Motorcycle Ideal Knock Sensor Placement

Hi,

I'm in the process of turbocharging my 2008 Kawasaki ZX10R. I have already purchased a few parts namely a Motec M170, which has knock control. I have watched webinar 203|Ideal knock sensor placement which has a wealth of information but would still like input from the tuning community as I am still very new to Tuning. The Japanese motorcycles don't come with an OEM knock sensor so there are limited options on where one can be mounted and would like to get feedback from more experienced people as I have never used a knock sensor before and don't know if it will even be viable for a superbike engine. The bike uses solid lifters with a rev limit of 14000rpm and the gearbox, clutch shares the same engine casings which creates a very noisy environment. Attached are a few pictures of the front and back of the engine with 3 possible mounting positions. Two on the hot side and one on the intake side where the big bolt is. I'm looking forward to hearing feedback from the community on how viable a knock sensor in this environment is and if there is a suitable placement for the sensor.

Kind Regards,

Wes

I would also be interested in the answer.

There is a reason the factory dont put knock sensors on them. I have no specific experience with ZX10R, but have tuned a few Hayabusa's, if you try to listen to one with audio knock equipment it sounds like someone has tipped his toolbox upside down inside the crankcase.

The short pistons, short rods, spur gear primary drive, spur gear dog box in the crank case, as well as chain driven cam, thin aluminium case etc all give you a nice symphony of extreme chaotic mechanical noise, but absolutely no chance of knock detection.

Just to play devil's, advocate. The BMW s1000rr comes with two knock sensors. So how does BMW achieve knock detection in a similar environment? I mean it is an indication that it is achievable. A possible solution would be to tune the engine with 0psi boost, do a dyno run with the knock sensor attached recording the noise and setting the knock threshold just above that. In essence, assuming anything louder then what the engine was with 0psi boost would indicate knock. This would, however, rule out a light knock which can't be detected above the mechanical noise.

Not disagreeing with you, just playing devils advocate to get as much information as possible. Have zero experience with knock sensors but would like to learn.

Regarding knock sensor placement; in between cylinders 2 and 3 on the intake side of the head ideally or as close to the combustion chamber as possible on the block. On Japanese bikes there isn't usually a provision so this would require a boss welding/bonding to the head or some form of bracket being made.

On the topic of mechanical noise; piezoelectric knock sensors are used in bespoke racing engines that tend to have straight cut, gear driven valve train with solid lifters so not dissimilar to a motorcycle. Using the formula =(1800/(PI()*bore dia)) we calculate and calibrate the frequency that the knock sensors listens to so everything outside of this is ignored. As mentioned above doing a pull at barro and logging the knock level will enable you to see if the mechanical noise from the gearshift is in anyway close to the range that the knock sensor listens to. its worth noting that 2 stroke GP bikes have used detonation counters (knock sensors) for decades however these do use a sparkplug based system.

Webinar 083 might be able to answer some of your questions https://www.hpacademy.com/previous-webinars/083-understanding-knock/

If you do find the audio knock detection method lacking, investigate what options are available that use a spark plug based sensor or alternatively April Systems in the UK produce a very simple m8 head stud based system intended for karting or 2T GP bikes that outputs 0-5V could something like this be repurposed?

@scotty88 Thank you for the information.

The bottom end has already been assembled and welding on it introduces warpage which is a risk I would like to avoid. Which bonding compound would you recommend for bonding the knock sensor boss to the aluminium block? Would bonding a boss to the block not dampen the knock sensors ability to detect knock?

I originally contacted Optrand to assist with a solution. They produce a pressure sensor that can be fitted to a factory spark plug which measures cylinder pressure but was informed that the device is limited to ~7K RPM and it is not of a high accuracy eliminating it for my application. The device will, however, be able to add a lot of value to vehicles with a rev limit of ~7k RPM to find peak cylinder pressure and knock detection via pressure spike detection. That only leaves the April systems if bonding a knock sensor to the engine does not work.

