Back on the road, but a few things to tweak

[jangell]

Hi guys. I finally brought my DMC back home on Sunday (3.0L with EDIS and MegaSquirt). It's been about five years and four months since I started this conversion, which began with water destroying my fuel system leading to the discovery of a hole in the engine block. Thanks to all of you, it's now running pretty well, but there are a few things to tweak still:

- Cold starts take a few tries. After 3-5 cranks it starts, then dies. Another couple of cranks and it either starts and runs, or starts and dies. The third one usually sticks. Warm starts aren't a problem. It's hard to figure out how to tweak this since I can do it only once or twice a day.

- When I tap on the gas or switch from stopped to accelerating somewhat quickly, there's a momentary loss of power -- you can hear the engine basically stop for an instant and then start again. I don't stall -- I just lose power for a moment. I have set up basic MAP-based acceleration enrichment (I'm not running TPS at all now, since the sensor keeps breaking on me), and in general it feels fairly good, but I haven't figured out what to do about this problem just yet.

- I have a weird vibration when at idle, especially loaded idle (transmission in Drive or Reverse). It's not always there -- on the way out today it was strong, but on the way home it was completely gone when I stopped at home. When at cruising speed the effect is greatly diminished or disappears entirely. The effect is that when in the cabin you can hear the driver's side door rattles (need to adjust the striker pin a little, I guess), something near the glove box is rattling, and I can hear the engine cover rattle a bit (I think that's what's making that noise). This isn't really an EFI thing, but I'm not sure how what it is. From outside the car things feel fine. The engine isn't as smooth as my 2015 Subaru Outback, but that's a very apples-to-oranges comparison. My wife's new (to us) 1972 Dodge Charger with a carbonated 440 six pack is smoother too, though. And by "smoother" I mean I don't feel like the engine's shaking around as much. My engine isn't shaking a lot, but it's hard for me to quantify without really knowing what "normal" is. It's been so long since I've driven the DMC I don't remember what it was like, although my wife says it's definitely vibrating more. I put in new engine mounts, and the transmission mounts are "new" (well, about 7-8 years old, but not many miles on them). The muffler is fairly close to the rear subframe, but it's not actually touching it, so I don't think it's that, but I'll have to take another look in daylight to be sure. The cold weather is making it hard to want to spend a lot of time outside looking in the engine bay vs. just driving.

I'm still tinkering with various settings. I have my idle set to 850 at the moment, which was an attempt to reduce the vibration and keep the car from dropping to 650 when loaded, but adjusting the Closed Loop Gain and playing with Idle Advance Adaptive Timing seems to have fixed the idle drop, so I'll likely bring it back to 790 again. While I have basic acceleration enrichment set up, it's very hand-wavey, and more of a "this seems to work fine" kind of thing than anything empirical (I noticed Owen's setup (http://dmctalk.org/showthread.php?2057-Software-amp-Tuning&p=33227&viewfull=1#post33227) looks quite a lot different from mine, although he's on a 2.8 and I'm on a 3.0, but I might try settings closer to his to see if they help with my brief power loss issue).

Anyway, I've attached a log from today and my latest tune. Thanks!

-- Joe

58989
58990

[FABombjoy]

Cold starts take a few tries. After 3-5 cranks it starts, then dies. Another couple of cranks and it either starts and runs, or starts and dies. The third one usually sticks. Warm starts aren't a problem. It's hard to figure out how to tweak this since I can do it only once or twice a day
You probably need more cold start enrichment. But the lefthand side of your VE table looks like the rocky mountains and there's no way you're ready to tune cold start :D

Did autotune do this? If so, I understand the temptation to use autotune but your lambda delay table must be way outta joint because there's no way any sensible VE table should have such sharp peaks. Without adjusting the lambda delay, autotune will knock values around based on the readings from the wrong point in time.

If you have to autotune solo I would reign it in with a formula so it only tunes during static operating states (MAPdot / TPSdot below a certain threshold). Then you drive around, hold a constant RPM over several courses of the drive at various RPM & load points, and interpolate between those high-confidence values. Then look at high load / transients and remove fuel as needed. I say remove because your table should be set too rich to start for safety.

