How Does Knock Sensors Work?

[NICAT]

Hi.
My question is, How does ECU detects knock based on the information provided by knock sensor ?
Does knock sensor simply send variable voltages to ECU, then ECU decide if there is or not knock by making some complex calculations ?
Or just knock sensor tells the ECU there either is knock or there isn't ?
Regards.

 

[dogememe]

yeah it just sends knock as a signal to the pcm the sensors are pretty dumb as far as i know and just have a little piezo-electric crystal or something

 

[mk378]

The sensors with one wire are simply a microphone. The ECU does all the signal processing and decisions.

 

[billt460]

Knock, knock.... Who's there......Unleaded 87..... Sorry, I told you to just do Unleaded 91.

 

[bdcardinal]

Originally Posted By: billt460
Knock, knock.... Who's there......Unleaded 87..... Sorry, I told you to just do Unleaded 91.

Knock, Knock

Who's there?

Con

Con who?

Conrod making a hole in your block.

 

[The_Jeff]

I think it's more of a frequency than a voltage.

It's part of a feedback loop on many cars, so the PCM listens for "just enough" knock, if you can imagine such a thing.

 

[Dave Sherman]

It’s usually a piezoelectric sensor. Apply a mechanical stress or shock to it, it produces a small voltage. The ignitor in your gas grill uses the same principle. The ECU knows when the spark plug fires, it should “knock” at about the same time. If that “knock” happens at the wrong moment when the crank sensor tells it the crank isn’t at the right point, then it knows it’s detonating and adjusts accordingly.

 

[SHOZ]

It's voltage, the louder the knock the higher the voltage. The ECU has a threshold voltage and above it cuts timing.

 

[Virtus_Probi]

As mentioned above, the knock sensors I was familiar with were piezoelectric elements that produced a voltage when they were subjected to vibrations. This voltage would basically have the same frequency content as the vibration itself in its usable region of operation.
There were multiple elements placed around the engine block in our customer's system, I believe they chose the sensors and locations but we were responsible for the signal processing.
One difficult part about detecting knock was ignoring noise from normal engine operation, hitting potholes, and other things that caused vibrations other than knock. The system that my company developed used windowing functions in the time domain and filtering in the frequency domain to isolate vibrations that were actually due to knock. There was a feedback loop in the system and the customer told us that the ideal condition was to knock very lightly, at a level the driver wouldn't even notice...this was supposed to be a power and fuel efficiency sweet spot.
I wasn't involved very much in developing the system, the team I was in implemented the functions that were needed in miniaturized hardware that talked to the ECU. The filtering and windowing was all programmable and the results were constantly polled by the ECU over a serial bus so it could make its timing decisions.
We got to see the system working in a car on a dyno and it was wicked cool!

I believe there was some discussion on here a few years ago about a system that detected knock by measuring the impedance across the spark plug gap after firing...

 

[clinebarger]

Originally Posted By: mk378
The sensors with one wire are simply a microphone. The ECU does all the signal processing and decisions.


Not true for all knock sensors, The BROADBAND RESONANT sensors on my Camaro(And all GM single wire knock sensors) Have mechanical amplification & filtering of the signal, Over a broad bandwidth in the knock range, provides high signal levels while allowing for typical sensor & engine variations. Broadband sensors are also less susceptible to engine noise and vibration (that would fool a flat response sensor) due to their ability to distinguish false signals with their high frequencies.

• FLAT RESPONSE sensors- Have a relatively flat response (4,000 Hertz). Universal in nature but requires a VERY specific configured ECM/Knock Module for each engine model.


Originally Posted By: eljefino
I think it's more of a frequency than a voltage.


It's part of a feedback loop on many cars, so the PCM listens for "just enough" knock, if you can imagine such a thing.


BROADBAND RESONANT sensors are read in Scan Data as Voltage, I'm certain that's what people mean when they say that the sensor produces a readable DC voltage. In reality.....They're just seeing the converted DC Digital signal NOT the Analog A/C Sine Wave signature that the sensor actually produces.

Another interesting fact.....Broadband Resonant sensors "ride" on a DC Line Voltage (Results of resisted bias DC Voltage produced by the PCM & the internal resistance of the knock sensor itself), The bias voltage is usually 5vdc & is knocked down by the sensor resistance (1.5vdc-4.8vdc), If you back-probe the circuit with a scope on a 10vdc scale you will see the DC line voltage as the switch point for the AC sine wave pattern, If you set the scope to "AC Coupling".....The line voltage will be read as zero & you can see more detailed amplitude.

Flat Response sensors ride on a zero volt line (Sensor Ground Circuit), You can only see half the AC sine wave signature. It kinda difficult to see this signature & not think is a just a noisy circuit!

First pic....GM Broadband Resonant.
Second pic....Delphi Flat Response.