Jump to content
Please check your junk folder for registration emails ×

DIY Knock sensor listening - detecting detonation


Recommended Posts


I started this thread due to finding a DIY knock sensor kit on the net here: http://home.netcom.com/~bsundahl/knock/sound/KnockSounds.htm

I got all the things i needed from jaycar.

Myself, i need to determine if detonation (det) is present simply to write off knock being an issue for my boost problem (of which is a work in progress)and i just dont want to pay a tuner $75+ to tell me.

Plus i have found with most tuners you have to talk to their belly button so they can hear you properly because their heads are so far up their own assss.

A knock sensor is in its simplist form is a microphone, youtube and the internet has loads of info on knock sensors and det, YT even has vids on what DET sounds like.

If you are one of these ball bags that bypasses the factory boost controller and cranks the boost way high with no other mods to compensate that extra boost, well firstly your are a ball bag and secondly, you can save some bucks and save your engine by using this DIY knock system to check for det.

Once i have proven the concept (maybe with kooms help as he has done this too) we will put up pics etc, but the above link will get you started.

Here is some info i found useful from threads when i first started learning about knock sensors:

FROM: http://www.iwsti.com/forums/water-meth-injection-nitrous-intercooler-cooling/162723-new-meth-failsafe-idea-using-knock-sensor-signal.html

-I'm not sure if the knock type of timing pull is exactly what you want. But, I'm not really the person to chime in on this. It may be that pulling timing in the manner that the other failsafes do it is gentler or something.

-There are essentially two types of sensors that the knock sensor could be. One puts out a voltage (probably from 0-5 volts) and the other is a pulse width modulation (remember morse code...it's sort of like that...sort of).

-IIRC the knock sensor on a subaru is a voltage (aka analog) signal of 0-5v.

-I think the ecu pulls timing only when it receives several knock events (knocks?). I'm not sure if a continually applied voltage to show one large knock would work. You may need to do several knock-on knock-off repetitions to get it to trigger a timing pull. I know on the hydra you get a knock level and can tune a knock table so this trigger value changes. Not sure about the stock ecu.

-http://www.northursalia.com/modifications/misc/wiring/04_STi_ECU.pdf says that the knock sensor is at 2.8 Volts at idle. I'm guessing this is your no-knock condition.

FROM: http://forums.nasioc.com/forums/showpost.php?p=3257963&postcount=4

I think the you believe there is some threshold voltage from the knock sensor that the ECU interprets as knock. This is not the case.

First, the ECU limits the times at which it montors the knock sensor to what are called "knock windows." A knock window is the period of time (measured in crankshaft angle degrees) during a power stroke when spark knock can actually occur. The ECU also varies the start/end of the window based on ignition advance and engine RPM. Only monitoring the knock sensor during knock windows allows the ECU to ignore the sounds of valves falling into their seats, which occur outside a knock window and are quite noisy.

Also, the ECU uses another sophisticated method to distinuish knock from noise. During the knock window, it computes two values: the average noise level and the peak noise level, at the frequency that is characteristic of knock in the EJ20 engine. The knock subsystem only "declares" knock when the difference between these two values exceeds a threshold, which can be varied based on RPM and load.

And finally, the ECU may sometimes retard ignition timing only after a certain number of knocks have occurred within some short time period. If no knocks occurred during that interval, the ECU might even advance ignition timing further.

Link to comment
Share on other sites

Originally this is the link i used http://home.netcom.com/~bsundahl/knock/listen/listen.htm#Recorder

I first tried using the knock sensor with no amp, it was just too quiet. I used an old knock sensor rather than tapping into my main one.

I wired it up ruffly for a start on my desk and used the free software called spectrogram to test it


Here is the amp kit you need from jaycar, as per this link http://home.netcom.com/~bsundahl/knock/listen/listen.htm#Recorder

Jaycar has all the stuff you need incl the wee box






Link to comment
Share on other sites

more good info from a thread here: http://www.honestjohn.co.uk/forum/post/index.htm?t=33615

Although the knock sensor is tuned* to pick up the vibrations caused by knock, typically around 4KHz, it is possible for a knock sensor to 'hear' other signals which the ECU interprets as knock.

Things that are loose, rattling or impacting together can confuse a knock sensor.

As the ignition advance map is adaptive, it is possible for the ECU to learn an incorrect mapping if the knock sensor is hearing things.

*The sensor may be tuned by torquing the attachment bolt is the bolt is mechanically in parallel with the piezo crystal. If the bolt is too loose, or too tight, the centre frequency of the sensor can be incorrect, so the sensor might pickup the valve train noise rather than knocking, for example.

Also, as the knock sensor signals are very small, they are easily swamped by electrical interference and noise. The noise may be interpreted by the ECU as knock. I have ended up fitting metal braiding over the knock sensor wires, effectively extending the ECU metal casing all the way to the sensor, in order to shield out electrical noise in one particularly difficult case.

Link to comment
Share on other sites

Well i went for a boosted drive with the my DIY system and spectrogram software.

Koom, you are right, the 9v batt is heaps better, no where near as much noise!

Managed to record the exact 30-40 seconds of driving where the boost kicked back to 10 PSI.