On a boring 2 valve, no timing, no turbo motor, the VE numbers will typically
1: Be lower at the bottom, higher at the top
2: Be lower on the left, get higher until peak torque, then lower again.

Since cold start & after start enrichments are multipliers of the VE table, any shenanigans present in VE will only be amplified by these modifiers and impossible to tune correctly.


- When I tap on the gas or switch from stopped to accelerating somewhat quickly, there's a momentary loss of power -- you can hear the engine basically stop for an instant and then start again. I don't stall -- I just lose power for a moment. I have set up basic MAP-based acceleration enrichment (I'm not running TPS at all now, since the sensor keeps breaking on me), and in general it feels fairly good, but I haven't figured out what to do about this problem just yet.
MAP-based accel will never be as good as TPS. It's purely reactive to past conditions, whereas TPS is proactive since it can gauge your immediate intentions before airflow changes significantly. MAP-based can only respond to changes it can detect. What I mean is that response will depend on which point on the manifold you're sampling, how much vacuum tubing you have running between the intake and Megasquirt, the diameter of that tubing, if you're only sampling 1/2 of the manifold (not sure how the 3.0 is interleaved but the 2.8 is basically 2 separate manifolds lightly joined by the cold start / PCV channels), etc. MAP-based enrichment on large transient events will likely range from "reasonable" to "sucks". Whereas TPS-based can respond nearly at the speed of light - plus a few milliseconds for calculations - and increase your injector duty cycle to match the wave of air approaching the intake valves.

Since you don't have TPS, it's difficult to read the logs and gauge what your intentions are. At approx 583 seconds it appears that you drive at wide-open throttle as MAP jumps to 98+, but RPMs don't change significantly, in fact they sort of bobble up and down strangely. It's almost like you're going up a steep hill in a lower gear? Or driving up ramps in a parking structure? I'm used to reading manual trans logs so maybe I'm seeing some effect of the torque converter or TCC (do auto D's have a TCC?)

At the end you can see what happens when conflicting parameters chase each other around. You have a target RPM of 850, but the spark advance table has you interpolating between 15 degrees and 21.7 degrees. So the base advance is actually operating between 17-18 degrees, but you also have idle advance correction enabled which is "givin' her all she's got" to pull the advance values back down so the idle isn't running away from you. The stepper motor is already on the floor and can't limit any more airflow in to the motor. The idle adaptive timing is limited to 4 in your tune - this could be increased a bit on both ends of the curve to help Megasquirt tame the motor down at idle. If you get that working correctly you can probably drop the lowest breakpoint in the ignition table as having 5 columns up to 1200 is probably a little finer resolution than needed :)

As a side note I hope you're enjoying this process, as frustrating as it is. I've been tempted a few times to tweak/post a new tune file for you but I'm getting the impression that you want to drive the software yourself. If not I'd be happy to post some test corrections for you.

[jangell]

You probably need more cold start enrichment. But the lefthand side of your VE table looks like the rocky mountains and there's no way you're ready to tune cold start :D

Did autotune do this? If so, I understand the temptation to use autotune but your lambda delay table must be way outta joint because there's no way any sensible VE table should have such sharp peaks. Without adjusting the lambda delay, autotune will knock values around based on the readings from the wrong point in time.

Yup, auto-tune. I did the first drive with it set to Very Easy, and it's been running on Easy since then, with the eventual goal of switching to Normal. I occasionally smooth the cells out, but I haven't done that recently. Most driving is on surface streets, with very little highway driving far. I haven't touched the Lambda Delay table.

I did notice the spikes in the 860-1600 range, but it's not clear to me if there's a good reason for the values to the higher in those cells or not.

I noticed that the logs show that I rarely get up over 3000 RPM, which strikes me as curious. I feel like I get over there pretty often in the Outback and the Charger (I have a bit of a led foot). Peak torque is supposed to be at 3750, according to the 3.0L manual. The DMC doesn't feel particularly sluggish or slow, although I don't have a working speedometer (angle drives fail after about 5 miles in my car), so I don't actually know how fast I'm going, but it feels fine...


If you have to autotune solo I would reign it in with a formula so it only tunes during static operating states (MAPdot / TPSdot below a certain threshold). Then you drive around, hold a constant RPM over several courses of the drive at various RPM & load points, and interpolate between those high-confidence values. Then look at high load / transients and remove fuel as needed. I say remove because your table should be set too rich to start for safety.

