Knock sensor level calibratition question

[OnegreatGuy]

Hey everyone, I've recently fitted a wideband knock sensor onto my SW20 MR2 instead of its stock resonant type. I have started calibrating the knock levels but noticed it was quite difficult to get them close together, and there isn't one that is consistently higher than the others. I just spliced the new connector wire into the old knock sensor harness (which is shielded from the factory) and ran the other cable to my transmission to the ground (this cable is not shielded). Is the variance I am seeing normal, and if not, what would the recommended ways be to fix it? I'd imagine the first step is likely to change the ground wire for a shielded one. Please note that I still have to mess around with the gain as my knock level global exceeds 1000. I am also setting up the Knock Threshold table by logging 100% throttle followed by 0, then adding 20% on top of the knock values recorded (as per David Vizard's thread). I have retarded the timing in the base map I received by 5 degrees to ensure no dangerous knock is occurring. I appreciate the help!

tuning v17 knock.pclx knock datalog.llgx

[cbDrift]

just for clarity - are you running a two wire wide-band donut knock sensor - with only one wire shielded to the knock input on the ecu - and the other wire is unshielded to your transmission as a ground?

if that is the case its not wired correctly for that type of knock sensor sorry and your guess for a first step is correct about shielded ground - but you need to run a twin core shielded wire not two separate shielded wires.

as per that thread you mentioned:
-------
1. Wiring
Run one wire to the knock1 or knock2 wire on the link loom, and one to sensor earth, polarity unimportant. must must must must use shielded wire. The knock sensor outputs a very low voltage signal that's prone to interference.
-------

there is a specific pin on the ecu called sensor ground - dont use a normal ground

and both wires from the knock sensor need to be shielded, not just one wire - use twin core shielded cable for the knock sensor and ground the ecu end of the shielding to the sensor ground as well

this should give a more consistent signal in your logs to work with all going well.

[OnegreatGuy]

6 hours ago, cbDrift said:

just for clarity - are you running a two wire wide-band donut knock sensor - with only one wire shielded to the knock input on the ecu - and the other wire is unshielded to your transmission as a ground?

if that is the case its not wired correctly for that type of knock sensor sorry and your guess for a first step is correct about shielded ground - but you need to run a twin core shielded wire not two separate shielded wires.

as per that thread you mentioned:
-------
1. Wiring
Run one wire to the knock1 or knock2 wire on the link loom, and one to sensor earth, polarity unimportant. must must must must use shielded wire. The knock sensor outputs a very low voltage signal that's prone to interference.
-------

there is a specific pin on the ecu called sensor ground - dont use a normal ground

and both wires from the knock sensor need to be shielded, not just one wire - use twin core shielded cable for the knock sensor and ground the ecu end of the shielding to the sensor ground as well

this should give a more consistent signal in your logs to work with all going well.

Thanks, I figured I had something wrong. I will purchase some twin-core shielded cable.

[Adamw]

I would first at least half the main gain and also try 7 & 8KHz frequency before jumping to the conclusion it is a wiring issue.   Your input is saturated and you may just be listening to a particularly noisy frequency.  

[OnegreatGuy]

9 hours ago, Adamw said:

I would first at least half the main gain and also try 7 & 8KHz frequency before jumping to the conclusion it is a wiring issue.   Your input is saturated and you may just be listening to a particularly noisy frequency.  

Just did as you said, and lowering the gain and changing the frequency definitely improved the quality, however there is still very clearly some noise. Also, this is being used on a 3sgte with bore 86mm, which should have a knock frequency a little under 7kHz. 8kHz seemed to have the smoothest results, but having this be more than 1kHz off seems unideal.

 

I also tried 13kHz to try to get the second harmonic as suggested by the online manual, and that seemed even better. However, there still seems to be an unreasonable amount of variance, though I won't pretend to know what is acceptable. Other logs are in this drive folder as I ran out of upload space, as each log includes a steady 4k rpm and then some light, medium, and heavier driving to cover all bases.

