inductive vs optical

[Andrew Dalgleish]

On Wednesday, 6 March 1996 9:55, owner-diy_efi-outgoing wrote:
[snip]

>
> PS. Angle sensing using optical light/dark coding on a wheel is   
_really_
> dark ages stuff. Missed pulse/Long pulse detection is much better.
> At least if the sensor fails the whole system stops - what do you think
> would happen with a binary encoded wheel? Ugh! - makes me shudder!
>
> Michael Fawke
> fawkacs at ozemail.com.au
>
>

Make sure you're not confusing hardware probems with system design.

Sensor problems give you the same problems regardless of sensor type
(inductive, optical, ESP).
If a sensor isn't working, you don't know what's happening. Period.

Assuming your sensors are ok, the basic idea is to get the most
information from the minimum number of sensors (hence $$$)

Missing/short/long pulse detection needs either time or frequency based
detection. (Any other's I've missed?)

Car engines run over a relatively wide rpm range, so time-based detection   

is out.

Frequency based detection is usually done by a phase or frequency locked   
loop,
either analog or digital.

Either way, you still have a fairly complex filter to design to ensure
under/over-shoot doesn't cause false (extra or missing) triggering.

Multiple sensors (inductive or optical) remove the need for any loop   
filter
design. You do have to allow for sensor failure, and yes, you now have   
more
sensors which could fail.

If you use one sensor for angle and one for index, the detection is   
fairly
simple. Assuming 1 per (cam) degree, 1 for index, if I get more than 720
degrees without an index, or less than 720 on the next index there is a   
problem.

If you use two sensors for angle (eg quadrature encoding), you can test   
for
either of these failing as well. Ok, so if the engine rocks backwards   
when
you switch it off will confuse this, but you don't care at this point.
I'd prefer to have the sensor checking than to know how far my engine   
rocks
back :-)

As others have pointed out if you encode your cylinder number in binary
you are asking for trouble, Gray coding is far less error-prone.

I think it is easier to use a state machine which is advanced once per   
cylinder.
This can come from an extra sensor, or (IMHO preferably) from the angle   
divided down.
Again, check the cylinder count is correct every time you receive an   
index pulse.

Generally, I'd say an inductive sensor is more reliable than an optical   
sensor.
The sensor itself isn't affect by voltage spikes as much as a LED,   
although the
supporting analogue circuitry may be.

Unfortunately, it's not easy to design an inductive pickup which will   
work from
cranking speed through to red-line. An A14 goes from 200 to 10,000 rpm.
(The push-rods start to flex at around 11,000 rpm.) If you add   
missing/short/long
pulses to the design requirements, it gets harder.

I can buy off-the-shelf optical sensors which provide a TTL signal I can   
plug
into a computer. I can print my own encoding wheel and reduce it   
photographically.
(The laser in a printer may be accurate to X DPI, but check the size of   
your toner
granules.)

Development time is the single most expensive part of any design.
Hence three optical sensors (1 for index plus 2 for quadrature angle), a   
printed
wheel, an EPLD, about 2 hours of programming, and I'm done.

Incidentally at my last place of employment we used a proximity sensor   
which was
placed *inside* the magnetic field of a 500MW power station generator to   
measure the
stator to rotor gap for pole shift, thermal movement, etc.
The sensor cost around US$5,000. Anyone want to buy one of these for   
their car?
Regards,
Andrew Dalgleish
Axon Research, Pty Ltd
6 Wallace Ave,
Toorak, VIC
3142
AUSTRALIA
Tel +61-3-9826-5538
Fax +61-3-9824-0083