More Data

[grippo]

I tested your msq on the bench last night and couldn't find any problems. I used a 60 tooth wheel simulator and it ran the rpm up and down from 0 to 6000 and back over and over and there were no missed ignition outputs. This was at a really fast rate, probably a second or 2 to go 0-6000. Also I didn't see any point from 0 to 6000 rpm at which the dwell started to approach anywhere near 100% duty cycle. It might have been 70% at most.

I am meeting Bruce tomorrow and we will start sequencer testing on his truck, although we don't have all the pieces to connect fuel and ignition. But once they come in he should be able to fire it up right away.

The only thing I did see was that you had trigger offset set to +12 deg. This number should really be 0 or negative, same as for microsquirt. You can compensate by more or less delay teeth. But it didn't cause any problems that I saw.

[TheMonkey]

...
The only thing I did see was that you had trigger offset set to +12 deg. This number should really be 0 or negative, same as for microsquirt. You can compensate by more or less delay teeth. But it didn't cause any problems that I saw.

this was sort of a legacy habit from MSII. in MSII, i had cranking set to 'trigger rise' instead of calculated, so that my advance during cranking would equal the offset; so i adjusted delay teeth to a point that let me get a good cranking advance. but in sequencer, i have cranking set to 'calculated' as instructed. i'll set offset to zero or negative and adjust teeth as you suggest since i'm using calculated.

As of late May, Marc said he cleaned up noise and was running ignition only. when he had noise, do you know if it was similar to this?

[TheMonkey]

Perhaps I spoke too soon that the sequencer was fueling okay, and that the problems were isolated to ignition.

the only way to have found this was to pore over fuel pulses and compare to the VR trace. remember, i am running without a cam sync in wasted mode, so it should get 1 fuel pulse to each injector for each revolution.

i noticed that when i would rev and listen to exhaust and look at o2, there was just the tiniest little miss every 1 or 2 seconds. on the o2 it was a super tiny little rich puff or lean burp, so i wanted to see the signals to injectors up close.

it seems that the cylinders sort of 'phase shift'. i probed one of the fuel injectors, and it should have had the injection occur in the same spot of the VR cycle every rotation. car is 6 cylinder, and 2 cylinders get fuel at same time at 3 different points on each rotation. PROBLEM: sometimes, it would move over to one of the other spots and hang out there for awhile and kind of bump around. I would have thought that it would correct itself the next time the code saw the missing teeth in the wheel. the little miss i heard was that during a phase shift' the timing of the valve opening would make it so that sometimes it would pick up an extra pulse in a cyle (3 injections instead of 2), or lose one (only 1 injection instead of 2).

this picture is one example of the fuel pulse shifting. the shift is always exactly into one of the 3 spots that a pulse belongs. if anyone wants to see the data file for hobbylab scope, i'll post it, it's full of these 'phase shifts':



it does the same thing with ignition, but couldn't see it before because i was looking only at IGN outs at one time. i compared only 1 ign out vs. the VR output and confirmed that they are phase shifting too.

i haven't tested to see if this happens on the stim. Al- perhaps your stim test would have revealed this if looked at with a scope (datalog does not give any hint of a problem).

i made a conversion harness to plug old MS2 box into sequencer amp seals, installed a flyback circuit, and configured an MSQ for this car; MS2 works great on this car. i did this just because i never had an MS on this car before beta testing sequencer so I wanted to make sure everything was good with the signals going in. there does seem to be some noise spikes on the voltage, clt, afr, and others.... which did not show up at all on sequencer.

[grippo]

When I test on the bench I always do it exactly as you did, one probe on the VR input and one on an injector or ignition output. I just don't see this shift. I also have a set of leds for injectors and another for ignition and I can look at the order of the pulses. At low rpm it is easy to spot anything out of order. Also, the ordering of the injection and ignition is totally independent of each other in the code, so it is strange that there would be a code problem in both cases.

Unfortunately Bruce didn't get all the parts he needed, so he still hasn't been able to test his engine.

[Bruce Bowling]

OK, I still do not have the engine tested yet on my end, but on this phase shifting issue above.. when this happens do you have any missing tach pulses or such? Also, I think this is a 6-cylinder engine, am I correct? Do you have the two reset lines b/w the processors tied together?

Reason I ask is that if for some reason there was a skip b/w the two processors the routing could be off. Remember there are two outputs that then are routed to 6 outputs, so if there was a skip b/w the two processors things could shift. Only way this could happen is if there was a reset on one and not the other - the resets tied together correct this.

We may need you to scope the intermediate signals between the two processors, stay tuned...

