Ok I'm just looking for some general info here but I expect this thread to turn into a larger discussion because from what I've read WB to NB conversion discrepancies as seen by the ecm haven't been adressed much... if at all.

The general consensus seems to be that although WB sensors and their associated hardware are much more accurate and process at much higher rates- therein lies the problem. Our ecm's just can't keep up with the WB's output and thus overcompensate. Furthermore, when a WB is used in place of a NB output can fluctuate beyond our ecm's preprogrammed limit, cauing the ecm too 'see' it as an o2 sensor failure. So what are the advantages of using a WB instead of a NB (except as a separate datalogging tool) if the ecm can't use the data it recieves? Converting factory NB to WB would be a huge advantage, has anything been done to fix this?

 

Anyone? I can hear crickets...

 

I don't know if this helps you, but I read in the 0-5v WBO2 signal and convert it to a NBO2 signal using the ECM. The conversion table can be tuned to anything you want. It allows the WBO2 to act like a "perfect" NBO2 sensor.

I think the WBO2 controller conversion to NBO2 isn't working the way you need it to. If you do it in the ECM then you can apply the correct filter and translation table. You can never have a too fast update/sampling rate.

 

I'm mostly repeating what I've researched so far (and probably not describing it correctly either) so if it's wrong correct me.

What I've found so far is that when WB to NB conversion hardware like innovate's LC-1 is used to simulate a NB signal it translates linearly. When the ecm compensates to correct a rich/lean condition the WB compounds it by registering the A/F change linearly instead of 'switching' like a NB does, so the ecm continues to richen or lean until it recieves the rich/lean signal it expects. The delta A/F change can be enough to make the ecm see it as an o2 sensor failure, negating the filter coeficient.

So you're saying I can change the filter coeficient and o2 sensor values to make it linear? That makes things sooo much easier! What are these tables listed as? I can't find them in tuner pro.

 

the stock lc-1 analog output 1 is a simulated nb and is not linear... the sensor itself is more accurate and will send more accurate data to the ecm. the analog output 2 is a linear 0-5v signal but you can change the voltage values of the output. so 0-5v linear can be changed to 0-1v linear. thats what i did to turn my nb air/fuel gauge into a wideband. you may have to slow the sample rate on the lc-1 if your going to do that. from what i hear though your supposed to use the analog out 1. one of the guys on austinthirdgen said its the best his car has ever run.
hope this helps

 

It does help, but what are the tables listed as? Something like 'O2 filter coeficient' and 'O2 mvolts vs A/F' maybe?

 

im thinking your talking about tuning the ecu itself to adjust the incoming voltage and that i have no idea about... i honestly didnt know that it was even possible. what i was talking about was on the lc1 itself you can change the voltage of the outputs. so analog output 2 is 0-5v you can change the lc1 programing to be 0-1v. its all point and click no tables or buffers. are you even using an lc1 or are you just trying to figure out a way to hook a wideband o2 sensor dircetly to your ecu?

 

junkcltr said that I could change the values in the ecm itself. I don't have an LC-1 yet, like I said I'm just researching to find out if it's possible to hook it directly to the ecm. If I can't change the individual values that presents a problem because the ecm will still be looking for the values it expects to see from a NB, which could be way off from the values the LC-1 uses.

 

I think you misunderstand the relevance of using the WB as a replacement for the NB.
The NB sensor is not good at setting anything other than a switching point around 14.7
It can be slightly offset to work with OK results as I and others have done.
The farther away from 14.7 (and not that far I might add) the results of sensor switching gets unpredictable or just plain non-repeatable.
The WB can output a "simulated" signal that is the same "range" AFR as the NB sensor but is more accurate in its actual switching point.
The issue then becomes, can the simulated signal "switching" be adjusted so that the fuel control will respond correctly to its signal.
The switching could be too fast or too slow for the control to be effective in making adjustments to the fueling.
To correct for the speed issue there are "filter coefficients" within the operating code that can make the input signal "change" faster or slower.

From there, any AFR (within reason) could be commanded and measured with the closed loop control.
Hope that gives you a better understanding of the problem.
Here is another thread on commanding different AFR's in closed Loop.
https://www.thirdgen.org/forums/diy-...-richer-f.html

Now on to Junkcltr's thing
I've always wondered about taking the input for the WB and then having a specific offset value added and subtracted from it in each loop. Pass that through a filter with its own coefficient to provide a switching function that would still work with the stock tables in $8D.
The stock operation needs to see the crossing of a window boundary to keep the Integrator from making fuel changes if the AFR is correct.
Make an extended RPM table (or use an existing one) to provide for the PE mV settings for each RPM range to make "windows" from.
Is this the kind of function you created?

 

I understand why a WB is better than a NB. NBs just 'switch' at stoich, allowing the ecm to find stoich but not much else. The WB on the other hand is more accurate at stoich and can register AFRs other than stoich with reasonable accuracy.

What you've said clears up issues I had with part throttle, but my problem is still the accuracy of the AFR the ecm sees at anything other than stoich. I assume there's a difference between the voltage the WB puts out at a particular AFR (even after conversion) and the AFR value that voltage holds to the ecm and I'd like to correct that.
Ex. Say the ecm's in PE and the WB puts out .5mv at 13.5/1 AFR, whereas the ecm is programmed that .5mv is 12.5/1 AFR. It ends up running leaner than intended so the voltage value needs to be corrected. I don't want to change the AFR the engine runs at, just the accuracy of the AFR it sees.

 

It isn't that bad to do it in the code:
Read in the WBO2 input
do a "coma" to invert the WBO2 value to read like a NBO2 value in terms of AFR
Subract the inverted WBO2 from a 2-D lookup table (it has the S-curve shape)
Now you have a NBO2 looking output
Do your filter on that.

Make the 2-D table an XDF table so you can "tune" it.

One of these days/weeks/years I am going to change the code to work off of the linear WBO2 for closed loop, but the translation works well for now. Been having too much fun tuning the turbo truck lately. Playing with a stock 305ci TPI engine with 130K miles on it, a big intercooler, and water/alky injection controlled by the $8D, and a couple of turbos has kept my busy lately. It is a lot of fun. Tweak here, tweak there, small code change here and there....87 octane at 12 PSI with 9.5:1 CR is a handful in a 5000+ lb rig. I will get to the WBO2 linear closed loop code eventually.

 

I believe that is exactly why the sensor simulated signal is not linear. Otherwisethe reaction would just not be the same "time" and "magnitude" that the control is setup to operate on.
The relationship in the code (I think) is linear in nature but has some skew to the response so that it runs richer instead of lean. Speaking on the error correction routines having a multiplier for one direction and not the other.

The only other way is like Junkcltr stated by making a lookup table with the "S" curve on the response you want and tweak from there to make it closer.

I guess now my questions are...
Does the code really take into account the "S" curve of the sensor?
The o2 target tables do but besides what I stated above, My guess is no.
So does following a switching "curve" or following a switching "line" have the same effect if you imput a "good" reading value?
Tired and thinking quickly so hopefully that makes sense to you.
Jp
----------
One just wasn't enough?
Remote mounts or right up in there?
I'd love to go boosted soon, Cost of air is still prohibitive

 

JP86SS,
The O2 switch tables kind of take the S-curve into account, but not much. You only see less than 100mv change in the tables for switch point vs. MAP & Airflow. The WBO2 -> NBO2 conversion uses a 2-D lookup table to make the linear WBO2 into a NBO2 value. You invert the RAW WBO2 A/D. Then subtract off a value to get a NBO2 value. I do a 16 bit subtraction.

I made a spreadsheet with a column of RAW WBO2 voltage, then NBO2 voltage. More columns converting to A/D hex values. I then do a multiply by 1.6 to the inverted (coma) RAW WBO2 A/D value and get a 16 bit value (FDIV in the code). I take that hex value and subtract it from the NBO2 hex value to make the 2-D table in the spreadsheet. The tables gets put in the code.

NBO2 = ((1 / WBO2_A2D) * 1.6) - 2D_table;

Sorry for the slow response.

Yeah, the turbo rig is fun. A couple of T3s under the hood. It wasn't that bad. I did all the work myself and the total was $1600-$1700 with a big IC & water/alky. It all sits on a stock 305ci TPI...the ones everyone chucks in favor of a 350ci.

 

I tried out a JAW WBO2 controller today. What a noisy/fast responding bugger it is. I had to put a low pass (lag) filter with a .125 coefficient on it to take the transients out. It reads very close to what the DIY-WBO2 read. Seems like it is a good kit for the money. LSU sensors are cheaper than the NTK sensors that keep going up in price. I figured it was time to move on since the NTK can't last forever with water/alky being sprayed. Oh yeah, another plus is the JAW is a DIY KIT.