PDA

View Full Version : Knock Knock



Wodjno
03-10-2011, 12:03 AM
Does anybody know the answer to this question?

If frequency was displayed on a graph as Voltage x Time! Would the graph show sharper peaks for higher frequencies and more rounded peaks for lower frequencies :uhoh2:

or would is this not even possible :rolleyes4

Cheers

Wodj

Kenneth
03-10-2011, 02:00 AM
It depends on how the voltage is being generated. I know that is a vague answer but your question doesn't lend toward a more specific one.

If the wave form is not square then increasing the frequency would, in the same time scale, cause the slope to appear more pointy. However all else being equal, the ratio of time spent at the top of the wave form will not change and so any perception of it being more pointy is an illusion. By changing the frequency only, a wave form will not be come more or less pointy.

If you have a ball point pen and say the ball tip is the pointy bit, does making the pen longer make the tip pointier? What if you have to be further away to see the whole pen? (it would sure LOOK pointer) This is basically the problem.

The Vee
03-10-2011, 11:08 AM
If you're looking at a wavepattern like that music stuff, I think the peaks will be of similar shape as they are showing level (amplitude?). So a high peak loud and smaller lower peaks quieter. What does vary is the time scale so the width on the baseline (reading in time) will vary. A sharp or lighter tick will show a narrower base than say a heavy bassy type thump. Not sure if I've understood it right,..............but hey!

Wobble
03-10-2011, 12:04 PM
i little thing i just made up , hope its relevent . /lol

http://www.doctronics.co.uk/signals.htm

Wodjno
04-10-2011, 03:53 PM
Thanx for replys..

Knock Knock.. refering to engine knock /yes

Was wondering if it were possible to decipher engine knock from a live line graph of voltage x time?

So from the above and what i have read. Engine knock will be a higher frequency, than that of say suspension or exhaust knock! So on a line graph would show
to be narrower at the base line and the stronger the knock, the higher the peak? Also with increase of speed there is increased noise, thus increased voltage. So the baseline will not be flat and will slope up gradually as the speed increases.
If all of the above is understood correctly, i should be able to work out which peaks on the live line graph correspond to engine knock? And if i can do that? Then i would have to then try do work out what height of the peaks correspond to what knock count.
I suppose if i ran evoscan, side by side against the live line graph. Then when knock is detected via evoscan, then look at the corresponding point on the live line graph to work out the shape and height? Then i could chart it for future reference.
Thoughts please... ??

Nutter_John
04-10-2011, 07:55 PM
knock on the vr4 should have a peak somewhere around 6-7khz , you could use a band pass filter so that anything outside of this range is ignored

What are you trying to actually achieve ?

Wodjno
05-10-2011, 03:32 PM
knock on the vr4 should have a peak somewhere around 6-7khz , you could use a band pass filter so that anything outside of this range is ignored

What are you trying to actually achieve ?

Constant knock monitoring via my E-01.

Nutter_John
05-10-2011, 09:22 PM
can the the E-01 only monitor via a voltage ?

if so then what you need to do is build a band pass filter and then feed the output of that into the E-01 this will then measure the amplitude of the signal being generated by the knock sensor , bigger signal means more knock ( you will have to work out at what point the background noise stops and the real knocking starts )

another way would be to use a DSP running a FFT on the inbound signal and output a voltage based on the signal in , so give 1 volt for normal , 2 volt for moderate noise , 3 volt for err little bit worried now , 4 volt ffs will take your foot off the gas , 5 v do you want the phone number of a pick up truck as my engine has shat it's load

Patryn999
07-10-2011, 12:25 AM
Yeah, I considered doing this at one stage also. Problem is the existance of so much noise within the same frequency decade makes filtering quite difficult - lifter tick, engine mounts, injector noise - all sits within the same band. Basically means you either have to use something like stacked switched capacitor filters, but given that attenuation is dB/decade this gives you a lop-sided filter when considered in the linear time domain - meaning its still very difficult to get the necessary sideband attenuation without having a filter order so high that the circuit has instability issues. Or create something like a superheterodyne receiver - I had a simulation of one, but never got around to building it - wanted it to make a set of very narrowband "tunable" audio det cans.

Kinda saying that to do it in an analogue domain, with a low false-positive count, I suspect would be quite difficult. TBH if people on the forum know the ringing frequency of det the easiest solution is to do light filtering on input signal, feed it through to a basic microcontroller and do a cross correlation for the desired frequency.

Or the audio samples I was using (actually of a honda, forgive me) might've had a considerably higher noise margin than the knock sensor out of VR4. Be interested to know if anyone here has samples of det on VR4.

paulg23
07-10-2011, 02:25 PM
From what I read knock sensors are made to be resonant at knock frequencies so little processing is required. I might be wrong ( i was once /Banana ).
Might be worth looking at the megasquirt websites to see how they detect it.

Nutter_John
07-10-2011, 05:58 PM
A knock sensor is just a piezo crystal mic

Turbo_Steve
07-10-2011, 10:25 PM
True, John, but they do tend to be tuned to the 5k - 9k band. It's worth mentioning for knock monitoring that the relative amplitude of detonation, against anything else, is huge.
With the above bandpass of a regular bosch knock sensor, you're looking at a gain differential of 7 or 8dB for a typical knock event - to replicate this for testing purposes (and illustrate just how much force a detonation event creates against the piston crown) you need to hit the block with a lump hammer....and even then most ECUs would register this as "relatively marginal" levels of knock.

If the E-01 can measure between 0v and 1v then it should work with a regular knock sensor without any additional circuity. If it isn't that sensitive, a basic amplifier circuit should be sufficient for most purposes. Unless you're trying to tune to the "bleeding edge" of knock, in which case you'll need things extremely accurate (and one sensor per cyclinder so that the differential can be used to establish which sensor is registering the knock) otherwise, just monitoring knock to pull timing in an emergency......just use a £35 bosch sensor.

Wodjno
08-10-2011, 07:05 AM
Oooh... Lots of interesting info here /yes Some of it a little too complimicated for my likkle brain /Help

Hey Steve, long time no see Bud /yes

The E-01 can monitor anything from 0 -16v in 0.01 increments. So i can set it for 0-1v. Is 0.01v increments fine enough for a fairly accurate display?
Can i take the feed direct from the ECU ? Or would it be best to fit extra knock sensor and feed the E-01 that way?