About the Wideband Lambda Sensor

TL;DR Yet another consumable

With the engine upgrade also came the upgrade to a wideband oxygen sensor. What's the difference you may ask. In simple terms, a narrowband sensor is only able to tell if the combustion is rich or lean, not really measuring how rich or lean it is. Its output will oscillate between two extreme voltage values close to 0 and 1 volt or so. With maybe a narrow band (hence the name) around 0.5 volt where it's able to work out a value when it's close to stoichiometric. A wide band one, on the other hand, is able to provide a reliable signal from 0.65 (lambda; rich) to over 10 (super lean). This allows for a finer tuning of the combustion by the ECU; that might aim at say 0.90 rather than 1.00 for high rpm and boost pressure. That strategy in itself limits the risk of knock and keeps the engine cool(er-ish).

The first installed wideband sensor failed after maybe 2 or 3 weeks, leaving us stranded 100km from home. An extra month of erratic engine cut-off later, the defective oxygen sensor was identified as the explanation. Truth be told it was solved swiftly once I dropped the car back for a check-up. Sensor lifetime : 1000km before the first failure, it was replaced after 5000km.

We got the car back with a new sensor and went straight to House of Speed for the inauguration of their new showroom and workshop. That day was topped-up with the crash into a deer in the afternoon. That to say there was no driving between mid-April and the last week-end of August.

The second sensor lasted for about 2 months (September and October) or 5000 km. A first hint that something might have been off was the unusually poor fuel economy when driving to Normandy; which I attributed to the strong front-facing wind. When I changed the oil and found it rather dark again, I thought it most likely came from the track day from a week prior. I got to drive for about 1000km after that 'vacation' oil change before more tangible signs appeared : the engine was cutting and jolting abruptly on the motorway at the slightest lift of the gas pedal.

We drove to Luxembourg and back on Nov 11th. Something was clearly off although not as bad as the first time in March. Everything seemed fine below 5000 RPM, at which some kind of soft rev limiter was triggering. The knock detection bars I've set on a page of the CANchecked gauge confirmed something fishy was happening. The issue was intermittent. At least we weren't stranded. 

Back home I set new pair of user-defined CAN messages in the ECU for the actual and target lambda values and added a new display page on the CANchecked gauge for air-fuel ratio monitoring. It didn't take long to see that the fuel mix and catalytic converter were managed properly at idle, oscillating nicely between 0.98 and 1.02 (1.00 means perfect stoichiometric ratio). A catalytic converter needs to see both rich and lean gases to be happy, and that's precisely what the gauge said was happening. I learned that from the very timely video Technology Connections published about a month ago.

Last Friday after work, I decided I'd hop to Spa-Francorchamps to see how the track day of the Lotus on Track group was going, a 45 minutes drive would make for a nice test of the oxygen (lambda) sensor and its monitoring.

It didn't take 5 minutes for the sensor to fail and report a value stuck at 1.00. A scenario that could only lead to an overly rich mixture as the map expects richer ratios when revving and pushing a bit.

Strangely enough, the old IT trick of "when in doubt, reboot" got me the car back in business for the evening. I could examine how the sensor was reading in various conditions. Like how it maxes to 10 with fuel cut-off when decelerating. That fact will become more important soon.

A small trip on the week-end began with everything looking fine for about 45 minutes; after which the car was barely able to run. This time a short restart didn't help, it took a 5 minutes stop and restart. I could feel not everything was back to 100%, and at the same time that deceleration cut-off ratio couldn't reach over 3 or 4 (instead of 10).  I assumed that could come from some kind of self calibration and compensation being applied at startup. It's actually not. This limited bandwidth is really a sign of a failing sensor.

So I got a new Bosch LSU 4.9 sensor. The swap took 10 minutes at most as it's readily accessible on the Caterham 170. With expert advice, I tweaked the heating startup delay with a small increase, but nothing dramatic; it just couldn't hurt.

Wondering about the surprisingly short lifetime of those two sensors LSU, I looked around the internet and could confirm that a rushed heating strategy vs humidity could crack the sensor. This has been addressed already, and you'll see on the next paragraph that it might not be the biggest issue.

Another interesting or concerning point of view from Tech Edge/Wbo2 is that a wideband sensor should be placed 1 meter away from the exit of the turbo. They state that while car manufacturers have no issues with narrowband sensors located right after the turbo, this is no place for a wideband one. Their website is really worth a read.

On my car, there is literally 9.5cm between the turbo outlet flange and the centre of the oxygen sensor. At the same time, there is less than 50cm between the turbo and the catalytic converter thus no way to relocate the sensor to a more appropriate position (>1m). Following Tech Edge's advice to recess a sensor that's too close to the turbo outlet (I paraphrase here), I placed an order for a short M18x1.5 extension. Keeping the new sensor out of the hellish furnace should/could/might help with reliability. 

The spacer I got is supposed to deal with error codes that come from a failing catalytic converter. I got rid of the filter and sawed off the entire front nose and only kept the 26mm extension part of it. I salvaged the metal o-ring from the freshly failed sensor and spent 10 minutes to get everything on the car. 

While the heating strategy could be the only cause for concern and there's no risk in delaying a few more seconds than strictly necessary, recessing the sensor might either help or be detrimental to the reliability of the readings and overall engine management, we'll see. I might have to remove the spacer before the end of the week-end. A short drive around the village didn't show anything to worry about in terms of values and responsiveness.

As it's now clear I might have to consider those LSU 4.9 as consumables on my specific setup (not a quality control or design issue), I did place an order for a spare one right away.

I hope I'll be able to get 20 000 km out of those. It should be possible to pro-actively replace them by monitoring the "decelerating fuel cut-off" lambda values that show the early signs of a failing sensor. Fresh out of the box, the new one jumps from a ratio of 1 to 10 in about a second. Any slugginesh or loss of bandwidth... and off with its head. 

Thinking about it, another clear sign of O2 sensor trouble was the erratic idling. In some instances the ECU was chasing idle between 500 and 2000 rpm for no reason at all, in spite of the closed loop control for fueling and closed loop PID control for idle itself.

The day I'm sick of that little game, I can still go back to a narrowband one. An older technology will always be better than a failure waiting to happen.