Knock Sensor Applications

[William Boulton]

Hi folks,

I read a lot of e-mail about "knock sensors" and get the distinct
impression that they are greatly misunderstood. This is especially so in
the area of adding sensors to vehicles not normally fitted with one. I
have an article on engine knock detection in the SAE magazine of December
'93 which really threw me when I read it. I'd love to quote large passages
from it but I don't want the SAE suing my butt off for breach of copyright
so I'll summerize in my own words.

There are 3 ways to detect knock in order of cost and complexity: auditory
(ear), accelerometer knock sensor and pressure transducer. The authors of
the article say none is without disadvantages regarding detection
certainty.

When refering to placement, they talk of "nodes" where knock can not be
detected and "anti-nodes" where background noise swamps the signal
entirely. You wouldn't want to place a sensor at either of these points.
To make matters worse, on some engines, these ??nodes are not stationary
but move around in some RPM related pattern. Peter reported once that GM
spent millions on the problem and no one seemed to pick up on the
significance of the fact. I doubt that corporations like GM would spend
that sort of money if they didn't have to.

But wait, there's more. There are 3 bands of knock noise to be considered
at 6-8kHz, 11-13kHz and 17-20kHz. The significant band can vary with RPM
and any number of other characteristics. Broad-band sensors are not the
answer since they have problems with knock detection at mid range (3000
rpm) and require non trivial DSP (eg, 256 tap floating point FFT) to
determine the presence of knock.

OEMs use an advance determination method called MBT (Minimum spark advance
for Best Torque), unfortunately knock can sometimes set in before best
torque is realized.

So far, things do not look very good. All of the above leads me to a
number of conclusions (all mine):

    Knock sensors are not magic. They are basically a microphone in the
    noisiest (electrically & mechanically) environment possible. A knock
    sensor should be considered only in conjunction with an amplifier
    which is designed to match its characteristics and only on the engine
    for which it was designed.

    If a motor was never fitted with a knock sensor, don't bother. Just
    set as conservatively or agressively as you dare. It's your engine.
    You have to rebuild it if you break ring lands or burn a hole in a
    piston.

    If the motor is fitted with knock detection then duplicate the circuit
    exactly. Most pizo knock sensors have some bias voltage or current
    applied which influences the sensor responce characteristics.

    Learn to read spark plugs for signs of knock. (I can't, yet)

    I believe the OEMs are probably using the output of the knock sensor
    at points where it is known to be reliable and otherwise being very
    conservative. It has been reported that GM units over retard under
    some conditions. This may be the reason.

Now that I have let the fox into the hen house, let's have some feed-back.
I've seen the results of undetected knock and it ain't pretty and I
wouldn't wish that on my worst enemy. I look forward to reading responces
to this.

Bill Boulton