Have any EcuFlash guru's made any more headway into the knock filters. Specifically reducing the sensitivity. My car has BS removed and is experiencing phantom knock. (it's definitely phantom and it's not the dp) There is speculation by others that removing the BS causes harmonics which affect the knock sensor. I do not want to turn the filters completely off; just reduce sensitivity at certain rpm/tps points. Any ex-dsm'ers have any tricks for the knock filter? I think this was a pretty common problem for the 1g cars. In another thread, someone mentioned that the old dsmlink would alter the knock sensor threshold... Anybody had any luck with the fluidampr cancelling out harmonics/phantom knock? Any helps/thoughts would be greatly appreciated.

 

DSMLink just added code to turn off the knock sensor based on values of TPS and RPM, so that you can tune out the unwanted phantom knock.

For example, you can have the knock sensor off for anything under 20% throttle and 3000 RPM.

I'm sure someone like jcsbanks or Bez on Aktivematrix could write some quick code to be able to do this. The location to turn off knock control was already found. So writing some logic to control should be doable.


Eric

 

I know DSMlink has it, but unless I want to socket the evo ecu for a 2g eprom that doesn't really help. Anybody brave enough to start experimenting with the knock tables? Tables (on my '05 VIII) 1,2,3,5 seem to be the biggest knock degradients on my ecu. Theoretically, if you know that it is phantom knock, could you edit each map one by one (reduce entire map by -5%) until the knock goes away. It seems that some of the knock that people describe (tip in, decel, etc.) could be avoided.

 

You don't have to adjust the entire map, just the RPM and load range that the ECU is detecting the phantom knock at. I've been running into the same thing on my Evo - built 2.0 with no balance shafts. Once the weather clears up a bit and I get some extra time I will be testing knock filter adjustments. Had some potentially good results the last time the car was on the dyno, but it needs to be tested in the real world to make sure that it works the way I think it does...

l8r)

 

I have tried many ends of the hex code spectrum and it does not seem that changes in low rpms knock filters make any difference. From 0-255, I have seen no difference in these results.

Straight to ****ing 25-36counts without bshafts seems to be common in the 2500-4000rpm range. After 4000rpm phantom sums turn off like a switch.

I have changed many parts to try to improve the results. Downpipe, hangars, motor mounts changed back to stock all for a bit more comfort. Unfortunately high phantom knocksums is uncomfortable at part throttle cruise.

I will likely wrap the DP today which probably not change a thing but maybe a little bit of exhaust twang.

 

Maybe a better solution would be to add some code to the stock ECU to either be able to control the knock sensor or just to turn it off below certain thresholds.

I remember on my Eclipse with DSMLink, I would get the part throttle phantom knock, too, but they added code to the stock ECU that enable control of the knock sensor with parameter such as TPS and RPM. So, for example, I could essentially turn off the knock sensor for anything less than 35% throttle and 3500 RPM, where I knew that I wasn't getting any real knock.

I think that Bez, John, or someone on AktiveMatrix that is good with disassembly and code may be able to write logic similar to this.


Eric

 

I thought from reading the 9417008 disassembly thread that the filters actually seemed to be timers or something of the sort.

One thought is to create a circuit to attenuate the signal, the exceed has this logic built in and attenuates the knock sensor signal by some percentage.

I've been wanting to play around with the filters myself, but my green is on order and I doubt I will have time to retune my car before the green arrives. The green cars I have tuned were much less knock prone than stock turbo cars.

 

I have played with the filter maps but I decided to go back with stock knock sensor map. I think stock knock senser filter map is very trustable. I thought I was getting phantom knocks but I was able to tune most out by spending more time with entire load range.

My car has totally different timing and fuel values on lower load and it looks nothing like the stock map. I had to change closed loop operation threshold values to avoid low load knocks. My lower load cells act more like higher load cells in the stock map especially around 2000~ 3500 rpm.

However, there are real annoying phantom knocks that doesn't reponds to timing or fuel change. If that happens, you can actually make it less sensitive. But, you have change all 12 maps. best thing to do is copy the higher load cell values to lower problemetic load cells.

 

I cannot see this as a feasible option. You would not have adequate ignition timing at low load and fuel mileage would suffer.

 

Glad to see that you guys are working on it. Can anyone explain the knock tables yet? Are the different tables for frequency? It would be nice if someone at MMNA could leak some info regarding deciphering the ecu.

 

This is why I wish the automobile would make it DSP technology, that way you can look at the frequency domain of the filter signal and therefore design your own digital filters to fix your problems. Matlab is a prime example of the power of it.

 

There's like 5 arrays that's not defined by the originator ..

It's something like this ..

for knockvolt = 1 thru 5
If knock voltage over this value 1 thru 5
compare rpm and preset voltage
if => preset voltage jump knock subroutine 1 thru 5
end compare
endif
end

initial voltage starts around 1.5 I think .. and last one is 5v ..

 

I wrote a matlab code to look at frequency domain of signals such as the filters and such and designed my own filters, convolution integrals to experiment ect...
but kinda sucks that you have to rewire tons of crap and have a laptop in your car at all times to use it.

 

Think you can junk the codes and redesign them .. hehe

E9's have only 2 filters .. compared to the E7s and E8s ..

Strangely E9's have alot in common with the E7.5s and the E9.5 autos ..

The 2 knock filters and the 6 knock reference maps

Oops .. forgot the referenced voltage .. there's 6 of them .. sorry .. later I'll add them in

 

Ironically, your gas mileage will suffer if you leave those phantom knocks since you will be dragging the octane value down most of the time.

Anyway, your gas mileage wont suffer very much. the knock sensor seems to be tunned for the stock power level. you are knocking at the part thorttle because you are making more power at the same part throttle load level.

Bringing down the power level to stock will make the stock knock sensor and your engine pretty happy. Your engine needs a cool down time and that should be the light part throttle time.

I suggest modifying your timing before you start blaming the knock sensor. I was able to solve it using the lower timing numbers or changing that region to open loop and add more fuel.