Knock Sensor Output 3

[NDSUFormulaSAE]

Does anyone know what kind of signal a knock sensor puts out?

I see MSD makes a cheezy box that you wire up to a knock sensor; it has a couple LED's, a beeper, and a sensitivity adjustment on it. And they can still sell it for $150 because nobody knows how it works.

I am wondering how difficult it would be to modify the knock sensor output to control an electronic advance distributor. Has anyone done it?

 

[Andy330hp]

Knock sensors are accelerometers. They are contantly putting out a nasty waveform, but the spikes become much larger when knock occurs. You may not have to worry about it. From my limited reading on the subject:

GM for a long time in the mid 80's through early 90's made external knock sensor modules that took the raw signal from the knock sensor, filtered and conditioned it, then set the ECU what basically amounted to a digital signal for Knock/No Knock. You can get one from a junkyard for $10. The olny problem with this is that different knock sensors have different natural frequencies just as different engines will have different natural frequencies when they knock. So, either you would need to determine the natural frequency of your engine (I'm not sure how to do this other than strapping an accelerometer on it and tappping it with a hammer, but I'm sure there are more exact ways), or by finding a GM car that is a 4cyl with as small a bore as possible, as I've read that bores affect the frequency knock sensor that should be used

As you can tell, I'm not much of an EE, but if someone on your team is sharp with electronics, they could back-engineer from the GM module what kind of filtration is done. It looks to be almost all hardware filtered, however the ECU steps the ignition differently under different conditions (ie, if knock is sensed at the same time as rapid delta TPS, a stock ecu will be under the impression that you are doing a very hard acceleration and will be a little more dramatic to protect the engine, whereas if you are cruising and one or two knocks are sensed, timing adjustment will be a little more gradual. I think it also counts knocks between fueling calculations, and figures the more counts=more timing adjustment needed. All these aren't "filtering" though)

For the reasons mentioned aboe the MSD setup probably wouldn't work for you anyways, since they are probably using large V8 style knock sensors.

To do it right yourself, determine the natural frequency of your engine (use an accelerometer and either tap the block, or purposely induce knock (be carefull with that one!!)). then find either a true knock sensor or a rugged accelerometer with the same natural frequency as your engine, and find an EE guru to filter it accordingly. OR just try a GM setup from a 4 cyl and see if it works!!

Andy

 

[Whittey]

Hrm. I dunno about that bore relationship to frequency thing. Here is a page right out of a Ford service manual:

http://acc-electronics.com/cloud/images/page3_55.jpg

Note that the 5.8L is a 4" bore 3.5" stroke V8 wherein the 4.9L is a 4" bore 3.98" stroke I6. Also note that the frequencies are vastly different.


-=Whittey=-

 

[Andy330hp]

Whittey,
Engine configuration is obviously the number one factor. A V8 and straight 6 will have diffenrent values fore sure. My point was that, given two 4 cyl cars from which to choose a knock sensor, you should go with the one that has a bore closer to the engine you will be mounting this setup on. If imports use the same basic piezoelectric sensors, it may even be possible to mix and match (wire a knock sensor from smaller disp import engine to a GM ESC module) to get the best match possible. That will require a little more research and time to make sure it will work (for instance, the modules may only be able to filter certain frequencies, and so they may form "matched pairs" with the sensors that are used with. I have no idea about that though). Sorry for the confusion

Andy

 

[GregLocock]

I'm not sure if the knock sensor is tuned to the acoustic resonance in the chamber.

Quick maths: say temp is 600 K (on average). c~500 m/s

First mode in 100 mm cylinder has lambda=0.2m , so frequency is around 2500 Hz.

Which is close enough to 5 kHz that the idea may be valid, or at least part of the story.

However I do know that selecting the tune for a knock sensor for a production engine is done by a highly sophisticated procedure involving an engine on a dyno, a long piece of brake pipe, and a plastic cup, and a handful of different knock sensors.

Cheers

Greg Locock

 

[grogg]

Hi,
I built a box that works similar to the msd box, the led flashes and the buzzer sounds when knock in the motor makes the sensor produce enough voltage to pass a predetermined setting.(both times I used the motors original knock sensor.)

All this helps, is it tells you when the motor is about to hurt itself so that you can back off.

Connecting it to a distributer to controll advance I could not get to work.
From what I could figger out you would need a small micro processer and an a/d converter to do the job.

Note: That little box came in handy quite a few times.

 

[techguru]

what kind of motor.
tell my the make and size and il ltell you if you can add knock sensors.

i hope its a small block chevy..

 

[JonieR]

I agree with Andy yep a very spikie wave form and yes " tuning " to the engine helps , however , given that the the frequencys generated by the engine when detonation occures can be "heard" in the range of the senser the conditioning is a faily simple amplitude trigger for a on off digital output. the threshold should be set on each installation. An analog output could also be generated based on the amplitude and frequencey of the spikes exceeding a calibrated threshold to indicate the severity and speed at which correction is required to protect the engine. and yes a microprosser , whether part of the ecu or an external pid algorithem controler , would be nessary to make it work automaticly with any timing or timing/fuel control system , of course the digital form could be used to turn on a led so the driver / tech could manually alter the timing in real time or modifiy the timing curve to prevent the detonation. Design of the electronics is not difficult , calibration could be interesting in a lot of ways , making it reliable and resonably packaged for use by most people in automotive applications with the inherent noise , heat . voltage variances , inatallers interpritative experiance and vibration would be a challange , but doable.

 

[jons999]

Knock sensor systmes on late model vehicles are very precise, they are tuned to the specific frequesncy of a knock on that particular engine. The ECU ignores knocks from bad rod bearings, loose a/c compressors, etc. So tapping on the block with a hammer no longer works. I'm not sure how complex the msd system is though.

