Oxygen sensors

[Dirk Broer]

>I found an interesting GM patent (no. 4,130,095 Dec. 1978) that describes
>how they can tell of the sensors are hot enough to be given reliable data.
>The model of the O2 sensor is like a battery.  The terminal voltage is a
>function of the CO or oxygen content, and the internal resistance is a function
>of the temperature.  These things are really high impedance devices.  GM uses
>a clever approach that alternates between taking samples with a high impedance
>input, and one that loads down the sensor.

According to one of the shop manual - GM biases the O2 sensor through a high 
impedance with .450V.  The computer sees this and assumes either the engine is 
cold or the engine is in tune - and therefor makes no changes based on the O2 
sensor.  When the O2 sensor heats up the voltage starts to vary - the output of 
the O2 sensor is enough to exceed the high impedance voltage source.  With this 
arrangement GM has three error codes:
Too long rich ( computer cannot lean out the engine )
Too long lean ( computer cannot richen the mixture )
Too long with changing ( the O2 sensor is at or near .45V and hasn't gone rich 
or lean for a while - indicating bad O2 sensor or broken wire )

This assumes that the O2 sensors output will vary - and judging from what I saw 
with a DVM at idle it does.  I wouldn't try to close loop tight control with an 
O2 sensor since there are just too many variables - different cyclinders may 
have different effective length runners, different rpms and throttle positions 
may have different time constants and last but not least - as the engine wears 
you'd have to recalibrate.

According to the shop manual, when the computer is running closed loop and sees 
a rich or lean condition it incraments or decraments a ajustment variable.  
This variable is 8bit.  When the variable gets to close to an extreem - (255 or 
0) The computer adjusts the learned table and resets the variable to 125.  The 
adjust should be done to the NVRAM and is a learning function.  Personally, I 
think the computer should differentiat between what was learned at idle and 
what was learned at part throttle.

Dirk