I'm mostly stuck tuning by myself. I just set up "mapDOT > -50 && mapDOT < 50". It's cold today, but I'll try to get out and do some runs. Hopefully this will clean up the cells a bit.

I'll have to figure out how to trigger higher loads. There aren't as many hills around here as I thought there would be, and I don't really want to ride the brakes to load the engine either.

Just to be clear, you're suggesting that the table will become too rich when tuning only at stable RPMs? Presumably I'd want to tune for the target AFR, which is what auto-tune should be doing, anyway, within the bounds of the inputs I'm giving it.


On a boring 2 valve, no timing, no turbo motor, the VE numbers will typically
1: Be lower at the bottom, higher at the top
2: Be lower on the left, get higher until peak torque, then lower again.

I have a boring setup like that. :)


Since cold start & after start enrichments are multipliers of the VE table, any shenanigans present in VE will only be amplified by these modifiers and impossible to tune correctly.

Gotcha.


MAP-based accel will never be as good as TPS. It's purely reactive to past conditions, whereas TPS is proactive since it can gauge your immediate intentions before airflow changes significantly. MAP-based can only respond to changes it can detect. What I mean is that response will depend on which point on the manifold you're sampling, how much vacuum tubing you have running between the intake and Megasquirt, the diameter of that tubing, if you're only sampling 1/2 of the manifold (not sure how the 3.0 is interleaved but the 2.8 is basically 2 separate manifolds lightly joined by the cold start / PCV channels), etc. MAP-based enrichment on large transient events will likely range from "reasonable" to "sucks". Whereas TPS-based can respond nearly at the speed of light - plus a few milliseconds for calculations - and increase your injector duty cycle to match the wave of air approaching the intake valves.

Yeah. I've considered jury-rigging something to the gas pedal, but that's not ideal due to throttle cable play and wouldn't trigger when pulling the throttle manually. I don't have a good solution to this, and the correct mounting point keeps causing the TPS modules to bind and break.


Since you don't have TPS, it's difficult to read the logs and gauge what your intentions are. At approx 583 seconds it appears that you drive at wide-open throttle as MAP jumps to 98+, but RPMs don't change significantly, in fact they sort of bobble up and down strangely. It's almost like you're going up a steep hill in a lower gear? Or driving up ramps in a parking structure? I'm used to reading manual trans logs so maybe I'm seeing some effect of the torque converter or TCC (do auto D's have a TCC?)

I'm thinking of getting a GPS so that I can map my logs in MegaLogViewerHD -- I get lost looking at the logs a lot of the time myself, even though I'm the one who was driving. It would help if I could add marks or notes to the log from TunerStudio (drive, stop, add a note, drives some more).

There's no TCC (no locking torque converter) in the DMC. I think being in gear on an automatic is like having the clutch part way out on a manual? There aren't any ramps or parking structures here, so it's probably just the way automatics work.


At the end you can see what happens when conflicting parameters chase each other around.

That's the biggest problem I have: try to reduce the interdependencies, while also need the car running well enough that I can tune it and not stalling out on acceleration or deceleration.


You have a target RPM of 850, but the spark advance table has you interpolating between 15 degrees and 21.7 degrees. So the base advance is actually operating between 17-18 degrees, but you also have idle advance correction enabled which is "givin' her all she's got" to pull the advance values back down so the idle isn't running away from you. The stepper motor is already on the floor and can't limit any more airflow in to the motor. The idle adaptive timing is limited to 4 in your tune - this could be increased a bit on both ends of the curve to help Megasquirt tame the motor down at idle. If you get that working correctly you can probably drop the lowest breakpoint in the ignition table as having 5 columns up to 1200 is probably a little finer resolution than needed :)

The 850 is from me trying to keep the idle from dropping too low, but I think I have that solved, so I'm setting that back to 790 (and the 810 column in the ignition table is going back to 790 as well).

I do have a lot of resolution in the low columns of the ignition table, which I only really noticed the day after I set them that way. I'm getting rid of the low column, and re-binning the columns between 1200 and 6000.