- Bruce

[TheMonkey]

hey guys... thanks for the replies. sorry, i missed them at first, my email did not pick it up.

i was able replicate this on the stim.

my MSQ changed a bit since the last one Al had. i changed the offset to a negative number, and adjusted the delay teeth as he suggested. i'll attach the new MSQ. Yes, 6 cylinder.

this is on a separate Sequencer unit. this happens on my car with one unit, and and on the stim with a separate unit, same MSQ. Bruce- if i understand your question about tying together reset lines, you mean bootloaders? YES, both bootloaders are tied to together to VREF on both units.

here is a pic that i just picked up off the stim (the very first ign signal is the correct one). VR signal against IGN #4; it shows both the phase shift, and the start of the flanking. it seems that the flanking occurs as soon as it starts to reduce dwell to avoid overlap. there are some missing dwells too. pic from stim session:



generally, i record on scope 30 seconds worth, rev up and down a few times, the scroll through and i can see all the missteps. none of them show up on a datalog as any kind of miss.

[TheMonkey]

i increased my 'next pulse tolerance' from 15% to 35%, and that seems to fix it on the stim. i cannot recreate the phase shift. so, does this mean that the code will shift a cylinder if a signal falls outside of tolerance? if i recall correctly, i had to tighten up that variable to 15% to get rid of some tach dropouts, or some trigger +/- on datalog.

phase shift on fuel is not a big deal, but could really cause a problem with ignition. couple instances scrolling through the log, i could see where the ignition output would creep up in timing a little bit more every cycle, and then would phase shift after it became extreme.

[grippo]

The pulse tolerance logic has only 2 possible effects - either the timing of a tooth that has just arrived is within tolerance of the last tooth or it isn't. If it isn't, it will add or throw out a tooth. But if two missed or happens more than one time (2 or more in a row), then it will set resynch(rpm=0) and it starts the synch process over again. Either way this has to show up in the trigger count. And there is the master check of the missing tooth. This has to occur at the right tooth or it will also cause a resynch(rpm=0). What I am getting at is that if you throw out a tooth and then you add in a tooth, and they cancel out before the missing tooth and there are never 2 missed or 2 extra teeth in a row, then so what - you will see one tooth (6 deg) worth of jitter. But that isn't what you are seeing.

Is the phase shift correlated with rpm accel/ decel activity on the car ? on the stim ? That is, if you let it run at steady rpm do you ever see the phase shifts ?

I will check this on the bench tonight - finding a correlation with Pulse Tolerance was a good find - thanks.

[TheMonkey]

Al-

no, i can't seem to find any kind of RPM correlation for the phase shift. for the ignition signal 'flanking', that occurs at overlap - but not everytime - i think it only does it when phase is shifted. (edited to add: it still flanks even if in proper position).

on stim, i could not get problem to happen in any V8 mode, or in any 36-1 config. only 6 cyl, 60-2.

on the stim, i had an interesting phenom happen.... i was changing the ignition settings to a different trigger style, and changed the toggle on jimstim accordingly. it found itself in a steady-state of shifting 1 position every ignition event. somehow, it satisfied some kind of condition for a consistent phase shift every time. this was at about 1500 revs. i have it on scope recording, and it was a short recording, so i was also able to save it as a txt file. i cycled power off-on and it was back in sync.

on the car, i swapped running next tolerance to 35%. car seems to drive the same, although the occasional lean/rich puffs are still happening. when i get a minute, i'll scope and see if it is still there on scope. i suspect it is.

[TheMonkey]

yes, it persists on the car with running pulse tolerance expanded to 35%. the phase shift happens pretty consistently, yet without any indication of what causes it. at any time during idle, not even touching the accelerator. i did a very simple datalog of about 15 seconds. datalog looks perfect, you can see in the picture that there are zero resets, and trigger +/- sits at zero the whole time. the time stamps on the scope pretty much start half second before the datalog starts. you can see that i let it idle for several seconds, then wind it up to 3,800 revs to get the flanking. scoped IGN #4 and VR signal.

when the flanking happens, the sum of the duration of the flanked noise, and the properly timed dwell sums up to the total time required if the dwell is not pared down. for instance, my dwell time at idle is about 5.5ms. greater than ~2300 revs, it starts to get pared down. if the size of the dwell should be pared down to, say, 3.5ms, then there will be a flanking noise of 2.0ms that overlaps with another IGN out.

scrolling through scope data, here are the phase shifts i found. the convention i'll use to label will be P+0 = proper Position for dwell, P+1 = late by one cylinder spot, and P+2 = late by 2 cylinder spots:

0.0 -> 5.1s = is proper dwelling in P+0
5.10s = shifts to P+1 at 845 RPM (1.33 revs later)
5.1s -> 9.2s = dwelling in P+1 position
9.2s = shifts to P+2 with 2 dwells immediately next to eachother (1/3 rev later) at 992 RPM
9.2s -> 10.4s = dwelling in position P+2
10.4s = shifts to proper P+0 at 1143 RPM (1.33 revs later)
10.4s -> 11.7s = dwells proper in P+0
11.7s = begin of flanking while proper dwelling at 2308 RPM
11.7s -> 13.0s = proper dwell location, with flanking through peak of RPM
13.0s = end of flanking at 2400 RPM
13.0s -> end of file = dwell remains proper in P+0

datalog (also attached):

yellow = RPM
light blue = MAP

red = PW
white = TPS
green = Engine (cranking, ase, wue, accel, wue, accel, decel, wue)

blue = SecL
yellow = tachCount
white = trigger+/-