Welloilbeefhooked

iTrader: (3)

My ESL ecu doesnt seem to be populating the knock control map when I can clearly hear det through my det cans. I was wondering how servere the det needs to be before the ecu pulls timing?

Also, IAM at 16.

The Rig

iTrader: (13)

Subscribed

Jolly Green Monster

iTrader: (4)

This is why mapping needs to be done with detcans.

Firstly are you rereading the knock control map after the event?

The ecu is listenning on a set frequency and the reason for the det does alter that slightly.

Bear in mind you are playing with an ecu thats 20years old so the knock control is not exactly advanced (pun intended).

Andy Stevens

Is the IAM control set correctly or learning will be turned off?

Log the cyclic knock control and see if it ticks/jumps up, that is the raw feed to the ECU.
If you are definitely getting audible det through cans you should see that jump up. If it is flat then there may be a sensor torque/loom issue etc.

Simon, IIRC you can read assembler? I'll show you the knock control algorithm sometime, it is clearly the result of a massive number of man hours of coding and is hugely advanced (pun etc), the sort of resource you only get from a heavyweight manufacturer (one that can custom fab their own ECU ICs for example).

The knock control is excellent on these ECUs and will pull timing, add fuel and reduce boost as necessary. Some running on stock maps in Kenya on 89 RON showed how good it was at defending engines on all sorts of fuels and climates.

Welloilbeefhooked

iTrader: (3)

Simon, I am viewing the knock control map after the event and refreshing the map. The funny thing is that I did have a few cells a while back (previous map) that showed 0.25 adjustment and I retarded ign to correct. At that time I heard no det through the headgear. This time I can hear det and there is no change in the kcm.

Andy, afaik everything is working correctly and the iam starts at 8 after ecu reset and climbs to 16. I will log the cyclic count next time out to check what you said.

I could force the issue with some 95ron!

Jolly Green Monster

iTrader: (4)

Yes Andy I used to write assembler to test the assembler thats keeps the eurofighter in the sky

Do often hear det and its ignored by the ecu during mapping write up to the newest ecus and as you say sometimes it pulls timing and there is no det heard.. but other times it is perfect and gets it spot on..

The Rig

iTrader: (13)

Just to add my tupence, on my 95 wrx i get exactly the same issue as the o/p,i can hear Det at high speed/rpm yet knock control map is all 0, yet my good old faitful knocklink lit up like a xmas tree, my timing has been retarded so no more audible det but was dissapointed the KCM didnt kick in to retard on its own.

My knock sensor is fully functional as well.

Welloilbeefhooked

iTrader: (3)

I not doubting the product, its more that I want to understand what the limitations are of the knock sensing and what it can and cant detect. I also need to know if I have faulty sensors.

Although joking about the 95ron I may have to provoke the system a little. Yes, I know this is dangerous and could cost an engine but hey ho, I want a new build anyway soon.


Wayne.

Andy Stevens

You and Spence have alot in common then.

Interesting what you say, suggests a frequency filtering window shift?

Dmacnz

I have been playing a lot with the Knock count vs the Phormula KS4 knock module. Seems to have extremely similar results.
This is using two different knock sensors in two different locations. I was initially worried as I also had zeros on the whole correction map and IAM 16. New Knock sensor had the same results!

Out of curiosity, how fast does the IAM change when knock is detected?! or no detected?!

Andy Stevens

You guys may have tried this already but I'll list the steps anyway.

1. Check the IAM control value and the highest value in the advance map to ensure knock control is turned on, although if the IAM is climbing it is in learn mode.

2. Log cyclic knock control. This is the raw feed to the algorithm and not particularly intuitive. It's an 8 bit counter that will wraparound, not a level, so the actual figure isn't a problem, it's how fast it climbs. Each tick up is an event registered by the sensor. Real det will give a big jump, and that is what you are looking for in conjunction with the other references. The KC algorithm takes a statistical view of this data and moves the IAM and values in the fine map to suit.

A few things may be happening, if the cyclic value is flatlined the sensor isn't giving info. Generally I would expect the diagnostics to throw a code 22, which I guess isn't happening. Knock sensors can also be sensitive to fastener torque.

If there is a signal then the KC algo has decided it isn't sufficient of the correct intensity or frequency. When I've seen major knock when I've been mapping the race car the IAM reacted instantly.

It is probably possible to identify and expose filter coefficients within the KC algo to adjust it's sensitivity, but going away from the factory defaults will require some pretty deep understanding.

Dan, the IAM goes down fast and up slow.

Cheers

Andy

Dmacnz

Thanks for the Info Andy, always deliver with the goods.
The IAM Ign Advance being default value of "9" is ok for all maps?! My Advance map is pretty well 4 all over the high load parts of the map!

Thanks
Dan

Andy Stevens

HI Dan, learning would be off in that case. It will only learn in parts of the advance map where the advance node > IAM control. I would try 3 in IAM control to start with.

It is useful to use this in areas of spoof noise to turn learning off where appropriate.

Dmacnz

Gottcha thanks, Like a filter almost! Thanks

Andy Stevens

Yea, a pretty brutal filter, but the analogy is good.

53

iTrader: (41)

Definitely Experienced this recently

The Rig

iTrader: (13)

So out of curiosity, how much lower does the IAM advanced threshold need to be than the highest figure in the advanced map ?, i.e the highest figure in advanced map is 7, do you make the threshold 6 or 5 or 4, does it matter how much lower you go than the highest advanced figure ?

what im trying to understand is, why isnt the threshold 0 then it wil learn in all cells with an entry of 1 or higher

cheers

Dmacnz

From ESL Manual. Could be your answer.


