NN training troubles

[GHuggins]

Does anyone know what controls the NN configuration to shape the new VE tables? I have been fighting this lousy NN training tool for a month now. I raise up and interpolate or smooth the tables from low load to high load and then run it through the trainer tool and it ALWAYS spits out new settings that are VERY sloped from low load to high load and flattened out which takes out all of the fuel I just put into it. I've tried putting in ridges going from low load to high load like the STOCK tables have - still flattens them out. If I raise the low load decel row it winds up being way too rich there, but halfway right in the mid range where I need the extra fuel.

I then took a different approach and started modifying the SD VE base tables and that "seemed" to help more than anything else, but still far from right and requires some serious mods for small changes. I'm basically up to 8% lean where it idles through wot in those same load areas because the stupid tool keeps sloping the tables. Is this just how the Dodge tool works vs the gm setups running the same thing? Or is it an issue with hpt's configuration tool? Should I be adding 8% all the way across these load zones including the wot area? This one had about 16% added into it and came back 8% lean and not only 8% lean but was also now lean where it wasn't previously before on higher up in the loads. Seems like the lowest load setting is controlling the whole table. I can't tune with Nn disabled on this one because apparently another tool corrupted some of the switch controls and it's impossible to disable it even with the cam changes that certain years models have issues with so trying to work with what I have.

NN Problems.jpg NN Fuel Problem.jpg NN VE table 2.jpg NN problem 2.jpg

[PurpleRam]

What year are you working on?

[GHuggins]

I think it's a 10 or 11. I already tried the cam settings to get the base VE table to work if that's what your asking. Pretty sure it was the previously installed diablo tuner that corrupted something as even the owner stated once he pulled it out hpt was asking to save changes even though he couldn't see any. It'll usually only do that if something in the background was changed. I wound up working the main VE tables to work around the automatic curvature that's put into the new VE tables upon calculation. Only way I could work around the dip that the "auto calculate" was creating and compromised slightly on the decel trims just enough to not cause other problems.

[rays04gtx]

I think it's a 10 or 11. I already tried the cam settings to get the base VE table to work if that's what your asking

that was going to be my thought also,

[GHuggins]

Yeah, it's just something weird about how the NN trainer spits things back out. Honestly you shouldn't have to use the main tables to force fueling a certain direction. This was the first time I used the NN trainer on a Dodge. I've always just killed the NN and used the main VE table before. It is running better with the NN being changed I believe.

It's also an odd setup. One I haven't run across before with the intake manifold. It's using an rpm switch based long/short runner manifold with the 6.4 cam, injectors and heads. I also confirmed the year model of the 6.4 injector donor. Required a ton of fuel to be pulled below 3600 with the short runner setup and then a ton put back into it above that. I've tuned plenty of 5.7's with 6.4 heads, cams and injectors before without a single problem, but this one definitely gave me a run for the money.

It also had a weird cylinder cut out issue - would drop 2 cylinders and limit how high the engine could rev until the key was cycled and even then it wouldn't let the engine rev above 5000ish - never seen that before and it was due to the VE changes even though I had the limits just about maxed. I'm almost certain the diablo unit corrupted something. Just wish the NN trainer didn't manipulate the tables the way it does. Tuning in the NN allowed it to rev back out and only every once in a blue moon still drop the cylinders. I think the cylinder drop was just finding the right balance between the main VE tables and the NN tables.