Injector driver?

[seijirou]

I have a sudden issue where bank2 is very rich.

As expected the plugs looked bad, so I changed them, no improvement.
I suspected funky injectors so I swapped bank to bank, no improvement.
I questioned cylinder filling so I inspected cam lobes, lifters, and reset valve adjustment, no improvement.
I questioned the wideband sensor health, so I replaced them both with new ones, no difference.
I questioned the wideband controllers so I swapped which sensor feeds which controller, no difference (the rich reading is simply read by the other controller now).

Now I'm wondering if it's in the ECU. I looked up DIY's injector driver troubleshooter (https://www.diyautotune.com/support/tec ... or-driver/) and on a stim I can get LED response for bank1 and bank2 by grounding the associated U1 pins 21 and 22 respectively. Nevertheless I verified the inverted output of U4 and that the anodes of D5 and D7 go low when the corresponding U1 pin goes low (which I believe is confirming that Q1 and Q5 are doing their jobs). The guide pretty much states that if the LED comes on, the board is good. Are there any other tests/checks I can do in a situation of a bank suddenly going rich that could be MS board related or did the above checks basically put this to bed?

What I have not done yet is swap the wiring for bank1 and bank2 to see if the problem follows the wiring or stays with the physical engine side, and I have not swapped the coil packs bank to bank yet to see if the problem follows the coils or not. A compression test is also on my mind.

[seijirou]

So I swapped the injector wires bank to bank, and the problem followed the wires so it's something in my wiring or it's something in my megasquirt. The basic troubleshooting I outlined above didn't indicate (to me anyway) any issues in the megasquirt itself but I think that's going to be a lot more likely than the wires.

Any ideas appreciated!

[seijirou]

Well I've mostly figured out what's going on. I don't have a scope and I've exhausted my limited electronics understanding but here's what I've got so far.

My MS fuel was configured for alternating events, 4 squirts per cycle.

I walked through the schematics and tested every component from U1 pin22 through to Q5 pin2 for proper resistance, diode function, +5 where and when it should be, and +12 where and when it should be when U1 pin22 is grounded, all of that checked out.

Then I reasoned that when running on the stim, U1 pins 2 and 4 should be receiving a +5 square wave with a duty cycle that increases with RPM, and that I should be able to roughly gauge the duty cycle with a DMM set to measure AC voltage. Similarly U1 pins 5 and 7 should be delivering a +12 square wave so I should see a correlation just at an elevated AC voltage. Importantly whatever I measure on U4 pins 2 and 4 should match each other, and the same goes for U4 pins 5 and 7.

I started by setting the stim RPM @ 1,000 and found that U4 pin7 read significantly higher voltage than pin5. I believe this means that I have a higher duty cycle on U4 pin7 which would correlate to longer injector pulsewidth and a richer mixture. I then turned the RPM up to 6,500 and then was surprised to measure significantly lower voltage on U4 pin7 than pin5. I then measured the input side of U4 and found that the voltage differences did correlate. At 1,000 RPM U4 pin4 was higher than pin2, and at 6,500 RPM U4 pin4 was lower than pin2. So at this point I believe U4 is working properly, doing as it's commanded.

I then changed the configuration of MS fuel for simultaneous events, 1 squirt per cycle. With this configuration tested at both 1,000 RPM and 6,500 RPM I measured the same voltage at U4 pins 5 and 7, and again the same voltage at U4 pins 2 and 4. It's too late for me to physically test the engine but I expect this should activate the injectors equally and would solve my unbalanced fuel problem.

Investigating further I found that measurements were equal and what I would call "good" for Simultaneous with 2 events, and Alternating with 2 events.

When I changed to Alternating with 4 events, the measurements became unequal again. I also tested Alternating with 8 events and saw unequal measurements.

As best as I can tell looking at the schematics I believe my board is good and is doing what it's told to do, the problem is with the signals coming in to U4. In my mind that leaves 2 possible causes which is either a problem with a component on the daughter card as there is an integrated circuit (U2 on the daughter card) between the CPU and U1 pin 22 so it could be this component which is faulty, or it's a bug in the firmware that's presenting itself only with certain configurations.

Testing the IC on the daughter card is above my head without specific instructions so I did burn every stable firmware revision back to 3.3.3 to see if the issue cleared up but it remained. At that point it's late enough for me so I'm calling it a night.

