Hi-resolution Crank angle sensor

[Paul E. Campbell]

Darrell A. Norquay wrote:
> > As a matter of interest the Nissan 300 ZX Twin Turbo distributor has 2
> > sensors in it. One reads crank angle from the 360 lines on the disc and the
> > other reads an encoded cylinder number from the 3 digit code. Japanese
> > engineering!
> 
> I was actually surprised that no one was doing this. (until now, that is) 
> This is a combination of a relative position sensor (the 360 lines) and an 
> absolute position sensor (the 3 digit coded portion of the wheel)  There
> are many examples of commercially available encoders of both types, but
> to my knowledge there is no combination of the 2 used in industrial
> position sensing.  For a DIS type system, should be easy to substitute an
> optical encoder for the guts of the distributor and get any timing accuracy 
> you want.  Optical encoders are notoriously fragile, however, and you may
> have a hard time finding one that will live a long and fruitful life in the 
> underhood environment...

Okay..I didn't bother to mention this because I figured it was more trouble
than it was worth but I was thinking in terms of optical encoders instead of
magnetic ones because optical encoders are infinitely easier to work with for
home construction because I can simply take a flat disk and drill hole patterns
in it with much better precision than I think I could get away with making
a toothed ring/gear.

As for survivability, I think the only way to get an optical encoder to
survive is to get it really good and sealed up. I'm thinking that the part
that holds the interruptor modules should be part of a shroud that totally
encases the optical encoder wheel except for maybe a hole at the bottom for
drainage and a hole for the shaft that is turning.

As for using the standard white/black markings, I don't think this is a good
idea. Putting a hole completely through a disk lets you put sensors on either
side and as long as the holes don't get clogged (easier to clean than
munging up a nice optical encoder pattern on plain white paper) is probably
just as easy to make. Also, standard emitter/receiver pairs are available for
about $3 (check a Digikey catalog) that are already set up for that
arrangement.

The advantage of the magnetic encoder is that there's much less opportunity
for the disk to get munged up since it is just a gear and will tend to
centerfuge off anything that isn't sufficiently sticky (assuming you have it
exposed). I don't know what your controller would think of having the
encoder immersed half in water or some really nasty mud though.

Now..to comment on a few other things that were said..the reason I was
thinking in terms of totally analog PLL's instead of digital ones is that
in analog, making up the various filters in the loops requires only a couple
fairly cheap op-amps. Reading the resulting output can be done with counters.
I was trying to come up with a way that the CPU could just read an external
8 bit register whenever it was ready and the hardware would handle most of
the filtering before it got to the CPU. I figured a simple 1 or 2 pole filter
on a PLL (which inherently cleans up noise and does the differentiation on
the input signal) would be a cheaper and simpler route than doing all the
data smoothing/cleaning in the CPU. You'll have to use an A/D on the error
voltage then (okay..you can do slope integration and such but it's still
A/D when you get down to it), but the loop does all the filtering. I was
still trying to think in terms of something that could be easily constructed
for the .1 kilobuck range. But..I just found out about the ADSP2181 this
week so I may be tempted back into the totally digital filtering realm (the
problem with the $10 ADSP2105 was the development system cost..the ADSP2181
development system is $90 which changes the picture significantly).