Testing the frequency response of a knock sensor.

[xwarp]

I've been trying to think of a way to figure out what the "resonant" frequency of a knock sensor I have is. It's from a GM TBI system and based on the cross reference at one of the parts shops, it is used on 3.1 and 3.4 engines.

Based on this formula:

http://www.phormula.com/KnockCalculator.aspx

these engines should have a knock frequency of about 6.2khz to 6.4khz.

With a very minimum of equipment available at home, coming up with a way to see how the knock sensor responds on a bench is kind of difficult, but I thought I might try something.

So, using a 2mhz sweep generator and a midrange speaker along with the Tex TDS 460, I tried something. (See attached picture)

Starting at a frequency of 200hz, I slowly increased it and followed the sweep. You can see where the output of the piezo in the knock sensor peaks at about 2khz, 4.25khz and then drops off quite a bit.

I'm pretty sure this setup is pretty crude and maybe placing a 1/16" thick square of aluminum over the driver and then tapping it to more securely attach the knock sensor to it would yield better results.

Thoughts?

http://youtu.be/5Tli4j7m5BA

[krivx]

I can't get the video to play (still processing?)

I would knock it with my hand and view the output on a scope. Set the scope to one-shot capture. It should look like a ringing with decreasing amplitude, the frequency of ringing is the resonant frequency.

I'm only guessing as I haven't seen the video but your method sounds like you will have additional resonances caused by the speaker...

[xwarp]

Had to re upload it....

http://youtu.be/5Tli4j7m5BA

[xwarp]

I can't get the video to play (still processing?)

I would knock it with my hand and view the output on a scope. Set the scope to one-shot capture. It should look like a ringing with decreasing amplitude, the frequency of ringing is the resonant frequency.

I'm only guessing as I haven't seen the video but your method sounds like you will have additional resonances caused by the speaker...

I have not yet figured out how to get this scope to do a "one shot", like a 466 for example. I'm not too familiar with DSO's.

When you see the video, you will see where the knock sensor output increases.

[BennVenn]

A knock sensors output is largely dependent on the engine block it attached to. The engines cylinder resonates at a unique frequency (varying with temperature) which the ECU filters and acts on. Knock sensors are usually matched to this frequency but not always. The RB25DET ECU has a daughter board ( a bunch of op-amps ) tuned to the engines knock frequency. The VG30DET (ECU often used on RB25’s due to re-programmable EEPROM but you loose the knock sensing) uses the same knock sensor though the knock filtering circuits are not compatible, it will cause false knock triggering or inability to detect knock at all (depending which way you go)

[mikerj]

The principal problem with this setup is you don't know if you are measuring the speaker response or the sensor response.  The speaker will certainly have a resonant frequency, which will depend on how the driver is loaded, i.e. sitting with the cone on a hard surface like that the resonant frequency will be a lot higher than it would in free air.

You could try just using an impulse (e.g. tapping the sensor with a screwdirver handle) and look at the spectral output.

[xwarp]

The principal problem with this setup is you don't know if you are measuring the speaker response or the sensor response.  The speaker will certainly have a resonant frequency, which will depend on how the driver is loaded, i.e. sitting with the cone on a hard surface like that the resonant frequency will be a lot higher than it would in free air.

You could try just using an impulse (e.g. tapping the sensor with a screwdirver handle) and look at the spectral output.

I was sweeping the speaker, (midrange), with a sine wave from a sweep generator.

Now, tapping the knock sensor did result in it producing a short "ringing" output, but what does that show other than that the sensor produced an output, because, what frequency was the contact between the two two? I did find that the variation of frequency was minor whether the sensor was smacked lightly or pretty hard.

But, if the sensor is somewhat specific to an engine, then it would seem that the sensor would be used to output a signal, that would peak around those frequencies, as the video above seems to show.

[Anks]

If you can put the FFT in averaging mode and just bang the sensor till you have a uniformed response. Not great but will give you a better idea than measuring the response of the speaker with a knock sensor.

