Wide Ratio O2

[garfield at pilgrimhouse.com]

On Wed, 18 Mar 1998 16:50:17 -0800, garfield at pilgrimhouse.com wrote:

>I'm waiting on paperwork; and
>either way, will be doing some eXperiments on the NTK gizmo this weekend
>(I've tried to check everything else out before firing up the "ion
>pump"). I have a nice trick display already made, and as soon as I know
>the calibration slopes for pump current (it's coming in the mail), I'll
>try it out.

Got more info from FrankP and thot I'd give y'all an update on what
we've found so far, by going over it together on the phone:

1.	Temp Dependency:
The SAE article Frank previously cited shows a +|- 0.2 AFR change over
the range of our interest, due to EGT AND battery voltage driving the
heater element (10V to 14.5V). Seems pretty astonishing, but it's there
in blackNwhite, with complete scatter diagrams. And if you want to
regulate the heater voltage (we're talking around 1.5A at 12V roughly, so
it can be done, but it's kinda a pain), then you can double this
resolution to +|- 0.1 AFR. You'll see from the next bullet item why I
think this is maybe wasted effort.

2.	Linearity:
The curves relating Ip (that's that pumping current thingee) to
AFR/Lambda are surprisingly straight on either side of stoich (each side
DOES have quite a different slope, though). Without using ANY multiple
piecewise-linear segments, just a straight line approx. for either side,
the bottom line is this: within a range of AFRs from 10:1 to say 19:1
(is that wide enough for most everyone?), the straight line is within
+|- 0.3 AFR of true.

Put another more "graphic" way, IF you used LED bar graphs and made each
element represent 0.2 AFR, you'd always be roughly within on element's
worth. For a range of 10 to 18 AFR, that would amount to two 10-segment
bars on either side of stoich, with a display resolution of 0.2AFR, and
an instrument accuracy fairly close to that. Not too shabby. [I have
something built like that right now that costs a pittance, uses 4
LM3914s and 4 ten-segment LED bars, fits in 3.5" X 5" X 1" palm-sized
box. Neat]. Just mentioned this for ducks; many other system design
possibilities, depending on the person's own particular app.

3.	Calibration:
even more astonishing is that the docs appear to say that the device
w/proper electrics, can be made to be "self-calibrating" because it
specs the Ip to expect when the sensor is operated at ATMOSPHERIC O2
levels!! (Yes, Virginia, it apparently WILL go THAT lean!). It even says
that the range around the expected Ip at atm. O2, which is 15%, is
exactly what is corrected for by the embedded CAL resistor that comes
from the factory with each sensor. This 15% range is reduced/corrected
to 1% when the CAL resistor is used in situ. What I think this MUST
mean, is that with the CAL resistor propertly used, AND the instrument
operating in ATM O2, you can expect the measured Ip to be within 1% of
the average/nominal spec'd value (happens to be 7.5ma).

4.	BUT, "Houston, we have a problem" ...
now the bad news (well, not bad, just a slowdown methinks). In the SAE
article, it reveals there is yet another pumping current, a "very small
constant bias current" applied to the Measurement Cell, called Icp (not
the same as the Oxy Pump's varying current Ip). All that the entire
literature says is it's "small" and "required". Beyond that, it's
obscure. Frank and I discussed this, and what Frank's gonna try to do,
since he has one of these sensors AND the NTK interface on loan, is to
see if he can't measure what this Icp actually is, since it's
essentially the current in the lead to the measurement cell's output
pin, when connected to the NTK interface box.

So we do still have a couple mysteries to solve, but it still looks VERY
promising.

5.	Comments:
If you're at all like me, you're thinking alot of this sounds a tad too
good to be true. All I know at this point is, if most of what we think
we know so far IS SO, then one thing for sure: Horiba and NTK are making
a HEFTY margin on them boxes they're selling!

6.	Futures:
I'm sufficiently optimistic about the outcome, that I'd venture it's
time for us to start individually thinking how our own needs/apps might
drive some other features we'd want.

I'll toss out two such examples to get things buzzing: those who will
need to data-log the sensor's/instrument's output might want to think
about how to go about this. The NTK outputs a voltage "curve" as an
analog of the Ip current; hence, it's sorta linear on either side of
stoich, but not a true linear analog of the mixture from say 10 to 20
AFR. Hence you'd need to do some lookup or interpolation whilst/after
logging the values. From what I can tell about the Horiba, it DOES do a
full-up linearization, providing a voltage out that's a true analog of
AFR. So think about how these two approaches might suit yous loggin
guys' needs. [BTW, the response time for the NTK sensor is spec'd in the
SAE paper for a couple of big AFR swings of like 12 to 18, 12 to stoich,
stoich to 18, and the response times are given in 100%->50% and 0%->50%,
as being around 300mS (yes, milliseconds), so it's probably too fast a
sensor to be converting directly to digital with a runOtheMill
dual-slope A/D. Just thot I'd mention that for your consideration, in
case you were thinking of using one of them mundane panel meters with
the digital outputs. Too slow, I'm afraid.]

Second thot is that many of us will want to have a UEGO as our only O2
sensor whilst testing, instead of having two bungs in our collectors
(geez that sounds disgusting, don't it!  B), so if that is one of your
goals, we'll need to consider "ordinary EGO emulation", to create/mimic
the output your ECU is gonna expect; ya know that El Capitan shaped
curve we've all come to know and 'love'. I don't think this will be at
all difficult, cuz detecting stoich point on the UEGO is bedrock solid,
since it's the point where the Ip current is ZERO, but it is something
to consider.

Fine'. OK, that's way long-winded enough. Gentlemen, start your engines,
er um brains (personal ECUs?) on what lays ahead. Always good to
anticipate the turns, dontchatink?

Garfield