I will confirm in the near future that using a setting of Alternating with 2 squirts, which gives me "good" measurements solves my fuel issue. If I can do any further data gathering to determine exactly what is going on I'm happy to help but I'll need some specific instruction on what to do at this point. If one of the devs sees this and there's a particular firmware of interest that would be worth trying to see if that solves the issue I'm happy to do that too. I'd rather not walk backwards through every firmware in the archives if I don't have to but if that's what it takes I'll do it if the devs are interested in figuring this one out with me.

[DaveEFI]

You could make up a circuit the same as the LED drivers on the V3 board. Connect that to the injector driver outputs in turn on the daughter card, or at the inputs to U1. At low revs from your stim, the LED should flash in the same way as the injector lights flash on the stim via the injector drivers.

[seijirou]

You could make up a circuit the same as the LED drivers on the V3 board. Connect that to the injector driver outputs in turn on the daughter card, or at the inputs to U1. At low revs from your stim, the LED should flash in the same way as the injector lights flash on the stim via the injector drivers.

Observing the LEDs on the stim was actually my first clue that I was on to something with the problem being in the MS and not outside of it. It was obvious at a low RPM with my original Alternating, 4 squirts configuration that the channel 2 LED was a lot more active than the channel 1 LED. Initially I thought to record it on video to post here but the arguing voice in my head said that's not good enough and I need to somehow measure it. Following that is when I came up with the idea of using the DMM set to AC.

[DaveEFI]

You need to ascertain if the injector driver circuits are OK - and that the processor is sending pulses to both. I find it easiest to check the hardware is good first. If it is, the problem is likely the software.

[seijirou]

You need to ascertain if the injector driver circuits are OK - and that the processor is sending pulses to both. I find it easiest to check the hardware is good first. If it is, the problem is likely the software.

I feel that I've already determined that the driver circuits are OK from the driver output at the large transistor Q5 all the way to U1 pin22 where the handoff to the MS2 daughter card occurs. U4 pins 2 and 4 are the inputs from the daughter card, pins 5 and 7 are the outputs to the driver transistors. I am getting signal in, and out of U4. The problem is the signal coming in on channel 2 is not equal to the signal coming in to channel 1 under certain configurations. Is there something about my method that isn't sufficient? I don't have a scope to measure duty cycle directly unfortunately.

[DaveEFI]

That's why I suggested the LED circuit driven from the processor direct. At slow rpm, any difference between the two would be obvious.

[seijirou]

Oh directly to the CPU pins, I didn't understand that's what you meant. I'm not sure if I can get something on pins that small but I'll have a look.

Approaching this as a thought-experiment though, I think the board testing I've done is already pretty conclusive.

So U2 is just a 4 way NAND gate with only 2 being used. Point being it's just another component that switches an output on/off inverted with on/off switching inputs (2 "on" inputs = "off" output and "off" output, being ground AFAIK, is sent to motherboard U1 pin22 and fires injector channel 2) - in this case the input source is the CPU.

If the duty cycle of the output switching of U2 is greater than the input switching, U2 is being "lazy" and it's bad. However that really should be repeatable under all circumstances. U2 has no "knowledge" that settings are for Alternating or Simultaneous or how many events are set. 2 events @ 4,000 rpm is the same input switching frequency as 4 events @ 2,000 RPM. So if it has a problem responding quickly enough to the input frequency it would be repeatable at such an equivalent point.

However, MS operates fine with 1 event @ 6,500 RPM and 2 events @ 6,500 RPM, but fails with 4 events @ 1,000 RPM. If 4 events @ 1,000 RPM was hitting a failure point of a flaky U2, it would also fail 2 events @ 2,000 RPM and 1 event @ 4,000 RPM, but that doesn't happen. So I really don't believe U2 is the problem.

We know the input and output frequencies are going to be the same, just inverted. If the CPU was having a problem creating an output frequency over a certain amount it would fail to produce that signal regardless of how we get there in terms of RPM * Events. Because the failure doesn't actually correlate with a frequency I'm disinclined to blame the hardware.