As you're in somewhat uncharted territory there's going to be an element of trial and error, but making the base of the boss as big as possible will help however bonding the boss on will still be a compromise. I've has some success in the past using Devweld and Araldite 2014 to bond fixings to engine castings but there may be others such as some of the Permabond products, familiarise yourself with the data sheets to be 100%.

Optrand rev limit is more than likely related to the signal processing in the amplifier, i have used Kistler in-cylinder pressure monitoring kit in the past but this is strictly a test bed exercise and the sensors themselves don't last very long, ask the F1 guys - its also the cost of a small apartment in most countries too.

So how does BMW achieve knock detection in a similar environment? I mean it is an indication that it is achievable. I suspect BMW might have had a slightly bigger development budget than you or I. For even a mild performance increase, knock control is the first thing you have to disable on the S1000rr. Same with the new harleys with ION sensing knock control, even on stock bikes that arent knock limited at all it is retarding all over the place.

A possible solution would be to tune the engine with 0psi boost, do a dyno run with the knock sensor attached recording the noise and setting the knock threshold just above that. In essence, assuming anything louder then what the engine was with 0psi boost would indicate knock. No, all of the mechanical noise level increases with torque output, so whenever you are tuning a knock threshold curve it needs to be done at a similar torque output as where you want it to work. Your knock threshold lookup would usually have some sort of load axis if there is variable boost levels used.

Using the formula =(1800/(PI()*bore dia)) we calculate and calibrate the frequency that the knock sensors listens to so everything outside of this is ignored. There are probably much smarter strategies used by some nowadays (for instance the OP's M170 can watch 4 different frequencies at once), but still I have used most of them and in my experience the human ear/brain when coupled with good knock detection audio is very good at differentiating knock from engine noise. And I have always found that if you have an engine that you struggle to determine knock from noise when listening with your ears, then there is absolutely no chance of consumer level electronics doing it.

Its worth noting that 2 stroke GP bikes have used detonation counters (knock sensors) for decades however these do use a sparkplug based system. Im interested in what you know about this. When you say spark plug system are we talking load cells or ION sensing? I cant imagine ION since they are all CDI. I have come across some vague references to sparkplug load cells used similar to ICPM before but never seen anyone actually using it. I was materials engineer for a 500GP team in 99 & 2000 which were the last years they done 2 strokes, that engine definitely had no knock detection and I dont even really remember detonation being much of a concern with anyone at the time. The ignition systems back then were crude as hell, I dont even remember it being 3D mapped, I think it was just 2D or possibly even static.

From running high end 4 cylinder touring car engines as well as single cylinder motor bike engines knock detection can be a head ache at the best of times. But when it works strategies on ignition advance can be implemented.

As Scott has said If possible to stick two sensors (if possible) between each cylinder depending on management the timing of the engine can be used to filter some of the background noise so that the sensors are only “listening” to firing events for each cylinder. So you are not just using a high and low pass filter to only listen to knock but also the timing of the knock window.

As for on plug sensors I have seen and used these in many applications; both wtcc/f1/dev work; they are very useful and give individual cylinder readings. Cost the earth and are fragile at the best of times being in a very harsh environment!

I do agree which ever option you can get to work on the engine I would have another sensor linked up to a pair of headphones to listen yourself to the engine; yes motorbike engines are noisy as hell but you can still hear knock over the top of everything.

kistler sensors have been used in WRC, Wtcc, moto gp and many many oem development situations. They do cost a small mortgage! And are extremely fragile! These are very rarely if ever ran in an engine for anything pass the development stage due to cost and the machining involved to have the sensors in the head. They are though the best way to monitor and tune internal combustion engines. As you can visually see.. injection/ignition/peak cylinder pressure/pre ignition and detonation.

Thank you for the information. I contacted Plex tuning directly to enquire about using the Plex Knock Monitor V2 with a High revving motorcycle engine. They advised that although they have customers that have used the knock monitor successfully on high revving motorcycle engines, it is difficult to say if it will work in my particular application. With high revving engines, it is probably better to look at the signal output from the knock monitor instead of the audio and configure the filters for the higher harmonic frequencies of the knock to eliminate the engine noise as much as possible.

