Ad: VTT N54 Built Motor Knock Sensors

Rob09msport

Major
Oct 28, 2017
1,919
Monroe CT
Exactly simpler way so everyone gets it is if you have noise at 10 khz and it follows a rhythm for very short interval it can pollute a 100 hz signal. Digitally if you isolate a 10khz signal that lasts 1 hundredth of a second that will grossly exaggerate the amplitude of an actual 100 hz signal which a dsp can do as long as the resolution is high enough however it is def better to eliminate as much garbage noise before it gets into the system. Gigo garbage in garbage out is the proper term for this situation I belive.

I am not arguing with the dsp when that's developed great cause if it's powerful as is said then can easily filter out alot of junk ahead of time.
 

Jeffman

Captain
Jan 7, 2017
1,395
Out of curiosity I decided to do a mini patent search to see what’s known these days about knock sensor design and dsp methods.

This US patent from 2007 (originally owned by Freescale Semiconductor and now owned by NXP USA) is directed to the noise problem by using a short term Fourier transform to extract the digital signal parameters:

This Japanese patent application (not sure of the company) is directed to a non-resonant knock sensor capable of flattening detection sensitivity in a vibration frequency detection band.

I’m not sure whether these knock sensor inventions were actually put into commercial use, but reading these patent documents do give the non-expert an appreciation of the technical problems that need to be overcome in sensing knock in ICEs.
 
  • 1Like
Reactions: doublespaces

noorj

10 Sec Club(N54)
Jul 12, 2017
137
Detroit, MI
@Jeffman nice research!

FT plots (fourier transformations) are sometimes used but only really needed when little is known about the engine so a wide range of data is needed. You get a nice residency plot of amplitude (acceleration g's) vs vibration frequency which you then apply a filter to only look at knocking points above X threshold. Comparing this to a non-knocking point to remove any engine induced noise, you now know what frequencies (generally second and third order) to use for knock detection.

In my field though, simulation will have a good idea on the knock frequencies so validation testing is all that is needed. Then you just need to calibrate the detection window, and sensor signal amplitude vs. knock amplitude (units of pressure). I've never heard of someone having a patent on the FT plot method though- it would be like having a patent on using differential equations to find stress in a connecting rod.
 
  • 1Like
Reactions: jyamona

Jeffman

Captain
Jan 7, 2017
1,395
Thanks, @noorj.
I've never heard of someone having a patent on the FT plot method though- it would be like having a patent on using differential equations to find stress in a connecting rod.
Patenting methods is actually very common. But that first patent doesn’t cover the use Fourier transforms as broadly as you suggest. The patent rights are written in the claims - and those method claims require a number of steps to be done to be protected by the patent. Happy to discuss this with you further in a conversation or off line.
 
  • 1Agree
Reactions: fmorelli

Rob09msport

Major
Oct 28, 2017
1,919
Monroe CT
Thanks, @noorj.

Patenting methods is actually very common. But that first patent doesn’t cover the use Fourier transforms as broadly as you suggest. The patent rights are written in the claims - and those method claims require a number of steps to be done to be protected by the patent. Happy to discuss this with you further in a conversation or off line.
Hats off to you I thought that but wasnt sure enough to make post. I feel like I needed to make a flow chart or something. It does seem like it is more applicable than it looks at first.
 
  • 1Like
Reactions: fmorelli

Top