Ecotec LS Coil Driver

[antus]

Yeah it can be really dangerous posting answers that look right, but are not. And if you can validate it then you didnt need it in the first place. I just wanted to get the ball rolling on calculating the answer, but I admit I did think its answer looked right. Good to have the crew here keeping it real.

[BennVenn Profile Website]

I was curious to see if it really did learn from your correction to its math, it didn't. I asked the same question you did and got the same wrong answer. I wonder how many corrections it needs to stick? or does it only correct your own login?

After a prompt, it got the correct answer

chatgpt_deg.JPG (90.74 KiB) Viewed 3860 times

[antus]

Ive been reading a book about AI, apparently they have multiple language models in the background, and as a human I provided a feedback loop for it to select a better one for the response I hit thumbs down on. Which was actually not the original correction, but the response where I gave it pman92s answer to prove it was wrong. It gave me 2 side by side answers and asked which was better. Neither were the correct answer, but one of them ran through pman92's and concluded it was 'also' correct and it might have made a mistake, and one just appoligised and that was it. So its probably weighted the algorithm to the other one when its corrected, but not done anything to fix the original fault or taught it anything.

Interesting that you can tell it where it went wrong and it'll have another go and get it correct.

[BennVenn Profile Website]

It sounds a lot more primitive than I thought it would be. I was expecting constantly evolving solutions and corrections. I guess comparing and weighting algorithms is just as good if it is right more often than not.

so is 20deg at cranking (3ms @ 250rpm) a problem?

[antus]

It would be, but 250rpm is half 500, so travel is also half what we calculated at that speed. So 5 degrees, not 20. 5 Degrees isnt a problem.

But like described above the charge limit is to do with current not time, so we are only ballparking it. And the RPM in the stock tune to hand over control timing control to the PCM is at 400rpm. Im still thinking its probably not a problem just for a moment for crank but it'd be nice to collect some real data. I would imagine a scope on the logic signal to the coils, would show us what is requested, and there would probably be an RFI spike in the signal when it fires, so we would get all the timing data if we measured it on a scope without needing to interface with the high voltage side. Someone has to have all the kit though.

[BennVenn Profile Website]

I have an LS coil here, I can pop it on the scope and track primary current vs dwell. I'll take a look tomorrow then we'll know what max dwell is and also what happens after

[pman92 Profile]

I asked the same question you did and got the same wrong answer. I wonder how many corrections it needs to stick? or does it only correct your own login?

I don't think its learning new stuff as such. Its just using the old data it was trained on.

It still says queen Elizabeth is still alive. And I'm sure there have been many thousands of people correct it on that fact.

[SilentBob Profile]

So how do I know which coils are not recommended ?
I do know of one or two who have used these IGN-1A Coils with a LSx Coil Driver box (Not your SilentBob) supplied through this site. Both had Kalmaker tunes applied & as far as I know no issues yet, apart from a bad earthing issue in ones wiring set up.

LS2 D585 & D514A not recommended.

LS1 coils (D580) recommended and most tested.

IGN-1A bench tested, but not long term in vehicle.

Most logic level coils should work fine, but due to the amount available I haven't tested them all.

If you're looking at any in particular, shoot me a pm and I can let you know if they'll work.

[OZ38 Profile]

So how do I know which coils are not recommended ?
I do know of one or two who have used these IGN-1A Coils with a LSx Coil Driver box (Not your SilentBob) supplied through this site. Both had Kalmaker tunes applied & as far as I know no issues yet, apart from a bad earthing issue in ones wiring set up.

LS2 D585 & D514A not recommended.

LS1 coils (D580) recommended and most tested.

IGN-1A bench tested, but not long term in vehicle.

Most logic level coils should work fine, but due to the amount available I haven't tested them all.

If you're looking at any in particular, shoot me a pm and I can let you know if they'll work.

Thank you.
I already have purchased the Coils. So I will find the specs & PM you.
Cheers

[In-Tech Profile]

LS2 D585 & D514A not recommended.

LS1 coils (D580) recommended and most tested.

IGN-1A bench tested, but not long term in vehicle.

Most logic level coils should work fine, but due to the amount available I haven't tested them all.

If you're looking at any in particular, shoot me a pm and I can let you know if they'll work.

Hiya,
I am curious why you don't like the D585 heat sink coils? The IGN-1A is a bad ass mofo. With enough relay and dwell, these things can almost weld