News:

New Users: Send [email protected] an email with your account name. I've turned off auto-approve to reduce the 100 new spam accounts a day. Thanks, -Jason

Main Menu

Playing with 1.1.88 firmware knock control... Some questions and observations.

Started by dnb, May 14, 2011, 01:09:07 AM

Previous topic - Next topic

dnb

I made a datalog using the latest vemstune (big improvement over last year's versions BTW) and configured my single knock sensor to be enabled for each of the firing events on my v8.  I would prefer 2 sensors, but I haven't got round to this yet.

Here is a histogram of my data.  The data has been normalised so it can be compared between cylinders.  The snag is that channel 7 appears to always read virtually zero.  Is this a Vemstune issue that hasn't been picked up yet?

I know that I am not using timing correction yet, but it does bother me that all 8 channels ("adjust0 to adjust7" do not have the same default value.  Is this another Vemstune logging issue?



I assume I tweak the knock signal gain to bring all the channels to as close to the same median value as I can, and change the time constant to get noise to be roughly 1/3 full scale.  Then I can set the threshold to declare knock and have the ECU take sensible actions.

Thanks for any pointers.  I'm also keen to understand the individual power stats.  Presumably these are based on counting trigger teeth per second?

GintsK

Are you sure those signals are knock values, not individual power? I was not able to get any signals on 1.1.88. But on 4 and 6cyl.

Gints

dnb

Not sure at all!  They are labelled as knock when the log is exported as csv but that doesn't mean much.

dnb

Looks like they are tooth timing difference data, not knock.  It makes sense now there is one that is always very small.

rossi46numberone

Is there in the future firmware some project to make the knock detection more easy to set and enable ??
At this time i'm the way to mount a "JAYCAR" knock detector circuit :-\....
Thanks,Michele.

JRD McLAREN

I'm make some tests yesterday.. (fw 1.1.81)
but when I have disconnected knock sensor, knock chanels 0-5 still get some values...
so I think, this is not real konock signal..

today I will make some tests on 1.1.88 fw..

GintsK

Quote from: rossi46numberone on May 23, 2011, 05:58:49 PM
Is there in the future firmware some project to make the knock detection more easy to set and enable ??
At this time i'm the way to mount a "JAYCAR" knock detector circuit :-\....
Thanks,Michele.
From firmware side everything is easy and straightforward now. Tuning software is little behind - mainly sensor chanel sequence.

GintsK

Quote from: JRD McLAREN on May 24, 2011, 09:44:31 AM
I'm make some tests yesterday.. (fw 1.1.81)
but when I have disconnected knock sensor, knock chanels 0-5 still get some values...
so I think, this is not real konock signal..

today I will make some tests on 1.1.88 fw..
I think better idea is to try something post 1.1.88.
Like 1.1.91 - downloadable by VT webtool.

z0tya


rossi46numberone

I think this function should be much more elaborate, and more simplified setting.
This could help everyone to find the best ratio between power and safety timing map...
Michele.

GintsK

Quote from: z0tya on May 25, 2011, 07:24:03 AM
It seems to me there is no knock_alt in 1.1.91 :/
Ouch!!! Really?!
I have no possibility to try. I am not at home for two weeks...

GintsK

Quote from: rossi46numberone on May 25, 2011, 08:28:33 PM
I think this function should be much more elaborate, and more simplified setting.
This could help everyone to find the best ratio between power and safety timing map...
Michele.
I think in knock_alt settings are very straightforward and simple. Please try to understand how it works!
If we want more simple - it means crowd some self-learning functions to ECU. That would take much of coding and probably even not possible because of CPU capabilities.
Knock-alt have adjustable gain for each cylinder (like volume) and 3D level map. If signal from sensor is higher - means knock.
What is needed to do - tune engine with no knock, equal "volumes" at WOT, find maximal noise values from log, and write it in table with added some percents more. That's it!

GintsK

BTW previous system was smarter. More automated. But finally knock recognition was unpredictable and not controllable by tuner. Now it is. Human mind is much more smarter than tiny CPU. So better is to give simple pretuned tasks for CPU than wait miracles from it.


JRD McLAREN

So ..
Can anybody tell me, if X and Y axies (in Knock tab) are MAP+RPM or TPS+RPM

cause, I have sometimes max retard, sometimes nothing..
and the marker is still on the top of the line...no browsing on map ..  :-\

rossi46numberone

Hi Gintsk,
I know your right, but sometime there is some frustrations when things are a little bit complicated :-\.
Yes it's true the better way is at first understand how does it work...
Do you think in the future the "F1" touch will give some basic help?
Michele.