J-series Acura/Honda

[Fred]

Looks pretty realistic :-)

Wildly rich half the time, wildly lean post stepping on it (needs more code) which will result in a slight bit of hesitation if you JUMP on the gas, which you'll barely notice because it'll be disguised by the sudden lurch in acceleration. Typically more noticable with sudden 50% throttle changes. The time lag between the stepping on it and going wildly lean such as between 5.91 and 6.187 of just 0.277seconds is pretty decent if you ask me. Alan will likely be interested in this log so I'll link him here.

You need to actually tune your VE curve to match reality. Right now it's miles away. You can figure it out from this log, at least enough to get it close. I would suggest hacking AFR into the bin decoder so that you can think about it easier. For example at 6.645 seconds you have this:

Lambda the EMS thinks it is/should be of 1.0
Lambda that stead state O2 sensor reads of 0.718964
VE is flat everywhere at 60%

So correct VE for that instant in time is given by one of these, which ever you prefer:

60 * (0.718964 / 1.0) = 43.13784
60 / (1.0 / 0.718964) = 43.13784

Be careful not to mess with anything where the throttle is changing suddenly, and I guess to grab the O2 reading from 0.25 seconds later than the other values. IE the above is wrongish, though not much. I'd recommend doing a log of some smooth driving at diff loads/rpms and then an offline source file tuning session, then go from there with MTX watching the integral lambda and exhaust lambda gauges and making VE changes to get them to match. When they match all the time under steady state, your VE is tuned. Don't go too crazy on fine tuning yet as things are going to change when your sensors are stable and the RPM is stable. Also don't do any tuning until it's fully warmed up. Also don't do any tuning until your sensor inputs are stable, which they're not right now:

You've also got HUGE (14* C) noise in your coolant temp, which is fucking with your fueling. IAT is better but still noisy. You can see the noise from these in the O2 signal, which is cool.

I hope that helps! :-)

Fred.

[Fred]

Questions:

Darlington fix installed? If not, that will affect your tuning by changing the dead time value.
Big injectors or small? Just curious. Might pay to get it running well on the small ones first, then you've got a better base to switch to the big ones.

[Peter]

Yep the darlingtons are installed. Right now I have the small injectors in it.

So for the noise on the sensors should I double up the size of the capacitors?

Thanks for the detailed explaination for me to start tuning it.

[Fred]

You're welcome! I would suggest that the wiring or grounding is bad, but it's hard to know. It could be a wire routing issue too.

1) Do your P&H drivers have a totally independent ground setup? IE, ONLY 6 signal wires connect it to the CPU system?
2) Do your P&H wires run in parallel with the temperature sensor wires?
3) Do your temperature sensor wires use twin + shield?
4) Do your temp sensors ground back to the freeems only?

Possibly other things :-)

You may be the first person to properly attempt tuning of a FreeEMS, all the rest have been pretty rough tunes so far, in terms of fuel anyway. So don't be afraid to tell us various things suck. Obviously starting and sudden throttle changes suck, but that's to be expected for the time being. I hope it's tolerable for the short term.

If you find that ending up with the correct VE and matching lambda numbers makes it not run well, then you should change the lambda to be what you want it to be. You may find you need to fatten it up temporarily off idle to make up for lack of throttle pump effect.

Fred.

[Fred]

PS, post 3D screenshots of your tuned VE curve when you get it closeish :-)

[Fred]

From Alan "For my current generation stuff, in real world around 200-250ms sounds about right." just FYI :-)

[Peter]

I had a shotgun shoot over the weekend, and even though I could have got some tuning done I sat around and BSed with the other shooters. Hopefully I'll get something done during the week. I have a shoot in Oklahoma this weekend, and then Kansas City the next weekend, and Raton, NM the weekend after that. So progress will probably be fairly slow for the time being.

1) Do your P&H drivers have a totally independent ground setup? IE, ONLY 6 signal wires connect it to the CPU system?
2) Do your P&H wires run in parallel with the temperature sensor wires?
3) Do your temperature sensor wires use twin + shield?
4) Do your temp sensors ground back to the freeems only?

No, 6 wires with one common ground for the whole board, Yes, No, No. So basically all wrong. I'm thinking I'm going to set the P&H stuff up with an off board ground, and then shield the wires on the harness a little bit first. Then see how it looks.

PS, post 3D screenshots of your tuned VE curve when you get it closeish

I've played with it a little bit, but I'm not very happy with the results. I'll have 43% VE in one place, and then two boxes to the left I'll have 85% VE. The difference being whether the turbos are spun up or not. I feel like the table is a little coarse.

From Alan "For my current generation stuff, in real world around 200-250ms sounds about right." just FYI

I was thinking about this, and I realized I don't even know where to change the values for the injector lag times. Using the bigger injectors have kind of fallen off the priority list for the time being.

[Fred]

Don't worry about a lack of progress! Just be gentle to the car until you know it's tuned right.

Fix the ground situation first, make sure both the CPU/Sensor ground and the high current injector current are good grounds with decent wire, and not connected at the PCB.

Don't worry about the parallel and unshielded wire situation yet, however if you can easily bring the ground side of the sensors back to the board, do that. If not, try with just the high current separated. That really is important

The firmware supports larger tables, however MTX does not. In any case, 16x16 is almost certainly sufficient for a good tune on your engine. What you may not realise is that you can shift the axis to where they need to be. Some parts of your map will be flat, give those parts less cells, other parts will have sudden knee points, give those parts more cells. This is the art of tuning :-) You can shift the cells around after you get it closer to what it needs to be too, before you get there, you won't really know what is required anyway.

The dead time table is in TunableConfig1.c in initialisers, and the data for it is likely in the matching header, ctrl click through using eclipse or search with grep or whatever tool you prefer. You need data on that though before you can do anything meaningful to it.

Hope that helps :-)

Fred.

[Fred]

Hey man, how's the tuning going, any progress? I'm NOT bugging you for progress, just genuinely interested! :-) Also, if you do have it running a bit better then you might be a candidate for some video footage of it up against a 200+ rpm hysteresis fuel cut limiter. Because you're sequential such a thing is perfectly safe even with boost, providing there is enough hyst. If you're up for that (you can set the limit low, 4 or 5k or whatever) let me know and I'll hook you up with what you need.

[Peter]

Sadly no progress, I've been being a bum lately. Feel free to harass me for progress, because I'd like to see some to. I'd be happy to make a video of a rev limiter test, but I'm in Stillwater, OK this weekend with the university shotgun team. So it'll probably be Tuesday at the earliest, because it's about 13 drivng hours back to Laramie. Also it looks like a storm might keep us from getting back on time.

I've been considering putting a freeems ignition setup on my truck so that I can pretend that I'll be able to start developing a missing tooth decoder on the parallax propeller. For the hall sensor I was thinking I could use an op amp and subract the hall sensor output from a 5v signal, and it would look like the max9926s output. But I don't know if I would need to. The MS people seem pretty proud of the BIP373 to drive the coils. Is that a good choice other than they're kind of expensive?