Flywheel/Encoder patterns

[RABBITT_Andrew at mv8.orbeng.com.au]

>> overflowing the counters all the time waiting for the next tooth.
>
>All my counters are sixteen bit.  But remember, the counters used for 
>driving the injectors and coils are not processor clock driven.  They 
>are driven by the crank pulse, so their count would never exceed 719.

I think someone may have covered this before, so bear with me;

In an ideal world your fuel injector pulse width is nominally constant 
for a given torque at any speed.  However if you're using your crank 
encoder to drive your counters, then your pulse width (in counts) will 
increase with engine speed.  Your fuelling resolution at low speed 
will be poor.

EG: at 600 rpm idle, you'd be looking at 'round 5-6 mg per cylinder 
per cycle.  You'll have a pulse width resolution of 500 us, where your 
idle pulse width would be about 2ms, therefore your resolution will be 
about 1 part in 4.  If you use a 1 MHz clock for your counter, you'd 
have a 1 part in 2000 resolution.  Much better!

How about instead, using a programmeable down counter which you load 
with your injector pulse width counter and trigger both the counter 
and the injector turn-on with your encoder/angle count.  Then the 
counter just counts to zero and turns your injector off.

Don't forget you'll might need some voltage compensation for your 
injector turn-on and off delays.  This will make resolution issues 
even more important to you.

Andrew Rabbitt

PS:  these numbers came from data just lying around.  Don't take them