[Success Story] Ford Festiva / Protege / Miata motor (#3)

[AndreyB]

Good news is we have some sweet numbers at the wheels, but bad news we were getting a couple of nasty misfires pretty much on every pull. That's with higher then stock compression but also we an MSD coil instead of a OEM coil. Open question if the issue is with the ecu/firmware or if it's somewhere between the distributor and coil.

https://svn.code.sf.net/p/rusefi/code/trunk/misc/logs/1993_escort_gt/2016.02.20.zip

[kb1gtt]

In your last couple pulls, it is really consistently at 6.4kRPM. I think it's more likely to be a real world issue than a firmware issue. Some kind of resonance, or similar.

I wonder if that could be a mechanical issue, like valve chatter, valve overlap, or perhaps valve overlap caused by valve chatter. I doubt valve overlap as I don't see a pressure spike in your MAP, but who knows, perhaps a short spike is simply smoothed out in the log. Remember a small back fire would consume the fuel in the manifold, which would then cause all sorts of problems for the next couple rotations. However I suspect it's failed ignition, do you see a blip in the exhaust as the spent charge burns in the exhaust? I see the AFR is starting to trail off at the higher RPM's. Perhaps you need a pinch more fuel. The pulse claims to be 7mS, and at 7kRPM the time for 2 rotations is 17ms, so you have room for making the pulse longer. Keep in mind that your ignition window will change with AFR. If it were at say an AFR of 15, you would have a different spark advance for peak power than you would have at your AFR of 13. Also keep in mind that the AFR might not be correct, as well AFR doesn't indicate much about peak power. You can change your AFR to 11 with almost no change on power, but if you change your spark by 2 degrees, you get a significant change in power. Also keep in mind that peak power isn't always max advance.

What's the general setup? Are there pictures of the engine? Where does the MAP connect in the manifold? Where are the injectors located in the manifold? Is this a fuel regulator with return hose, or does it just create pressure? At 7kRPM and 6kRPM how much can you change the advance before you get misfires? At 6kRPM does it start to misfire at 35 deg, 36 deg, or other?

I'm assuming you were pushing the advance around and didn't see it get better as you backed off the advance. Also It looks like the table has a max advance of 33deg, but the log notes almost 34deg. I'm not sure what's up with that, or if your tuning table was simply backed down at the very end. Also I see you were running off of 12.1V battery. It might be worth while to get that to 14.4 via alternator voltages. The internal resistance of a battery can cause some short term fuel pulse issues.

[AndreyB]

for individual coils in protege or Miata you would need something along these lines:
bottom W6 to top W32
bottom W9 to top W29
bottom W12 to top W28
bottom W13 to top W27

[AndreyB]

Making progress on individual LS coils

[AndreyB]

Good news:
[video][/video]

News: pretty random MAP-based acceleration enrichment helps
https://svn.code.sf.net/p/rusefi/code/trunk/misc/logs/1993_escort_gt/2016.03.13_DataLogs.zip

Bad news:
Still one misfire about once every three minutes even on idle. I need to dig through logic analyzer logs first to find it: https://svn.code.sf.net/p/rusefi/code/trunk/misc/logs/1993_escort_gt/2016.03.13_misfire.zip

[stefanst]

Talking about the missfire: I just read in the "Poncho" thread that the injector drivers can go into thermal protection mode. I haven't studied the datasheet, but they may also have over-current protection. An injector driver cutting out for a short period for over-current or thermal protection could cause the very issue we're observing- couldn't it?

[kb1gtt]

Thermal overload when tripped would stay off until the chip has cooled down. This would typically take several minutes. Short pulses are not likely caused by such an hardware safeties.

[AndreyB]

top three channels are three spark channels, fourth channel was not recorded for some reason, two other channels are cam sensors.

https://svn.code.sf.net/p/rusefi/code/trunk/misc/logs/1993_escort_gt/2016.03.13_misfire.zip
I've looked at the .logicdata files using http://support.saleae.com/hc/en-us/articles/201589175 and I've looked at the .csv files using OpenOffice - I do not see the misfire, at least the spark is perfectly periodic.

spark rise
=IF((C2=1) & (C1=0);0;1)
current spark timestamp
=IF(J2=0;K1;A3)
duration between sparks
=K3-K2
unexpected duration
=IF(OR((L9=0);AND((L9>0.1) ; (L9<0.13) ));0;1)

[stefanst]

So spark triggering is doing OK. This leaves us with hardware or fuel as a problem. Since most of the spark hw has been replaced without any notable change in result, I'd be looking at fuel next. We do usually see a brief lean-spike during the stumble, so fueling failing is definitely an option. The ECU thinks it's applying unchanging amount of fuel, but sniffing the pulse output with a logic tester may make sense.
Just for my education, are you running sequential or batch fuel?

[kb1gtt]

Do you see the blips in the LED's on the frankenso board? That would help show is the problem is real world hardware, or frankenso hardware.

Can you wire an input or something, such that when the blip happens you toggle a pin, which would then leave marker in the logic analyzer, showing when the miss-fire happened, +/- a second or two.

Can you use a timing gun, and see if the timing gun fails to blink during the blips? AKA it would help identify a spark issue.

Those 2 plug wires resting on the block, raise concern. Can you put a larger gap between the block and the wire and see if that helps.

[AndreyB]

are you running sequential or batch fuel?

OEM batch wiring at the moment

[stefanst]

Is there maybe anything funky going on with the battery voltage during those events?

[AndreyB]

Is there maybe anything funky going on with the battery voltage during those events?

It did not occur to me to record a log file while idling, all we have is the logic analyzer log of camshaft sensors and three ignition control channels.