7kHz 3 gain knock log.llgx

[Adamw]

On 7/7/2024 at 9:22 AM, OnegreatGuy said:

Also, this is being used on a 3sgte with bore 86mm, which should have a knock frequency a little under 7kHz. 8kHz seemed to have the smoothest results, but having this be more than 1kHz off seems unideal.

The calculated frequency isnt all that relevant, it is better than nothing, but all it tells you is the theoretical prominent knock frequency, it doesnt take into account all the other mechanical noise an engine makes.  What is most important is the signal-to-noise ratio - i.e there is no point listening for knock at 6KHz, if your mechanical noise is also very prominent at that same frequency.  An example frequency analysis below from an EJ205, the calculated knock frequency for it's 92mm bore would be about 6.2KHz (frequency is the Y axis), you can see there is certainly plenty of knock noise at that calculated frequency, but there is also plenty of similar looking noise at that same frequency on the left hand end when it isn't knocking.  The best signal to noise ratio (best distinction between knock and no knock) is about 8-8.5KHz on this engine. 

Looking at your logs, I would try more testing at 8KHz, balance the cyl gains a little better, then make it briefly knock by putting a large advance value in an ign table cell where loads arent huge and that you pass through fairly quickly (say 3000rpm, 120kpa) to confirm you get a spike when knock occurs.  

 

[DenisAlmos]

Does all the EJ205 engines have the same mechanical noise frequency?

I'm also running an EJ205, can I be sure to filter knowk at 8kHz?

[Adamw]

Not necessarily, even things like exhaust system and engine mount stiffness can change what frequency gives the best signal to noise ratio, but it will quite likely be in the same ballpark.  

[DenisAlmos]

So best way to identify knock frequecy is to use a sound spectrogram software and test the knock sound with a large advance value?

From HP academy, the double of the calculated knock frequency (about 12.5kHz) is a good starting point.

[mapper]

Fft analysis function is built into pclink. 

[Adamw]

9 hours ago, DenisAlmos said:

From HP academy, the double of the calculated knock frequency (about 12.5kHz) is a good starting point.

If you look at the spectrograph in my post you will see 12.5KHz would be useless in this case.  I would start with 8K, do a log of a pull, then try the same at 7K and 6K, and compare logged knock levels, it is generally quite clear in logs which is giving the best signal to noise ratio.  

[OnegreatGuy]

Sorry to revive this, but I've been struggling with this knock stuff for a minute now. I ended up making a DIY knock box out of another knock sensor and a microphone amp, and have been recording the output produced to try to see if I can find the frequency with the best signal to noise ration. However, to my untrained ears I can't hear any knock, even with 45 degrees of timing and 8 psi of boost at 3000rpm. Even when I slow down the audio, It's hard for me to pinpoint whats normal and what could be knock. Would someone with better trained ears mind taking a listen to my audio, and confirm or deny whether I have in fact induced knock? I have included pictures of the waveform of the two recordings, as well as the spectrograms in this post and the following post. Audio recording, and the 11 degrees of timing spectrogram can be found in this drive.

Waveform of 45 degrees timing, 3000 rpm, 11.7 afr, and 8 psi of boost

Waveform of 11 degrees timing, 3000 rpm, 12 afr, 8psi of boost

 

Spectrogram 45 degrees timing, 3000 rpm, 11.7 afr, 8psi  boost

 

 

[Adamw]

Your audio isnt great, I think possibly the MP3 compression has lost some quality or perhaps sound card settings were a bit low, but it is definitely knocking heavily in the 45deg recording and Im pretty sure still knocking in the 11deg one, just less severe.  I always stick with .wav.  In future I would also suggest you change the Ign timing while recording so you have an example of knock and no knock side by side under the same conditions so you can see what background noise looks like when it is not knocking.  I usually set up a DI like the AC switch or rear demister to switch a 4D ign table on/off to make it knock.  

I reckon about 6.5kHz looks good for this one, those sharp vertical lines are where I hear crackles of knock. Since you cant set a centre frequency of 6.5 in the software, I would go for 6.0kHz as the SNR is better on the low side than it is on the high side.  there is a decent SNR all the way down to ~5.5K, but at 7kHz there is more noise and less signal.    

 

The settings I used to generate that spectogram: