DIY LTCC or similar system for LT1s

[spfautsch]

kur4o your PMs are full again.

I think I might have mis-interpreted your most recent one.

On hard decel event after cruising, the spark advance drops from 40* to 2-3* and than when you apply throttle it jumps again to 40* in no time. I guess we need to take some logs to see if it affects controller to lag some 2-3 spark events.

It was also felt in car like sluggish pedal, not sure if it is fuel or spark, since the injectors are also cut out at that point.

After re-reading this I think I better understand what you're saying here. While I would still like a better explanation of what you're talking about with "lagging some 2-3 spark events", I have noticed that transitioning out of DFCO is somewhat abrupt. As I have no way to do an A : B comparison (diy-ltcc vs distributor), I can only suggest you try that to see if the sensation improves. Personally, when DFCO is active I know the injectors are off and I don't expect a smooth transition back to combustion versus air pump mode. But I've never heard a pop or perceived what I felt to be a misfire.

I finally got mine back together and ran it some today. It's been sitting for over a month so the passage of time makes the sensations all that more intense. It was refreshing to remember that when I'm cruising at 15-25 mph in 2nd gear, or 20-30 mph in 3rd gear, a quick stab at the accelerator reminds me that I need to lose about 25 lbs (as my beer gut and fat head get sucked into the seat). If there's any lag or sluggish pedal here, I'm not capable of perceiving it with my butt-dyno.

Anyway, I'm going to try to do some logging + testing this week to see what, if any interpolation is needed and other fine tuning tweaks. I'd also like to scratch a few items off the to-do list like time-based failover / limp mode and to finish an assembly + install manual and document what I learned back in December on voltage sensing accurizing.

[kur4o]

I make some indepth log analysis. Finding the biggest positive jump of about 16*. On high rpm converted to ms is not much and can be compensated. On very low rpm around 1000 it can be a little enough to be felt a bit, since it converts to something about 1.5ms reduction of dwell. I still needs to find the exact condition when this can happen and to fatten the dwell in that range.

Now on the decel conditions, the issue seems to be fueling. The dfco have some spark blending recovery, so you don`t get a 40* jump but slightly increasing the advance in couple of jumps of 3-5*.

Also plan to increase the high rpm dwell to compensate for the 2 inbuilt Gm timers. What do you suggest as the max on-off time in percent that will not decrement spark output. I am sure some visiual testing can be done on your rig, with different on-off settings for 5-10 minutes constant running.

Not sure if you can make some DYI pressure chamber. A thick plastic box with a thread for a spark plug and a hose for pumping some 10-16bar pressure.

[spfautsch]

Finding the biggest positive jump of about 16*. On high rpm converted to ms is not much and can be compensated. On very low rpm around 1000 it can be a little enough to be felt a bit, since it converts to something about 1.5ms reduction of dwell.

I'm thinking that instead of some table interpolation maybe a simple damping / smoothing function on the millseconds to degrees conversion would be better.

Now on the decel conditions, the issue seems to be fueling. The dfco have some spark blending recovery, so you don`t get a 40* jump but slightly increasing the advance in couple of jumps of 3-5*.

My morning test drive (haven't driven it in a month) seemed to agree. I have on occasion felt the fuel resume from DFCO present a bit of sluggishness. I do have some code in the profiling logic that can help determine whether it's missing a firing event or two in this scenario, but I'm not sure it's worth spending time on at this point in the project.

What do you suggest as the max on-off time in percent that will not decrement spark output. I am sure some visiual testing can be done on your rig, with different on-off settings for 5-10 minutes constant running.

I think 50% duty cycle is more than safe unless you're planning on trying to qualify for a NASCAR event where you're going to be running > 5000 rpm for several hours. But that's just my gut impression of how these coils work. A safer answer would be I don't have enough information to answer that question.

Not sure if you can make some DYI pressure chamber.

I'd love to build some more complete data for these coils, but I also have other projects I need to / would like to tackle. Fixing the rotten garage door trim, the bathroom windows, etc.

Anyway, I do have some good news. I just got back from the first of what I'm sure will be many test runs, and I think I reproduced your on/off issue with a hard WOT pull (no deceleration was involved). The issue continued on for a while just like you described. Felt more or less like a 2-step that had been left on after launch and was getting random, incorrect sensor data. At the first cross road I pulled off, shut it down and restarted, and the issue was no longer apparent. This tells me it's most likely some boneheaded picket-fence error (unchecked bounds) or something similar. I'll have to spend some time analyzing the logs, but it's probably something incredibly simple + stupid. That's not to say it won't be hard to find, but my fingers are crossed.

Also, I've been messing with fattening up the cranking dwell, and I think a 160 multiplier for the 100/200 rpm rows is already too much. Will post more observations after some experimentation.

[spfautsch]

This should be worth a laugh or two. Didn't have a chance to analyze the logs yesterday afternoon, dug in just now and found the issue straight away.

[runtime],[rpm_index],[volt_index],[ect_index],[map_index],[dwell_ms],[dwell_multiplier]
533,8,7,11,7,3.33,120
534,10,6,11,6,3.76,119
534,11,7,11,6,3.27,118
535,15,6,11,15,4.05,128
535,21,6,11,15,3.96,125
536,24,5,11,13,4.19,113
536,25,5,11,14,4.19,113
537,201,6,11,0,4.65,147 < unchecked bounds returned from the rpm index function gives unknown dwell multiplier
537,201,6,11,17,3.54,112
538,201,6,11,17,3.54,112
538,201,6,11,17,3.54,112
539,201,5,11,17,4.16,112
539,201,5,11,17,4.16,112
540,201,5,11,0,5.46,147
540,201,5,11,15,4.75,128
541,201,5,11,14,0.04,1 < multiplier lookup likely getting data from a changing location in ram. this was causing the on / off sensation
541,201,5,11,15,4.75,128
542,201,5,11,0,5.46,147
542,201,5,11,0,5.46,147
543,201,6,11,9,6.36,201
543,201,6,11,0,4.65,147
544,201,6,11,0,4.65,147
544,201,6,11,0,4.65,147
545,201,6,11,2,0.03,1
545,201,5,11,13,8.28,223

Stupid mistake made when condensing and modularizing the table index lookup functions.

I'll have a fixed, sweep tested version ready soon. In the mean time just don't spin your engine past 6800 rpm. Edit, make that 6600 rpm.

[spfautsch]

Found and fixed improperly handled array bounds logic in all three table lookup routines that could have impacted any min/max transition adversely. Sweep test framework has been added as a conditional. Somewhat embarrassed to have put this out for consumption by the general public, but then again it's code that makes a v8 run to beyond 6500 rpm without a mechanical distributor and it cost you nothing to download.

Drove the fixes home over my normal 63 mile "back roads" route and the butt-dyno was not disappointed. Idle seems to be slightly more stable and overall torque was more authoritative / bordering on brutal in the mid-2000 rpm range.

Planning to add a dwell base adjustment - thinking "-" globally decreases the map vs rpm multiplier table, and "+" increases it. So ten "-"s fed into the uart would adjust the base multiplier from 128 to 118 and vice versa.

Will try to get this ready for public consumption early next week.

[spfautsch]

0.9.43 tested, sig updated with link.

Disclaimer - I was not able to test above 6600 rpm in-car because I didn't want to come to a complete stop and shift into 1st. Next time I'll wait until 80mph in 2nd to let off. But I'm fairly certain the fixes will resolve any unwanted behavior.

Also added in this version (quite some time ago) is a 5.xxx volt constant in config.h for fine-tuning voltage sensing. kur4o if you have a 7-18v variable dc power supply and a lab quality multimeter, fine-tuning this to the nearest millivolt (3rd decimal place) gives best precision. If not, the default value will put you close enough.

The multiplier offset seems to work, though I've been unable to push it far enough to see an increase in map / decrease in responsiveness. This lets the user increase or decrease the entire map vs rpm multiplier table by a fixed positive or negative number. It resets to zero with loss of power.

After giving it some thought, it wouldn't be impossible to turn this into self-tuning routine that could find ideal dwell in cruising ranges, though it would require some damned careful steady-state driving.

[spfautsch]

Just wanted to update with more info. Fully tested to 6800 rev limiter, no abhorrent behavior.

Also, I tested the multiplier offset fairly extensively. Based on your observations (kur4o) I decided to lower dwell to the point I could sense a difference in idle. It didn't happen until about -84 (base multiplier 44) or about 70% [edit: below] of the test data "max" dwell target. Somewhere in the neighborhood of 1.1ms. Even drove on interstate @ 75mph as low as -78. I ended up scaling the 8183 dwell table by 0.6 across the board, and adding some minor tweaks to the map vs rpm multiplier table. Seems to put the average dwell target around 2.0ms in normal driving.

I put about 150 miles on this setup with another 50 already on the trip odometer since last fill-up and came out at 22.1 mpg even with several hard wot pulls, a 1/4 mile pass to 130-ish and 6-8 aggressive merges onto interstate (15-75mph @ 60-90% throttle). No noticeable change other than feeling some "surginess" around 1700-2400 rpm occasionally that seems to be a product of using too many floating point logging pids at 500ms intervals.

kur4o if you've got time I'd love to hear your impressions of using the multiplier offset and how the engine reacts to less / more overall dwell. I'd really like to figure out if it's possible to use one map vs rpm multiplier table or not. Though if I have to I can probably just buy a complete set of D514a coils. Maybe a set of D585s too. Though I'd really rather spend that money on paint and fixing the dying window regulator motor.

I'd also like to discuss some ideas for implementing multi-spark - i.e. what type of tables would make the most sense, when it should be disabled, etc. Obviously it will have to work within lower rpm ranges, based on coil charge / recovery time.

Meanwhile I'll try to plug away at the documentation. Have I mentioned how much I hate writing documentation?