Although I'm happy to check the hardware, there isn't any room left for a hardware cause as I see it. But this is not my specialty so I could very well be evaluating things wrong or missing something entirely and if so I'd love to know what the theory is. I admit I don't fully understand the active flyback clamp or the flyback PWM circuits on the "back side" of the injector drivers so perhaps something wrong inside there is manipulating the duty cycle to the injectors but that's just such a weird problem for hardware to cause. Phase-shift? Sure. Inverted signal? Sure. Signal clipping? Sure. But maintain phase, frequency, and manipulate only duty cycle? I can't come up with what would do that.

It's really fascinating and frustrating at the same time, but I'm leaning towards a bug.

[seijirou]

I have confirmed that a configuration of Alternating - 2 squirts does solve my unbalanced fueling problem. I think then I have a reasonable correlation that my psuedo-dutycycle measurement by means of an AC voltage measurement of the injector channels coming from the daughter card does indicate what will happen at the injector.

AFAIK regardless of simultaneous or alternating, and regardless of the number of squirts per cycle, when the general settings define single table the activity on both fuel channels should be identical. I can absolutely say that with some settings they are, but with some they are not.

[slow_hemi6]

This is mine on ms2 extra 3.3.1 with 4 squirts untimed injection. 36-1 wheel with injection over the top. 2 pulses per 360 degrees of the same size. You can see the missing tooth just after the first injection pulse.

[slow_hemi6]

[seijirou]

Is that 4 squirts simultaneous or alternating, and are you getting the same results for both channels? It looks like that's just one channel.

I didn't try the 3.3.1 firmware. I tried 3.4.2, 3.4.1, 3.4.0, and 3.3.3.

3.3.1 isn't available in the archive section that I can see unfortunately. If you can attach it here I can test it.

Attached is an MSQ that has the problem, and an MSQ which does not have the problem.

[slow_hemi6]

It is only a single channel and configured as 4sq Simultaneous. I was testing a fair while ago for a different issue but thought we might have had something in common. I now think we do indeed. Please turn off sequenced batch fire and redo your tests. I had nothing but trouble with sequenced batch fire and very little interest from anyone about getting it working. Feeel free to search for my post about it. Not sure how you would go if you went back to 3.3.1 as it is old serial and you have been on new serial firmwares. Still I doubt that would help.

[seijirou]

It is only a single channel and configured as 4sq Simultaneous. I was testing a fair while ago for a different issue but thought we might have had something in common. I now think we do indeed. Please turn off sequenced batch fire and redo your tests. I had nothing but trouble with sequenced batch fire and very little interest from anyone about getting it working. Feeel free to search for my post about it. Not sure how you would go if you went back to 3.3.1 as it is old serial and you have been on new serial firmwares. Still I doubt that would help.

I see. Are you running all your injectors on a single channel now?

I'll test with sequenced batch fire turned off.

[slow_hemi6]

No always 2 channels, 4sq and Sim. I usually run staged injection primaries 1ch and secondaries 1ch, That restricts you to Sim. It's a big 4bbl TBI unit. For a while now I have had staged off and just run 4sq Sim as my secondaries were not getting used enough when set up as staged.

[seijirou]

Gotcha. I don't believe I saw a problem with Simultaneous 4 squirts when I was testing but I'll test every option tomorrow with and without sequencing and write it down this time.

[slow_hemi6]

This was my issue. There is something not right with the sequenced and 4 squirts.
http://www.msextra.com/forums/viewtopic ... 90#p477890

[seijirou]

The issue is definitely related to batch fire sequencing. I ran through every combination of Simultaneous and Alternating with 1,2,4, or 8 squirts with sequenced on and sequenced off. Measuring and comparing AC voltage at U4 pin 7 and 5, the voltages matched for all settings except Alternating with 4 or 8 squirts with Sequenced On.

This of course does not tell us the timing of events in the way that you saw the problem, only the equivalence of the events of bank1 and bank2. So our problems are kind of similar but not exactly the same, but the common denominator is batch fire sequencing.

Here is a spreadsheet with the results of the test.

https://docs.google.com/spreadsheets/d/ ... sp=sharing

[slow_hemi6]

Given 2 different issues have been found with it, I think at the least, 4 and 8 squirts should not be allowed to be selected with sequenced batch and a config error raised. Fixing it would be better, but there seems to be no interest.