accuracy of the stock o2 sensors above stoich???

[smslyguy]

Does anybody know how far off the stock o2 sensors can be the farther away from stoich you get?? I can't really see it being all that far off, otherwise why would they use them in a stock vehicle. And i do understand because of cost, but they do have to be somewhat accurate, right???

[dfe1]

The long version--

Standard narrow band O2 sensors are essentially switching devices that are designed to determine whether the exhaust is richer or leaner than 14.7:1. They do not provide particularly reliable readings for ratios that are very far off of stoichiometric. You can play some games by altering reference voltage, but that takes a good bit of testing to determine value/ratio relationships.

As for the use of narrow band sensors in stock vehicles, their purpose is to monitor air/fuel ratio when (and only when) the system calls for a stoichiometric mixture. That's the reason closed loop operation is only used when a mixture of 14.7 (actually 14.63 or 14.68, depending on the calibration) is commanded.

The further the mixture moves away from stoichiometric, the less voltage change you'll see per percent of mixture change. During a dyno test session, I once made a change of one full air/fuel ratio (as indicated by wide band readings) and the in-car O2 sensors showed a voltage change of only 60 or so millivolts.

If you're thinking that you can use narrow band O2 sensors to tune wide-open-throttle air/fuel ratios, revise your thought process. What value are you going to tune to? You'll find information stating that wide open throttle O2 voltages should be 850, 900, 1000 millivolts, or anything in between. But the fact is that it will vary from car to car and from sensor to sensor, and regardless of actual voltage readings, won't be accurate or repeatable.

The short version-

Get a wide band.

[joecar]

+1... I agree get a wideband, they don't cost that much and now they're dead easy to hookup (serial comms).

[mr.prick]

Narrow and Wide Band Linear Oxygen Sensor Theory (http://www.gonefcon.com/whatifdyno/theory.htm)

[swingtan]

Last January I broke my WB sensor while on holidays. I had a large amount of log data with me that had both the WB and NB data in it, so as I had nothing better to do, I made a calc-pid to convert the HO2S11-Voltage, to a pseudo WB-AFR reading.

The result proved that it's a waste of time.

The NB O2 sensor swings much too quickly between max and min voltage to give any indication of a true ratio between the air and the fuel. In the calculations, I had to "average" the readings as the margin for error for a single given WB-AFR compared to the NB voltage was huge. The very best you could use this for is to tell if you were.....


very rich.
rich.
close to stoich.
lean.
very lean.


hoping for anything more than that when just using the NB O2's will just end in pain.

Simon.

[mr.prick]

Out of curiosity, what was the formula?
I thought of this once but I could not find what the NBO2 voltage to AFR is
"supposed to be", this varies by source.
After reading an article on NBO2s and how they're "accuracy"
is dependent to certain temps I saw the futility of this.
I am curious as to how low voltage converts to high AFR and vice-versa.

[swingtan]

I used a "look up table" in the calc_pids.txt file..... crude, but it was good enough for an "indication".

Basically I exported the log data where I had both WB-AFR and NB Voltage. I then imported that into Excel and used that to average "NB-Volt cells" at given "WB-AFR" readings. I then made the lookup table for the calculated PID. Like I said.... it's not pretty.....

If you want a copy, send me a PM.

Simon.

[mr.prick]

I see.
I was trying to find some way to convert voltage to AFR and could not figure an
equation to go the opposite way.
After I read the article (that I can find now :doh:) I realized it had no hope of ever being accurate because of sensor temp, besides the fact that I could not
find a consensus on voltage to AFR for NBO2s.
I still would like to figure out how one would convert with math tho.
Look up would be the easiest provided you can nail down volts to AFR.

[swingtan]

The problem with the lookup table, is that as you mention, the OP Voltage changes too much for the same AFR conditions under differing conditions. That's why I went the "average" path, which can help to "normalise" the error margin. The big problem is that you may have more "error" than accurate data, so the average process actually introduces more error instead of removing it. If you had both a WB and NB running at the same time, you could possibly "tune" the lookup table for a given NB, but as they are not calibrated at anything other than Stoich. The table probably wouldn't work for any other NB sensor.

Simon.

[smslyguy]

so what you all are saying is you can tune your car with the stock o2 but only around stoich, is this correct

[swingtan]

yes. In this case though, you would have the tune in CL mode and use the STFT's to gauge the current state.

[mistermike]

There is a legendary fellow on LS1GTO who shall remain nameless. He steadfastly insists that WOT tuning can be safely accomplished with a NB sensor, and that all of our insistance that he use a WB is merely a cruel plot to relieve him of his money. He's on his third motor, not that I'm implying any correlation here.

[joecar]

Consider how the closed loop algorithm works: the PCM drives the AFR in one direction (crossing the "switch point") and looks for the NBO2 voltage to go in that direction by a sufficient amount; then when the PCM sees that swing it drives the AFR back in the opposite direction (crossing the "switch point") and again looks for the NBO2 voltage to swing back in that direction... this is then repeated as long as the PCM stays in CL mode...

the NBO2 voltage:AFR relation is approximately linear for about 100mV on either side of the switch point and then rapidly ramps up/down until it bangs the "pegs"...

the only way to accurately know wheather the AFR is stoichiometric is to perform the closed loop algorithm... if the voltage can be made to cycle then the AFR is very close to stoich... if the voltage pegs up or down, then the AFR is not stoich... i.e. either yes or no with no in-betweens...

this tells me the NBO2 sensor might be accurate (and this may even be questionable) only at the switch point (typically 450mV) and inaccurate elsewhere.

:)
$0.02.