Engine Knock
TJHJ
Posts: 243
Hey again all,
I was hoping if anyone has had experience trying to sense engine knock, they could point me in the right direction here.
So I was considering using the Piezo Film Vibra Tab Mass, mounted to the block, but the 1st experiment melted off the block as soon as it got warm. What I was planning was a a kind of normalization to sense it, The engine makes sounds in this range, anything greater produces knock warning. But since Im not to sure about the Tab mass now and I am stuck at a cross road. The next thought was to use a small microphone set inside of a bolt/tube filled with high temp epoxy. But using this method I would need to be much more specific to the sound of knock, Here in lies the big problem, I have been scouring everything I can find on the subject, and I cant find the frequency that knock occurs, everything says "Knowing the knock frequency, we can sense accordingly", Incredibly frustrating, I cant seem to find a hard number any of what frequency knock occurs at. I have even gotten desperate enough I have been the libary, and been through a ton of books and papers on the subject. (Real paper... Its weird
) I know that each engine has a slightly different frequency, and ideally I would take an engine advance the hell out of the timing, record all the data and blow it up, but thats just not realistic for me. But proably would be fun. 
So Does any one know the frequency at which knock occurs?
If not maybe a true knock sensor that would interface well to the propeller chip, to avoid the microphone problem?
Hell if anyone has anything please let me know, Maybe I am just a failure at searching, I hope not.
Thanks for the time, I cant thank you all enough.
TJ
I was hoping if anyone has had experience trying to sense engine knock, they could point me in the right direction here.
So I was considering using the Piezo Film Vibra Tab Mass, mounted to the block, but the 1st experiment melted off the block as soon as it got warm. What I was planning was a a kind of normalization to sense it, The engine makes sounds in this range, anything greater produces knock warning. But since Im not to sure about the Tab mass now and I am stuck at a cross road. The next thought was to use a small microphone set inside of a bolt/tube filled with high temp epoxy. But using this method I would need to be much more specific to the sound of knock, Here in lies the big problem, I have been scouring everything I can find on the subject, and I cant find the frequency that knock occurs, everything says "Knowing the knock frequency, we can sense accordingly", Incredibly frustrating, I cant seem to find a hard number any of what frequency knock occurs at. I have even gotten desperate enough I have been the libary, and been through a ton of books and papers on the subject. (Real paper... Its weird
) I know that each engine has a slightly different frequency, and ideally I would take an engine advance the hell out of the timing, record all the data and blow it up, but thats just not realistic for me. But proably would be fun. So Does any one know the frequency at which knock occurs?
If not maybe a true knock sensor that would interface well to the propeller chip, to avoid the microphone problem?
Hell if anyone has anything please let me know, Maybe I am just a failure at searching, I hope not.
Thanks for the time, I cant thank you all enough.
TJ
Comments
http://delphi.com/manufacturers/auto/sensors/et/flat/
Another possibility is that you could use or modify an oscilliscope or a fft object(It seems like I have seen them here somewhere) and feed the input directly into the prop adc setup to detect the difference. You will probably have to induce your engine to ping lightly somehow, which is not hard if you have an old engine you can adjust the timing on.
It's been twenty years since I studied this, but as I recall, the frequency is 17.5 Khtz.· I would suggest just using a OEM·knock sensor.·· They are $20-$50 on Ebay.· One thing to keep in mind.· They can generate 30-50 volt spikes.·
Daryl
www.max-boost.co.uk/max-boost/internet_articles/BigMooses_Ramblings_Knock.htm
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
Aka: CosmicBob
Thank you all for your help, now comes the fun part
In one of the books, I can't remember which one, there is a good chapter that explains why the exact frequency of the knock differs by bore size, and the pressure and temperature of the combustion gases in the chamber at the time the knock is happening. OEM knock sensors therefore, are built to sense specific freq's for specific engines and may not work when applied to the wrong engine. Since engine knock is characterized by the echoing of sound waves back and forth inside the cylinder, and since extreme temperatures and pressures make significant changes to the speed of sound, you must first calculate the exact speed of Mach 1 inside your cylinder during the knocks, and only then can you use bore size to determine the frequency of the ringing from the knock. The book I mentioned shows in great detail how to go about calculating all of this, as well as a thousand other things about engine design that you'd never think of. It's the best book I've ever seen for learning about designing engines.
