Knock Sensor Sounds

The knock sensor is basically just a microphone screwed into the side of the engine block. You can actually connect an audio amplifier (or in my case, a tape recorder) to the output of this sensor and hear the pinging caused by detonation, along with all the other noises that an engine makes.  All of the sounds on this page were recorded from the knock sensor, using a conventional cassette recorder.

Index to this document


Engine Knock

Here is an excerpt from a .tlg file which I would consider typical (for my car anyway).  It shows some engine knock under mild acceleration.TMO Log

This log show second and third gear. If you play this sound file, which corresponds to this log, you can hear the same sounds that the ECU hears.  The pinging is audible just after the shift to third.  It has a fairly distinctive sound, that I would describe as similar to the "twang" that you can get from flexing a piece of sheet metal.  You will also notice a rather loud crackle around the top of second.  I hear this a lot, and I am fairly sure that it is electrical interference, rather than a mechanical sound from the engine.  I can't be certain that the ECU is actually hearing this at all... it could just be due to how I have my recorder wired.  You will notice that it does NOT show up as knock, even though it is quite loud.
 

Knock Frequency

There has been a bit of talk about what frequency knock is at. Or put another way, what exactly is the ECU listening for?  The more we know about this, the better we can predict what other sounds are likely to fool it.  I have tried looking closely at the sound file, to try to see some pattern in the audio wave that I could call knock.  But I could not find any such distinct pattern.  What I needed was a spectrum analyzer.

A spectrum analyzer is a tool (may be hardware of software) which shows the frequency content within a sound (or radio wave or whatever). Spectrogram is a freeware spectrum analyzer for the PC.  It shows a "3 dimensional" graph, where the X axis is time (same as an audio wave form viewer), the Y axis is frequency (as opposed to signal amplitude in a wave form viewer).  The "3rd" dimension is color, and is used to show amplitude or intensity.  A dark color (black or dark blue) indicates little or no content at that particular frequency and point in time.  A light color (mostly light blue in my images) indicates a stronger content.

Spectrum Plot

The spectrum shown here covers about one second of time, corresponding to roughly 1:18 to 1:19 in the TMO log above (right where the knock sum starts).  You can click here to listen to this portion the sound.  If you listen carefully to the "ping" sound, you will likely realize that it is not a single frequency.  You can hear it drop in pitch over the fraction of a second which it exists.  In this spectrum this would show up as a diagonal line, because the frequency starts out high at one point in time, then drops in frequency as time increases, then fades away in amplitude.  Look at the spectrum above, just to the left of the "+" which corresponds to 6KHz, 7.6 sec.  There is a light blue, slightly diagonal line there, and other similar lines in the first half of this image.  These lines correspond to the pings in the sound file. This is what knock looks like.

I should say that this is what I think knock looks like, because I have not been able to completely correlate the audible ping sound, to the knock sum from the TMO logger.  I often can hear knock (which is visible in the spectrum) which does not show up on the logger , and the logger often records knock that I cannot hear or see in the spectrum.  I have yet to figure out exactly what the ECU knock filter is looking at.
 

Fuel Turn On

Knocking isn't the only thing you can hear through the knock sensor.

TMO Log Most fuel injected engines will shut the fuel off completely when coasting.  This log file shows such a case.  The RPMs are in the normal driving range, but the throttle is closed and the injector pulse width is at zero.  At about 6:48 on this log, you can see the throttle tip in and the injector pulse width comes up.  This is what it sounds like just as the fuel comes on.  You may need headphones or good speakers to here it, as the relevant frequencies are fairly low.

This is what that sound looks like on the spectrum analyzer.Spectrum Plot

Look at the bottom half inch of the spectrum.  The left half of the image is almost empty in this region, but the right half shows strong "spikes".  Each of these spikes is a cylinder firing.  You can see almost exactly where the fuel turns on!  I measured the spikes at 11.6ms apart, which would correspond to 2586 RPM.
 

Accel Enrichment

I stated above that the spikes near the bottom of the spectrum were caused by a cylinder firing, implying that this is a sound produced by the combustion.  But I am speculating that the sound is really the injector ticking.

TMO Log

Here is an example of a gentle start from idle, and what it sounds like.  The injectors normally fire (open) in sync with the engine, at 5° before TDC on the compression stroke (note 1).  During accel, fuel is added simultaneously into all four cylinders every 10 ms, in addition to the sequential injection (note 2).  I don't normally log accel enrichment, so unfortunately that does not show in my log.  Listen to the sound (recording starts near the middle of the log above).  Hear the "zzzzt" staticy sound just as the throttle first starts to tip in?  Those seem to be our accel pulses!

And we can see the extra pulses in the spectrum!Spectrum Plot

This is the portion of the sound that covers this image.  This spectrum here is much cleaner than the previous examples because we are at idle, and the engine is much quieter at idle.  We happen to be idling at 1000 RPM, so that means the crank rotates once every 60 ms, and a cylinder fires once every 30 ms, or roughly 3 times for each 1/10th second shown on the chart.  But there is an obvious region here where there are three times as many pulses as normal, that is, once every 10 ms!  Not only that, but the pulses appear to be much stronger than normal, implying that all four injectors are firing simultaneously, just like the book says they should!

Also notice that there is something going on at the end of the image (right hand side).  The last three pulses look stronger than normal, and the very last pulse is only 10 ms from its neighbor.  My interest in this is that it is audible!  Even though it is only a few very short pulses, I can hear it in the sound file!


Spectrogram

Spectrogram is a freeware spectrum analyzer for the PC.
If you want to try this your self, you can find Spectrogram at: http://www.monumental.com/rshorne/gram.html  You can use it to analyze the .wav files that are contained in this document, or any other .wav files for that matter, including ones that you digitize yourself.
 

Recording your own knock sensor

All of the sounds on this page were recorded with a cassette recorder, by connecting the microphone input into the knock sensor wire at the ECU.  I have a 100K resistor and a 0.1uF capacitor in series to prevent the cassette recorder from influencing the signal that the ECU sees.  (Click here for more details.)  I then digitize the portions of interest on my PC, using the standard Microsoft Sound Recorder.
 

Microsoft Sound Recorder

This program comes with Windows 95 (and I assume other Microsoft operating systems).  Look under: Start/Programs/Accessories/Multimedia/Sound Recorder, or look in your Windows directory for Sndrec32.exe.  I am using a 22,050Hz sample rate, 16 bit, mono.


Notes

    1. This is roughly one full crank rotation before the intake valve even starts to open.
    2. See the "Laser/Talon Technical Information Manual", 1990, page 14-38 for timing diagram.  This manual is available on CDROM from http://manualcd.dsm.org.

Written and maintained by Bill Sundahl
Last modified on December 9,1999