7EVO7

iTrader: (1)

stop the arguing, by now everyone knows the end of this, let's help the OP.


thanks

l2r99gst

iTrader: (2)

I believe mrfred gave the OP's answer in post #20.

mrfred

iTrader: (50)

mbartel - if you are interested, I know the table where the value of "9" is set (at least for 88590015). I can give it to you for your ROM if you want, but obviously the better solution is to find the reason for the bad knock signal.

03whitegsr

iTrader: (8)

Actually, knock sum DOES NOT give the severity of the detonation event. It gives the number of knock evets. But if the knock level is above the threshold, it is counted as an event, no matter how hard or soft that event is.

Can we log the raw knock data? DSMLink had it and you could log the knock level per cylinder as well as what the ECU was using for the knock threshold level.

l2r99gst

iTrader: (2)

Are you sure about that? I will have to re-read John's thread. But, I thought the knock count was determined by the amplitude of the signal beyond the knock threshold. Which, in my opinion, would correlate to a knock severity. If you haven't read that thread, I can find a link for you...everything is loggable...the threshold, the knock voltage, etc, etc.

Even if not, and it was purely a knock 'amount', I would think repeated knock events are more severe than isolated events. And the amount of timing retard that is needed to reduce or eliminate the knock, to me, correlates with the severity of the knock.

I know we don't have pressure sensors for knock sensors, so we can't see the absolute pressure spike of the detonation event, but I think we have enough data to determine the 'severity' level of the knock indirectly.

infini_IV

iTrader: (4)

I realise that but what this is doing is working the sum backwards, allthough the scaling is set to correct this why not just log it correctly from the beginning instead of taking the knock retard figure and dividing it by .3 to get the 'knock sum" either way you have confirmed my point that yes the data being pulled is the knock retard but evoscan scales and sums this to read "knock sum"

Thanks for confirming this

l2r99gst

iTrader: (2)

OK, here is John's thread where the fun begins.

Here is one of his graphs:


And his explanations:
So, it is what I remember. The increase in knocksum is the amplitude of the signal above the knock threshold, which to me, correlates to a knock 'severity'.

l2r99gst

iTrader: (2)

Evoscan or whatever data logger isn't using the knock retard scaling and dividing by .3. We are logging the value directly, with a scaling of x, meaning, just show us the raw output.

To log knock retard, you would have to use a formula of x*90/256.

This, by the way, is why mrfred posted a way to log timing in 1/3 degree increments. That way, we can see the true effects of knock retard.

Read the whole thread I linked to above by John and you will have a clear understanding of how the Mitsu Evo ECU handles knock control.

03whitegsr

iTrader: (8)

Ah, my mistake. I remembered the knock base stuff, just thought it was a direct "if knockadc > knockbase -> add 1 to knock sum" type of setup.

You are correct, there is a proportional amount there that would detect "severity."

fireroasted

iTrader: (3)

7evo7 - yeaaaahhhh.

l2r99gsr - I did understand mrfred's response in #20, but I would follow with questions about the logs I trapped in that the knock goes from 0 to 9 and pegs there and back to 0. That behavior doesn't match his gradual sounding scenario of octane hitting 154. Is that an indication of something specific based on those complex formulas. More simply; I was expecting to see more curve values, instead of Off = 0 and On = 9, like a step. Getting to a theory that this would prove or disprove the knock sensor being bad.

mrfred - Thanks for the input. I have to be up front here, in that you guys are having discussions a click ahead of my understanding. I wonder about how knowing the table the value was in would help me identify or eliminate causes for the knock values. I guess my skill level was at a point where I figured there must be more sophisticated logging or interpreting of logging I could/should be doing with evoscan to help me identify the root cause.

For instance, if I wanted to prove that there was or was not a problem with boost leak, what values juxtaposed to the JDMMAP value would prove/disprove that. Or confirm that the boost solenoid is working as expected based on my maps, etc.

I am slowly ripping the systems (IC plumbing, exhaust) apart looking for 'something'. I feel kind of like a dumb@ss, because I am blindly hunting around for whatever. With the sophisticated tools I learned to tune the car with, I am sure there is a better way to resolve this. So far I am too much of a blockhead to have figured it out on my own.

I suspect something failed and now i am looking for the failed part. Just hoping that intelligent logging would give me better guesses for the hunt.

Either way I appreciate all the input, even or especially the stuff over my head.
M

l2r99gst

iTrader: (2)

P0325 is a knock sensor malfunction. Looking at your logs it appears that anytime you are above a certain load where the knock sensor becomes active, it simply jumps up and defaults to a knock sum of 9.

I *think* that may be the default value when an error in the circuit is detected (your cel code). I don't know for sure, but I would also assume that the ECU would pull boost as well during that particular cel. If the ECU thinks the knock sensor or circuitry isn't working, then it will most likely use several failsafes to compensate for it, such as pulling boost as you have seen.

Since mrfred mentioned that during normal function boost is decreased at a certain sustained knock level (octane level), that it would make sense that it would be an immediate correction during a knock sensor/circuitry malfunction.

So, what I would suggest is to check your knock sensor and wires/harness, etc. You may simply need a new knock sensor. That is my guess as to what your problem is. As long as you are getting that cel, you will most likely continue to have this problem.

Depending on exactly what is triggering the CEL, you may be able to simply disable the knock sensor as a test. Obviously don't run like that, but as a test. The instructions are somewhere in this forum.

mrfred

iTrader: (50)

M - I think l2r has the right idea. Its got to be something in the knock sensor circuit. I think you can put the rest of the car back together.

The sustained knock sum of 9 will not immediately drop the octane value below 154. It will take some time, but once it drops below, boost will be pulled fairly quickly (a few seconds). The octane level resets itself pretty quickly (to 255), perhaps within 30-60 seconds of going off boost (its really quite different than learned knock retard on subies where it apparently can persist for quite some time), and so you could end up going through this scenario over and over again.

fireroasted

iTrader: (3)

OK, Thanks. I will concentrate on the knock sensor. Initially I x'd that off my list because I could clear the code and it would behave as expected with only 1 and 2 counts of knock (I mean retarded timing/3 and a bunch of other math)

However it makes sense that load is a factor in even making the sensor output applicable. From a driving perspective I need load before I can incur the CEL.

Does anyone have the resistances that I could check the sensor with? I figure I can eventually find the schematic I had once for the ECU so I an trace the harness, but then I want to test the actual sensor too.

Back to that other post about how to remove that burried thing. Thanks
M

GST Motorsports

iTrader: (8)

I see infini_IV still hasn't read any of the disassembly findings.... and also still doesn't understand "logging raw data from the ECU".

- Bryan

infini_IV

iTrader: (4)

Are you some kind of jcbanks groupee or something???

Raw data. clearly you dont understand it, if we were logging raw knock sensor data we would have a consistant "knock sum" directly proportionate to knock voltage

you have offered zero constructive input. your child like ad hominem wont prove your argument either, Why wait for everyone else to post then add your comments?

from what i have come to expect from your posts, will we see a little "haha i told you so" from you? i know your dying to get it in somewhere