Widebands That You Would or Wouldn't Buy

[TeckniX]

Do they document their serial data format ?

I wish....

[toalan]

Ok, I will take that unit, your current unit. I will just use the linear output.

you got pm

[bmotorsports]

Considering the AFX is not a NGK made item, and is a ECM CO dervied product I would disagree there
http://www.ecm-co.com/

Great wideband discussion! The NGK Powerdex AFX was indeed designed and manufactured by ECM. The NTK AFRM was not designed or manufactured by ECM. ECM is the high end calibration hardware company whose hardware is ubiquitous in the professional calibration and tuning fields next to Horiba. The NTK AFRM is using a new sensor design that has no present replacement availability. The AFX is kept alive as the Ballenger Motorsports AFR500. The only distinctions are some additional sensor options (LSU 4.2, multiple NTK sensors) and another harness length option. The AFR500 and AFX before it have an excellent reputation with professional tuners and the NTK AFRM seems to be developing a positive reputation thus far. The NTK AFRM and AFR500 / NGK AFX are unique designs with different strategies towards the same goal.

[HelmutVonAutobahn]

I see that PLX has bumped their version up to "GEN4". Looks like it's just a firmware change to support the LSU4.9 sensor. I guess I'll order one to check it out.

[adamw]

This is possibly a question Toalan or Helmutvonautobahn can answer. I need the correct Ip/Lambda curve or table for a LSU4.9 sensor. Im not building my own wideband controller or anything silly - I need this info only to interface a Techedge 3H1 controller to a Land & Sea dyno via serial. The 3H1 only broadcasts the Ip over serial and not an actual AFR or lambda value (actually what they call "Ipx" which is Ip standardised with the calibration values applied etc). To get the dyno to read AFR I need to program it with a look up table but in my search I have found two datasheets from bosch that have completely different curves - 1 is similar to the LSU4.2 curve, the other is completely different. I might be reading these wrong but would appreciate any advice on which figures I should use:
https://www.rbracing-rsr.com/downloads/ ... _lsu49.pdf
http://www.bosch-motorsport.de/media/ca ... 659pdf.pdf

[toalan]

The curve is slightly different between the 2 documents for the rich side, for the lean side both curves look identical (within rounding error). I would go with the first document as it is closer to what I consider as a datasheet, the 2nd document is more akin to promotional literature.

What I would do is convert the pump current from the table into Ipx units, then you can curve fit lambda vs Ipx via excel/matlab/mathcad/etc.. break the curve up into 2 sections and do 2 curve fittings. One curve for rich, pump current <0, another curve for lean, pump current>0.

For both fittings a 2nd order polynomial will give you a RMS error of ~0.01 lambda, a 3rd order should net you very close to 0 RMS error.

If you are using a PC then just calculate lambda with a 2nd or 3rd degree polynomial using floating point precision. A look up table is only really needed for low powered devices without a FPU.

Now you might run into the problem where your calculated lambda is different that what is shown on the TE display (if it has a display), this would be due to the fact that the TE unit uses a lookup table and your PC is using a formula. In cases like this, if you are confident that you did the curve fitting correctly, I would take the calculated lambda value as more accurate than the lambda value based on a lookup table.

Edit: I think RMS error is not the proper term, it should be LSE I think.

[toalan]

I see that PLX has bumped their version up to "GEN4". Looks like it's just a firmware change to support the LSU4.9 sensor. I guess I'll order one to check it out.

To support the LSU 4.9 you need a firmware change and also you need to add a resistor, ~100k, between the feedback signal and Vdd.

[HelmutVonAutobahn]

They are at VDD=3.3v so, the resistor needs to be about 60k
Let's see if they remembered it

[adamw]

The curve is slightly different between the 2 documents for the rich side, for the lean side both curves look identical (within rounding error). I would go with the first document as it is closer to what I consider as a datasheet, the 2nd document is more akin to promotional literature.

What I would do is convert the pump current from the table into Ipx units, then you can curve fit lambda vs Ipx via excel/matlab/mathcad/etc.. break the curve up into 2 sections and do 2 curve fittings. One curve for rich, pump current <0, another curve for lean, pump current>0.

For both fittings a 2nd order polynomial will give you a RMS error of ~0.01 lambda, a 3rd order should net you very close to 0 RMS error.

If you are using a PC then just calculate lambda with a 2nd or 3rd degree polynomial using floating point precision. A look up table is only really needed for low powered devices without a FPU.

Now you might run into the problem where your calculated lambda is different that what is shown on the TE display (if it has a display), this would be due to the fact that the TE unit uses a lookup table and your PC is using a formula. In cases like this, if you are confident that you did the curve fitting correctly, I would take the calculated lambda value as more accurate than the lambda value based on a lookup table.

Edit: I think RMS error is not the proper term, it should be LSE I think.

Thanks for your comments Alan. The Dyno data acquisition software only has an option to use a lookup table for RS232 devices so that was my only option. I used the Ip curve from the data sheet that you suggested and feed back from the dyno user suggests his lambda readings are somewhere in the ball park.

[Hentai]

Picked up 2 AFM1000 units recently.