As for the idle advance, it should be interpolating between 14 and 15 degrees at 850 RPM, given the that my MAP is around 30-40 at idle (more like 45 under load at idle, which gets us up to 15-17). Or is 1200 still considered the idle range?

I'll boost the idle adaptive timing. Yesterday was the first day I turned that on. Before I had adaptive timing on, the car would idle at 790 RPM, though (or 810 or 850 or whatever I had it set to), but would drop down to 650 or before creeping back up to 790 when I put it in gear. Adaptive timing helped a bit there, as did noticing the Closed Loop Gain slider and properly adjusting that form 500 to more like 2500.

I can close the idle screw a bit so that the IAC has some more room to reduce the air flow. I set it close to be fairly close to closed at idle from your suggestion in another thread, but that's before I turned all this other stuff on. :)


As a side note I hope you're enjoying this process, as frustrating as it is. I've been tempted a few times to tweak/post a new tune file for you but I'm getting the impression that you want to drive the software yourself. If not I'd be happy to post some test corrections for you.

I won't complain about having a tune to look at for reference, but yeah, I do want to make sure I understand why I'm making the changes that I'm making, and you learn more through experience and making mistakes than you do just by being given the correct formula.

Time for a test drive. I've got most of the settings changed (except the idle screw), but I need to leave now. :) Thanks!

-- Joe

[jangell]

Ignore that thing about "never getting above 3500 RPM"; I drove the Subaru today and noticed I generally don't do it there either unless I hit the transmission kick down, and I usually don't hit that on the DMC either (it is working, though).

So I left my driveway, turned onto the road and the engine cut out. Couldn't get it started at all. After a few minutes it did start, though. I realized that what changed to make it start was that I turned the key all the way off instead of just trying to start it again from "run". I think that MegaSquirt and/or the TunerStudio MS Beta broke during communication, and since MS was waiting for the deadlock to end it couldn't do anything else. Makes sense, but weird.

Anyway, I did some basic tuning with the new settings via autotune, but not too much. My wife can probably help me next weekend, when it will be warmer. I'll try some more auto-tuning by myself in the meantime.

I have autotune set to "easy" but I still have those high peaks in the 1300 RPM range when the throttle is fairly wide open. I'm wondering if I should smooth the whole block from 71 kPa/860 RPM through 50 kPa/2200 RPM just to get back down to some more realistic values. It looks more like hills than mountains when I do that, but there's still that green zone in centered near 10 Kpa/1300 RPM that doesn't make much sense, given how rich the areas around them are.

I'm also wondering if I should add a column at 3750 RPM, since that's when peak torque is. ...and I just noticed I have a column at 2500 between 3000 and 4000 (as well as one before 3000). That's probably confusing things. I guess that's my 3750 column now.

I also have to properly mount my old power inverter and ignition-switch it at some point -- my test drive (plus a few chores) was cut short due to the laptop battery dying.

Thanks again!

-- Joe

[FABombjoy]

I have autotune set to "easy" but I still have those high peaks in the 1300 RPM range when the throttle is fairly wide open. I'm wondering if I should smooth the whole block from 71 kPa/860 RPM through 50 kPa/2200 RPM just to get back down to some more realistic values.
That's a sign that your lambda delay table isn't correct and autotune will never function correctly. It's changing your VE table based on examining AFRs at the wrong point in time. If it can't change the right part of the table it will just keep adding fuel to the same spot over and over again but not actually changing the correct cells in VE. If it does this but leaning thing out you could do some damage.

You can use autotune even with a totally jacked up lambda delay table, just not with transient load/RPM changes. You would keep a stable point in the VE table, turn on autotune, let it fix those cells, turn it back off, then move to another point in the table and repeat. With enough "good" points you can interpolate and have a sensible base map.

Constraining autotune with MAPdot is a good start but unfortunately RPMdot (or TPSdot) is probably a better, just because you're getting >50 MAPdot readings at idle. I'd just read up on tuning lambda delay and pore over your logs and adjust your table. Then autotune will be less likely to make quick, nonsensical changes everywhere in your VE table.

[jangell]

Oh yeah, RPMdot makes a lot more sense; that's any easy change. I'll do some research on setting up a lambda delay table. My initial searches earlier today suggested that the default values were good for most people, but I may be an exception here. While I get what the table does, it's figuring out what values to put in it is unclear.