"The ECU determines “high load conditions” from the amount of ignition advance specified in the ignition advance map. If this is greater than the threshold specified in this table, then the opportunity is taken to “learn” the IAM. Note if the threshold is greater than any value in the ignition advance map, the IAM will never “learn” its correct value and the closed loop knock control will not operate."


Wish i read this before. Seems IAM advance should be less than your adv map. If it can be 0?? I may try tomorrow and see what the outcome is!

Dreep

Am I jumping the gun here ? Is this a fault due to

a) the mapper not mapping with detcans
b) user not familiar with setup
c) ECU knock control isnt great?
d) faulty knock sensor

Dont meant to open a can of worms by asking about ECU Knock Control 'not being great', just that I plan to change from a Power FC to the ESL setup on my 98' Type R. All info on these ECUs are new to me but so far everything has been positive.

Also a question maybe a little off topic, but would you advise to run a Knock Monitoring system such as Phormula or Knocklink with the ESL or just any ECU in general?

53

iTrader: (41)

Any modified /mapped car should have a knock alert system of some description PFC Commander is great IMHO I also have KS3 in my Newage

Andy Stevens

Yes, it does bother me that people might make that inference that the knock control is lacking when the reverse is actually true, it was ground breaking stuff.

Like I've said in this thread, the algorithm that controls it is clearly the result of heavyweight manufacturer resource and testing , so I think I can say with some confidence that is isn't (c).

Here are the patents Subaru filed to protect it which will give you an indication of the scope of the project. The exact same strategy is carried on to later cars and gives you a feel of the difference between the resource associated with heavy industry and that of an aftermarket manufacturer.

https://www.legionpatent.com/patents/5000150/

http://www.google.co.uk/patents/US5038736

I suspect that it is actually to do with reducing the number of learning cells by reducing the advance timing and not modifying the IAM control to suit.

For example, the IAM control on a z4 or 6k map is set at 9.



Any value in the advance map greater to or equal to 9 is a learning cell. You can see on boost and load it will learn. Off boost and at very high engine speeds the cells don't contribute to modifying the IAM.

The reason that IAM control isn't zero is it would drive up the number of false alarms. You only really want to be listening for det on boost as with a low compression turbo engine you won't get any off boost.

HTH

Andy

Dmacnz

The knock control feature is half the reason I went with esl on my classic. Something no other after market ecu is going to do so well.
Fantastic engine saving feature. And all automatic out of the box!

Welloilbeefhooked

iTrader: (3)

So in theory, with everything working as it should and knock control enabled correctly, you should be able to map the ecu and steadily increase timing until the ecu detects this and then retard from there? Although, as I have seen and from what has been mentioned by Simon, amongst others, det sometimes occurs without the ecu detecting this.

My next run out is to log the cyclic knock count and check for sensor operation and the climb in values when I hear set through the cans. I do suspect that there is a fault with the sensor or its feed to the ecu. When I built the engine back up I put my hand in the box of crap and chose from one of the 4 sensors I had. I'm thinking I picked up a dud.

Checked my knock control value which is at 6 and the advance map is the standard map from the 6s. I'm adjusting the base map only. Does a high advance map with a low base value make for better control than a low adv and higher base value?

Wayne.

The Rig

iTrader: (13)

I like the theory to use the knock control map as a guide to retard timing when the knock control map starts making adjustments, but theres a reason Det cans are used :-)

Dmacnz

Really interesting stuff here. Learnt a lot!
This ECU is clever!

Andy Stevens

Until you check the cyclic knock count, you can't say the ECU isn't detecting it. The IAM finds a level and locks, then moves to the fine control map, If the fine control map goes out of limits, it transfers back to the IAM.

Talking to Spence, there is another fast response correction always in operation that I can expose as a log parameter.

I would stick with adjusting the base map.

The only exception is if there an area of resonance that is spoofing the knock sensor at a particular engine speed and load, then you can turn learning off in that area by transferring some of the advance to the base map and putting the advance map under the radar for IAM learning. We have seen race gearboxes and slappy engines do this, but it is quite rare and obviously not happening in your case.

The old software, whilst DOS based and old school, did have a feature where it painted the cell a colour if the cyclic count ticked up so you could knock a degree off that cell.

You can do the same with the log feed of course, just log the axes and cyclic knock and if you see a big jump there and a correlation with a det monitor then drop a degree or two from that cell.

Andy Stevens

The Mines EPROMS certainly rely on that method. I've seen 5 deg added across the entire base map and the KC is left to sort it all out. Scary stuff!

joz8968

iTrader: (13)

Ha. That bears out the widely held belief that MINES boards runs an ultra aggressive map.

Such an appropriate name, methinks.

Dmacnz

If Mines believe in the Subaru knock control is clearly must be bloody good!

john banks

I learned to tune on MY99/00 JECS and it scared me to death forever with knock not being detected or corrected on modified cars (still stock internals though) but back then the details of how knock control worked, or any limits involved were not documented. However, it just seemed not to stop some pretty loud knocks that could be heard once you were used to them even without det cans.

When I later disassembled Mitsubishi and Nissan knock control and tuned those cars I found I could rely on their knock control very heavily and hung up the det cans most of the time. Any similar experiences?