Or should I go to a 5v zener to ground, and a voltage divider to drop back to 3.3v?
So I found it all, I cant thank you all agin for the help, I thought I would post the formula here so if anyone else ever comes across this post and needs it
Fundamental Knock Frequency = Fn = (Kc*c)/B = (Kt*(t)^(1/2))/B
Kc, Kt = Contestants
c= Local Speed of sound
B= Cylander Bore Diameter
T= Combustion Chamber Temperature
Then the Fn*Mode= Knock Frequency
Mode is dependent on cylinder shape but is 1.65, 2.07, 2.26, It has been stated that it is best to use the 1.65 for most engines as to delay knock to the higher frequencies special head shapes and designs are required.
Thanks again all.
Post Edited (TJHJ) : 6/16/2008 7:27:02 PM GMT
Also if you wanted to get crafty and pave a slightly new path, you can measure the electrical resistance of the spark plug after the spark has fired. I am not sure how it all works, but you can "watch" the combustion happen thru the resistance of the spark plugs gap.
Also if you cannot get he detonation sensor working you could always use a fast thermocouple on the exhust to sense the rapid rise in EGT, showing that you have either detonation or a preingition
If you use a hall-effect continuous rotary angle sensor, you can tell exactly how fast the crank is spining, and which cylinder has contributed to the output (using the engine's firing order table as a reference).
-Tim
I am actually already using the fastest response thermocouples I could afford to watch each cylinders temperature and for any abnormalities as well as during "Cruise" Conditions to alter the advance to more evenly distribute the load to improve cruise economy. But I fear that with the 11,000 to 13,000 rpm operation range observing the temperature change in each pulse is not possible, I don't think the thermocouples can compensate fast enough to see each pulse at those engine speeds. As for Big moose tell him thank you for me, his work so far has helped me out a lot.
I am reading for inputs, EGT, CHT, Throttle Position. I have calibrated all the intakes to be with in 3% of each other so I can make a pretty confidant statement that each cylinder would receive the same amount of air, for any calculation ideas anyone has.
I was considering using this as a safety method, I am actually using a hall effect sensor on the engine already as the "Crank Pick up wheel" is actually part of the crank shaft, I went to the hall sensor to help get more accurate readings than I was getting with the magnetic system(My guess really was interference), but the hall effect has been some much easier than the magnetic style sensors anyway.
So here is the question, the pickup wheel, if you can call it that, is two notches every 164 degrees, that are 16 degrees wide. Short of re machining a case Im pretty much stuck with it, as I have no other external drive reference I could use to mount a multiple tooth sensor to. My original plan was to use a 30ish tooth wheel and look for any acceleration in the time measurements before the spark event, indicating pre ignition fire, to bad now.
Being stuck with this pickup wheel, I can only see a time for the majority of the stroke, being a non static engine, I can really find a good way to differentiate between road condition based acceleration/Deceleration, or knock acceleration/Declaration. Maybe I am over looking at this wrong, but short of putting a limit that if last reading is < 1.2(New Reading) = Danger I cant think of a way where external factors wont mess it up.
The method of using the plug to watch seems very interesting, time to do some digging.
Thanks again all,
TJ
So this a stand-alone ecu project?
In case you have any other questions I know quite a bit about ECU systems and would love to see the project when your done. I have thought about doing it myself for making my car run off of injected propane, but i have other projects way ahead of it.
What did you end up using for counting the air? Mass(hot wire) or MAP.
How does any project start out as the simplest thing ever and go to way more complicated than it ever should.... Because it can. (thats my theory at least)
A device like this needs a small magnet mounted to the end of some shaft (ideally the crankshaft, near a thrust bearing).· It can provide an "absolute position" of the magnet at 10-bit resolution (so it can tell you where the crankshaft is to within +/-0.175degrees, for this part).· Once·your microcontroller·has that data, the high speed pulse outputs can give you a quadrature-type output to allow you to detect movement in either direction at basically any speed from 0-10K RPM.· This means that the micro should read the position before the engine is started - then keep tabs on it.
