knock sensors

[sofadriver]

In Generatorshovels bike build diary "one for the 250 folk" the talk has turned to knock sensors. Can you tell me all about them? Power requirements, mounting, sizing, etc., etc. Sounds like a great tool to help avoid frying my little 2smoker! (didn't want to hi-jack his thread)
Thanks.

[maj]

good stuff is not cheap
http://www.linkecu.com/products/TuningTools/knocklink_g4

[Milwaukee Midget]

While I like the idea of a knock sensor, I found in my application that they were rather ineffective. 

The Electromotive ignition set-up I have has provisions for a knock sensor, which pulls back ignition timing when knock is detected.  On the XDI-2 system, the knock parameters are adjustable in the software. 

Essentially, a knock sensor is a piezoelectric microphone that sends a signal to the ECU. 

The problem I had was that the ECU would see a signal from the knock sensor that wasn't necessarily pinging.  I was running solid lifters, stiff pushrods and about .020 clearance on my valve train.  The MLS head gasket didn't have the sound deadening characteristics a standard gasket has, and further transmitted noise into the iron block from the head.  Before I changed over to a Gilmer drive on my cam, the chain also created noise.

So there are a lot of extraneous noises occurring that the ECU was unable to discern from knock.

After about 3 or 4 pulls on the dyno where the ECU kept pulling back the timing, we did a plug inspection and determined we weren't getting any knock and disconnected it. 

We were able to run 37 degrees total advance WOT for two passes at Bonneville with no indication of any engine damage. 

So I guess the lesson I learned - at least in my case - is that while a knock sensor might be a nice tool to prevent a total failure, using it for a backstop can cost you power, and there's no substitute for dyno testing to optimize tuning.

Chris

[hotrod]

For the average user putting a knock sensor on an engine it was not specifically designed for is a bit beyond the level of complexity most are able to deal with.

Each engine has a characteristic noise signature which as noted above changes with modifications for performance. In the major manufacturers case, they can spend hours and hours on the dyno with a specific engine build and sophisticated noise analyzers and perform knock detection testing to find out what to listen for to identify knock in that engine.

They can then build into the engine management system electronic filtering to only listen for the right sound frequencies to detect knock. They can also figure out where on the engine block or cylinder head to place the sensor so it most effectively hears only knock. In the case of the Subaru engines for example they place the knock sensor near the drivers side rear cylinder which is most prone to knocking, and they turn off knock detection at higher rpms. The inherent engine noise that engine produces too much false knock at high rpm and the sensor can no longer discriminated knock from mechanical noise.

That engine being a very over square turbocharged design is quite sensitive to knock and fuel octane requirements (tends to break piston ring lands) . Knock is most problematic at low to mid range rpms but seldom an issue at higher rpms since there is less time for the knock conditions to develop in the cylinder. Honda discovered at high enough rpm (F1 engine) knock tendency shows a noticeable reduction and octane requirements actually go down as engine rpm goes up.

[sofadriver]

Well, now I know.

I checked maj's link (thank you) and, despite their claim that it can distinguish knock from noise, I found this in the manufacturer's description. "Not suitable for use on MECHANICALLY noisy engines. This includes engines with high clearance pistons and two-stroke engines".

Just not feeling any love for my little 2 stroke. 

Thanks guys.

[Jack Gifford]

... figure out where on the engine block or cylinder head to place the sensor so it most effectively hears only knock...

I've read about some engine builders having success with a trial-and-error approach to sensor location (V8 engines), under somewhat controlled conditions of engine-dyno sessions.

[Milwaukee Midget]

They can then build into the engine management system electronic filtering to only listen for the right sound frequencies to detect knock.

True - essentially an audio spectrum analysis can eliminate and notch out the superfluous frequencies, and the chip is programmed to only let "knock" frequencies activate the retarding circuitry.

On the Electromotive setup, what I discovered was that there is a trim function which pulls back the entire spectrum of frequencies, but it is not sophisticated enough to isolate only the characteristic signal of knock alone.  It can be trimmed back to only let the strongest signals pass to the ECU, but if the engine is a clanker, you're likely to trim the signal past the point where the ECU could save the engine. 

