True Flex Fuel Tune

[WildHorse]

It is accurate enough for Ford to warrant the engines using this logic.

Good for FORD. And there millions, AND MILLIONS into the development of there flex fuel system. Far cry from little johnny running a flex tune from 5-star. Lemmie know if anyone can do better. Oh wait, GM.

Sponsored

 

[mejohn50]

Just thought I’d leave this here. This wasn’t on their site a couple days ago.

For the record, this isn’t the full “workshop” level suite, so the custom OS isn’t supported, but for $100 it might be worth it to see the parameters available. Plus it still gives the ability to upgrade later if you want to spend the coin.

I’m in no way affiliated with PCMTEC, I’m just really interested in their platform.

https://www.pcmtec.com/pcmtec-editor-protuner-edition-usvp1

Edit to add: Just heard back from PCMTEC. Currently still no plans to offer a single car workshop level package. So if you want the custom OS the ~$2000 full workshop package is the only option for now with no public plans to offer the single car version anytime soon. I am optimistic seeing as they’ve just made the ProTuner package available that I linked above so I guess time will tell.

 

[WildHorse]

https://www.pcmtec.com/pcmtec-editor-protuner-edition-usvp1

Definitely gonna buy it.

 

[shogun32]

Timing needs to be increased pretty substantially to make real power on e85

So, what's so hard with the computer starting over (zero'ing) at every keyon (or upward fuel level) and slowly increasing the timing limits ONLY after a dozen greater than 70pct throttle runs that breach 6000 rpm? If during these runs knock sensors can arrive at how much E is really present and can clamp said upper bound until the next fuel level change.

This can't be that hard to implement some rudimentary logic.

 

[ZXMustang]

So, what's so hard with the computer starting over (zero'ing) at every keyon (or upward fuel level) and slowly increasing the timing limits ONLY after a dozen greater than 70pct throttle runs that breach 6000 rpm? If during these runs knock sensors can arrive at how much E is really present and can clamp said upper bound until the next fuel level change.

This can't be that hard to implement some rudimentary logic.

ford does it quite a bit better than that. The tune is setup to relearn after a certain level of refuel. Then it will relearn for about 10-15 minutes of normal driving.

 

[engineermike]

I believes there's more to it than HPT gives us, I think there's likely a delay that we don't have access to. Ford has a patent for this, and in their patent they indicate that the system waits a set amount of time before it starts learning after a fill for the lines to purge.

However, I've seen our system basically immediately starts learning once you put in enough fuel. I'm now wondering if that register is set to 0 and needs to be increased.

If you ever find it it would be an amazing leap for inferred flex on this car.

If you can find this parameter in any ford cal, I have a friend who can usually find it in our OS and create a custom parameter. But without that, it would be nearly impossible without a “key” from ford.

 

[junits15]

If you can find this parameter in any ford cal, I have a friend who can usually find it in our OS and create a custom parameter. But without that, it would be nearly impossible without a “key” from ford.

Thats kind of what I figured it would be, someone with inside knowledge would need to tell us what to change for this to work.

Sounds like a great friend you have lol

 

[engineermike]

Sounds like a great friend you have lol

He’s a real whiz on hex and control systems and has been a great deal of help learning tuning. He designed and built the “humimod” I described in other threads. He’s also found lots of errors in hpt’s definitions. Just yesterday he found a mixup in the CSER and Starting DI blend tables that’s causing rough running/mis-fires on cold start when using e85 on a flex tune with CSER timing retard disabled.

 

[engineermike]

Ok, I wanted to comment on a few things in this thread since I've been experimenting with this for the last couple of months.

Accuracy:
The Ford method is as accurate as the MAF calibration, Injector data, and lambda-to-alcohol% transfer table. If these data are accurate and there are no physical problems, then the learned ethanol can be dead on.

Regarding how it works:
I initially spent a lot of time trying to speed up the learning (not "maturing") process so that it would learn faster and be safer on transitions. After trying all sorts of changes, I concluded the ONLY change that sped up the learning rate was the "max update rate" parameter. I did observe that the rate of learning is a function of the max update rate, current error magnitude, and also airflow. It learns faster if the error is greater and if the airflow is greater. I haven't tested the WOT learn rate yet, though.