With a 2-stroke this type of input is perfect, since you don't need a 50%-drive-speed input (like from a cam shaft) to get the "cycle" hemisphere (recall that with a 4-stroke it goes: 1-intake, 2-compression, 3-combustion, 4-exhaust; with two compete up-and-down actions [noparse][[/noparse]4 "strokes"] each firing - and thus two complete turns of the crankshaft).· For a 2-stroke each rotation of the crankshaft causes one complete combustion sequence for each cylinder.
To check for engine load vs. "knock" should be fairly easy - engine load is more steady (even things like running over rocks and the like on a motorcycle will cause a deceleration over multiple combustion cycles), while a miss-fire is a single-shot event or a continuous deceleration·on the same cylinder (indicating something like a bad spark plug or wire for that specific cylinder).· Knock should cause the engine to decelerate a bit earlier than a combustion acceleration was expected (based on the instantaneous timing advance angle for the cyclinder).
Similarly, an engine that is beginning to "lug" will have a higher exhaust temperature (EGT) than one that is running on a fairly light load.· An engine that is having more throttle applied will have a higher vacuum after the air filter (the throttle isn't restricting the flow of air as much when it is opened more).· A throttle position sensor (TPS) change should pre-empt any vacuum increases and indicate that the fuel-map may need to change columns.· An engine that is running rich or lean is also easy to detect (using off-the-shelf O2 sensors).· Account for these items·and the engine's temperature (giving it a "cold engine" fuel map until the engine reaches temp)·- and you basically have an OBD-1/2 ECU...
-Tim
I have ordered some of the sensors so it will be a week or two before I get them and then I think I will have to split the case to mount it up.
One question, I feel that there would be to much interference but I am unsure how sensitive these are to outside factors. Would this sensor be able to be mounted in the center of the stator (Magneto), or would the magnets moving around there just mess the whole thing to all hell? It would be really nice to be able to mount it there because it means I do not have to, pull the motor, split the case just to mount this sensor.
Thank you for your time, and all the awesome help.
TJ
I think it depends on your stator. My 'Berg has magnets all around the circumference to supply power for charging the battery. But, my ignition only uses 4 magnets consuming about 60degrees of the total 360degree revolution.
If you are going to add magnets (aka counterbore some holes and drop in some strong magnets potted with epoxy) that will end up being very close to your sensor, then the interference may be cancelled out by the sensor itself.
My instinct (which is iffy at best most of the time) says that you'd be fine just sticking the thing in there, as it will (should) recognize a magnetic field change in polarity. If you have both a charging and ignition magnets, the ignition magnets should cause the field to change, while the charging magnets should be a higher frequency change.
Not sure if that all came out right, or if it was even correct. But I thought I'd add my 2 cents.
-Parsko
Check it out:
video.google.com/videosearch?q=myt+engine&sitesearch=#
angellabsllc.com/resourse.html
I am no way affiliated with them(wish I was), but this thing everybody should know about!
How do they seal all those flappy things? The combustion chambers are the wrong shape. How do they stop the flaps from overheating and burning? I first saw that idea many years ago, no-one has actually produced a working version.
Leon
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
Amateur radio callsign: G1HSM
Suzuki SV1000S motorcycle
I once tried my hand a rotarys, 6 motors later I decided to sell that project.
Parkso, Yeah Im just going to get it on there and see. This has 3 sets of magnets, goofy I know, the internal set act as·power to main electronics. A set on the wheel itself 6 total, and then a·fixed outer set, High voltage for ignition. I hope it works out.
Edit* To clear up a bit. I plan on mounting the magnet that is speced for the sensor in the center of the stator wheel, Im just concerned about the magnets spinning around the outside of it, and how large the field of dection is, eg, the wheel is about 4 inches D. So guestamating, would the field of the sensor see and be affected by that?
Thanks again for all the great advice and help.
TJ
Post Edited (TJHJ) : 6/26/2008 4:34:01 PM GMT
Hey, watch what you're sayin there!