Printable View
under certain conditions, I get a lean mixture. I can?t understand what he?s missing. The PCM is trying to enrich the mixture. Ignition misfires appear on the 8th and 7th cylinders. I associate misfires with the lean mixture.
5.7 maggie's supercharger 2300
Swap it over to speed density and tune it the better way ;) Right now your building boost - no PE - misfiring - loosing boost due to misfiring and then it's defaulting to spark safety tables. Hopefully the engine isn't harmed hitting those lower boost levels without fuel where it's misfiring. Also looks like your tuning for boost in the PE tables - set these to what your target is at the desired air loads and tune fueling via VE Maps.
how did you understand that there is no PE mode? right now installed 2.9 volts on the pedal.Quote:
Originally Posted by GHuggins
Desired mix in log 10.5 afr or 0.75 lambda. The injection time is 11.6ms, but this is still not enough for some reason.
Is the whole problem in VE?
I was going by map and commanded lambda in the log. It's still commanding a lambda of 1 after hitting 120ish kpa map if memory serves. That's when it starts missing and then commanding a richer lambda much later in the run. Others do, but I don't ever tune a heavily modified Hemi using the neural network.
At least that's what I saw in the log I looked at unless the data was jumbled some how?
Here is what I'm seeing in the log you posted and why I'm saying to tune this via SD/VE - In the first photo you will see your going into boost and never commanding PE - look at commanded EQ ratio - Pedal stays at 40ish% the whole time during this entire rpm pull with varying engine load - after it goes into boost it starts misfiring as represented by the lean O2 spikes - you eventually hit around 5200ish rpms with roughly the same throttle position at which time it finally starts going into PE... Again - tuning via VE/SD is the best way to fix all of this...
This is another screenshot using a dodge layout where you can clearly see the boost building section/misfiring section of the log where it clearly states it's in CL ;)
I understood you. Thank you. I tried to enrich the mixture using the AFR sensor, but reached 300% in some cells in the VE table, and realized that it was useless, nothing really changed. Problem with NN? When NN is turned off, it gets better, but until the engine is warmed up, it works very unstable. How to setup Gas Burn fraction?Quote:
Originally Posted by GHuggins
You have to disable NN to tune VE and to run VE only. You should also be able to tune other fueling tables for colder start enrichment multipliers. It's a little harder on a dodge platform admittedly to do this, but it's possible. This is how I tune all add on blower jeeps, hemi's and so on... Did you at least get a base tune for the blower and injectors from the company you purchased it from? You might even want to consider going that route as they usually have a lot of this already in the tune. Could save you a lot of time and possible engine failure.
I have a stock setup. the engine runs on it very richly.Quote:
Originally Posted by GHuggins
I ran into a problem while editing VE tables. When you try to decrease or add fuel, there is no reaction to the action. If the cell has gone through several stages of change to increase the%, then the mixture in this condition becomes even poorer, as if I were decreasing and not increasing the%. NN network is disconnected, CL is open for lambda probes before catalysts and lambda probes after catalysts. In some cells, the values reached 300%, the mixture did not change. Mixture control by AFR sensor
Really sounds like it's not disabled. Well you can always put the VE tables both back to how they were in the base map provided to you and then tune the fuel mass table for the NN and use it.
This is how you tune the ann by fuel mass written up by Jay.
https://forum.hptuners.com/showthrea...&highlight=ann
I used this guy's technology. it helped me a lot. Thanks for the advice.Quote:
Originally Posted by GHuggins
I turned off the oxygen sensor control and the NN network.
I share the firmware and record the log. In some loads, the desired mixture is too rich. I can't understand why.
Attachment 116794
Attachment 116795
If you tune with VE and disable the NN you will have to "Fix" Some driveability issue also the old solution of which is described by "Jay" dialling in the injector flow rates to achieve the correct mixture is tuning the injectors into the airflow not the airflow to suit the engine modification. If you disable the NN you will always have quirks due to the the variable cam altering the VE for which a static VE cannot compensate. HP now has the ability to retrain the NN through a series of VE tables that relate to various com positions, it is a fair bit more work but offers a far superior result as you enter the correct injector data and adjust the NN which generates the airflow/airmass. It is time consuming and with a V8 which has a fixed single cam requires you to tune 5 VE tables in different cam positions from minimum to maximum angle.
Which version is this supposed to be in? I have the latest beta and none of the hemi tunes I've opened show any sort of NN VE tables. Just blank unfinished looking nn trainer page that nothing works in.Quote:
Originally Posted by 4wheelinls1
Honestly if your cam is set to be at a specific position at a specific rpm and load then your VE table should keep to the same fueling. Yes, it can wind up looking very weird, especially boosted setups, but should stay to correct fueling. I don't recall ever having drive-ability issues on any mopar platform I've tuned. Granted I have had to rework some cam tables on boosted setups for owner "preferences", especially those who constantly drive in and out of throttle all of the time.
share the firmwareQuote:
Originally Posted by GHuggins
This has different cam tables for you to try for your setup. I also changed your VE tables to support the changed cam tables. You'll need to dial you VE back in, but they should be close. This will help with your rough VE tables and should help smooth things out for you.
As for the built in 5 VE tables to tune - I would like to see a file that has this operational...
Sorry I have been away. It has been in Beta for 12 months. You need to generate a file to train and then submit it for training.. https://www.hptuners.com/products/neuralnetworktrainer/ it takes tokens and gives you a month on the server. If you move the camshaft you move the trapped air and so the efficiency even though the manifold pressure and RPM may be the same.Quote:
Originally Posted by GHuggins
thank you for your work. Can I use these tables without the involvement of the NN network?Quote:
Originally Posted by GHuggins
I have done about 40 5.7 Ram vehicles with superchargers with NN retaining, fortunately there are only 2 significant OS variation so once they have been done the same NN values can be used for the same modification over and over. I have also done small camshafts in 5.7 Ram with full variable motion and 6.4 Hemi with larger restricted camshaft movement (limiters). The vehicles calculate airflow correctly so drive better and deliver better economy.
The image is of the NN trainer when tokens are purchased, the cam angles are selected in the top Left and due to the one piece solid cam in the V8 only 5 need to be trained as the inlet cannot move independent of the exhaust. If you were to do a Pentastar there are 25 potential positions, the more of which you train the more accurate the NN output for airflow will be.
Attachment 117015
Guess I should have researched it more. Thanks for the information and I do understand what your saying, but at the same time even using just the VE in the tune only, I've gotten extremely good running and driving vehicles. SOooo, doing it the more proper way if you will, should theoretically yield an even better result. Really wish HPT didn't make you have to purchase other credits to use it :/Quote:
Originally Posted by 4wheelinls1
To 88nec - yes you can tune it how you choose. IF you do the one VE only, just dial it in how you were - just try to keep the table transitions smooth...
The cost is much lower, you need tokens which are only $2USD, 25 for one vehicle for 1 month use of the HP processor.Quote:
Originally Posted by GHuggins
Having done all three methods I can say very honestly this is the best by far and although more time is spent on the NN VE tables being multiple it means less time on the idle and transmission as the airflow and Tq calcs work correctly.