The timing retard value that we log doesn't care what the "source" was. If there is a value shown, then that is what is pulled from the calculated timing value.

Why can't you guys just go outside and try the experiment? Is it really that hard or do you just prefer to throw ideas around without having to be responsible for them?

I did do the experiment, and yes there is a knock detection when hit hard enough.

but when my car pulls timing at 6000rpms there is no knock detection!!!

How do you know it is pulling timing?

some things just aren't meant to be ....

http://www.chapso.de/galpics/2010/12...1291868783.jpg

The datalogs, that are posted in this thread....

you continue to miss the point, entirely ...

Enlighten me,

Btw I emailed you my timing map and u never replied....

edit: apology my bad, sent you email

.

You show the timing being pulled, but you don't show anything that indicates it is knock. Because it isn't.
I already explained to you what is happening back on the 18th.

Whats wrong with the IAT?

Those tables are looked up based on the difference between measured RPM and desired RPM. Desired RPM looks to be set by the the idle tables, or when off idle it is set to RPM itself. I don't know it this ever actually amount to anything like a delta. No reason to speculate though, you can log desired RPM with any scan software that has the Mazda enhanced parameters. If anyone uses Torque for Android, it is mode 22 pid 0x96e. Scaling is x/4. I broke my Elm327 dongle or I'd do it myself.

In any case, this correction seems to be mainly for controlling hunting idle.

Desired rpm isn't used for much else, but it also looks up base idle timing. I'm not really sure why you can cause feedback knock retard at idle, but it is tangential to the candidate issue. It does highlight one important point: 0.5 load is the maximum for feedback knock retard. This is mislabeled in my version of AccessTuner.

Meanwhile, one of the knock learning strategies only starts at 0.5 load. There are also IAT limits for this stuff, so you could probably disable the whole system by miscalibrating it...

I think we need a utility like Learning View to look at the learned parameters. If you're satisfied with what has been divined from crescent wrenches and defended with hand-waving, I have not a fuck to give.

What FazdaRX_8 have to show in order to indicate that it is knock and not just timing being pulled ?

At the end of the day, that is all that matters to anyone that is tuning these things.
If you are only interested in parametric masturbation, I have an equal quantity of fucks to offer.
The actual knock decrement value, which he has already said shows zero.

so anything above .5 load would not be logged in knock retard?

? Bump

I have been experiencing this very issue and based on all my available data, the knock sensor is the culprit.

Looking at the code oltmann posted, there does appear to be an error present with regards to the logging capability. If I'm reading it correctly, any timing retard value less than 10 will always show as zero. It just so happens the max knock retard value is -10.

I have been running a very conservative timing map so my issue would compound itself. The sensor would think it hears knock (due to unknown reasons), pull timing which would then actually create a misfire causing further timing to be pulled. IAT's were all normal during this time and the RPM delta tables were zero'd which did not solve the issue.

After reading a post previously made by Kane regarding the knock tables, I zero'd all of them and the problem has gone away. No misfires and my timing is being set to what I have in my base table.

For all I know I have a failing left motor mount that is causing the sensor to freak out.

I think the jury is still out on this issue but removing the knock tables from the equation completely fixed the issue for me.

Good work Flash .
Is this under boost or NA ?

I always limit the knock retard value by about half anyway and don't see any misfire either .

I am keeping mine stock,

I mean if you flat line them all and you get bad gas, what keeps your motor from blowing?

Proper tuning ;)

Yeah, where do you get one of those?

This is under boost.

Something has clearly changed in my setup as I never had this problem previous to doing the Turbo rebuild. For all I know I damaged the driver side motor mount with all the moving of the engine and the knock sensor is picking it up.

While removing the knock sensor's capabilities might not be ideal, no one should depend on it to save your engine. With the sensitivity to knock that the rotary has it is very possible you would destroy something before the PCM had time to react in the first place.

I have been looking into the mazdaspeed 3 ap, and knock stuff. Cobb gave up early on us. The knock sensor I think works fine, we just cant see what is acually reported. Other then logging timing and compare to comanded.

I do think that the kr system can save an engine from predetination, due to bad gas, that is why mazda has the system, incase someone put in 87octane.

The only way to get a "proper tune" for it is to tune with bad gas...

Not at all. You can build in safety margins within the tune to account for poor quality fuel. That said, the best thing you can do is ensure you use quality fuel with octane ratings that can be accounted for and stick with it.

I've made a habit of carrying a couple gallons of Xylene with me when I make road trips to areas of AZ that I know have crap gas. In the end, that stuff is cheap while engines are expensive.

You could zero the table out for a test just to see what the result is :dunno: just make duplicate map with the tabled cleared and do a simple back to back comparison under the same conditions. It doesn't get any easier than this.

So, sorry to bring this up from the dead:

Am I to understand the cobb does not log knock retard because it has filtered out all values below 10? And our knock retard is negated? Is there a workaround for this? I was thinking of getting an AP but am not very comfortable tuning w/o being able to monitor knock levels.