An Oxygen Sensor Too Far

TL;DR : a bad idea, don't do it

With the recurring failure of the wideband oxygen sensor (4 killed fair and square and 1 DOA), I was ready to try about anything to increase reliability.

The change in November consisted in increasing the heating delay in the ECU, and I did add a 25mm spacer to get the sensor out of the direct exhaust flow. As a reminder it's located 8 cm from the exit of the turbo.

That (already third) sensor lasted for about 6 months, only to fail at the track day in March. I had a spare on hand that didn't even last 2 kilometers.

I got yet another one, that was installed straight on the pipe this time and didn't even last a full month.

What I think happened is that re-starting the car when the engine was still warm-ish (~40-50°C coolant) was the scenario that led to all those failures. That was a clear thing on track where I drove 20 minutes sessions followed by equivalent pauses. It was also the case when I only used the car to commute to work and back in the afternoon, with the coolant still warm at the end of the day. I even suspect the "DOA" one didn't survive being tested idling and being restarted moments later when we left.

The heating delay is now set to 45 seconds for any coolant temperature above 10°C and even longer when it's really cold. This should tackle the issue for good. I also got the 25mm spacer back.

As usual, those are my personal notes; what I actually want to document here is how the 45° extension was a bad idea.

The 45° thing was supposed to prevent condensation pooling on the sensor tip, so I got a angled adapter designed to fool a post-catalyst sensor on cars with a failed catalyst and an engine check-light.

I did discard the small filter cartdrige, saw off the thread for the discarded small cap; I drilled all openings to 12mm; used the 25mm spacer instead of the provided 40mm part they provided; and I filed-down part of the triangular body to save on weight. 

It looked OK overall. All that remained was some testing.

It didn't take long to confirm that it was way too long, lacking the required gas flow. With the engine warmed up, idling was unstable; there was a very long lag in the readings that did oscillate slowly and widely. Deceleration fuel cut-off (DFCO) readings took ages to reach about 6;  where they reach 10 in a snap otherwise.

So I got that extension out and reverted to only keeping the 25mm one as it worked well since November.