I’ve read in a couple of different posts that “the factory O2 sensor is not good enough to tune with. You need an aftermarket wide band set up.”. I started thinking about this and some things do not add up.

I’ll take the European stuff as an example. Audi, BMW, Porsche, VW and the like have all used the Bosch LSU 4.2 wide band sensor. The only difference between a $50 VW sensor and a $275 Porsche sensor is the connector! Each manufacturer has its own connector and that is one of the reasons you can not buy the sensors from Bosch directly. In addition, the connector has a trimmed resistor inserted in it that is used in the individual sensor calibration scheme so you can not just swap out the Porsche connector on a VW sensor and run with it (unless you can carry over the VW calibration resistor to the Porsche plug). All of the above is useless information until you consider that at least a couple of the aftermarket wide band O2 sensor companies use the VW LSU 4.2 sensor! Yep, it is not good enough to tune with when it is put on the car by teams of OEM engineers but is just what the doctor ordered when a gage company sells it to you. I do not know about you, but I see some humor in that.

I’ve had a chance to compare dyno mixture data with OEM wide band data on the same run. Every time I have looked, the data overlaid nicely so I never gave using OEM data a second thought. That being said, I have by no means done a though evaluation or comparison of the two. Has anyone out there looked at this stuff in depth?

I did see one area where a little extra knowledge might be helpful. This is kinda a nerd thing, but there are two control loops that run a 4.2. The first is the heater circuit that keeps the Nernst cell at a specified temperature and the second is the Nernst DC loop that keeps a specified voltage across a portion of the Nernst cell. When it came time for me to rely on 4.2 data to tune, I wanted to see that both of those control loops were on target before I trusted the data. You can easily imagine how exhaust transients could blow either loop off target. Drastic changes in mixture (whacked accelerator pump tuning) could easily have the Nernst cell chasing equilibrium (Nernst DC off target) and drastic changes in exhaust gas temperatures would most certainly affect heater equilibrium. Sure enough, when we built our own 4.2 controller we could easily see when the PID (proportional, integral and derivative) loops were knocked off target and we disallowed using that data for tuning.

I guess the moral of the story for me was that the OEMs know a lot about what they are doing. Mazda is no exception to that. I would not necessarily trust their wide band data under extreme transient conditions but I would give it good consideration on something like a dyno run.

Does anyone out there have more experience with these things and have some information they would like to add?

 

Indeed

I would love to hear the verdict on this. The problems you describe sounds like they would affect an aftermarket sensor just as much?

 

I'm sure there is another "table" in the ECU tuning that corrects the WB02 signal..

I do know the Bus only transmits above 10.5/1 AFR.....

 

It works just fine within the range that the average person would need. I've seen people complain that it doesn't go X low or X high and that it gets inconsistent and inaccurate at these levels. I would argue that a properly tuned car never hits those a/f ratios anyways. That's just my opinion. Yes there are aftermarket units that are better. I personally don't think the average person needs them though.

 

From an OEM "insider" perspective; the existing federal legislation is so burdensome with regard to emissions-control that the OEMs find it mandatory to use the best O2 sensors they can find and afford in mass quantities. How could Mazda have one of the best OEM PCMs and expect it to do its job with less-than-stellar sensor inputs? How about the resolution of each sensor the RX-8 uses? I'd bet they are all pretty decent. With one exception perhaps; the knock sensor(of ALL things on a rotary!).

 

Is there any knock sensor that works on a rotary?

 

I think the one I did have sufficient pump current to go down to about .74 ish lambda (.74 X 14.7 = 10.88 A/F gasoline). I think we based that on roughly what the 4.2's data sheet said the sensor was good for.

 

The only person I have seen that really wants/needs to go lower is a buddy that does the 50psi boost methanol drag race stuff. Inlet air temperatures are over 250 F and he is running so rich the sensor just stays saturated. I do not think many are trying to cool their motor with fuel like that on the street.

I've used some really expensive ($2000 ish) NGK based units but really can not tell the difference. Maybe if I got a bunch of calibration gas and looked at sensors side by side I would see something.

 

Charles,
Thanks for the insider perspective. I've always pictured a bunch of really smart engineers at Mazda (and Porsche, BMW,,,) doing stellar work somewhere hidden away in a back room. It sounds like you just confirmed that.

 

The very best of luck to you in your endeavors. I like your approach!

 

I delved into this a few months ago by reading a lot of the various patents online. One that proved especially enlightening was this one:

http://www.freepatentsonline.com/5106481.html

You need a login to get into that site, but its free and IMHO very worthwhile.

I'm fairly confident this patent describes exactly what Mazda is doing for the factory setup. The patent shows an analog control circuit minus actual resistor and capacitor component values, but its enough to get the general idea of the control loop used if you can read analog circuits. I would imagine its much simpler to implement and describe digitally -- but obviously the guy writing the patent loved the opamp and creating discrete RC oscillators. Digitally, you'd have more freedom to tweak the control loop variables to increase and optimize bandwidth and response time.

The basic premise of the patent is a way to use a regular single Nernst cell O2 sensor to get reliable wideband data that otherwise requires the more expensive and fragile dual-Nernst cell "wideband" O2 sensors. It does this by switching many times per second between a sample/hold of the Nernst voltage, and then alternately sink/source current to electrochemically "pump" enough oxygen to maintain 0.45V (stoich) in the voltage sampling phase. The amount of current pumped corresponds linearly to AFR. Theres a sub-circuit that overlays a small high source impedance AC voltage signal over the current source/sink action and by measuring its amplitude, you can determine resistance of the Nernst cell and therefore sensor temperature.

After reading various patents, I was curious as to why aftermarket wideband sensors still seemed fixated on the fragile and more complicated 2-cell Nernst O2 sensor. My best guess was that design cost is less an object in the aftermarket and that a 2-cell control circuit would be a lot simpler and less risky to implement and get right-- but really I don't know. I'd love the opportunity to design this control circuit digitally someday-- it should be at most about $10 in parts cost in a box similar to the LM-1.

Another interesting note is that there is a circuit used on the E-Manage Ultimate kits for RX8 Greddy turbos that play around with the O2 sensor to force the stock ECU into open-loop. Greddy is obviously protective of the circuit as there is a layer of epoxy over the components of the circuit board. Some simple reverse-engineering revealed that all it was was a resistor and a solid state relay. The resistor makes the AFR reading as read from OBD2 go very lean, but what I really think may be happening is that it is confusing the RX8 ECU in thinking the O2 sensor is not yet heated up thereby forcing the ECU to ignore AFR data. I'm not positive on all this, but is fun to speculate on nonetheless.

 

I'll grab the analog portion of our LSU-4.2 and post it just for grins.

There are two parts. The first is a PWM controlled current source. The second is the feedback element. There is a third which is small that allows injecting an AC component into the Nernst Cell (above the natural frequency of the pump) to measure cell impedance (the cell acts like an NTC thermistor and thus impedance is how you measure cell temperature for regulation).

The funny thing about it is that the total cost is one inexpensive quad op-amp and a few passives. Heck, the precision instrumentation amp to measure Nernst current (using the factory calibration resistor thus avoiding the "free air" calibration) is more expensive than the analog stuff and fet to switch the heater combined (and probably the ATTiny that does the PiD loops as well!).

The best part about doing the PiD stuff yourself is you know when the senor control loops are on target and thus when you can trust the data!

I think the reason everyone uses the 4.2 or later units is cost. You can get them from VW for under $50 (I think, it has been a while since I bought some) retail.

 

Here is the analog section-

 

Wow, what a great thread! I use the LM324 quad op-amp a lot in my little pet projects. Its a great little op-amp and its 26 cents each even from high-price online distributors like digikey and mouser. You can even get it at the local radioshack in DIP form if you're desparate for about $2.50. I had no idea the 2-cell O2 sensors were so cheap ($50) -- I guess since the patents I've been reading are all about 15 years old, their claims of 2-cell sensors being expensive may no longer be as applicable.

I'm curious if you do ever find the bandwidth of the loops. Post some scope traces during transient conditions if I you can! I imagine the heater control loop is much lower bandwidth than the Nernst loop. I also imagine the nernst loop depends significantly on the age of the O2 sensor. I'll bet things depend less significantly on the electronic control loops as much as they depend on just the size and aerodynamics of getting a representative sample of exhaust gas in the Nernst sample chamber. I have no idea though really -- which is why testing any of this would be a fun experiment.

Also, FYI, that control loop circuit isn't doing PID control. You need a lot more op-amps or a MPU/DSP to do that type of fancy control-- another reason someone needs to move this to digital. In digital, you could use even neater and more fun control loops like fuzzy-logic!

This schematic is for monitoring a 2-cell O2 sensor. I'm pretty sure the mazda factory sensor is a single cell. That is what requires the slightly more complicated control circuit described in that patent. You could compete head to head with all those aftermarket O2 sensor manfacturers though with this circuit though.

 

The ATTiny (8 soic) is doing the PiD stuff and it spends most of its time idling. What I posted was just the analog front end. Kallman filtering might do a better job of predicting then adapting to senor response but the project hardly seemed worth that much effort.

You are right on the heater, the thermal inertia is reasonably high while the Nernst response is very quick.

As for competing, the circuit I posted is an evolution of the VEMS circuit. The one in the Innovate product seems to be an exact copy although I did not do a part for part comparison. Our mods were to speed up the Nernst current generating portion as we found the sensor would respond much faster than the control.

We really did not have any desire to build wide bands. The project was done to gain access to the real time data so we could evaluate if we were on target for both loops (or more accurately, identify when we were not on target thus disallowing the data).

 

They were up to $80 last year and I just paid $100 for one a few weeks ago.