[Diy_efi] [Fwd: Crank trigger resolution]

[Phil Lamovie]

Phil Lamovie wrote:

> Hi Rod,
>
> A good discussion would be why there was a difference
> i suspect that the code in the Link may be to answer.
>
> > (at the moment I've got spark scatter you can see with a timing
> light
> > above 5000 rpm). If someone can tell me why I shouldn't bother
> > I'd like to learn why before I beat myself up over this :)
>
> OK good observation. possible causes.
>
> Inlet manifold resonance
> exhaust manifold res.
> turbo back pressure
> fuel pressure fluctuation is sympathy with regulator
> spring frequency
> onset of knock
> exhaust valve leak
> inlet   "     "     "
> Spark plug tip temp on edge of pre ignition
>
> The point of the list is to encourage you to get to the point
> where you know why it's happening.
>
> Otherwise all are equally at fault.
>
> If the spark scatter comes in abruptly at 5,000 rpm it's a
> resonance. If it comes in from say 3,500 and gradually gets worse
> that's more like time related.
>
> Does it still occur at 6000 rpm part throttle ? If it doesn't then
> what's the cause?
>
> Looks like we need some more DOE work to be done before
> we go any further.
>
> Remember that combustion is a random cascade of infinite
> variability. I have data from toyota (who do the most amazing
> research)
> showing the cycle to cycle variation on a std 4 valve 1600 cc under
> test.
>
> The 300 events had 2 that were negative power that is 2 combustions
> that cost energy and slowed the engine. ABSOLUTELY every variable
> was tracked and fixed. ......except "chaos" which didn't attend any of
>
> the planning meetings.
>
> > The Link runs on a M68HC711E9 at 12 MHz and has capture-compare
> > inputs capable of 333 ns resolution. Although interrupt latency
> > can mean the result is not immediately available, the actual
> > period measurement looks like it's not affected by latency
> > (did I miss something?).
>
> The 68hc11 is fine for the task. But the maximum executable loop time
> or MELT is unknown. So latency could be a factor. But what happens
> if it's a long division to calculate. does the ECU use the Last Known
> Good
> Advance or LKGA which has an error proportional to the rate of
> acceleration?
>
> Or does it wait until it's finished ?
>
> > The setting of (output) spark control will have a latency of 12
> > cycles, but this is constant (unless some kind of really bad
> > event happens ... reset/watchdog etc.) so is just an offset
> > rather than a jitter. Let's say there's error in the period
> > measurement at both ends, and the spark control, of 333 ns each,
> > giving a total of 1 us
>
> You say error in period measurement. But you need to quantify it.
> What is the resolution of the input capture timer ?
> Is it 16 bit or 24 bit ? How have they scaled the internal clock.
> Perhaps all the jitter is within the LSB that's discarded after the
> LDiv.
>
> Messy when analyzed isn't it ?
>
> > (how did you know this?).
>
> Short answer is I design engine management systems for a living.
>
> Long answer including source code also available.
>
> > To my mind,
> > at 6000 rpm (10 ms period), I have ~27us/degree, so I thought
> > the 1 us jitter was irrelevant. Sooooo, I'm thinking, TPU is not
> > the issue. Did I goof?
>
> You are not adding your errors like an engineer should.
>
> Each part of the event is variable. As the engine spins it's rpm
> is sinusoidal with two acceleration and two decelerations per
> revolution. So RPM is a fairly wildly inaccurate measurement to start
> with
> It's an RMS. of RPM that your seeing.
>
> It may be that the CPU and code are getting caught in the rise and
> fall
> periods because of the triggering pulses coinciding with a speed up
> and then a speed down event. Now who's going to take the millions
> of dollars of rework blame on that one ?
>
> > What do you think? Could I get away with 4 teeth?
> > Thanks for your input. I really appreciate it.
>
> YES. I am familiar with 68hc11 devices that run to 24,000 rpm
> with a maximum error of 1.6 degrees of crank when accelerating at
> 6,000
> rpm/sec. They use 4 teeth for a 4 cylinder.
>
> Try your timing light on some real cars and see what you get.
>
> HTH
>
> phil


_______________________________________________
Diy_efi mailing list
Diy_efi at diy-efi.org
http://www.diy-efi.org/mailman/listinfo/diy_efi