AFR oscillations at idle that stabalise when fans turn on

[RichardB]

I have AFR oscillations at idle. These seem to disappear and AFR completely stabalises every time the coolant temp reaches the threshold and the fans kick in. When the fans turn off, AFR's start to bounce about again. I would like to have the nice stable idle all the time ... so if anyone can diagnose what is happening here, that would be huge help, as it's got me stumped.

Below is a log screenshot, showing what I'm describing ... you can see both AFR and RPM oscillating when the fans are off.

IdleAFR.JPG (71.15 KiB) Viewed 22741 times

Edited to add - S6Pnp on a modded 2005 STi

[TimH]

Not got any magic answers...but is it a new phenomenon, which might suggest something's failed/changed, or has it always been there?

[secure]

I am not expert but it is normal wideband sensor correction. Even the OEM ecu on would jump between 14.7 to 14.0 .
If you don't want "oscillations" at idle
Go to ;
Run Mode Fueling > Corrections > Closed Lambda > Wideband Lambda Sensors > Base Target 1 or 2,3,4 (check calsw assignment)
Check the highlighted idle cells and make them(and 1 step surrounding cells ) zero.
By doing this you are telling the ecu that you don't want wideband sensor corrections for stoich target at idle areas.
So now the ecu will read from base fueling map.
Than go to Run mode fueling > Base injection time 1 or 2,3,4 and set the injector opening time +/- for stable AFR around the idle cells.

The other way while keeping wideband sensor correction ON
Run Mode Fueling > Corrections > Closed Lambda > Wideband Lambda Sensors > Integral minimum and maximum
These values there will restrict the ability of removing and adding fuel at a percentage.
Also proportional gain and integral gain are related with this but I don't think I can explain these values with my poor english =))

[pavlo]

Your fuel map and/or voltage compensation is imperfect, and that is the cause of your symptoms, there isn't anything "wrong".

I can see that your lambda goes up and down in time with your RPM fluctuations and this will all be initially caused by greater airflow required to deal with the not insignificant extra load caused by the fans on the engine, via the alternator.

Is your battery voltage also fluctuating? And if so is your injector battery compensation correct at the varying voltages?

Is your injector battery compensation even correct at the nominal voltage? If it's out by a chunk it will exaggerate or dampen changes to the fuelling at idle due to multiplied corrections.

Is your closed loop multiplier going up and down? The closed loop response is something that needs to be "tuned" especially when you look at the integral which can start stepping up and down if you're not careful.

Do more tuning, get it right.

[RichardB]

Thanks all for your advice and comments.
Sorry for the delayed reply - in the midst of selling our house and moving and all the madness that entails so not had a chance to respond until now.
Changing to version 1.25 of the firmware and substantially increasing the closed loop transport delay in the new 3D map available has really improved closed loop operation at idle. I also enabled the new Integral Term Scaling function and reduced the integral values as advised in the firmware release. The oscillations are gone and the closed loop function now works as I would have expected.

Paul,to answer your questions, yes the injectors are ID1000's and I was extremely careful when entering the latencies. Battery voltage fluctuations are small. When the fans kick in, the voltage dips initially, but has recovered within 2 seconds back to 13.4V. The closed loop multiplier was all over the place before, but since making the changes described above it is smooth.

[pavlo]

The 3d table for service interval is not directly related to transport delay, you've been looking at too many MoTeC webinars!

It is the cycle time for the control loop, in particular the time base over which the integral loop is conducted. So you should not assume that this table needs to be time delay for the exhaust gas to reach the sensor and for the sensor to read it.

However you are on the right track, you should consider tuning your fuel table at low end with the closed loop turned off if you're having problems. Then look at using a combination of the Service Interval and Integral response curve to get a nice smooth closed loop response at low load. Then you can drop your service interval much lower at high load to get the integral term to work for to your best advantage. If you use something akin to this strategy you will probably have more success by also turning off the "Integral Term Scaling With Service Interval" option.

You will also find that less than ideal injection timing will make this harder.

I am also sceptical of IDs published lag times for their injectors, also be aware they will change with fuel pressure.

[RichardB]

The 3d table for service interval is not directly related to transport delay, you've been looking at too many MoTeC webinars!

I've not seen any of the MoTeC webinars, I was referring the 1.25 firmware release notes as below.

// Cll service time calibration

Allows users now to adjust Service time of Closed loop Lambda based on Rpm vs Primary Load, very useful for increasing service time on engines which have long exhaust manifolds and UEGO sensors post-turbo as before the limit of 125 on earlier firmwares made some engines struggle with closed loop lambda control at low mass flow rates due to the transport time between the cylinder and the sensor.

With additional option to disable integral gain scaling by service time so that the integral gain becomes percent per service rather than percent per second. Note that this will require integral gain values typically 10 times smaller than currently used. The option default to enabled to retain existing behaviour.

It is the cycle time for the control loop, in particular the time base over which the integral loop is conducted. So you should not assume that this table needs to be time delay for the exhaust gas to reach the sensor and for the sensor to read it.

However you are on the right track, you should consider tuning your fuel table at low end with the closed loop turned off if you're having problems. Then look at using a combination of the Service Interval and Integral response curve to get a nice smooth closed loop response at low load. Then you can drop your service interval much lower at high load to get the integral term to work for to your best advantage. If you use something akin to this strategy you will probably have more success by also turning off the "Integral Term Scaling With Service Interval" option.

Thank you for the advice.

You will also find that less than ideal injection timing will make this harder.

I'm still running 540 as the end angle across the rev range. I know others have seen some improvements tuning the end angle and I did a little playing with changing this, but generally it made the vehicle run poorly.

I am also sceptical of IDs published lag times for their injectors, also be aware they will change with fuel pressure.

That's dissapointing to know given how much they market their injectors based on matched latency. I did enter the latancies for the fuel pressure I'm running.

[TimH]

FYI I tweaked my injector end angles as I don't think the generic 540 works well. I put mine down to 360 at low rpm rising to 540 by 2500 rpm, simply to sort poor idle that only "worked" at 0.7 lambda.

Mine really needs remapping for more suitable end angles across the rpm range, but that was a quick fix to get it to idle better.

[pavlo]

There is a transport time Richard, but that is not what the "Service Interval" time directly relates to. It's not important in as much as you shouldn't get caught up in the notion that the gas "couldn't' possibly have a delay of X in the service interval table".

[Benkku]

I am also sceptical of IDs published lag times for their injectors, also be aware they will change with fuel pressure.

Sure the latency varies as a function of fuel pressure and also injector driving - voltage when opened and back-emf voltage when closed - affects it's opening and closing characteristics. Even the switching component, has some influence to latency, be it dedicated injector/solenoid driver or ordinary switching component. Hence - published values, may not reflect your application latency.