Certainly sounds like det at about 26 secs, then you can hear i simply slowed down cause my handy LED thats hooked up to my Boost solenoid stopped flashing and my boost gauge went to 10 PSI so i knew i was in safe mode. So i pulled over and stopped the recording.

Its the same thing thats been happening all along, at least i can replicate it at will i spose.

Any experts out there i can email the wav file to have a listen?

Link to comment
Share on other sites


Three techs have said frm the WAV file that its more than likely not det.

EDIT: i now know why you cudnt hear it that well, the in-put volume was too loud and recording to much other noise, once i turned it down and recorded again the det was very clear, as per later on in this thread

I have Got a torque wrench to torque the knock sensor to 24nm, this apparently is critical as is the 60 degree angle from the back of the engine.

Relocated knock sensor to up on the manifold on the IC bracket

Will reset ECU and test drive, if that doesent work, i will wire the following urber shielded wire straight to the ECU using the twin inner cores as one.


Link to comment
Share on other sites


Right, after putting the knock sensor up on the Intercooler mount and torqueing it correctly etc i reset ECu went for a drive and its no different, it still detects knock and puts the car into safe mode.

So, i took the knock sensor off completely and cable tied it to a hose, reset ECu and went for a drive.

Well it Deted its ass off, i recorded the whole thing on with my DIY knock sensor listener and laptop. This was a one off test and i wud not do it again after hearing all teh det.

So this tells me:

(1) with the knock sensor on it is doing a great job at protecting my engine from the det

(1.5) relocating knock sensors still work very well

(2) i have real det issues and now know/suspect the compression ratio on my EJ208 is well wrong for the V5/6 STI ECU and the timing is too advanced casusing the det. Mine is 9:1 and a v5/6 STi is 8.5:1

(3) i didnt thrash my engine so saved myself thousands from taking it easy right from the start

(4) my DIY knock sensor listener is worth its weight in gold

(5) my wife hates me

(6) i need to get an after market ECU

(7) the whole TT to single T conversion is a right prick and rather costly

If you want to hear what det sounds like i can send you teh wav file, it plays in windows media player, but its 12mb

This is what the det looks like in the spectrogram software


Link to comment
Share on other sites

 Koom said:

Could be cheaper to lower the compression via thicker headgaskets rather than ECU. Although ECU is a "better" fix. Was it you that was saying recently that your car is running quite rich? Although is it still rich after a re-set?

Yea, rich all round, reset or no reset. The chap i got the STI ECU off said it was rich ass for him too, suspect he wud had had similar issues or not even known he was in safe mode coz he had a MBC.

ECU is the best thing for me, i may end up being single after this LOL, and she might end up getting the car LOL (i bet she'd burn it)

Link to comment
Share on other sites

 Koom said:

What condition are the coils in? Sparkplugs good? Right gap? Tried Closing them down a bit? What sort of boost were you making when it was detting?

Not running coil over plugs any more, running single 4 point coil pack, prob still there with both setups.

Running heat range 6 tried 7 gap .712

Anything over 9 PSI and DET was happening

When i had that snapon engine analyser on, it was even detecting knock in safe mode!! and the ignition always looked far too advanced

Link to comment
Share on other sites

what area u from bro..

and maybe host the wave file somewhere so can be downloaded easy .. i wud like to hear what det is lol..

PS> just a theory but what if the factory knock sensor is buggered and causing false det noise - just wondering thats all

maybe try the wrx ecu and injectors and afm - i hear thay run higher compression than the sti

Link to comment
Share on other sites


[quote name='Koom said:

Did the analyser tell you approx how much advance?


i saw 40 at cruise

Seems high but did it drop back when loading it up?

Tried another ecu?

Didnt drop back as much as i thought it wud i remember thinking, and there was constant knock voltage.

Yea man, caspros kindly lent me his 00 STI RA ECU, no difference.

I have done the works, new AFM, Fuel PUMP, new o2, x2 cans upper engine cleaner, crank and cam sensor, another ECU, two knock sensors, 4 Boost noids 2 & 3 port, new vac hoses, etc etc etc god help me!!

Link to comment
Share on other sites

i just rang up Mitchell motorsport who built my motor, asking about what my compression would be and he said he didnt work it out and wouldn't know?? do you not have to do this to build a engine?

and then i asked about my low boost/boost pills in the right place and he reackons 7 & then 9psi is standard for a b4. pretty sure its not unless every other b4 ive been in wasnt standard even tho the owners thought is was...

told him what i was thinking sti ecu, single turbo etc and he said the difference between 9.0:1 to 8.5:1 wouldnt change to much and cause a det issue, more the ecu itself being more advanced or higher tune or some crap.

Glad i didnt pay for the rebuild or id be even more pissed off how he cant answer afew q's without being a wank. wont be taking my car there...

Link to comment
Share on other sites

Join the conversation

You can post now and register later. If you have an account, sign in now to post with your account.

Reply to this topic...

×   Pasted as rich text.   Paste as plain text instead

  Only 75 emoji are allowed.

×   Your link has been automatically embedded.   Display as a link instead

×   Your previous content has been restored.   Clear editor

×   You cannot paste images directly. Upload or insert images from URL.


  • Create New...