[stefanst]

Is there maybe anything funky going on with the battery voltage during those events?

It did not occur to me to record a log file while idling, all we have is the logic analyzer log of camshaft sensors and three ignition control channels.

Maybe there's something in one of the older logs?

[kb1gtt]

Where was the LA connected? I'm assuming it was the digital signals coming from the STM32. If so there is some potential (not much but some) for a hardware issue. Just because the STM tells the chips to do something doesn't 100% mean the chips do it. For example, that thermal protection of the injector drivers is external to the STM's control. The LED's will are driven by the driver chips outputs, and would indicate what is happening pretty much at the harness connector. If those are blinking as expected during the miss fire, that would indicate the problem is outside of Frankenso. However if the logs don't claim a problem, but you see funky blinking lights, then the problem would be frankenso hardware, and of course if the logs claim the issue, it's frankenso software.

I suspect it's an external issue, probably spark jumping a gap some where not expected, or a poor crimp connection on an inductive circuit.

Do you have O2 on it? Do you see a blip in the O2 readings? AKA a back fire would potentially show as rich blip, while skipping fuel injectors would show as a lean blip. This probably won't work well, as O2 sensors are very slow. Do you hear a pop in the exhaust? Are there any indicators of backfire?

[AndreyB]

EE question:

The car now runs on individual coils, this is sweet. Problem: tachometer wants the 4xRPM frequency signal. Ok, I've added a 5th output logic into the firmware and we have the 4xRPM frequency output on the OEM wires alongside the four 1xRPM signals on the conversion wires.

Now the EE part of it, all OEM harness
ECU -> igniter -> coil+tachometer

igniter output wire goes both to tachometer and to coil. Tachometer does not work without the coil. We've installed the (unneeded and unused) coil back and tach is working! Only problem we now have an unneeded coil being charged! And this just does not sound right.

So, how do we eliminate the coil while still having tach run? Does the coil work as a pull-up or a pull-down? what kind of resistance would we use to replace the coil? I guess not just a wire to gnd or +12? any other ideas?

[kb1gtt]

I potentially see a couple options. The coil is connected to GND on one side, and signal wire on the other. If you remove the coil, there is nothing pulling that signal low. So you could potentially replace the coil with a resistor. I would say you can start by trying a 1k ohm, and work your way down to 100 ohms if you need to. Keep in mind that 1k ohms will dissipate up to 0.144W while a 100ohm will dissipate up to 1.44W. It it doesn't work with a 100ohm, then you could measure the resistance of the coil and insert a resistance that's closer to that.

Alternatively you could wire a small pull up onto the rusEFI board, then connect that directly to the tach wire, eliminating that igniter as well as the coil.

[Old Grey]

blue line is stock ECU, red line is rusEfi

Did you ever find the missing high rpm power? From what I understand the only change was the ECU, so it has to be the ECU hardware, software, or tune.

From the graph the extra power under 5250rpm can be explained because you gave the engine the AF it wanted, not a figure in someone's head that someone THINKS should be right. An engine that has the ports polished to a mirror will need a AF of 10:1 to make max power, but a port with a rough carbide burr finish will only need 13.2:1 to make that same power - it will also use less fuel -.

Have you thought about leaning out the top end?

If you look at the VE curve of an engine the VE curve will practically over-plot on top of the Torque curve, which means that the engine will fill more per cycle at peak TQ and decrease as rpm goes up.
Say at 3500rpm the VE is 100%, due to factors the VE at 7500 might only be 75%, so even though air demand doubled because the rpm doubled, at the higher rpm it only uses 75% of the supposed rpm doubling of demand. Basically no engine wants a flat AF ratio to run at max power.

[kb1gtt]

I seem to recall it was the air valve not being activated. It's some resonant valve that needs to change at a certain RPM to keep the air flowing well.

[AndreyB]

That screenshot is from a year ago, @ do you have the dyno log / screen shot from this year?

[stefanst]

I believe the coil is being pulled low by the igniter and the other side is connected to +12V (IGN), essentially acting as a pullup. So a pullup resistor (start at 1kOhm as suggested by kb1gtt) should work.

[Old Grey]

I seem to recall it was the air valve not being activated. It's some resonant valve that needs to change at a certain RPM to keep the air flowing well.

That will do it every time. My internet is flaky and the VICS pic didn't come up so I thought it was only conjecture.

Just to give you an idea of the forces of wave tuning here is a video of John Kaase putting his finger into the port of a running race motor.

[video][/video]

[AndreyB]

That's quite an amazing video, thank you for posting!!

[AndreyB]

Friday test day - the car & the ECU worked flawlessly!

[kb1gtt]

Sweet, and good news.

[AndreyB]

rusEfi endurance racing debug is officially a success. I believe Festiva made more then 200 laps (x2.25 miles) and it was FUN to drive!
Drew would publish the logs later.

[AndreyB]

Race day logs attached

[stefanst]

Awesome!

Needs more data though
A few observations:
- Looking at TPS and RPM data, it's quote obvious that you -unfortunately- don't have a traction problem
- The regular noise on CLT and MAT follows the same pattern: Check your sensor 5V and GND
- CLT calibration is off
- CLT spikes every once in a while which is not realistic

Where's the oil-pressure data?

[AndreyB]

Where's the oil-pressure data?

Oil pressure was the green Miata and these logs are lost Festiva does have oil pressure sensor so we should wire it into rusEfi before next time.

Is there enough data to plug into VirtualDyno? Please note that time could be about 10% off since real time osc is not installed on this board.

[stefanst]

I'll see what virtual dyno says. Does anybody know the weight of vehicle + driver?
Gear and diff ratios are also needed for somewhat precise VD data....