🚀 go-pugleaf

Howdy.
I am involved with a control system design-an anti-skid
system for an aircraft.  This system depends on the output
from a DC tach generator located in the axle to provide an
indication of wheel speed.  These generators produce DC in
direct proportion to wheel rotating speed, and also produce
(undesired) AC noise, most of it proportional in frequency
to the speed of rotation.

The output of most generators is not too bad (noise levels
of less than 1 VAC compared to 10 VDC RgoodS output at 80
knots), and can be cleaned up with some fairly simple
adaptive FIR filtering.  

Note-The filtering is FIR because linear phase simplifies
the control algorithm and because of the stability problems
with adaptive IIR.

Some sensors, due to exposure to heat (up to 400 degrees
Fahrenheit) and wear and tear, output as much as 5 VAC.  The
output of these can be filtered using FIR or IIR methods
but the amount of filter roll off required and the low
cutoff frequency causes excessive delays and the control
algorithm breaks down.

Even with these bad sensors, the signal is highly periodic,
i. e. the output for each rotation is the same as the last
if rotating speed has not changed, and the DC part of the
output is still linearly related to wheel speed.  When the
wheel decelerates (goes into a skid) both the amplitude and
period will of course change.

It seems reasonable that the periodicity of the signal could
be used to advantage  and a valid representation of wheel
speed extracted without much time delay if the right
technique was applied.  Some possibilities could be
(1)  Using an adaptive filter which stores previous output
and compares it with current output.
(2)  Some type of Kalman filter.

We are early in the design and can throw whatever amount of
CPU resources at the problem that we need (CPU power is
cheaper than new tach generators), but of course simpler
solutions are better.

We have had dozens of accidents due to malfunctions of the
old system, and your help is appreciated greatly.  Please
reply by E-Mail.  If a good working solution is developed
and proven, I will post it.

Send e-mail responses directly to 

kingj@oodis01.hill.af.mil

-davone sends for john king