Actually my experiance is that they pick up less flase knock beacuse of the higher density of the steel compared to the aluminum.

 

Has anyone had any success in reducing the sensitivity of the Knock Filter for a 2.3L forged Motor - it would beat switching the Knock Control off altogether...

Also doesn't make sense to buy an aftermarker ECU to get rid of knock control if you can just switch it off on the stock ECU...

.

 

Without going into too much depth (and seeing the maps for the first time) the knock sensor is basically a ruggedized microphone. It picks up sound vibrations. I suspect the computer is doing a rudimentary FFT (Fast Fourier Transform) into the frequency domain and looking for key amplitudes on certain frequencies. When the amplitude at a certain frequency gets above a specific threshold, it counts that as a 'knock' event / count.

If anybody wants me to delve a bit more into frequency analysis I'd be happy to expound, otherwise I need to bury my head in some maps (as I've finally got my Tactrix cable) and see if the information here correlates with what I think is happening.

~j.

 

I wouldn't waste too much time on the "knock filter" maps. I've played around with them, to the point of inputting ALL extremes for knock values, and it made absolutely no difference in how much knock was being picked up. This was one of the main reasons why I ended up switching over to a standalone...

l8r)

 

once someone bothers to understand them I think they will become useful.

but the need just isn't there yet

 

True that, like markus says most just end up going stand alone - I might eventually also go Autronic...


I was actually going to try copying over the Evo 8 Knock filter maps to the Evo 9, as someone said on one of the threads that the 8's are much less sensitive to knock than the 9's...

.

 

IIRC, the only one that has made any inroads to the stock ECU knock routines is Bez. And I believe even he just ended up disabling it, since it was a PITA to try to tweak.

l8r)

 

This is very interesting. Does anybody know which of these is used? If they are octaves, it is possible that the bits add up to make the final frequency. Like if it was 0,0,0,0,1,0,0,0,0,0,1,0 this could be 110Hz + 7040Hz for a 7150Hz center frequency on the filter. With a broadband knock sensor you NEED to define a bandpass filter (or filters) to look for the frequency knock actually occurs at.

What is the OEM setting on the knock sensor filter entry?

Kevin

 

For those with built engines (no balancer shafts, strokers or noisy valvetrain etc) having problems with part throttle low load knocks that can't be tuned out no matter what, turning off the knock control of the stock ECU is a cheaper option than going to a standalone with no knock control. Obviously, the best option is to have a tunable knock control based on LOAD/RPM. If this option becomes available one day, I intend to turn off knock control under the load of 100 and 3500RPM.

 

My comment on that above is once fairly large mechanical "stuff" goes into the 4G63 the frequencies exhibited by the motor -- change.

It would be very useful to understand the actual code that is run in order to do knock sensing.

Of course everybody and their dog can get the maps out of the EVO, but really only Mitsubishi has the runtime code.

Any comments on runtime code? i.e.

My first contribution is the EVO uses the SH7052 - SH2 Based SuperH Hitachi Processor. I have experience with the SH3, and hope the SH2 is about the same.

~j.

 

Actually, it shouldn't. Knock frequency is determined by bore diameter and combustion temperature. Combustion temperature is, pretty much, just a function of AFR.

I'm most curious if Mitsu went after the primary knock frequency or an upper octave of it. In the 1g DSM's they went after an upper octave. I've only seen knock at the primary frequency though.

Kevin

 

I suppose I don't fully understand why the frequencies would not change?

My understanding of frequency "generation" in the mechanical world in this application is this. The cylinder comes up the bore, the A/F mixture hits a hot spot and pre-ignition occurs at a point where mechanically the piston doesn't want to go back down the bore. The pressure goes very high and all of the mechanics in the system feel that pressure spike. If its strong enough, we as humans hear that as "pinging" but if there is a microphone buried in the motor, the microphone electronics can pick that up at a much lower level, if interpreted right, be able to react to save the motor.

That rise in pressure is pretty close to a step input (low pressure to high pressure in a very short about of time). Step inputs excite the "frequencies" that make up the motor.

Think of the motor as a literally a bell, bells resonate at certain frequencies. If you get say a vibrator out, hook it up to the motor and adjust the frequency of the vibrator over time, at some frequencies you'll find the motor "resonates." My thoughts are stroker and serious modifications change these "sympathetic" frequencies, and the stock electronics are not "tuned" to find these frequencies.

This step input is also in part akin to taking a hammer and tapping on a motor and going, "wow ... this thing "rings"." Finding where it rings is the same thing to finding the sympathetic frequencies.

Therefore if there was a desire to use the stock electronics to find "knock" with new innards, I don't think it would be done through a table, but rather we'd have to be digging in and around the code that looks for these frequencies, and change where they look.

0.02

~j.

 

Thanks j, you made half the explanation for me!

The 'thing' that is ringing is the combustion gases across the chamber. Every conventional knock system I've seen is tuned for the wavelength that fits across the chamber diameter. It is the resonant frequency of the major dimension of the chamber during which knock occurs. Since knock happens near TDC, the largest dimension is the bore diameter.

Knock sensing does not measure the frequency of any metal part of the system because things like oil clearances greatly influence the noise levels they produce during normal operation. Those frequencies are also always there. To effectively measure knock, you need to pick out something that only exists during knock. That turns out to be the resonant frequency across the chamber. Now there are a few frequencies that show up during knock events, 1/4 wave, 1/2 wave, etc. My question is, which one is mitsu listening for?

If somebody knows the contents of the OEM knock filter, I'd appreciate if they would post it.

Kevin

 

Good insight Kevin. I suppose when the piston is upwards of some thousanths before TDC you're right that bore diameter would look like a resonant 'pipe' to the shock pressure wave of a detonation. Good call ... it sounds like I need to pick up a book or two on motor frequency analysis.

Back on the algorithm at large. I don't have my digital signal processing book in front of me, but if the knock system is being implemented via a SuperH Hitachi processor (the SH2) series, it probably doesn't have tons of computational power for a fancy dancy frequency finding algorithm. My guess is they're using a simplified FFT that looks for a few select frequencies. Has anybody been able to start cracking into the runtime code at all on the Mitsu. EVO computer?

Thanks,

~j.

 

So, if I take down the units of measure from between 30 and 80 TAR, and 2500 to 4000 RPM to zero for instance, will it help me to see less of that P0300 error code? Do the TAR units measured equal load?