Best Knock Sensor Position ?

[V8RX7]

Hi, I'm playing with a supercharged MX5 with an aftermarket ECU.

The knock sensor was fitted as per the instructions (low down on the block) and wasn't detecting any knock even though I could hear it.

I can't understand why it isn't working in the usual position as it's a very popular and highly regarded kit but I have tried a spare sensor with the same results.

I spoke to a Bentley mechanic and he told me that on Bentleys they are mounted on the intake manifold and the best place for them was the centre of the head.

I moved mine to the centre of the camcover with a bolt that connects into the head.

The problem I have now is that it's detecting too much !

IE it's hearing "noise" rather than knock - it's pulled about 20 degrees more than similar spec cars.

I can adjust the sensitivity but a quick google shows people saying the knock sensors mounted on the head have to be deactivated over approx 4500rpm due to too much noise

Has anyone had experience of this ?

[stevieturbo]

The problem isnt the sensor or location.

The problem is you havent calibrated the sensor within the ecu for your engine.
You will need to set the noise thresholds manually to let the ecu know what is noise and what may be knock.

[V8RX7]

It's a fairly old ECU the only adjustment is the threshold level.

It was previously set at 20 and I raised it to 35.

Obviously I can raise it further but the instructions were to freely rev the engine and set the level above that noise.

The highest I got was 20.

I suspect if I set it much higher it won't be detecting low levels of knock.

[stevieturbo]

That's the thing though, it is a huge variable. The original manufacturer would have spent plenty of money and time to make it be functional on a standard engine.

Make changes and you can run into problems.

[V8RX7]

As Std there was no knock sensor (nor supercharger)

[Evoluzione]

V8RX7 said:

I spoke to a Bentley mechanic and he told me that on Bentleys they are mounted on the intake manifold

Try this maybe?

Surprised it isn't picking up from the block, they are quite often there or on a solid part of the head (not the cam cover) as standard.
I was having one mapped last week and I noticed he fixed his to the intake manifold.

[V8RX7]

I guess I'll try moving it around.

I was hoping there was a more scientific approach but with cars a lot seems to be suck it and see.

Thanks

[dblack1]

I don't see what is wrong with the original mounting location (between the IM and the oil filter)....
Mazda engineers put it there for a reason (I don't profess to know what goes on in their heads, but they are way better at designing engines than anybody in this forum). Unless your cool new supercharger setup is interfering with the sensor I wouldn't move it.
I wouldn't disable the sensor while using a supercharger either, that's a recipe for disaster.

[V8RX7]

dblack1 said:

I don't see what is wrong with the original mounting location (between the IM and the oil filter)....
Mazda engineers put it there for a reason (I don't profess to know what goes on in their heads, but they are way better at designing engines than anybody in this forum). Unless your cool new supercharger setup is interfering with the sensor I wouldn't move it.
I wouldn't disable the sensor while using a supercharger either, that's a recipe for disaster.

V8RX7 said:

As Std there was no knock sensor

Mine is a Mk1

I was really trying to find out why you can't put it on the head or if there is an optimal position.

OEM frequently have their hands tied by manufacturing / cost constraints that the aftermarket don't.

[stevieturbo]

V8RX7 said:

Mine is a Mk1

I was really trying to find out why you can't put it on the head or if there is an optimal position.

OEM frequently have their hands tied by manufacturing / cost constraints that the aftermarket don't.

OEM spend a hell of a lot more money fine tuning their systems than any aftermarket system will ever come close to.

They will site the sensor where they deem best, and then they will also spend hundreds if not thousands of hours fine tuning the entire setup so it works.

Aftermarket just give you some parts and let you work away.....as you're trying to do now.

[V8RX7]

stevieturbo said:

V8RX7 said:

Mine is a Mk1

I was really trying to find out why you can't put it on the head or if there is an optimal position.

OEM frequently have their hands tied by manufacturing / cost constraints that the aftermarket don't.

OEM spend a hell of a lot more money fine tuning their systems than any aftermarket system will ever come close to.

They will site the sensor where they deem best, and then they will also spend hundreds if not thousands of hours fine tuning the entire setup so it works.

Aftermarket just give you some parts and let you work away.....as you're trying to do now.

They don't.

With a cheap ECU I can get more mpg AND more power than OEM

Because I don't have to worry about fuel standards in India or Emissions tests etc

Similarly OEM count every penny - if it's 10p cheaper to mount it on the block instead of the head - it will be mounted on the block.

They DON'T make the car the best they can - they make the best compromise they can within their budget, time and legal constraints.

[stevesingo]

V8RX7 said:

They don't.

With a cheap ECU I can get more mpg AND more power than OEM

Because I don't have to worry about fuel standards in India or Emissions tests etc

Similarly OEM count every penny - if it's 10p cheaper to mount it on the block instead of the head - it will be mounted on the block.

They DON'T make the car the best they can - they make the best compromise they can within their budget, time and legal constraints.

If you want to discuss the cost/benefit balance of the OEM's choice of knock sensor position, then consider warranty claims. Sub standard positioning of the knock sensor could render the knock control less effective than optimal. The OEM would then forced to have a sub optimal ECU calibration and miss those emission targets you are able to avoid. Or, they push the limits to meet the emission standard and risk warranty issues down the line and this is far more costly than doing FEA study on a model engine, refining the positioning of the sensor correctly and calibrating the engine correctly.

I have been out doing some preliminary work for the mapping, setting the centre frequency for the Motec SKM knock control module.

This is what it looks like on the frequency analysis.