I'll reset my table to the "clean" version, switch to RPMdot for the filter, and see if I can figure out the lambda delay table and see what that does.

Thanks!

-- Joe

[jangell]

I loaded up a clean table and ran an errand today with autotune and of course the laptop died before I could save the changes. After recharging it, I drove home and tried again. I had to set the RPMdot filter to +/- 100 to get autotune to make changes, which is fine.

It was about a 10-15 minute drive home on surface streets, and I tried to hit some specific RPM ranges and MAP values, holding at those values. I shifted to 1 and 2 to put more load on the engine and hit a bit more of the table. In the ed, auto-tune gave me this table:

5900059001

I also ran it through VE Analyze in MegaLogViewer HD, which gave me this:

58998 58999

I used those to build a new VE table via lots of interpolation, which more closely matches what I think a VE table should look like:

5900559004

The downward tapering on the right is mostly due to not having any tuned data there, so it's just interpolating across it. I'll probably have to drove around in first gear to hit that part of the table in autotune, but it seems unlikely I'll hit it in practice. I should probably extend the higher fuel values based 4000 RPM into that top right area, though, I imagine.

Once while driving with auto-tune I hit an area on the edge of a high value cell and low value cell, and got the rapid change in RPMs you'd expect from such wild differences in fuel. It wasn't enough of a difference to do any damage, but you could feel and hear it. I'll take it was a warning to be careful with the fueling.

I'm going to do more runs (probably tomorrow) with autotune set to "normal" resistance on this table and see what that gives me.

I haven't figure out the lambda delay table yet. Something about looking for AFR transition events at specific RPMs and then looking back in the log to see when the pulse width changed, thus getting the delay between when the PW changes and when the matching AFR is read, but I'm not quite good enough at finding things in MegaLogViewer to do that reliably yet. I mean, it looks like at about 50 kPa and 2200 RPM there's just one record (~60ms) between the PW changing and the AFR changing, but that seems awful short -- less than half of that same cell in the default table.

Also, to answer a question from before: the MAP line is plugged directly into the throttle body, immediately below the butterfly. The 3.0L manifold doesn't split into two halves like the 2.8L one does, so you can probably tap it later as long as you don't mind an increased delay. If I ever get my 3D printer working again I'm going to just build a TPS sensor adaptor of some sort that doesn't bind, or just make my own TPS. A potentiometer in a housing just isn't that complex. I wonder if I can find a magnetic TPS and retrofit it somehow...

Thanks again

-- Joe

[jangell]

Quick update: So that fuel table was WAY too rich. I forgot that MegaLogViewer's autotuner was going to use the lambda delay table, which I still hadn't set up. I drove for about an hour on Monday, with about 30 minutes of that on the highway, and TunerStudio's autotune leaned it back out over the course of the drive, and from that I interpolated a new table. I hit a few of surging instances on the highway where it needed to interpolate from a 115 cell to an 85; not a big deal, and not enough of a difference to do damage, but I made sure to smooth out all of those transitions in the new table.

I also analyzed the logs and tried to create a somewhat better lambda delay table, so we'll see how that works going forwards. However, I still haven't found particularly good information on creating one, usually just a sentence or where people talk about creating their own, but not with enough specifics about how to actually do it yourself. There's mention of creating a "shelf" in one of the tables that makes the engine really rich, then seeing how long it takes at a particular RPM to hit that AFR after a MAP change, but it's not clear to me exactly how to set that up or how to run it safely. At the moment I'm just looking for such transition events in my normal driving logs.

-- Joe

[FABombjoy]

You can strategically make a few cells unusually rich. Drive, log, hit all of those cells.

Then look at the PW values in the log -vs- EGO. The logs already record the time. Look for the jump in PW and see how many milliseconds it takes to register as a drop in EGO and that's the delay at that RPM / load.

Lambda delay should ship as blank and not work until set. Or it should have a calibration mode. The default table seems to cause people grief and give VEAL a bad name.

[jangell]

Cool, thanks -- I'll try to do this on Saturday. It'll be cold, but the rain and snow shouldn't start until Sunday. I have a 2.5 hour (each way) trip I next Saturday, so it would be good to get this table figured out before then.

