Anybody ever put a scope on their knock sensor?

[amichie]

I looked at the signal on mine today and it seems very noisy.

[Jasper]

I have been recording mine (1UZ) with a notebook soundcard, and post processing the signal with adobe audition. This allow you (me) to isolate the characteristic knock signature. When compared to the datalogs from my autronic SM4, it is a useful tuning tool. I also have a Link knockblock to provide a 0-5V analog input to the ECU based on knock intensity, which I can use for closed loop timing or boost control.

[Jasper]

Some info here:

http://forums.nasioc.com/forums/show...0&page=1&pp=20

[amichie]

I have been recording mine (1UZ) with a notebook soundcard, and post processing the signal with adobe audition. This allow you (me) to isolate the characteristic knock signature. When compared to the datalogs from my autronic SM4, it is a useful tuning tool. I also have a Link knockblock to provide a 0-5V analog input to the ECU based on knock intensity, which I can use for closed loop timing or boost control.

Thanks Jasper

I will have another look tomorrow with the scope.
Just realised I had the scope on DC coupling. I will try again on AC coupling (since sound card inputs are AC coupled) and post up some signals.

[Jasper]

Here are some spectral graphs of the knock sensors' output on my 1UZ using a 180hp shot of nitrous and water injection. The first is the raw signal, note the fundamental frequency at 6800Hz, and the weak 2nd and stronger 3rd harmonics.

The second is the result of a bandpass filter, and discarding the lower intensity portions of the signal. This isolates the knock events so the can be clearly seen.

[Jasper]

Here is the filtered signal overlaid on a log of some of the engine parameters. Note: the time scales do not exactly match, but is is close enough to verify that the knock events shown in the filtered signal do indeed occur when we would expect them to.

[IN 05 NT]

alot of mumbo jumbo 2 me!.....but foook...good technical skills...

[oldcorollas]

that looks good, i was going to reply before saying the knock sensor is just a piezo microphone, so it DOES just pick up noise... lots of it... but your analysis shows it perfectly!

how did you determine at what particular frequency the knock events occur? did you induce them under low load to see?

[Jasper]

In an attempt to find the resonant frequency of the block, it made some recordings of it being tapped with a hammer while the engine was stopped. This gave a fairly sharp signal centred around 6800Hz.

From what I've read, there are two types of knock sensors. Broadband knock sensors pick up all frequencies, and it is up to the control system to isolate the frequencies of interest.

The other type is designed to have massive gain at the frequency of interest. I'm not 100% sure, but I think the 1UZ has this type, because of the amount of output at the 6800Hz range and the first two harmonics. The research I did indicated this (6800Hz) is pretty close to typical knock frequencies.

Because of the signal intensity at this band of maximun gain, I actually got better results from using a bandpass filter at a slightly higher frequency that this. There was a much better contrast between knock events and the background noise.

If you look at the raw signal, you can see a faint trace of the rpms rising and I took off from a standstill, like a little bit of a sine wave in the bottom lefthand corner. The peak of the wave is where the rpms dropped as I fed in the clutch.

After that there is too much noise to really notice anything.

I'll post some of the sound files so you chaps can hear what knock sounds like...

[1jzracing]

i have played with logging the voltage output from knock sensors to a logger and there was a lot of just random noise lines at first and they sort of followed the rpm, if knock occured they rocketed up but it was all i bit messy

i then tried using a diod to clean the signal and the outputs went up so much (7-8volts) they were outside the range of my 0-5v logger

but importantly it now has a nice clean graph line, i still need to lower the voltage with a couple of resistors to get it back to within the range of the logger and then try to destinguish knock from engine noise

so in conclusion i should be able to identify knock as any significant changes in the graph line outside the normal line which mimmicks the rpm line

[amichie]

I found these instructions on a Supra MKIV website.
Gave it a try and my knock sensor signal looked exactly like the two graphs in the instruction except for a marginal increase in noise.
I have loosened and retightened my Knock sensor and got a slightly cleaner signal.
I'll have to wait and see if it makes any improvement.

[Jasper]

http://home.people.net.au/~indecent/2006-12-23/


The .wav files at the URL above correspond to the two spectral graphs I posted earlier.

[Jasper]

Does anyone have access to that section of the TSRM for the 1UZ?

[MacroP]

Hey, cool thread.
Got my Fluke 192 in the mail the other day and have been playing with it a little bit on my 1UZ - ignition, o2 sensors etc etc. The knock sensors is one thing I haven't looked into yet. Once I've figured out the Flukeview software I'll try and post some results. Cheers.

[oldcorollas]

there are some REALLY good sites out there about detecting knock.. and a couple of Theses iirc..
goes into the origin of the knock signal, what kind of filters are needed (both frequency, and time... since knock only occurs in a certain time period for each stroke)
i will have a look later for that particular paper, but if anyone wants to look for me
basically, interpreting knock signals aint easy