Inspired by the Autospeed articles on using a piezo microphone with amplifier and headphones rather than pipe, rubber tube and stethoscope and speaking to Sam Elassar, I thought that since I use a laptop when mapping it would make sense to use the microphone socket on the side and windows sound recorder or similar to capture the audio from the knock sensor.

I am concerned that the input impedance to the laptop microphone stage is too low and may load the signal too much so that the ECU cannot see the knock, so I thought about using a separate knock sensor. You could then get an audio-visual readout from the knock sensor and later play back the sounds and look at the traces in detail along with ECU logs.

Apparently the Subaru sensor works quite well, so I want to get hold of one and try it.

Alternatively a microphone pre-amp with high input impedance might be able to be used with the original sensor, but I didn't fancy building one because of the split rail power supply required for a bipolar amp.

Any comments/suggestions?

 

Input impedance is typically 100k from memory.

Check out maplin for a range of cheap pre-amp kits.

Active bandpass filter is easy, op-amp based, inductor, capacitor etc, check out the "Loudspeaker design cookbook" although I am sure there are plenty of other references out their.

Above book also talks about using a piezo unit to measure speaker wall vibration.

I also believe you can get knock sensors with 0-5v ouput, how they are tuned/filtered I do not know.

I also have an FFT based spectrum analysis as part of a loudspeaker measurement system, which gives real time readout, but only in dos with my calibrated microphone. I believe there are a couple of free software tools that do similar with most modern soundcards.

Sorry I can only come up with vague comments, but hopefully they are some inspiration.

Paul

 

John,

talk to pat about this, he went there and did that a fair while ago.

Moray
bbs.22b.com

 

John,

Haven't we been here before?

Andrew...

 

A simple stethoscope style listening device for knock is fairly good. I tried the electronic approach as per autospeed, and whilc it does work, it seemed to pick up a lot of noise from everything else engine related. It was a bit tricky to actually distinguish/filter out knock from other noises.

 

Thanks Andrew, was looking for that link again to help things along - but they used a cassette recorder - why not just plug into your laptop direct and get a live display is what I am thinking now?

 

When I get chance I'm going to try the 0.1uF and the 100K off the factory sensor, and the map I am presently running runs occasional knock correction - I will leave it like that to see that it still does so and that I am not loading the sensor.

 

Been there done that, Pat H built up this sort of devive a couple years ago but we soon decided there is no substitute for your own ears, don't forget there are more than the one type of det !!

 

Has anyone tried the HKS device for listening. I think it is a mic placed in the engine compartment, which can be transmitted to the stereo speakers, through a filter unit, to filter out most frequencies other thatn those around the det ones. ( ie high frequencies only )
Sounds like a good idea, but it costs over £200. Your ears are free...and havent let me down yet.. It may have been detonating just before it blew up, but i did hear it..lol
A stethoscope is handy coz it does block exhaust noises etc, so you are just listening to the engine.

 

Just did a search for knock detection on yahoo....
Anyone ever heard of this thing??
http://www.adrenalineresearch.com/race.htm
Plasma igniton system?

 

If you need to test it John give me a call - I had Pat (with electronic setup) and Bob (good olde fashioned stethescope) in my car listening to my 22B...they both went deaf, but managed to hear all sorts of different det before that happened. I started using Millers very shortly after that LOL (as an aside and with reference to the 'octane boosters just 'fool' the knock sensor' theory, I took Pat out again after starting to use Millers and problems gone away)
C

 

Trouble is the det sounds can be chords of various frequencies, unless you can tie down exactly what it is you're going to hear it better than see it on a spectrum analyser.

On the website with the .wav file linked above, you can clearly hear the det.

I wonder if looking at the electronics of the ecu would yield any clues as to how the ecu filters the noise.

Paul

 

2 things:

1. JB commented above that he believes the OEM knock sensor is quite good, however, when fitting a knock link an alternative (bosch?) sensor is used in a different location, is this because the replacement sensor is better or because the OEM location is poor?

2. How sensitive is the ECU to det? I assume this will vary from MY to MY but does the ECU hear and indicate even mild det but only retards ignition/reduces boost depending on other conditions such as load, RPM, knock count? or does the filtering only allow quite severe det to be recorded by the ECU which then does something about it every time? I suppose this would need someone who has looked at the code to answer - Paul!

I could just monitor the knock flag and ignition correction in the ECU but this might take alot of time as my car runs so well I don't get any det

Andrew...

 

Edited to say:

I just found a toyota based diag technical doc on http://www.autoshop101.com/ in which it descibes engine knock sensors as being tuned for 7kHz.

Paul

[Edited by Pavlo - 7/25/2002 10:46:05 AM]

 

The microphone may need changing but how about this ?

QN35Q FM-1 Bug £29.99 from Maplin.

Install under the bonnet and tune car radio to desired frequency...

I'll get my coat...

Thanks

Gavin

 

How about a sensor linked to a box of LED's which illuminate in ascending order proportional to detected knock....... could call it... er... 'knock-link'........

Sorry, couldn't resist

 

Hmmmm not a bad idea... but ive noticed it also flashes at cats eyes etc..., and similar bumps in the road..
How about a device linked to 30,000volts, and when it detects det, it sends 30,000volts up your right leg... This would cause you to let off, and prevent engine damage. ( or maybe cause you to put your foot down, crash, and blow the engine up )

 

You must have BIG cats eyes in Belfast !!!

Muscles tend to contract with electrical impulse so your power reduction idea is sound

 

The idea was to listen just using the laptop as an amp, but also to record and display the waveforms and compare with what the JECS was showing, particularly over 6000 RPM, not to make another "automatic" system.

I'll try it at the weekend.

 

I did a search a while ago, and found a program, which did listen, and analyze the sounds from the engine, but specifically lookig for knock. Cant remember where it was though. I did download the program, altho it was only a demo I think. I never had a mic for the laptop to get it to work, so just deleted it.

 

Not an ideal result. I seemed to pick up a lot of interference and crackles using this method, along with alternator whine - all a bit like a ground loop. The stethoscope and pipe is much better.

 

I was following this with interest

 

http://www.johnbanks.dsl.pipex.com/knock.wav

Interference - see what I mean? Retarding the ignition 5 degrees made no difference to the sound. Poorly named file because I couldn't hear any knock

[Edited by john banks - 7/26/2002 10:47:17 PM]