Sensor Linearization

[Darrell Norquay]

At 08:04 AM 6/13/96 -0600, P.P.F. wrote:
>> Subject: Sensor Linearization
>> The inclination is to accept the Map sensor data as well as the Tps
>> as they are linear even though there is a loss in range. 
>> One problem remaining is the MAT and the Coolant Temp Sensor.
>> Going by service manuals and using a 1.2 K pullup this gives quite
>> a nonlinear temp vs voltage.
>> The question I have for the good folks is how do you linearize the
>> sensor. Or is it necessary?
>> or a lookup and then interpolate. 
>> What I am trying to get is linear numbers between 0 and 255 for
>> a temp range of about -40 to 250 degrees.
>> Tnx again everybody: peter


First question to answer is: Do you really need to linearize the CTS sensor
at all?  If I remember correctly, AFR requirements increase rapidly with
decreasing
temperature below 10 degrees C or so, and are relatively constant around
full engine operating temps.  This, coincidentally, is exactly the way an
NTC thermistor's curve looks.  It's resistance changes more rapidly the
lower the
temperature, and it has a dynamic range of 10^5 ohms from -50 to + 100 C or
so depending on the initial resistance and thermistor material.

Secondly, if some linearization were necessary, do you really need the accuracy 
of a 255 point lookup table?  Changing the injector duty cycle every 10
degrees C would probably be good enough for calculating AFR's, especially
since it is
only important for cold starting and becomes moot one you reach operating temp.

As far as doing polynomial calcs on an 'HC11, forget it.  It's much too busy
handling interrupts to be doing complex math just for temp. comp.  Someone else
mentioned the BOSCH linearization routine, this seems like something that you
could use as a guide.  Of course, unless you get their exact sensor, the numbers
don't mean much, since thermistor characteristics and curves vary all over the 
place.  

MAT is another story, albeit a similar one.  Since the air temp used for air
density calculations is in degrees K, it probably has a much smaller effect
on the overall formula than you think, as well as the fact that air temps
usually don't reach more than +50 C.  -40 to +50 C is really 233 to 323 K, a
much smaller (percentage wise) change.  You could contact companies like
Fenwal Electronics or Yellow Springs Instrument about their linear thermistor 
networks, or for even more accuracy,use a linear temperature sensor IC like
the Analog Devices AD590 (bonus: reads in degrees K) or a National LM34/35
series, 
but now you've got to use an opamp circuit which has it's own temperature
stability problems.

Just my $0.25 ($0.33 CDN) worth   :)


regards
dn