Knock Sum discussion
#18
Evolved Member
Thread Starter
Join Date: Mar 2008
Location: Lattitude 48.38°, Longitude 17.58°, Altitude 146m = Slovakia, for common dude
Posts: 730
Likes: 0
Received 3 Likes
on
2 Posts
Any new understanding? I do believe this thread would benefit: https://www.evolutionm.net/forums/sh...=1#post6330128
#19
Evolved Member
iTrader: (1)
The system described in that patent is essentially a microphone that feeds an analog bandpass filter and the output of the filter is integrated over time. This analog signal is then passed through an A/D converter and used directly as the knock signal. If this is how the Evo's knock detection works, there's no FFT going on in the code, which wouldn't entirely surprise me. Processing power is at a premium in that controller and doing an FFT and then all the digital processing to filter out the signal takes up a lot of processor time.
#20
Evolved Member
iTrader: (9)
The posted patent is over 15 years old, I wouldn't assume same approach alltogether. Interesting thing is all subsequent related mitsu patents (later than the posted one) included a form of ion sensing within spark signal, beside the knock sensor signal which ultimately acted upon ignition retard duty to coils.
#22
Evolved Member
iTrader: (9)
Maybe it's set not to pull spark at rpm level where knock is harmless.
What do we know about the 12 filter tables? I notice #7 and several others show distinct lower values at 1500rpm and below. Would be interesting if pasting the values from 2000rpm into 1500rpm would allow knock indication when it's heard.
What do we know about the 12 filter tables? I notice #7 and several others show distinct lower values at 1500rpm and below. Would be interesting if pasting the values from 2000rpm into 1500rpm would allow knock indication when it's heard.
Last edited by C6C6CH3vo; Nov 13, 2008 at 10:18 AM.
#23
Evolved Member
Join Date: Apr 2008
Location: Sydney, Australia
Posts: 604
Likes: 0
Received 0 Likes
on
0 Posts
I would be fairly confident that the ECU is only looking for a particular frequency in amounst the knock voltage signal. The same way in which you can hear the distinct pops when you use headphones to listen for knock.
I have a couple of 2D maps that I am going to have a play with today if time permits that I have a good feeling have alot to do with knock sensitivity. The more I play with this the more I dont believe there are any 3D filters. If you have a good look around the early Evo7 roms which are quite bare of 3D maps, there is very little there that can be a filter map, yet they still use knock control to pretty much similar specs to cars that came after it.
I have a couple of 2D maps that I am going to have a play with today if time permits that I have a good feeling have alot to do with knock sensitivity. The more I play with this the more I dont believe there are any 3D filters. If you have a good look around the early Evo7 roms which are quite bare of 3D maps, there is very little there that can be a filter map, yet they still use knock control to pretty much similar specs to cars that came after it.
#24
Evolved Member
iTrader: (9)
If I get ambitious enough I might change some of the values in the 80% 2000rpm range to see wich maps might cure the knock readings I get in the area just for testing.
No scientific theory used, just the dreaded shoot-from-the-hip style I'm proficient with. It's just that there's 12 of them, discouraging.
No scientific theory used, just the dreaded shoot-from-the-hip style I'm proficient with. It's just that there's 12 of them, discouraging.
#25
Account Disabled
I disagree, if you take the DFT it is a WHOLE lot faster and is VERY efficient.
With todays techonology im almost certain Mitsu is using the DFT for knock control.
With todays techonology im almost certain Mitsu is using the DFT for knock control.
The system described in that patent is essentially a microphone that feeds an analog bandpass filter and the output of the filter is integrated over time. This analog signal is then passed through an A/D converter and used directly as the knock signal. If this is how the Evo's knock detection works, there's no FFT going on in the code, which wouldn't entirely surprise me. Processing power is at a premium in that controller and doing an FFT and then all the digital processing to filter out the signal takes up a lot of processor time.
#27
Evolved Member
iTrader: (6)
Join Date: Jul 2005
Location: Charleston, SC
Posts: 1,228
Likes: 0
Received 0 Likes
on
0 Posts
Sometimes this method is engineering at it's best!
Oh, and you need to swing by and check out the new motor, she is a pretty one!
Once I get my Sparktech ignition, she will be a monster!
#29
Evolved Member
Join Date: Apr 2008
Location: Sydney, Australia
Posts: 604
Likes: 0
Received 0 Likes
on
0 Posts
C6, I mean the address locations of the maps. Just interested to see if you are playing with maps that are really related to knock, or mislabelled as most xml's are. PM me if you would like some assistance.