In Soviet Russia Factory Wideband Tunes You

Oh aren't I just clever with that title...

But on to the point I wanted to make: we already know that the factory primary wideband (WB) oxgyen sensor (O2 sensor) is not very accurate in engine operationg conditions that require a rich air/fuel (A/F) mixture. Generally this requirement is seen on vehicles equipped with forced induction (superchargered or turbo vehicles). Fortunately, Hondata found a pattern to this nuance and has graciously provided us with the "Air/Fuel Correction" (AF.Corr) sensor value in FlashPro Manager that does a great job of correcting this discrepancy.

Unfortunately (yes, there is some bad news), the factory O2 behavior isn't limited to just this correctable discrepancy. I wanted to share some of behavior I've seen the factory O2 sensors exhibit as they fail. In the vast majority of these cases you will not even get an MIL (check engine light), and in some cases the sensor even behaves as if it is working and there is nothing to worry about. Read on for more details after the break...

Failure #1

The first and easiest behavior to spot is what a simple "flat line". I do not have a datalog screenshot to demonstrate this as it is very simple: the A/F read out in your datalogs stays at 14.7:1 and will not change from that value no matter what kind of driving you are doing. The STRIM sensor value also stays at 0%. In the majority of these failures an Error Code will register in the datalogs (Error Codes button on the FlashPro Manager toolbar), and usually the MIL indicator on the vehicle gauge cluster will illuminate. However, I have in some occasions had the A/F sensor fail and give a flat line 14.7:1 read out without the MIL illuminating, but this failure is still easy to detect in a datalog.

Failure #2

The second type of failure I like to call the "unwavering A/F plot" and is demonstrated in the following diagram (click to enlarge in new window). This sensor failure is also very easy to detect. In the diagram I've put two third gear pulls that a customer did back to back with the SAME calibration. The vehicle in question has an intake and an aftermarket exhaust. Mind the plot on the left: the first thing that catches my attention is the the fact the A/F plot (grey line) is essentially "unwavering". It makes no dips, no hills or vallies. If you datalog using the FlashPro unit itself, the datalogs it creates are very high resolution, so even if you tune the fuel to a perfectly level read out on a working A/F sensor, you still normally have a "jagged" plot (for reference, see how the DUTY/INj sensors are wavey in parts? The same thing should be happening to the AF.Corr sensor).

The other thing of note from the left graph is the DUTY cycle of the injectors (OEM 310cc) are at 90% or more at redline. With a bit of experience it is obvious that you should not be seeing a flat 13.74:1 A/F read out with that high of a duty cycle on a car that only an intake and exhaust (taking into account the ambient conditions... as it is possible to see that high of a duty cycle if the car was seeing really cold ambient temperatures, but this wasn't the case on this datalog).

Yet another observation of note is on the right plot that was done literally minutes after the left, the A/F read out has changed completely, even though there has been no change to the calibration!

Now if you are in doubt there is a very easy test you can to do confirm the behavior. If you were to see your A/F sensor doing what happened in the left graph, you could add about 6% fuel to "attempt" to bring the A/F ratio to 13.1:1 at full throttle. With a sensor that is behaving like this example, when you add fuel the read out will not change according to the fueling changes you have made in the calibration. Chances are it would still be stuck reading 13.7:1. Or maybe it will decide to go to 13.4-13.5 and not waver... or maybe it'll do what the right graph shows. I think you are probably understanding the point I wish to make by now.

Here's another annoying fact about this type of failure: the sensor behaves as if everything is fine in "closed loop" operation (when you are not in full throttle acceleration). It will try to target 14.7:1 and the STRIM sensor value will behave according to the changes the ECU recommends based on the A/F sensor input in closed loop.

Final note: I have yet to see this behavior illuminate the MIL or throw an Error Code in a datalog.

Failure #3

Now this type of failure is almost impossible to detect without external A/F sensor readings, or by trying a new sensor. Best way to diagnose this kind of failure is to install an aftermarket wideband sensor (this is also a very good example of why absolutely ANY forced induction application should be using an aftermarket wideband A/F sensor for tuning) and wire it into the ECU (instructions in FlashPro manual). I do not even know what to call this type of failure other than the sensor is simply... wrong. This example comes from a customer that has a turbo kit on his Si. At the time he did not have his aftermarket AEM UEGO wired to the FlashPro. He was concerned about all the knock counts he was generating in boost and was worried that he was running lean since his factory wideband A/F sensor was reading quite a bit different than his AEM UEGO.

Well knock (if it is legitimate) can be caused by several factors, among them being ignition timing, running too lean of an A/F ratio, or even too rich of an A/F ratio (beyond the scope of this article). I asked him to wire the AEM sensor into the ECU.  The diagram to the right is a plot of all three sensor value read outs the ECU was now datalogging. Blue is the Hondata supplied AF.Corr, red is the stock A/F, and teal is the AEM UEGO. As you can see, there is a huge discrepancy between the AEM and the stock sensor, almost 2.0 A/F points between the aftermarket wideband and the stock corrected value.

With this behavior it's almost hard to decide which sensor to trust since there's no other odd behavior other than the large discrepancy in values. He even called up AEM and tested the AEM sensor, finding no fault. I was convinced the factory wideband was to blame and urged him to acquire a replacement after the procedure for giving the OEM WBO2 a new lease on life had no effect on his factory O2 sensor.

He did purchase a new factory WB O2 sensor, and the graph to the left is after it was installed. Note how the A/F.Corr value Hondata provides is almost identical to the AEM WB O2 sensor.

Now you say, well the factory O2 sensor was only reading a bit lean, so it would only run rich if tuned to 12.0:1 A/F, no harm done right? Well..... In this scenario the car would run really rich, and without an aftermarket wideband to confirm you would be none the wiser that you are just soaking the cylinders in fuel (running pig rich isn't any better than running lean).

How about this, what if the the stock WB O2 was reading richer than the car was really running? The next graph (right) displays that exact behavior (on my own car even!). Note that the car is tuned to ~13.0 on my AEM WB O2 sensor, and also note how much richer the AF.Corr value is. Without the AEM I would have never known and I would be running 13.7-13.8 if I tuned to 13.0 on the factory WB O2 sensor (and infact, that's where it was until I corrected it using the AEM sensor). Imagine how catostrophic this scenario could be if you tuned a vehicle with forced induction without an aftermarket wideband.

This third type of failure is more of a horrible sensor error than complete failure, as it continues to function just fine in closed loop scenarios (part throttle, idle, cruise), without any negative impact on fuel economy. However, it does render the sensor completely useless for any kind of full throttle tuning, especially forced induction applications. I cannot stress this enough: anyone looking to supercharge or turbo their vehicle, purchase an aftermarket wideband. Without this extra bit of security, you could be tuning with a stock sensor that is simply... wrong! You would have no way of knowing since the ECU believes the sensor is functioning and will not illuminate an MIL.