What I was looking for is a frequency of the knocking which stands out the most against the background noise. As you can see, about 6kHz.

As you can see on the Motec log, the SKM picked up the knock, initially on knock 3 (Cyl 4 as it is in firing order). The SKM did not apply any correction as the threshold for a response was set deliberately high so the knock wasn't turned off before my ears could pick it up.



An internal adjustment to the SKM sets it to 6kHz.

When on the dyno, we can now log background noise during mapping and set the threshold for response appropriately to background noise.

If your system doesn't allow you to select the frequency, you are just listening to noise, not the specific noise you want to hear.

This is on Motec ECU, which is way way less sophisticated than OEM ECUs.

Perhaps Max_Torque can chime in...

[stevieturbo]

V8RX7 said:

They don't.

With a cheap ECU I can get more mpg AND more power than OEM

Because I don't have to worry about fuel standards in India or Emissions tests etc

Similarly OEM count every penny - if it's 10p cheaper to mount it on the block instead of the head - it will be mounted on the block.

They DON'T make the car the best they can - they make the best compromise they can within their budget, time and legal constraints.

You seriously underestimate what the OEM do, and how much money they spend.

On the flip side...if you think they do everything so cheap...how come they can make a knock sensor work s well and you cant with your cheap ecu lol.

So perhaps of more importance...exactly what ecu are you using, and what parameters within knock control can you alter ?

[V8RX7]

stevesingo said:

If your system doesn't allow you to select the frequency, you are just listening to noise, not the specific noise you want to hear.

Yes I'm aware that's the issue I have.

However I've been running 3 MX5s with this old Link ECU at close to double std output without issue for 5+ yrs so I'm not going to upgrade just yet albeit I'm aware there is much better kit out there.

As engines are available for £100 I'm not to concerned.

[stevieturbo]

V8RX7 said:

Yes I'm aware that's the issue I have.

However I've been running 3 MX5s with this old Link ECU at close to double std output without issue for 5+ yrs so I'm not going to upgrade just yet albeit I'm aware there is much better kit out there.

As engines are available for £100 I'm not to concerned.

Some of the old Link's are very primitive indeed. They may be able to detect knock, but IMO any control they offer is too slow for engine protection.

[Mr2Mike]

The knock sensors themselves usually have a fairly narrow bandwidth that needs to be matched to the frequencies produced by detonation, which depends on the construction of the block, i.e. one sensor does not fit all. The ECU has to window the sample period to the range of crank angles in which detonation is likely, and filter the specific signals from the noise. It really is a far more complex system than simply looking for any signals above a certain threshold. I suppose you could always stick a spectrum analyser (or digital scope with FFT function) to see what you are actually getting from the sensor during det and use an external analog filter to remove other signals.

[Richyvrlimited]

stevieturbo said:

Some of the old Link's are very primitive indeed. They may be able to detect knock, but IMO any control they offer is too slow for engine protection.

You can't really compare an old LINK to the one Mark has installed on his MX5.

The ECU he has is the FMLink, this had the code re-written specifically for the Miata/MX5 for a company in the states called FlyinMiata - the guy who wrote the code was called Ray Alya IIRC. It's a much *much* more capable ECU than the standard LINK of the same era.

The knock chip was on a seperate board from the FMLink ECU housed with if I remember properly the MAP sensor too.

The frequency that knock occurs as I understand it is a function of the bore size, this is hard coded into the FMLink ECU - they were year specific and in the states only 1989 to 1993 got a 1.6 all other cars were 1.8, all you need to adjust for knock detection is the threshold.

Mark I think your best bet is to try the OEM knock sensor position of the later cars and see where you get. it could be that this particular engine is very noisy.

[Richyvrlimited]

Mr2Mike said:

The knock sensors themselves usually have a fairly narrow bandwidth that needs to be matched to the frequencies produced by detonation, which depends on the construction of the block, i.e. one sensor does not fit all.

That depends entirely on the knock sensor, Flat response sensors have a very wide range, but need more sophisticated hardware to analyse the data coming in - it's this sensor that comes with the FMLink, and it has complementary hardware for it.

Resonant sensors are as you state, they have a narrow range that they work in and are very application specific.

[V8RX7]

Richyvrlimited said:

The frequency that knock occurs as I understand it is a function of the bore size, this is hard coded into the FMLink ECU - they were year specific and in the states only 1989 to 1993 got a 1.6 all other cars were 1.8, all you need to adjust for knock detection is the threshold.

Mark I think your best bet is to try the OEM knock sensor position of the later cars and see where you get. it could be that this particular engine is very noisy.

AH !

You might have just solved my issue.

This was a 1.6 link that was converted to a 1.8 I wonder if there are any knock differences.

The starting point was that in the std FM position it detected virtually nothing (1 maybe 2) I moved it to the cam cover as a quick test really and there it detects up to 20 revving in neutral - which is about what it should get.

Problem I've had is that I'm still playing with my house and since posting about the issue I haven't had a chance to experiment with other positions.

[Richyvrlimited]

V8RX7 said:

AH !

You might have just solved my issue.

This was a 1.6 link that was converted to a 1.8 I wonder if there are any knock differences.

The starting point was that in the std FM position it detected virtually nothing (1 maybe 2) I moved it to the cam cover as a quick test really and there it detects up to 20 revving in neutral - which is about what it should get.

Problem I've had is that I'm still playing with my house and since posting about the issue I haven't had a chance to experiment with other positions.

The bore size is slightly different so yes that'll have an effect