3.6L light throttle engine surge

[goose77]

Awesome! Thank you. I will go in on Monday. Yes, automatic. My fingers are crossed..

Sponsored

 

[vavaroutsos]

This is no learning algorithm. This is a crappy tune. No ifs, ands or buts about it.

I have recently been looking into the inner workings of the GPEC2A PCM, and it uses an Artificial Neural Network (ANN). My understanding is the ANN is constantly learning to compensate for changes with aging parts. So I do think it's quite possible that depending on driving style, individual vehicle, etc. the ANN is learning some bad behavior that causes this issue to varying degrees.

I have purchased a cheap second hand PCM that I will unlock. Once I understand the PCM algorithms well enough, I'll try and fix the tuning.

 

[Litfuse]

I have recently been looking into the inner workings of the GPEC2A PCM, and it uses an Artificial Neural Network (ANN). My understanding is the ANN is constantly learning to compensate for changes with aging parts. So I do think it's quite possible that depending on driving style, individual vehicle, etc. the ANN is learning some bad behavior that causes this issue to varying degrees.

I have purchased a cheap second hand PCM that I will unlock. Once I understand the PCM algorithms well enough, I'll try and fix the tuning.

If you fix it, you will be considered a god amongst us mere mortals.

 

[vavaroutsos]

If you fix it, you will be considered a god amongst us mere mortals.

I just took the first log, so all my channels weren't setup correctly. However, I think I see the problem with the light throttle surging. The spark timing is bouncing back and forth between about 40 deg and 0 deg. Look at the 3rd chart from the top, green trace. I suspect the spark source is switching back and forth between two sources with a very different idea of what the spark timing should be for the given conditions. Further investigation is required.

 

[Gorilla57]

Curious why both LTFT are negative. If the computer is learning correctly, LTFT should be 0%. So, the computer is trying to lean out the motor. Did you recently disconnect the batteries or reset something?

 

[vavaroutsos]

Curious why both LTFT are negative. If the computer is learning correctly, LTFT should be 0%. So, the computer is trying to lean out the motor. Did you recently disconnect the batteries or reset something?

No, didn't disconnect the battery and haven't done anything to the PCM other than read it, and log data. It's been my experience with some Ford PCMs that the LTFT vary as they update when running open loop. I haven't logged enough on this PCM yet to know how the fuel trims operate.

 

[Gorilla57]

No, didn't disconnect the battery and haven't done anything to the PCM other than read it, and log data. It's been my experience with some Ford PCMs that the LTFT vary as they update when running open loop. I haven't logged enough on this PCM yet to know how the fuel trims operate.

Gotcha. I haven’t done any logging on a PCM in a number of years. In my tuning days, we knew when we were getting a bigger set of injectors on a “street tune” close when the LTFT would approach 0%.

Keep it up on logging, I’m interested in seeing what you can find out about these PCMs.

 

[vavaroutsos]

After more research, I'm pretty confident the Artificial Neural Networks (ANN) are only being used to estimate volumetric efficiency (VE). The normal ANN looks to be a three layer feed forward network with 10 neurons in the first and second layers and one in the output layer. There are four inputs which I suspect are intake cam angle, exhaust cam angle, RPM, and PRatio (MAP/barometric). The activation function appears to be tanh like.

There is a second ANN that is used when VVL is activated. This one appears to have 9 neurons in the first and second layer and one in the output layer. I'm not sure why it has less neurons. I have seen another calibration on an earlier 3.6l that had 9 neurons for the normal ANN and 11 for the VVL ANN, so I'm not sure what determines the number of neurons.

I don't think any adaption is happening in the ANN, so I was wrong about that. I think it's only trained during the factory calibration process as I haven't seen any of the synapse weights change.

 

[slamin81]

guy and gals, we bought our 2021 rubicon jlu in November of 2020, put 8000 miles on it, and have never had a problem out of it. Today we took it in for the first service, everything went well, after we stopped and filled up. no issues until we were half way home noticed the same issue everyone else is having pulsing/misfire/trottle. racking my brain till i came across this thread, all i know is that it sucks..

 

[bwright1818]

Just brought my 2019 in yesterday for a clutch line leak. As I had mentioned before, this Jeep had had the light throttle "surge" issue since day-one, which was sometimes so bad that it would clank the whole drive line!! They told me they could do nothing about it, in those early days. Then, last August, I had a mirror replaced; and apparently they reflashed the ECM, because it ran much better after that visit. They would never tell me what they did. I asked the service writer again, yesterday. They DID, in fact, do a reflash in August, right when there was a very noticeable improvement. I think they are on to this issue, finally....

 

[kelleymark]

*** TORQUE CONVERTER BAD ***

 

[mwilk012]

*** TORQUE CONVERTER BAD ***

What?

 

[wolf]

If you fix it, you will be considered a god amongst us mere mortals.

If you fix it, you will be considered a god amongst us mere mortals.

I have recently been looking into the inner workings of the GPEC2A PCM, and it uses an Artificial Neural Network (ANN). My understanding is the ANN is constantly learning to compensate for changes with aging parts. So I do think it's quite possible that depending on driving style, individual vehicle, etc. the ANN is learning some bad behavior that causes this issue to varying degrees.

I have purchased a cheap second hand PCM that I will unlock. Once I understand the PCM algorithms well enough, I'll try and fix the tuning.

come up with a permanent pcm fix to get rid of ss and graduate to god status.

 

[kelleymark]

What?

All these issues with surging and studder at the low rpm are bad torque converters on the 8spd. All ZF transmissions with the 8spd are having these issues. Aston Martin, BMW, Fiat Chrysler, Jaguar Land Rover, and the Volkswagen Group, as they put their own TC in these transmissions.

If the lockup clutch inside the torque converter is starting to malfunction, you may experience shuddering at around 30-45 mph. The sensation is very noticeable and typically feels like you’re driving over a rough road with many small bumps. As the converter switches over to direct drive, a worn lockup clutch can make the transition difficult, resulting in this sensation. The feeling may start and stop abruptly and may not last long.

 

[mwilk012]

All these issues with surging and studder at the low rpm are bad torque converters on the 8spd. All ZF transmissions with the 8spd are having these issues. Aston Martin, BMW, Fiat Chrysler, Jaguar Land Rover, and the Volkswagen Group, as they put their own TC in these transmissions.

If the lockup clutch inside the torque converter is starting to malfunction, you may experience shuddering at around 30-45 mph. The sensation is very noticeable and typically feels like you’re driving over a rough road with many small bumps. As the converter switches over to direct drive, a worn lockup clutch can make the transition difficult, resulting in this sensation. The feeling may start and stop abruptly and may not last long.

This is not even close to what is being described in this thread.

Sponsored