-ish exhaust and sensor) it happens within a few seconds after the sensor has warmed up and the values drift off much quicker (goes to full lean in just a couple seconds).

First thing I did was move the O2 sensor. It was sitting in the stock position in the manifold (similar distance from the head as the stock position on an NA Miata) and I thought that could cause it to overheat or something, so I welded in a bung further down the exhaust. Maybe half a meter from the flange between manifold to exhaust pipe, and about a meter away from the engine. This made no apparent change, and the weird values still happened. This is when I did the logs attached from a cold start.

Next guess was that I had killed the sensor somehow, so I swapped that out. It's a sensor I had in my Miata NA until last year, where it functioned perfectly. This made no difference at all; readings started out fine, and then drifted off into "crazy lean".

So I figured maybe the Innovate MTX-L was busted. I have an LC-2 controller as well (also came out of the same Miata), which has an analog gauge wired to it. Kept the same sensor (the one that unit always has been running with). This time, if anything the values drifted off even faster.

I did get some logs during one of the cold starts (when I had just moved the sensor further downstream). I started it from cold and didn't touch it as it was warming up. I was fooling around with the logging, and just as I thought "problem solved", AFR values drifted off. I've attached a couple log files of this, they're from the same time but with a 2 second gap or so in the middle as accidentally stopped it. Put together, they show the AFR drift from 13-ish to full lean (max value in TS comes out at 22.3, gauge shows 22.4) over about a minute, while everything else stays basically flat except me pushing the throttle a little bit once. Other readings/values (MAP, RPM, fuel PW, ignition advance) stay steady throughout.

A screenshot of some of the logs:
Attachment 228574

Not sure how terrible Speeduino logs and tunes are to read (it's on r201806), but the phenomenon is pretty visible in that picture.

At no point was the wideband controller giving any error codes (or for the LC-2, error light). Every part of the wideband O2 setup (sensor, controller, gauge) has been changed out, so I either have two sensors broken in the same way at the same time and/or two broken controllers, or something else is causing this. I've had sensors go bad before, but that always caused it to signal an error and never this "work fine for a while and then give weird readings".

The wideband has power and ground from the same point as the ECU. Supply voltage measured at the wideband controller is steady in the high 13's (forgot the exact number, but 13.8-ish) while idling. I did recalibrate the wideband each time I moved or changed out sensors or controller.

I'm grasping at straws a bit and trying to come up with some rational hypothesis that can explain this, and I have a few ideas all of which sound bad:

• Engine is actually running crazy lean for whatever reason, from getting lots more air or lots less fuel. I feel this should cause the engine to at least change RPM a little bit, and it isn't. I could measure fuel pressure, injector supply voltage and signal pulsewidth to doublecheck, but it seems implausible.
• One or more cylinders aren't getting any fuel and just pushing air through (which would mean lots of oxygen in exhaust). Should also cause noticeable RPM drops, be more of an "on/off" thing and not gradual, and I can't see how that could happen so consistently anyway.
• All my Innovate wideband O2 stuff is broken. I have a third unused unit, a 14point7 SLC Free I got specifically for this truck. Will test using that (probably in parallel with an Innovate unit) and see what gives. Can also see what kind of signal the stock narrowband has.
• The exhaust is sucking in air through some leaks. I know the exhaust isn't perfectly tight, but there's just some light "puffs" around a couple of flanges. I've seen other cars suck in air through leaks and throw off AFR values slightly, but never by nearly this much. Adding some backpressure (perfectly sized potato in exhaust) would increase the amount of exhaust coming out but also prevent air getting in, so I can test this.
• The same leaking exhaust is getting in to the sensor from the back. It apparently gets "reference air" basically leaking in by the wires. Maybe it's seeing exhaust mixed in there, throwing it off. Again, the leaks are pretty small and I'm outdoors, I don't believe the small puffs of exhaust can "poison" the air quite as much. There was also no noticeable difference in the two sensor positions (the lower part of the manifold, or on the exhaust pipe further downstream). Could stick a fan somewhere blowing fresh air at where the sensor is to test this.

In short, I am rather lost at what causes this. Will keep probing, but I would appreciate any input in what could be going on here. Can a wideband O2 fail in this manner? Can exhaust leaks really cause this skewed AFR readings? Can an engine really run this lean without affecting idle RPM in the slightest?