DIY_EFI Digest V4 #699

[Greg Hermann]

>
>Date: 16 Dec 99 21:28:36 +1200
>From: "Tom Parker" <parkert at ihug.co.nz>
>Subject: Re: DIY_EFI Digest V4 #696
>
>Bobby Radomirov <supra90t at prodigy.net> wrote:
>
>>The O2 sensor varies the voltage constantly. Abouth 8 times per sec on a
>>good one and 4 times per sec one old one.
>
>It is my understanding that the computer varies the mixture, which the
>sensor detects and responds by varying the output voltage. This is quite
>different from the sensor varing the voltage on it's own!!!
>
>>The perfect 14.7-1 mixture is
>>about .480V anything that is under is Lean anything over is Rich.
>
>"Perfect", "Lean" and "Rich" are open to interpretation.

The ecu uses what is known as a "Bang-Bang" control algorythm with respect
to the O2 sensor, as opposed to a proportional (P), proportional-integral
(P-I), or proportional-integral-derivative (P-I-D) scheme.

What this means is that when the sensor output voltage deviates a
pre-programmed amount to the rich side of the stoich sensor voltage, the
ecu begins to lean out the mixture (at a pre-programmed rate), and then
when the sensor output varies a pre-programmed amount to the lean side, the
ecu begins to richen the mixture, again according to a preprogrammed rate.

The programmable variables include how FAR the sensor signal must deviate
in each direction before the ecu reverses its corrective action, and the
size and rate of the corrective action taken in each direction.
It_IS_DEFINITELY possible to change the frequency at which an ecu
oscillates the mixture about the stoich point (and how far it swings each
way) when the ecu is in "closed loop" mode by changing these control
variables in the program. The limit as to how tightly and quickly the
mixture can be controlled is the response rate of the EGO (or HEGO) sensor.

The advantages of bang-bang control are that it is a very simple
algorythm--and thus it takes up very little code/memory capacity, and that
there is no control "offset"(error, or deviation) from the setpoint as
system load varies. The usually quoted disadvantage of bang'bang control is
the inherent oscillations of the controlled variable about the set point.
However, in these (vehicle ecu) applications, that disadvantage has been
turned into an asset--the oscillations are said to improve the performance
of a three way catalytic converter, and now OBDII uses counting of the
oscillations as a strategy to detect system malfunctions.

Proportional control gives a much smoother control, but has the
disadvantage of introducing a control "offset", or error--the controlled
variable will err in the direction of the load on the system in an amount
proportional to the amount of the load on the system.

PI control eliminates the "offset" error that one ALWAYS has with P control
by correcting the system input based on the Integral of controlled variable
error over time--the longer an error in the controlled variable exists, the
greater the correction that is applied.

PID control adds another correction to PI control so as to avoid the
tendency for the controlled variable to oscillate about the set point
(something that looks sort of like a deteriorating sine wave shaped
oscillation around the set point when the controlled variable returns to
the set point from an upset. In order to do this, the Derivative correction
adds a correction factor to the control input based on the RATE at which
the controlled variable is returning to its set point.

The behavior of P, PI, and PID controls can be "tuned" by changing the
control constants which are applied to each of the three components of the
control scheme. There are even "self tuning" programs which are commonly
available for these type of controls.

However--the processor and memory demands of these types of control schemes
should be pretty obvious.

Regards, Greg