Build my own a few days ago with LM324, and followed the usual methods of epoxing the electret condenser mic onto a 20A clamp. Primary purpose is to replace the mechanic's stethoscope, but in this case, I'm testing it's functionality to detect knock.

Problem is, it keeps picking up outside sounds no matter what,
Sound clip attached shows that it picks up the sound of my ITBs and IACV whistling, and perhaps the LMAS, and when it goes into vtec, the sound actually diminishes a lot.

Was a piezoelectric mic a better choice after all when making an stethoscope to monitor for knock?

Or is it that I still haven't sealed the mic enough on the clamp?

Help appreciated in suggesting better way of monitoring knock or creating a better pickup as alternative to the electret condenser mic for knock monitoring.

I am prepared to build another seperate box with a piezo mic if it's confirmed to be more accurate or less sensitive to outside noise.

https://dl.dropbox.com/u/33131116/engsounds.mp3

 

The OEM knock sensor is piezoelectric, so it is pretty effective. Some noise can't be filtered obviously, but it sounds like you haven't properly filtered for the frequency you need. In which case you should install a band pass. You can purchase a kit or parts for pretty cheap and designate your own frequency.
You can also add noise gates to clean up the unfiltered range, but that can get pretty technical.

edit* you might also try insulating the device everywhere other than the immediate pointa of contact, including the wires. rubber paint, foam dip, shrink wrap.

 

For a cheap DIY amp, I think that sounds fairly decent. An actual knock sensor will definitely be better than a cheap condenser mic.

I think you can get rid of a lot of unwanted noise by isolating the mic more from the engine bay. It also sounds like the gain is too high.

On the electronics side, even a simple high-pass filter will make a big difference. A basic RC filter might even be enough depending on how obsessive you are.

Mine started off with the usual det can path using a crushed copper tube attached to a Teflon hose going through the firewall. However, instead of attaching earmuffs on the other end, I attached a mic. The sound quality was ****, but I could pick out knocking fairly well, and it really encouraged me to pull timing out of my maps...

Here's a sound clip: http://s59.beta.photobucket.com/user...rive1.mp4.html

That was just using a Radio Shack personal amplifier with a 3-band equalizer. With better filtering to get rid of the injector noise, it should be easier to hear. It sounds better in person, without having to further butcher the sound by passing it through my laptop's crap sound card.

Like you, I've been working on a design for a better listening device, but it's still in progress (and at this rate will still be when I'm 60).

 

Isolate your sound to the frequency of the knock or come as close as you can. The unwanted noise from the other sources should be avoided by shielding the wires to and from your sensor. Copper wires are terrible.

 

it's pretty simple to calculate the frequency you need to isolate, which should also dampen the injector noise.
http://www.phormula.co.uk/KnockCalculator.aspx

 

Thanks for the input guys, I am on the way to building a band pass filter and using the stock knock sensor as pickup.

Problem is figuring out the noise to filter out ATM.

 

Pretty sure you'd have to convert to digital and program duration and amplitude conditions, which is really outside the scope of knock ears to filter that heavily.
It would be easier to just reverse engineer the knock board, but that's a topic for PGMFI.

 

A simple bandpass filter can be entirely analog and wouldn't be that hard to implement. It might not perform as well as an extensively developed digital alternative, but it would still be a big improvement.

 

Ran across this thread that uses a JY VW KS with a set of ear phones.

http://www.turbo-mopar.com/forums/sh...light=VW+knock

Mike

 

still noisy though, I don't believe the sensor is filtered.

I think adjustable bandpass filtering is ideal, given every engine has varying levels and frequencies of noise.

 

It isn't filtered. Those folks are using it as an electronic stethescope to listen to what's going on. They report the ability to hear knock before it's even picked up on the KS. Lifter noise, piston slap, and everything else is picked up. I just thought it was cool what they were doing.

Mike

 

since knock sensor is a microphone why not wire it in to the laptop's MIC input and use a real time spectrum analyzer while tuning. also could be routed to your speaker output and filtered with bandpass settings to isolate the knock frequencies.

 

I like! This group says the VW sensor is plentiful in the bone yards and cost like $5 each or so. Would be an inexpensive experiment.

Mike

 

challenge accepted.

I'll try to develop a standalone app.
TBC

 

one thing i dont know is how it is processed through the ECU.

does the knock board provide waveform output to an RTP or is it just knock or no knock and autmatic ignition retard?

if it did then a program like crome or neptune could in theory develop its own software based knock analyzer .




or vegas if you do develop it then speakers could be used to make some kind of warning signal when knock is detected.

 

I've been toying around with this myself for the last few months actually. I have a couple of knock sensors that I've characterized in labview. The stock Honda sensor is actually pretty good, but I think I'm going with the vw one. Trying to come up with an arduino setup that will log the knock along with cam and crank position so I can output a signal to Neptune to log knock along with what cylinder it is occurring in. Sample rate seems to be the biggest hurdle.

 

What u mean about sample rate?

 

has anyone try to modify stock knock board?

i thought i saw one of the guy sucessfully did but he said to calibrate it requires alot of measuring equipment and has to be done for each individal engine. he could have made some money lol.

 

You have to convert the analog outputs to digital to process them. You have 200 ignition events at 6000 rpm in a 4 cylinder, and the knocks happen quickly. If you sampling rate isn't high enough, they'll fall between the cracks.

Another simple solution I'm considering is just using a knocksense for megasquirt. It will output a zero to five volt signal based on intensity that you can run to one of the analog ins on a demon or whatever.

 

i was wondering if a knock can be heard through exhaust manifold.

we use EGT to detect knock.

why not mount microphones on ceramic pads( or other heat resistant material) directly on each manifold runner at the flange.
since the flange is not the hottest spot on the manifold.
a heat resistant pad will protect the microphone and they can be used with the electronic stethoscope and a switch between 4 of them

 

How do you detect knock from EGT?

 

exhaust gas temperature

 

Right, how do you use EGT to detect knock?

 

ok sorry mr wikipedia. it helps you to detect knock.

 

How does it help you detect knock?