@junits15 I too have a log where I switched fuels and it matured within about 3 minutes at the incorrect concentration. The new fuel didn't reach the engine until about 4 minutes in. I'm currently trying what I believe @K4fxd suggested and extended the "min time to infer" and "min ego switches" from 10 to 20 to see if that will prevent maturing too soon.

@K4fxd my understanding of the flowrate thresholds is that it has to be within that range in order to lock in/mature. I've observed mine learning at idle, which is well below the 2.4 - 8 lb/min airmass min/max window.

@stang17 it automatically clears the KAM and restarts the learning when you add fuel to the tank. However, for testing you can use VCMScanner to clear the KAM as well and it will trigger learning again.

Cold Start Roughness:
I mentioned above this issue but I'll explain more. My car would start good but run rough and generate mis-fires for the first 30 seconds to a minute. From logging, I realized the issue was that it was attempting to run 60-80% GDI blend when cold. On gasoline, it had been only running 40% blend. Cold start idle smoothness is more challenging on GDI and even moreso on E85. It probably works ok when the timing is super retarded and throttle open, as is the case in CSER mode for cat light-off, but I have that disabled. Anyway, I changed the tune to use 40% blend in CSER mode on either fuel and now it starts and idles really smooth even on E85. HPT didn't help things by mixing up the GDI Blend FFV Starting and CSER tables. Initially I was modifying the CSER table since that's what mode it was in when the issue occurred, but it was really the Starting table since HPT labeled the two backwards.

Safety/Learning:
One thing I've done to improve safety is to toggle the spark timing at an ethanol concentration of 55%. Basically, at anything up to 54% you don't get any additional spark advance, but you get all of it at 56%. This way, when you switch to gasoline, it won't take much learning to turn off the additional timing. Secondly, you can use the A/F ratio gauge on the dash to know when this took place. 55% is 11.5/1 A/F, so when the gauge hits >12/1 at steady state then you know you don't have the extra timing anymore.

One option also is to just disable maturing. This would ensure it's always learning and driving fuel trims to 0 and couldn't possibly lock in to an incorrect concentration. The downside, I think, is that it would disable certain diagnostics since the fuel trims could never drift.

And finally, another option I've considered is to change it to not use NVRAM lambda when learning commences. You could set the initial value to something like 12/1. I believe the post-fueling re-start would idle a little rough on E85 since it will be 20% lean, but it might be ok since the engine is hot. This method would force it to learn down to 11.5 every time you fill up before you get any additional advance, when coupled with the timing strategy I mentioned above. The only concern here would be if you switch from E85 to 93, it would immediately start learning the E85 (since that's what's in the line) and might toggle the advanced timing before 93 hits the engine and you're left with the same situation. But, you could play with the settings to see if the bad situation can be avoided.

 

[junits15]

He’s a real whiz on hex and control systems and has been a great deal of help learning tuning. He designed and built the “humimod” I described in other threads. He’s also found lots of errors in hpt’s definitions. Just yesterday he found a mixup in the CSER and Starting DI blend tables that’s causing rough running/mis-fires on cold start when using e85 on a flex tune with CSER timing retard disabled.

I've been watching that project closely, if you or your friend decide to sell kits or even plans I'll be first on that list. Its a very nice addition to the platform.

Its funny you mention misfires, I know I had been complaining about that exact issue over on the HPTuners board with my e85 tune. I had a feeling there was something weird going on because by the book everything was right. I hadn't looked into the blend as a cause, I had figured people ran e85 on full DI cars and full PI cars with no issues, I didn't consider that the blend could be a problem.

Ok, I wanted to comment on a few things in this thread since I've been experimenting with this for the last couple of months.

Accuracy:
The Ford method is as accurate as the MAF calibration, Injector data, and lambda-to-alcohol% transfer table. If these data are accurate and there are no physical problems, then the learned ethanol can be dead on.