[JimL]

Knock detection is best performed within a pretty small window of crank rotation (to avoid looking at extraneous noise) and within a frequency range appropriate for the bore diameter.  For instance, older Toyota shop manuals used to identify the knock sensor frequency (marked on the sensor shell), wherein the smaller bore car engines might have knock sensors for 7.7 kHz or even 8.2kHz, and larger bores could need knock detection down in the 7.2 kHz range.

The problem of knock detection at high rpm is somewhat affected by processor speed.  Our older KNK detection typically wouldnt work above about 7200 rpm on a 4-cylinder, about 6200 on a 6-cylinder, and about 5200 on a V8.  Gentlemen... I am growing old and may have some errors in my recollections...and newer KNK systems are using much better planar type sensors.

Mechanical noise can be a big problem, certainly.  I used to check KNK sensor detection (where a car was showing high knock output signal when there was no reason) by connecting two 2.1V, 1 watt, Zener diodes "back to back" parallel.  Small zeners blow open at .5V reverse bias.  My simple rig was wired with one end to ground and the other to the KNK signal.  This method took everything above .5V (positive or negative of the sensor sine wave) away from the ECU. 

If the car ran correctly with the trimming gadget, there was something mechanical going on at KNK frequency for that bore diameter.  One time it was a cracked alternator lower bracket, another it was a bad ridge in a cylinder.

This is also how we ran Dan Wagners GBGMR with a solid mounted engine, back in 1998-2000 using a pretty stock Celica All-Trac engine.  It gave us an extra 30 mph.

One successful option for spark control of a stock ECU is to take the crankshaft rotation signal through a small processor to offset the trigger output "x" number of degrees.  This method has been used on various aftermarket blower kits, where replacng the original ECU is not practical or emissions legal.  It is a pretty simple gadget for a hobbyist programmer of PICs.  Look for the robotics hobby folks....they are doing magic with small, simple chips.  You can probably come up with a cheap and serviceable solution for timing control.

They will need to know your signal durations and RPM to calculate the clock cycle of the required math processes.  This may mean youll need to pick a max advance to target (maybe 20-22 degrees?) and then reprogram chips for changes.  You will need to calculate the crankshaft TDC indication position and provide that info as well.  A decent oscilloscope will tell all.

Dont forget...if your current ECU has a piezo sensor, you will need those two zener diodes along with a signal delay device.  Try to keep it simple.

[fordboy628]

Knock is most problematic at low to mid range rpms but seldom an issue at higher rpms since there is less time for the knock conditions to develop in the cylinder. Honda discovered at high enough rpm (F1 engine) knock tendency shows a noticeable reduction and octane requirements actually go down as engine rpm goes up.

The reason knock is more of an issue at low to midrange engine rpm's is because this is exactly the range where the highest cylinder filling efficiencies (Volumetric efficiencies) occur.   Time available is, of course, an issue, but so are things like gas speed though the ports.

Once you are though the maximum torque "range", V/E starts to drop and with it the likelihood of knock.

Engines needing to run at wide open throttle through the peak torque range need to be very carefully tuned, OR, they need a very sophisticated knock sensor with a multi-layered spark map.   Turbocharging or supercharging adds another layer of complexity . . . . .

 
Fordboy

[wobblywalrus]

This is something I saw this year.  Three runs were made with no changes other than gearing and environmental conditions.  The run with the 38 tooth rear sprocket had the motor turning higher than the peak power rpm.  No indication of combustion problems.  The run with the 36-tooth rear had the engine turning at maximum power rpm.  No signs of any troubles.  The run with the 35-tooth rear was at an rpm closer to peal torque.  The knock light flickered.

[fordboy628]

This is something I saw this year.  Three runs were made with no changes other than gearing and environmental conditions.  The run with the 38 tooth rear sprocket had the motor turning higher than the peak power rpm.  No indication of combustion problems.  The run with the 36-tooth rear had the engine turning at maximum power rpm.  No signs of any troubles.  The run with the 35-tooth rear was at an rpm closer to peal torque.  The knock light flickered.

And there you go.  Perfect example of what I was talking about.

 
Fordboy