Closed loop lambda gain

[434josh]

What are "normal" numbers for closed loop gain?

I am having an issue where closes loop is adding a little fuel under wot but once it needs to stop adding it takes a while to do so and the log shows it going rich for several log frames. It seems to add fuel much faster than it takes away. Is that normal? 

https://app.photobucket.com/u/434josh/a/31f86a2d-d161-4109-bf45-85813e99d7c1

 

 

[Adamw]

Can you attach the tune and log.   I dont have any typical G4+ examples to share as all the cars im working on and driving at the moment are G4X.  But there are really no "typical" settings, as exhaust flow, distance to the probe, lambda controller etc will all have a big impact.  

I generally like to run high gain and lower rates as opposed to the oppsite.  And I generally find you still want a small gain in the zero error cell.  I will attach a couple examples from G4X later but not sure how relevant they will be to G4+

[434josh]

22 minutes ago, Adamw said:

Can you attach the tune and log.   I dont have any typical G4+ examples to share as all the cars im working on and driving at the moment are G4X.  But there are really no "typical" settings, as exhaust flow, distance to the probe, lambda controller etc will all have a big impact.  

I generally like to run high gain and lower rates as opposed to the oppsite.  And I generally find you still want a small gain in the zero error cell.  I will attach a couple examples from G4X later but not sure how relevant they will be to G4+

I have a log example and a current tune file but the tune isn't what was used during the log.  

https://www.dropbox.com/s/y9mbkrvujhma27j/Barra curremt.pclr?dl=0

https://www.dropbox.com/s/71sslfpezzueq4x/Log 2021-03-29 8%3B25%3B48 pm.llg?dl=0

 

[Adamw]

In the area where it goes sligthy rich you have a lambda error of about 0.04lambda so I think you could probably increase the gain a fair bit in the 0.033 and 0.066 cells.  You will probably need to lower the conrol rate in the 0/1000/2000 & 3000 cells at the same time as your rate especially at 2&3000 is probably way faster than the time it takes for the exhaust gas to reach the sensor and the sensor respond.  

 

Here is an example from our WRX test car (lambda just after turbo about 1M from exhaust valves).  This is a G4X so I dont know how well the gains translate across but update rate should be relevant.

[434josh]

2 hours ago, Adamw said:

In the area where it goes sligthy rich you have a lambda error of about 0.04lambda so I think you could probably increase the gain a fair bit in the 0.033 and 0.066 cells.  You will probably need to lower the conrol rate in the 0/1000/2000 & 3000 cells at the same time as your rate especially at 2&3000 is probably way faster than the time it takes for the exhaust gas to reach the sensor and the sensor respond.  

 

Here is an example from our WRX test car (lambda just after turbo about 1M from exhaust valves).  This is a G4X so I dont know how well the gains translate across but update rate should be relevant.

Thanks Adam. That gain example is extremely different than mine so I can definitely see that working better. The help file didn't give a "typical" example like it does in a lot of other places so I didn't know really what was too much it too little. 

 

Thanks again. 

[434josh]

12 hours ago, Adamw said:

In the area where it goes sligthy rich you have a lambda error of about 0.04lambda so I think you could probably increase the gain a fair bit in the 0.033 and 0.066 cells.  You will probably need to lower the conrol rate in the 0/1000/2000 & 3000 cells at the same time as your rate especially at 2&3000 is probably way faster than the time it takes for the exhaust gas to reach the sensor and the sensor respond.  

 

Here is an example from our WRX test car (lambda just after turbo about 1M from exhaust valves).  This is a G4X so I dont know how well the gains translate across but update rate should be relevant.

Well oddly enough, the g4+ maxes out at 20 on the gain table. What does the g4x max at for comparison?

[Adamw]

G4X allows up to 200.  As I mentioned earlier I dont know how relivant that is.  But I suspect you can get away with a lot more than you currently have at small errors since your log shows it is only taking quite small "bites" out of the error with each correction.