Timing compensation per cylinder

[Tor]

Extracting this topic from http://www.ft86club.com/forums/showt...=83945&page=38

Originally Posted by Tor

Searching for timing compensation per cylinder led me back to this thread. I wonder why Shiv zeroed it out in the most recent OFH tune but kept it in the regular stg 2 UEL tunes?

Otherwise, Wayno uses the AVCS from the the OFH tune. Total timing (base plus KCA) on the 102 tune is not significantly different - give and take (some areas more, some less). Except above 6800 rpm and load 0.9 and in 0.6-0.7 load, lower rpm range, where OFH has significantly more.

- I'll might add the timing difference above 6800 rpm to the Wayno 102 tune, but I am not sure if I should leave the TCPC or zero it out?

- Or I could start with zeroing TCPC out (without adding timing) and see if it makes a difference first.

- Of course, I could also just leave it all alone and enjoy hassle free timing.

Decisions, decisions. The bite of the damn optimization bug is annoying. Does anyone have the antidote?

Originally Posted by steve99

I believe the timing compensation per cylinder is because some cylinders may run slightly hotter or be more knock prone than others. So theoretically you may be able to run more base timing and just retard the problem cylinder or cylinders slightly, but going to be difficult without getting individual cylinder knock data

Originally Posted by Tor

I was thinking to zero it out and add timing and if I subsequently get knock then to remove it from TCPC A+B first?

Somewhere I read it could be because of where the knock sensors are placed and to ensure that C+D knocks first so the sensors can actually pick it up. I doubt that is the case since the sensors are located between the cylinders each side on the FA20.



Besides, the tables are already zeroed out up to 6000 rpm anyway.

A is probably cylinder 1 according to RomRaider. B is 2 or 3? If 2, those would both be the best cooled cylinders facing the front of the engine? If 3, could the right side run hotter maybe (exhaust cat below)?

Could it rather be because of flow in the stock header? In which case, the compensations may not make any sense with an aftermarket header? Are there any changes to these tables in the MY17 with the revised header/intake?

Originally Posted by Kodename47

Have you considered differences in fueling between cylinders due to unequal airflow? Heat management as cylinders will always run different temps within the block? Then you have all the external factors that you mentioned.

AFAIK cylinders A-D are 1-4 in firing order.

I would expect that the changes in the MY17 are more to do with the intake manifold than exhaust manifold, but there are also internal changes in the block that may also have an impact.

Originally Posted by Tor

Thinking out loud, from that logic, maybe they shouldn't have been zeroed out at and below 6000 rpm either?

The TCPC A+B are retarding timing in the same areas as the pre-MY17 as well. Even a bit more above 6k. Could it be because the stock tunes run so rich at high rpm (why btw?)?

From a logical perspective, the tables should either be left fully intact or be possible to zero out completely. Since we happily accept to run without these compensations below 6k, where we are at most of the time, I am zeroing them out completely next time I flash and will see how it goes.

[Tor]

So the update:

- So I added most of the total timing difference between the OFH tune and Wayno's 102 UEL stg 2 tune.

- I zeroed out the TCPC A+B.

- I copied the OFH IAT compensation for safety.


The reason for flashing and logging was to check a slight MAF adjustment and hence it's without learning.

Surprisingly I have no knock above 6k. The added timing at 3600 rpm is maybe too much (earlier in the log), but I know it will go slightly richer when it learns.

I know this doesn't say much because no knock is detected. So it could be that I could run a bit more timing and still A or B cylinder might be the ones to knock first. Or they could run leaner without any chance to know as per Kodename47's suggestion. Or, worst case it could be that it knocks without it being detected (unlikely as per the knock sensor locations depicted above, I would think).

I'm going to keep it they way it is for now and recheck this in the spring when we get higher temperatures. I also consider buying a knock sensor and an amplifier to make a DIY det can and see if it's possible to hear anything that the ECU doesn't.



http://datazap.me/u/tor/tor-201?log=...-31-42&solo=17

[Tor]

The above post didn't get any feedback, but maybe someone will find it interesting later on.

I kept driving the car with the TCPC zeroed out. On the street with higher temperatures I still don't get any noticeable FLKC.

On track it's another matter though. I drove 3x 15 mins stints and combined the log into one. So in 45 mins of combined track driving, this is the FLKC I got:

First a count. This is the amount of times FLKC was activated (yes, almost 1200 times at 6800 rpm load 1.2):



The mean values:


And the max values:


The high values (even the -4.47) didn't cause IAM to drop. I did have 3 instances of IAM drops. Once to 0.98 with a quick recovery. And once to 0.98 followed shortly after by a drop to 0.96. This, however, happened at 4000, 4200 and 4600 rpm where the FLKC values were low comparatively.