Maybe you could watch the output of the sensor on a scope and advance your timing till it pings and see what kind of output you have?

-Jon

 

[GregLocock]

"If I'm not mistaken, alot of factory ignition modules have an analog input for advance(like 0-5v for 0-30degrees or something)."

Maybe some do it like that, but none I've seen. (which is not saying much).

Our advance is calculated from a lookup table based on ECT, rpm and MAP, with some adders in for other things, roughly. WOT spark is open loop at high rpms but may be controlled by the knock sensor at lower speeds.

For a track car you could probably get away with intercepting the signal from the crank position sensor and delaying or advancing that, conceptually. That seems nearly as much work as buildinga your own system.



Cheers

Greg Locock

 

[Warpspeed]

I have not had any personal experience with this myself yet, but am very interested. As a retired electronic hardware design engineer, and an incurable car nut, here are my ideas.

The engine is going to have a lot of background noise which is going to change with engine RPM, and there will also be a variety of mechanical resonance points. The knock signal will have its own natural characteristics as well.

So what you might need to do is look at the output of the knock sensor on a spectrum analyzer, and try to find at what frequencies the knock signal is most visible above the background noise.

You would then experiment with different sensors in different locations in an attempt to maximize the knock signal amplitude with respect to the background noises.

Hopefully this has already been done if you are lucky enough to already have a factory knock sensor fitted.

When you have done that, you then apply some frequency selective filtering to further enhance the signal to noise ratio. You then accept any signal over a set threshold as genuine knock. That would be my approach to the problem.

But I can see it would be far from easy for the average person to do at home. You would also need to deliberately induce knock, which will have its dangers.

Because of all the variables, I cannot see a generic one size fits all black box knock sensor working very well.

 

[GregLocock]

One thing we look at is the knock window. We find which part of the knock signal (in crank degrees) is the knock part, then we look at the level in that window ratioed to the level outside of that window to decide whether knock has occurred. This means that the system is self calibrating for sesnitivity, and takes account of the variable background noise.

I agree that a general purpose knock sensor system would be unfeasible. In effect if one were available we'd use it and then clone the settings into a production unit. That is not how it is done.

Cheers

Greg Locock

 

[Andy330hp]

In order to take your approach, is it necessary to have multiple knock sensors (or, inversely, fewer cylinders per sensor) to ensure that the non-knock window is long enough to establish a baseline, or would this approach work even on say a V8 with only one sensor?

 

[GregLocock]

We usually get away with one knock sensor even on a V8, I think, certainly on an I6.

I would emphasise that tuning a knock sensor is very much reliant on the engineer involved, both for positioning, tuning and calibrating. That is why I am suspicous of a 'universal' plug and play knock solution.



Cheers

Greg Locock

 

[grogg]

Hi,
I have long since binned them but I could redraw them. Tell me, can you or do you know anyone who could build it for you. You will need to make a board from scratch.

The way it works is (this could be wrong - but it works for me)
1) the knock sensor it tuned to the motor by the guys in the factory, placement of the sensor and the type of sensor is critical. there is a tuned weight in the sensor which only vibrates at a predetirmined level(the guys at the factory.)
The frequency of a knock and that of a bearing rumble is very different.
2) the output signal is very rough and very fast. The voltage produced by the sensor changes from a few mv to over 12 volts depending on how how hard the knock is.
3) The current output of the sensor is tiny ( less than 1 MA) So an op-amp is needed to do most of the work.
there is a lot of noise produced by the sensor, but it is fairly easy to filter out.

4) The problem is that some of the components might have to be replaced with components that are more suited to the sensor you are useing.
5) when setting the trigger threshold the motor must be made to detonate (most wellbuilt motors can take a small amount) If anyone knows of a better way to set it up please tell me!

tell me if you want the drawings.

cheers grogg

 

[Andy330hp]

Where could I find a good selection of knock sensors with different natural frequencies? Or, how can I find info on what frequency knock sensor is installed on what vehicle, so the correct one could be purchased from a parts store?

 

[nagy]

dear Grogg,

I am present having a problem to find an simple equipment to show if the engine is knocking or no. Could you send me your drawings to construct this knock meter.I would like to evaluate this idea.
My email is Luis.nagy@magnetimarelli.com.br

thanks in advance

nagy

 

[Andy330hp]

Another question: Just how critical is placement, really? Going back to what GregLocock said, a single knock sensor can be made to sense all 8 cylinders. Now obviously this sensor cannot have what we are imagining as an optimal acoustical relationship to every single cylinder. So, do we really need one? Or, if you put in enough effort to measure the noise at whatever spot you choose, can you pick "pretty much" anywhere in the block? I imagine that you would want to stay close to the stiffest parts (near reinforcment ridges, etc) and avoid being in a large unsupported plane that could set up it's own frequency. Beyond that, though, it does not seem as if there are any strict rules for being "such and such" distance from a cylinder or so far up along the bore. PLEASE correct me if I am wrong, these are my speculations and NOT fact.

 

[clshore]

Knocking on the block with a hammer does still work for testing knock sensors. It introduces a single excitation step input, exactly as happens when knock occurs. Step input represents all frequencies, it is the engine assembly (block, rods, crank, etc.) that resonates in response to a step input, at it's characteristic frequencies.
Same thing as is used in Time Domain Reflectometry.
A wideband transducer connected to an audio spectrum analyzer will reveal the characteristic resonant frequencies. These will of course vary somewhat with temp, but the filter bandpass can be broadened a bit to cope.

Carter Shore

 

[heuf]

How feasible do you think it would be to use knock sensors in an alloy-block engine with wet liners? Would it still be possible to detect the knock signal?