[HighVoltage]

Most of the knock sensors used today have a range to detect frequencies of 1 to 20 kHz
The main resonance frequency usually is higher and outside the range of detection.
Most Bosch knock sensors have their resonance frequency at around 25 kHz.

To test a knock sensor, I would also just use an oscilloscope in a singe shot mode and bang
the sensor. Use a few different materials to bang it with (Plastic, wood, metal)  and you will
see a change in frequency output.

[xwarp]

I apparently did not make it clear that sweeping the speaker across the range of 1khz to 8khz revealed that the sensor's output increased in voltage at certain frequencies.

The sensors I am specifically interested in are for the GM OBD1 system, which were designed to be specific to the knock frequency of the engine.

[xwarp]

To test a knock sensor, I would also just use an oscilloscope in a singe shot mode and bang
the sensor. Use a few different materials to bang it with (Plastic, wood, metal)  and you will
see a change in frequency output.

I can do this, again, and suspect that I will see the same info as stated previously, a signal, but a signal probably not relevant to the design of the spark control system.

[xwarp]

Let's try a different point of view....

think like working with an accelerometer for a shaker and you don't know what the frequency/voltage output is of the accelerometer.

[max_torque]

Unfortunately, your method is, as other posters have pointed out, rather flawed.

You have no idea of the true excitation amplitude of you "shaker" vs input frequency, and even more important, no idea of the coupling ratio between the exciter and the sensor.


Luckily, none of that really matters, because, broadly speaking, what you care about isn't the knock sensors natural frequency, but that of the knocking pressure wave in your engine.

Knock sensors fall into two categories:

1) resonant
2) non resonant.


In early EMS systems an resonant sensor was used, that was carefully matched to F(Kp) as this did not require "high speed" crankshaft position synchronous signal processing in the ecu.  The ECU could just low pass filter the input signal to get a roughly RMS knocking voltage, and when that exceeds a certain value within a certain crank angle window, could flag up "knocking" and retard ignition etc. These sensors are highly dependent upon their installation environment, and use a spring/mass system that puts their resonant (and hence highest output) voltage at the corresponding F(Kp).

As the power of EMS system processing has increased, all modern systems moved away to a non resonant sensor architecture, in which the frequency response of the sensor is broadly flat across the measurement spectrum (4 to 16Khz typically).  These sensors require some analogue or Digital signal processing to maximise the signal to noise ratio of the knock intensity signal, but are much easier to use (as you don't need to develop a specific sensor hardware specification per engine derivative)


Now, assuming you want to use this sensor to measure engine knock, rather than some other kind of signal, you simple need to install the sensor on your engine, record it's output with a PC sound card or similar (needs a high impedance Opamp front end really), run your engine without knock (retard spark, or run super high grade fuel and log the frequency spectrum, then run deliberately into knock (at low speed, lowest load possible to avoid engine damage) and re-record the output spectrum.  By taking A result away from B result, you will see the various (fundamental + harmonics) knocking frequencies of your engine.  You will also  see, especially if you are not crank angle filtering the signal, a load of noise from things like valve events or piston slap etc. 

Vehicle OEMs spend literally millions optimising their knock sensing and response calibration, and it becomes very complex and quite subtle.  Even silly things such as road gravel hitting the sump (in the days before cars all had "aero" undertrays/guards) could trigger a false knock detection event!  ;-)

[mikerj]

I was sweeping the speaker, (midrange), with a sine wave from a sweep generator.

Now, tapping the knock sensor did result in it producing a short "ringing" output, but what does that show other than that the sensor produced an output, because, what frequency was the contact between the two two?

What do you mean by the "what frequency was the contact between the two"?  You are simply exciting the sensor by hitting it, and it will continue to ring at it's natural frequency.

I apparently did not make it clear that sweeping the speaker across the range of 1khz to 8khz revealed that the sensor's output increased in voltage at certain frequencies.