Regarding how it works:
I initially spent a lot of time trying to speed up the learning (not "maturing") process so that it would learn faster and be safer on transitions. After trying all sorts of changes, I concluded the ONLY change that sped up the learning rate was the "max update rate" parameter. I did observe that the rate of learning is a function of the max update rate, current error magnitude, and also airflow. It learns faster if the error is greater and if the airflow is greater. I haven't tested the WOT learn rate yet, though.

@junits15 I too have a log where I switched fuels and it matured within about 3 minutes at the incorrect concentration. The new fuel didn't reach the engine until about 4 minutes in. I'm currently trying what I believe @K4fxd suggested and extended the "min time to infer" and "min ego switches" from 10 to 20 to see if that will prevent maturing too soon.

@K4fxd my understanding of the flowrate thresholds is that it has to be within that range in order to lock in/mature. I've observed mine learning at idle, which is well below the 2.4 - 8 lb/min airmass min/max window.

@stang17 it automatically clears the KAM and restarts the learning when you add fuel to the tank. However, for testing you can use VCMScanner to clear the KAM as well and it will trigger learning again.

Cold Start Roughness:
I mentioned above this issue but I'll explain more. My car would start good but run rough and generate mis-fires for the first 30 seconds to a minute. From logging, I realized the issue was that it was attempting to run 60-80% GDI blend when cold. On gasoline, it had been only running 40% blend. Cold start idle smoothness is more challenging on GDI and even moreso on E85. It probably works ok when the timing is super retarded and throttle open, as is the case in CSER mode for cat light-off, but I have that disabled. Anyway, I changed the tune to use 40% blend in CSER mode on either fuel and now it starts and idles really smooth even on E85. HPT didn't help things by mixing up the GDI Blend FFV Starting and CSER tables. Initially I was modifying the CSER table since that's what mode it was in when the issue occurred, but it was really the Starting table since HPT labeled the two backwards.

Safety/Learning:
One thing I've done to improve safety is to toggle the spark timing at an ethanol concentration of 55%. Basically, at anything up to 54% you don't get any additional spark advance, but you get all of it at 56%. This way, when you switch to gasoline, it won't take much learning to turn off the additional timing. Secondly, you can use the A/F ratio gauge on the dash to know when this took place. 55% is 11.5/1 A/F, so when the gauge hits >12/1 at steady state then you know you don't have the extra timing anymore.

One option also is to just disable maturing. This would ensure it's always learning and driving fuel trims to 0 and couldn't possibly lock in to an incorrect concentration. The downside, I think, is that it would disable certain diagnostics since the fuel trims could never drift.

And finally, another option I've considered is to change it to not use NVRAM lambda when learning commences. You could set the initial value to something like 12/1. I believe the post-fueling re-start would idle a little rough on E85 since it will be 20% lean, but it might be ok since the engine is hot. This method would force it to learn down to 11.5 every time you fill up before you get any additional advance, when coupled with the timing strategy I mentioned above. The only concern here would be if you switch from E85 to 93, it would immediately start learning the E85 (since that's what's in the line) and might toggle the advanced timing before 93 hits the engine and you're left with the same situation. But, you could play with the settings to see if the bad situation can be avoided.

Excellent info, this will be extremely useful!

EDIT: would you mind calling out the specific table numbers that you've discovered are swapped? I'm curious to play with this now.

 

[stannypack]

Ok, I wanted to comment on a few things in this thread since I've been experimenting with this for the last couple of months.

Accuracy:
The Ford method is as accurate as the MAF calibration, Injector data, and lambda-to-alcohol% transfer table. If these data are accurate and there are no physical problems, then the learned ethanol can be dead on.

Regarding how it works:
I initially spent a lot of time trying to speed up the learning (not "maturing") process so that it would learn faster and be safer on transitions. After trying all sorts of changes, I concluded the ONLY change that sped up the learning rate was the "max update rate" parameter. I did observe that the rate of learning is a function of the max update rate, current error magnitude, and also airflow. It learns faster if the error is greater and if the airflow is greater. I haven't tested the WOT learn rate yet, though.

