Query about close loop lambda PID control.

[CRacing]

Good morning Syvecs forum,

I would like to set up the lambda PID strategy in a LifeRacing F88 GDI4 V2 ECU. If this message it is deleted, because it is related to a LifeRacing ECU I underestand it, but I think that this item could be helpfull for more people.

What is the complet process that you follow to achieve a good lambda PID?

I see that the stock PID of the LifeRacing only corrects small changes and the P term is very low and the integral term is very high. I see that the values are corrected mainly due to the PID term that is very slow respect the PD control.

I was thinking on delete the Integral term of the PID control to the Lambda and control first the lambda with the P term (thinking in have a fast response) but I don't achieve a good result. I am trying all at idle because I can't use the dyno this days. I suppose that for a good tuning I need to use steady steate dyno tests.

Is it better to control, in the case of the lambda control, use a lot of integral term in the lambda?

Is a bad control the P term in the lambda control?

I send a video attached, after I see how the default PID works, where you can see that I try to correct the mixture only with P to add after the integral term, but I see that the behaviour of the control is not good, with the VVT, DIP and other systems I see that the behaviour is realle different.

I am thinking, due to the facts that I see, that the PID control in the lambda is only destinated to small corrections and it isn't optimal for correct big changes but I have the doub that what could be happening if I have a problem like a problem in the fuel pump and the car is in an important race and I don't want that the car stops but I would like that the car achieves the better mixture possible. Do someone know what it is the real purpose of the close loop lambda control? How do you achieve a good control?

I have attached a video in the Syvces page in facebook where you can see how I try to control the Lambda:

https://www.facebook.com/10000143909234 ... 532967489/

Thank you very much in advance.

Kind regards,

[beetos]

Hi
Suggest zeroing integral gain and using a decent amount of P, but scale the gain of P accordingly. This takes Int wind up out of the equation. Limit the P gain at low load / low rpm. This allows cllfuelmult to be a good reflection of the direction your map needs to go but watch actual lambda difference versus target. Once you get it close you can experiment with how much P can be added at larger errors and calibrate the gain accordingly, then introduce I. Since cllfuelmult is the sum of I and P, if you have Int wind up without enough P the cllfuelmult can show the opposite of what the fuel table actually needs. Takes a while to get your head around it but the gain tables are key to getting a flat cllfuelmult log.

[johnsanchez]

Hi geometry dash lite,
Suggest zeroing integral gain and using a decent amount of P, but scale the gain of P accordingly. This takes Int wind up out of the equation. Limit the P gain at low load / low rpm. This allows cllfuelmult to be a good reflection of the direction your map needs to go but watch actual lambda difference versus target. Once you get it close you can experiment with how much P can be added at larger errors and calibrate the gain accordingly, then introduce I. Since cllfuelmult is the sum of I and P, if you have Int wind up without enough P the cllfuelmult can show the opposite of what the fuel table actually needs. Takes a while to get your head around it but the gain tables are key to getting a flat cllfuelmult log.

What an amazing idea! Thanks for your answer.