RomRaider

There is no knock sensitivity table.

I'm almost possitive it sets a bit physically rather than through data.

I've been looking but its set with either frequency or voltage (Similar to a switch)

For those of you interested.

I found:
Idle Timing Compensation for RPM Error:
Basically it is timing added or removed based on Current RPM vs Target RPM at idle.

Idle Timing Compensation for Load:
Basically it is a table (currently 0'd) used above 176*F ECT that the value is subtracted (or added) to Idle Base Timing.

I also have the threshold for maximum WGDC used by the ecu (which is capable of up to 100%) but currently set to max at 90%.
^ For SD Rom only. More or less if WGDC Current Target is @ or above 90% WGDC the ecu automatically forces 90% WGDC and no more.



IF anyone has any other suggestions on things that will help tuning, please do so soon. I have some free time and currently going through tboth the 02 Maf Rom and the Speed Density Rom.

I do hate to break it to you that there is no Knock Sensor "control" that I can find anywhere in any rom...

-Jerod

Ok just to confirm I found "read knock sensor" - It is just a ton of and/or/if and bit sets.

Basically its taking a byte or sequence of bytes (most likely set in ram) and manipulating them and deciding if the knock sensor byte is set or clear.

Set=knock
clear=no knock

I don't think in any sensible way it could be manipulated...
-Jerod

A couple of teasers...

This is an 02 Rom (E01C)



Those are Idle Timing Comps.



These are alternative OL Fuel Tables used under a condition that I have no verified yet. I'm almost certain its Maf Failsafe related (IE Maf Disconnected) or similar.

it was hypothesized years ago that the knock detection is actually hardware based DSP on a separate chip.

it is curious after all these years neither cobb nor ecutek nor OS has found any knock sensitivity mapping, and would seem to reinforce that hypothesis.



i've had my hands in this table for a few years now.

think of it like a short-term ignition correction based on delta RPM feedback.

i had hoped its cycle was long enough to provide a kind of real load-based ign correction. ie, reduce advance when the rpms are changing more slowly, as they would be in higher gears, a feature found in some advanced standalone emss. alas the clock is too quick for that.

those of us with lightened flywheels/crank pulleys can safely reduce the values a bit. reduce them too much and you'll get idle rpm hunting.

it retards when rpms rise and advances when they fall. this is a basically instantaneous action and works in concert with the IACV (which is a much slower control) to provide a nice stable idle.

Interesting - Did you define the table?

Wow this is bloody fantastic ....

*I* did not.

a little bit of code did it for me.

Any sugestions for things that you feel could exist but are undefined?

Dammit
so if a knock sensor works like a Mic and sends the ecu a signal base on 0-5 voltage, that at one point
i guess was how knock was detected, but the stock ecu doesnt use such a strategy then denso is tryng too hard to fix what wasnt broken.
oh well back on topic.

BTW cobb has a missfire treshold table base on RPM and load

that sounds like the a knock learning table to me.
Jsarv is good to have you helping out sick job dude

I haven't seen a misfire threshold but I'm sure its there.

The knock sensor is most likely frequency style sensor rather than voltage, which technically could be much more accurate.

There is enough undefined stuff in the 02 rom to take me a year to define.

That is why suggestions for observed or known comps or weird actions are always usefull.

-jerod

Does it look something like this? The numbers are probably different as I haven't applied an expression to that, but I'll bet it follows that pattern huh?

Andy

Yes you have

You're right, it's based on frequency. It is a piezo-electric sensor that detects vibrations, i.e. a speaker. It then converts that frequency into an electrical signal which is then read by the ecu. Somewhere, there has to be a conversion, as well as a threshold, i.e. a sensitivity setting. There isn't much info on the Subaru knock sensor, but from this article, http://endwrench.com/images/pdfs/ModelsInfo.pdf , we can assume that it has the same properties as any piezo-electric sensor(reads vibrations, turns them into electrical signals). In other knock sensors, such as http://www.arduino.cc/en/Tutorial/KnockSensor , there is most definitely a threshold to be met.

YES!

Andy

Can someone post a screenshot of the dwell table? I'm curious...

Thanks!

There is one here topic5577.html

Andy