I (and everyone else) apparently didn't make it clear that your speaker will have it's own resonant frequency.  Was the increase in output of your sensor because you hit the resonant frequency of the sensor, or of the speaker?

[joe 90]

Hi, I know this is an old topic.
Nobody has posted the correct answer yet.
Knock is one of the most misunderstood things about cars.
Car mechanics all fail at electronics and the industry as a whole can't be trusted.

First of all.........the formula was taken from some kid's thesis, that in turn was taken from some other paper.
A lot of the original meaning was lost. English as a second language and mis translation along the way.
The original statement which is true.................is that the lowest frequency you can pick up from knock is based on the bore size.
The truth about it is that knock is a single explosion and you hear it as a series of reflections off the walls of the combustion chamber. Not just the bore but the cylinder head and piston top. It's a band of different frequencies. If you can hear it with your ears, it's very bad but you hear it as a descending pitch which is the piston going down the bore.


The frequencies involved will change with the piston position.
The signal strength derived from the bore size will also vary with piston position.
Peak cylinder pressure is normally about 15 deg ATDC, that's where you'd expect knock to begin. As it gets worse as it always does quite quickly, it'll happen sooner, when it happens sooner, there's more unburned fuel available to explode and the piston is closer to the top. Detonation heats the piston quickly and if not eliminated will lead to pre ignition. The only real difference is that pre ignition happens before the plug fires instead of after, so it's BTDC. When the piston is at the top as it is with extremely bad knock, there's no bore for the waves to reflect off. All the reflections are between the curved or flat piston top and the curved roof of the combustion chamber.
Think on that for 5 minutes?



Pretty much all of these aftermarket devices for picking up knock, they're pretty useless and will never work as well as the factory designed devices.
Do a google search for bomb detectors ....that was a massive fraud too.

The commonly accepted theory is all wrong. Useless as tits on a bull.

As said, there's a band of frequencies.
To detect a signal you want the best signal to noise ratio so what you do is use a resonant sensor tuned to a frequency which has minimum noise but is in the knock band.
A resonant sensor always works best because all you have to do is bang it and will ring like a bell.

Modern computers all use a microphone instead, they have a flat response, often 20Hz to 20KHz but you need fancy software.
The reason they're now used is because they don't cost anything, you can use a single computer across a whole range of models and all you do is load the right software.
A resonant sensor puts out a nearly pure sinewave so it's a single frequency....no filtering required. If you reverse engineer say a 4G63 ROM and find tables of filters........it doesn't actually mean that they're being used.
The common GM sensors used in V8s are usually 10.5KHz. Toyota eg, the 4AGE, that's about 8KHz, Mitsubishi, 11.8 ish KHz for the 4G63, 11.2KHz for the 3.5 V6.

There's a vid on U tube testing them.The description is on U tube.



Check out my other vids too.

Knock detection is THE single most important aspect of tuning a car and making a lot of power. If there's no knock, you don't need forged pistons, often stock conrods too........a well tuned modern 3 litre engine can make 1000HP on stock internal parts.
If your knock detection system works properly it can protect the engine.
A typical jap turbo car will drop the boost, richen the AFR and reduce timing advance, usually in that order.
As soon as you fit any aftermarket boost controller, you've eliminated the first line of defence.
Most aftermarket computers, all they do is reduce timing advance which is ummm.......good for engine reconditioners.
If your knock detection system doesn't detect knock ....either there isn't any.........or it doesn't work.
Ears never work.
Only when the signal level is far too high.

Someone mentioned RB25....that's a resonant sensor too...I'll find one and check the frequency.
The VG30, that's a flat response sensor.
That explains why swapping the computer you lose knock detection....Easy fix, use the correct sensor.
External detection is really easy to build though....like everything is if you know what you're doing.


BTW I've just signed up because I fell onto this while looking for something else.
40 year hotrodder and 40 year electronics/ telecomms repair pro.