Just to be clear about how this works: The initial table does seem pretty useless. It appears to set the low end of MAP and RPM to the lowest parts of my table, which I will never reach, and thus won't ever get good values for.

I'm thinking of re-mapping the table to have MAP at 50, 75 and 100 and RPMs at 1000, 2200 and 3500. Then I'll set those specific cells (50 MAP and 1000 RPM, etc) to 130-160 units (which is about 30-50 units too rich for my table) on the VE table and drive around trying to hit them. Does that sound right? How rich is too rich before I start breaking things in my engine?

I was also looking in my logs at PW vs. AFR, not EGO, which was likely part of my problem.

Thanks!

-- Joe

[FABombjoy]

Sorry, I meant AFR. I was thinking "exhaust gas oxygen" when I said EGO.

[jangell]

Ok, so I wasn't as far off as I thought I was. :) I'll compare the time from PW change to AFR change. Thanks!

-- Joe

[FABombjoy]

Here's an example. I tripped overboost protection which cut fuel.

Since this was a high load situation and high exhaust velocity, when PW goes to 0, it only takes a few records for AFR to increase (0.134 s)
https://imgur.com/a/lhrOLxw


Closer to idle, it takes a little more time for PW changes to be reflected (0.2 - 0.266 s):
https://imgur.com/a/Z0wbVbm

Note that those delay times aren't perfect as they are multiples of the data sampling rate, but they're probably reasonably close enough to use for VEAL.

[jangell]

Thanks! I suppose I should turn off some logging options to increase the number of records per second. Right now my resolution is about 60 ms. Which is probably fine for auto-tune, but more resolution wouldn't hurt.

-- Joe

[Ras12]

This is all WAY above my pay grade. But Congrats on getting your Delorean back on the road. I'm sure I will see you driving around southern MA! Best of luck with the tuning.

[cpistocco]

This is all WAY above my pay grade. But Congrats on getting your Delorean back on the road. I'm sure I will see you driving around southern MA! Best of luck with the tuning.

You guys are pretty deep into the tech! Amazing... :)
I think it would be nice if we all joined a sub-group... I invite you to join The Delorean club of New England ( I just set up the group) If you are interested.
It has been defunct for some time....but it would be nice to resurrect it....just so we can share plans for cruise nights, shows, etc.
-Thanks
Charlie

[jangell]

There's a lot of "making it up as I go along" here. :) Lots of research on forums and docs, but some of it only really makes sense once you're actually doing it. Taking it step by step keeps it manageable. All the people here have been hugely helpful, too, and it helps a lot to know there's a community to help out when you get in a bind. And after you've done it and have it working, it doesn't seem like it was really all that hard, but actually doing it the first time can be a bit daunting.

I'm in favor of re-starting the DMC NE club. I think we had one meeting about 10 years ago where a 8 or so of us met up in the southern MA area, but we never did anything else after that.

-- Joe

[jangell]

I did build a Lambda Delay table on Saturday, but I think I needed to set the test VE table cells higher than just +50. I also forgot to turn off acceleration enrichment, which made it harder to tell tell if the pulse width change was because I hit the cell or because I was accelerating. I'll have to try again next weekend or the one after that, but the new table is certainly better than before.

-- Joe

[FABombjoy]

For measuring delay I don't think any enrichments will affect the delay measurements.

[jangell]

My problem was more that the spikes weren't big enough to easily spot the change. The other problem was actually hitting parts of the table at all, but I mostly figured that out. I'll try bigger spikes (like +75 or +100) and see if they stand out more in the logs.

Even so, the table I have now has about 2/3 or less of the delay of the original table, so there is definitely a change there. I also found that reducing the kinds of data written to the data log didn't affect the records per second rate, but ~67 ms per record is probably sufficient for this.

Thanks

-- Joe

[Ras12]

You guys are pretty deep into the tech! Amazing... :)
I think it would be nice if we all joined a sub-group... I invite you to join The Delorean club of New England ( I just set up the group) If you are interested.
It has been defunct for some time....but it would be nice to resurrect it....just so we can share plans for cruise nights, shows, etc.
-Thanks
Charlie

Post the link and I will join for sure. You should also connect up with the Facebook New England DMC group.