@junits15 I too have a log where I switched fuels and it matured within about 3 minutes at the incorrect concentration. The new fuel didn't reach the engine until about 4 minutes in. I'm currently trying what I believe @K4fxd suggested and extended the "min time to infer" and "min ego switches" from 10 to 20 to see if that will prevent maturing too soon.

@K4fxd my understanding of the flowrate thresholds is that it has to be within that range in order to lock in/mature. I've observed mine learning at idle, which is well below the 2.4 - 8 lb/min airmass min/max window.

@stang17 it automatically clears the KAM and restarts the learning when you add fuel to the tank. However, for testing you can use VCMScanner to clear the KAM as well and it will trigger learning again.

Cold Start Roughness:
I mentioned above this issue but I'll explain more. My car would start good but run rough and generate mis-fires for the first 30 seconds to a minute. From logging, I realized the issue was that it was attempting to run 60-80% GDI blend when cold. On gasoline, it had been only running 40% blend. Cold start idle smoothness is more challenging on GDI and even moreso on E85. It probably works ok when the timing is super retarded and throttle open, as is the case in CSER mode for cat light-off, but I have that disabled. Anyway, I changed the tune to use 40% blend in CSER mode on either fuel and now it starts and idles really smooth even on E85. HPT didn't help things by mixing up the GDI Blend FFV Starting and CSER tables. Initially I was modifying the CSER table since that's what mode it was in when the issue occurred, but it was really the Starting table since HPT labeled the two backwards.

Safety/Learning:
One thing I've done to improve safety is to toggle the spark timing at an ethanol concentration of 55%. Basically, at anything up to 54% you don't get any additional spark advance, but you get all of it at 56%. This way, when you switch to gasoline, it won't take much learning to turn off the additional timing. Secondly, you can use the A/F ratio gauge on the dash to know when this took place. 55% is 11.5/1 A/F, so when the gauge hits >12/1 at steady state then you know you don't have the extra timing anymore.

One option also is to just disable maturing. This would ensure it's always learning and driving fuel trims to 0 and couldn't possibly lock in to an incorrect concentration. The downside, I think, is that it would disable certain diagnostics since the fuel trims could never drift.

And finally, another option I've considered is to change it to not use NVRAM lambda when learning commences. You could set the initial value to something like 12/1. I believe the post-fueling re-start would idle a little rough on E85 since it will be 20% lean, but it might be ok since the engine is hot. This method would force it to learn down to 11.5 every time you fill up before you get any additional advance, when coupled with the timing strategy I mentioned above. The only concern here would be if you switch from E85 to 93, it would immediately start learning the E85 (since that's what's in the line) and might toggle the advanced timing before 93 hits the engine and you're left with the same situation. But, you could play with the settings to see if the bad situation can be avoided.

I'm at work and now my head hurts. I'll just pray juggernaut knows what they're doing lol

 

[shogun32]

I'm at work and now my head hurts. I'll just pray juggernaut knows what they're doing lol

Or forward him the post...

 

[engineermike]

EDIT: would you mind calling out the specific table numbers that you've discovered are swapped? I'm curious to play with this now.

Tables 3439 and 3440 are swapped. The starting table is for cranking and it immediately switches to CSER once it fires. You can tell they’re swapped because the CSER table row axis shows “starting events” which means it’s really the Starting table. The one to modify to fix e85 flex cold idle is 3440 and I’ve found 40% to work well as a max here.

 

[K4fxd]

@K4fxd my understanding of the flowrate thresholds is that it has to be within that range in order to lock in/mature. I've observed mine learning at idle, which is well below the 2.4 - 8 lb/min airmass min/max window.

Mine has learned at idle also but does need the time in range to lock.

Thanks for the write up.

The gen 2 does not have the "max update rate" defined.

 

[junits15]

Mine has learned at idle also but does need the time in range to lock.

Thanks for the write up.

The gen 2 does not have the "max update rate" defined.

Ohhhh you have a gen2? I missed that entirely, I've heard that its better on g2's because of less fuel volume without dual injection.

Sponsored