In a separate discussion, Turbogrisen mentioned that using the plex knock monitor on a noisy engine causes a crackling sound at higher rpms. Plex advised if the signal from the knock sensor at high rpm is so high that even when reducing the volume and gain of the device, it still saturates the inputs producing a crackling sound. Then a resistor can be used to reduce the amplitude of the signal.

I'm going to follow up with Turbogrisen on fitting a resistor to get his input before I purchase a Plex knock Monitor V2. It's a big investment to risk if it doesn't work as my focus is predominantly superbikes and won't have any other use for the device.

Adam@Link_ECU which knock monitoring system did you use? Plex Knock Monitor V2 has a high pass filter and allows you to choose second-order frequency which helps with engine noise. Andre.Simon also mentioned second-order frequency in webinar 083.

As a side note there are three options for detecting knock in no particular order:

1) Doughnut knock sensor, popular and most common.

2) In-cylinder pressure sensing, which is very expensive.

3) Ionization Current Sensing, which turns your sparkplug into a pressure sensor for in-cylinder pressure sensing. Currently used by Harley, Saab and BMW.

There is limited information on Ionization Current Sensing and I don't see any aftermarket support. If anyone has used or has experience with this technology input would be appreciated. It does sound like the way forward but with no aftermarket support, it is not applicable to me. Harley tuners have not had any success with this technology as far as I know but that could be related to the lack of tunability of the sensor similar to a normal knock sensor.

Adam@Link_ECU which knock monitoring system did you use? Plex Knock Monitor V2 has a high pass filter and allows you to choose second-order frequency which helps with engine noise. Andre.Simon also mentioned second-order frequency in webinar 083.

Yes I have a plex V2 and several other consumer level knock devices. I suggest you connect it to a laptop and record a steady state test at a couple of different RPM's with both no knock and mild knock occurring, then do a frequency analysis spectrogram using gold wave or similar, you will soon find for yourself on these types of engines there is no "good frequency" to detect knock, it just looks like a scatter plot of white noise with sharp dots of high intensity noise scattered everywhere through the whole frequency range both when its knocking and not knocking.

There is a good explanation of a basic frequency analysis process in the old Motec SKM manual using Goldwave or if you want a free tool there is a similar frequency analysis tool in PC Link (Link ECU software) that does basically the same thing with a lot less options to set.

Im certainly no expert in terms of resonance, vibration and natural frequencies etc, but my theory is it is possibly something to do with resonance of all the different spur gear ratios that are working inside these crankcases. Its been a while since I last listened to knock audio on a bike engine but my memory of it was the extreme harsh noise is kind of cyclic - like a handful of rocks in a tumble dryer that is changing speed.

I suspect maybe the wide operating RPM and all the different spur gear ratios that are all working at the same time means you pass through several bands of RPM that are highly resonant and everything starts clattering around. Scott above suggested it should be no different to a race engine with gear driven cams - but of course these are driven at exactly half crank speed so they will resonate at the same frequency as the crank and you can gate it out with crank position based widowing. In a superbike engine you have a cam drive at 1:2, a primary drive at something like 1:1.85, then the whole range of different ratios in the gear box that are all engaged at the same time, that means there are a whole lot of different frequencies all crossing over each other and none of them are direct orders of crankshaft speed. I know there are also various other frequency based phenomenons in spur gear design such as hunting tooth frequency etc which I dont have a good knowledge of which may play a part too.

Adam@Link_ECU Thank you for the feedback. How would you then go about tuning a turbocharged motorcycle which is severely knock limited without being able to detect knock? Kinda sounds like a seemingly impossible task without destroying a few engines in the process to determine how far you can go or leaving power on the table.

I did notice that the Kawasaki H2 and H2R use knock sensors as a safety measure but that again falls into the budget of an OEM Developer which I don't have.

There is an engine builder/developer - RLM Racing (rlmracing.com) that specializes in motorcycle engine powered race cars. I have seen videos of their dyno runs and they run dual knock sensors while tuning. Perhaps they will share some of their knowledge.