Wide Ratio O2 meter

[garfield at pilgrimhouse.com]

On Wed, 11 Mar 1998 09:24:52 -0500 (EST), Frank F Parker
<fparker at umich.edu> wrote:

Hey Frank. Hey dudes.

I wanted to add a few centavos to these issues separately from me last.

> 4. What is not so simple, at least for those of us with no micro
>  design, is how you get a microprocessor to do this: you got to put
>  resistor in parallal, measure resistance and do math- I guess on a
>  rapid sampling basis so do not disturb sensor.
> 5. Once you have the sensor temp, you can now go to your 2D map of
>   air fuel vs temp of which we have 3 curves from above. Others have
>   to be calculated from interpolation if temp is not at 650,750 or 900.

I'm still not sure this can't be done in analog simpler than in micro.
Hope springs eternal. Humor me, and suppose we treat the complete
Lambada curve as one verticle shelf, with two logarithmic legs (hey,
remember it's the Nernst equation, "logs" ya know). Make the electrics
with a dead band for the stoich shelf (i.e., within a range of V around
0.45, it's all stoich to us). Outside this band, use two separate log
curves fitted to the two lean & rich sections of the LSM-11 curves
(ignore Temp for a sec, OK?).

Well, as much as we've all come to disdain them wayTooSimple LED meters,
remember NatSemi makes a LOG curve equivalent driver (LM3915) that can
be cascaded easily, so use a bar graph set of LED segments for each of
these log legs (and a single light for stoich). Knock yersef out,
neither the chips nor 10-segment LED bar graphs are expensive; use 20-30
segments on either leg! If ya cain't borrow a Bailey, you can always
hook up and try your friendly smog check guys for a run upNdown the
Lambada curve. Yeah, it ain't stunningly accurate, but it might get you
to within 0.5 AFR. I dunno, that's what I'm wiring up now; I'll keep you
posted on Garfield's Folly.

Now, back to Temp. In possession of three curves for temp, and IF (maybe
a big if) we can current sense the heater with say 0.5 ohms in series
(not gonna disturb nuthin) and derive Temp roughly from that, then use
this error term to bias the logs upNdown/inNout (actually, the rich side
should be the only sloppy log leg you'd need to tweak; usually the lean
side is tight with T). Easy enough for an eXperiment, anyhoo. I'm goin
for it.

Garfield