Narrow Band Support

[Fred]

I know I said I would never do it, BUT, I was talking to Abe just now and he pointed out that it could be done with a very small table, so I'm going to.

8 cells
16 bit contents
16 bit axis values

Narrow band curve :



My point selection (must stay within 0 - 2.0 lambda) :



A nice addition to the code base!

Fred.

[jbelanger]

I don't get it. What are you going to do with this table? The only real information you can get from a narrow band sensor is if it's lean, stoic or rich. Anything else is false information. Having a specific lambda for a specific voltage is only true for one specific sensor at one specific temperature and pressure.

I'm not against supporting a narrow band but this gives the (false) impression that you can actually tell a lambda value other than 1.0 using one.

Jean

[Fred]

Well, I know it's grossly inaccurate and that you are 100% right. Can you suggest a better way to map from an analogue curve to a set of realistic Lambda values or some sort of state mechanism? The intent here was that anyone using narrowband understands that it sucks but can log the output from it to see where the 1.0 crossing occurs and tune idle/cruise with it.

MS waste data space with a lookup table for it, this also gives a false Lambda value doesn't it? Or not? I've never looked closely enough and can't remember what I did see anyway. This seems like a compact way to achieve the same lie.

I'm open to a better solution soon if available but I'd like to log the narrow band on my volvo as I drive in reference to the other parameters that I'm connecting to.

If you've got something better, I'll code it up :-)

Fred.

[EssEss]

could be an opportunity for Jared to craft up a comparator. use a spare DAC to feed in your switch point, and hook the output up to a capture input and you can get ints/timestamps on state change and simple watch it in the log ? if no spare captures (I'm not thinking there is) feed it into a gpio. I'm not much interested in nb myself either. I'd rather feed in my wideband.

but, another cool trick of this is that you can tune for 1.0 (since you're doing real ve) - and then simply adjust the lambda table and it (should) be right on for that load point

[jbelanger]

Actually, MS uses a single voltage value as a target which usually is the voltage for lambda=1.0. This is basically the only thing you can do with a narrow band. And if you don't have a catalytic converter on your car there's very little point.

Actually, that's not quite true because you can use what ever fueling you get for lambda=1 and scale that to a better AFR value for either economy or power (neither of which will be at stoic).

Jean

[jbelanger]

I guess I type too slowly and didn't see EssEss post.

Jean

[jharvey]

I guess I could draft up something, I suspect the hall OP-amp wouldn't oscillate in a bread board. Would it need a DAC, or could a POT wiper work. Might be low tech, but I think Fred would be looking for a pile of readings at a set threshold. A screw driver is quick and simple to adjust.

I'm not sure I'm seeing what good data can be gathered from a Volvo using an existing ECU. I can see how one might use narrow band correlated to the injected fuel to make an estimate of the VE at a specific RPM. But with out knowledge of how much fuel is being injected in, I'm not sure I see what the goal data is, or what it would be good for.

[Fred]

Hardware is no good because I'm planning to use this tonight :-)

It is an analogue signal, let's read it that way and deal with what it means later(ish) I'll be recording the ADC values too, so no data will be lost, it'll just be easier to see on the tuner screen while driving.

Actually, MS uses a single voltage value as a target which usually is the voltage for lambda=1.0. This is basically the only thing you can do with a narrow band. And if you don't have a catalytic converter on your car there's very little point.

Exactly, they generate a value from 1kB of wasted table space, then use it with a target to control the AFR under certain circumstances. All I'm suggesting is using code or a small table and some math to do the same thing in *possibly* less than 1024 bytes. The MS way :

Code: Select all

many,
of,
these,
one,
seventies,
170,
170,
170,
170,
170,
165,
160,
155,
150,
147,
147,
140,
135,
130,
125,
120,
120,
120,
120,
120,
etc,
etc,
etc

Actually, that's not quite true because you can use what ever fueling you get for lambda=1 and scale that to a better AFR value for either economy or power (neither of which will be at stoic).

Thanks! I hadn't thought of that :-) Good idea.

Fred.

[AbeFM]

OMG missing out on some awesome mint julips/mojitos

You should be able to get a stoich mix, then, add 5% or 10% more fuel and hit a ROUGH ROUGH number. But, light lean burn for mileage you'llknow is right and somewhat rich you'll know it right

Good enough for government work.