[Land-speed] Knock sensor placement on flathead!

Kirkwood saltfever at comcast.net
Tue Jun 2 18:42:04 MDT 2009


Greg: Interesting idea and thanks for the posting. 

You have correctly identified the real issue and you have a much better
grasp on this than I do. My $00.02 is only the logic of your effort. Knock
sensors are seriously problematic in that they report all ''noise''. Noise
being defined as ANY signal other than the PRECISE SIGNATURE you are looking
for. Because of this you will spend copious amounts of time trying different
locations or positions, and hair-pulling sessions trying mechanical
isolation or dampening of extraneous signals. Why bother? I would just
assume there is noise in the system and go into partnership with a PhD who
understands Fast Fourier Transform (FFT) signal processing. Seriously, a
smart software packages (as you mentioned) that is industrial strength with
robust signal processing is more practical. You can chase mechanical issues
forever. Just accept them and filter them out.  Put a copper strap
longitudinally along each head and use two sensors (one on each side of the
engine). Filter the noisy data, looking for the exact signature. There is a
wealth of information on detonation signatures, one being Heywood pages
450-465. The problem with OEM sensors is that they don't have the computing
horsepower to be extremely sophisticated. With your experiment it takes
little effort or cost to improve on the OEM stuff by orders of magnitude.

http://www.dspguide.com/ch12.htm

 

Mayf: Thanks for the rule book point-out that brought a subsequent
clarification. 


More information about the Land-speed mailing list