In theory, I have no problems to continue driving with the current tune. The ECU pulls timing as it should and I get the most performance when driving the car on the street.

However, I want to test if I can get better performance on track by reintroducing TCPC. The idea would be that perhaps the ECU would need to pull less timing overall if it's sufficient to take out some timing on cylinder A and B.

From load 1.0 and up I will add the following compensations, based on the knock values seen above and looking a bit to the OEM TCPC tables:



Will report back the result. Hopefully, I will have a chance to drive the car under similar track conditions in the near future. The intention is not to make FLKC free on track but to see if I can get it to pull less timing.

[renfield90]

How do you know which cylinders are the slightly more knocky cylinders? Do you have knock count per cylinder?

My understanding of TCPC is that Subaru has a single knock sensor for the whole motor and it's placed closer to one bank than the other. Since it's possible for the far cylinders to knock lightly but not get noticed by the sensor, TCPC ensures that the first cylinder(s) to start knocking are closest to the sensor.

Shiv probably removed TCPC as an optimization step for the aggressive OFH tune. It sacrifices an increased risk of undetected knock in order to encourage the ECU to be less aggressive about pulling timing. The onus is on the user to verify the tune and use good fuel.

[Turdinator]

Originally Posted by renfield90

How do you know which cylinders are the slightly more knocky cylinders? Do you have knock count per cylinder?

My understanding of TCPC is that Subaru has a single knock sensor for the whole motor and it's placed closer to one bank than the other. Since it's possible for the far cylinders to knock lightly but not get noticed by the sensor, TCPC ensures that the first cylinder(s) to start knocking are closest to the sensor.

Shiv probably removed TCPC as an optimization step for the aggressive OFH tune. It sacrifices an increased risk of undetected knock in order to encourage the ECU to be less aggressive about pulling timing. The onus is on the user to verify the tune and use good fuel.

I am fairly certain there are two knock sensors. One for each bank.

[Kodename47]

@Tor don't forget that your cell count is not instances of knock detected retard but how many logged rows are less than 0. There is a difference, for example how many rows are there for just one decrease in timing? That will depend on logging and settings, but also you are including where the ECU is actually adding timing in after a retard event.

[Tor]

I made a mistake with the data. When looking at the first log I noticed rpm spikes when the engine was at idle. And looking at video from the session, sure enough, I forgot to turn off the A/C. I wasn't aware that the A/C would cause actual FLKC. From what I read here previously, I only though I could cause false FLKC when clicking in and out.

Anyway, looking at log 2 and 3 only, it has a lot less knock. Perhaps so little, that I won't bother correcting it. The important take-away for me is, besides for all the other obvious reasons, never ever run with A/C on at the track.

Count:


Mean values:


Max values:


This is, by the way, a video of the 3rd session with Datazap log overlayed:

[ame="https://www.youtube.com/watch?v=hC93-8wFo9E"]New season and personal lap record... Nürburgring GP circuit (GT86) - YouTube[/ame]

Originally Posted by renfield90

How do you know which cylinders are the slightly more knocky cylinders? Do you have knock count per cylinder?

No. Hence, the idea with the test would have been to see if it would have reduced the overall knock only to correct A + B.

My understanding of TCPC is that Subaru has a single knock sensor for the whole motor and it's placed closer to one bank than the other. Since it's possible for the far cylinders to knock lightly but not get noticed by the sensor, TCPC ensures that the first cylinder(s) to start knocking are closest to the sensor.

See the picture in the first post. There two sensors and they are located between each cylinder on each side. So I do not think this is the reason.

Originally Posted by Kodename47

@Tor don't forget that your cell count is not instances of knock detected retard but how many logged rows are less than 0. There is a difference, for example how many rows are there for just one decrease in timing? That will depend on logging and settings, but also you are including where the ECU is actually adding timing in after a retard event.

Yes, also I suppose it could as well just keep a value it set during a previous run through the same rpm/load? I any case I am aware that it doesn't mean knock. Only that there was ECU activity at that point to prevent it.

[Kodename47]

The AC shouldn't cause FLKC/FBKC. The clutch will disable at high load, this is no different to having it switched off. The frequency of it disabling the clutch shouldn't trigger the sensor either.

[Tor]

Here are the logs. The only difference is the A/C on. Otherwise ambient conditions the same.
http://datazap.me/u/tor/tor-2141-track-logs