Ignition Jitters and Spark Scatter

    This one I am still looking at.  I have always noticed that the RPM reported by the control module seems to jump around quite a bit.  Other owners of this system that I have talked to have expressed the same concerns.  Wasn't sure if it was real noise or just the normal jitters of a digital display.  On most of the data logging runs I only sampled the RPM at 10 second intervals, but even at that low sample rate it looked like the timing periods was moving more than expected or explained.  Remember that RPM is just the derivative of the timing period.  I then ran a couple of runs collecting nothing but MAP ADC and the timing period (L0040 - the predicted next timing period) on the first and nothing but the timing period (L003e - current timing period processed) on the other.  The fastest that you can get data out at 9600 baud using the standard request/send/verify format from the ECU is a little over 52 responses per second.  Still way under sampled but a little better.  What was interesting looking at this data was its magnitude and also it clearly had some associated period.  This is more than just random noise.  There were places that reported an RPM change of close to 200 RPM pp, maybe 10 times within a second !  Well I guarantee that cruising steadily down the freeway with a manual transmission I would notice the motor changing 200 RPM !  Because the data is so under sampled you can't determine the actual period but it is obvious that there is one.  What all this means is that as the timing period changes between each cylinder firing the calculated fuel pulse width is going to be modulated also.  Probably not huge problem because the fuel burn will tend to get averaged out.  Of more concern is what all this jitter was doing to the spark.  With this magnitude of timing period change you can have 4 to 6 degrees of spark scatter.  To prove if what I was getting out of the ECU was real or not I ran another experiment.  I took two inductive timing lights and put each one on a cylinder 180 distributor degrees apart (like 1 and 6, or 4 and 7, etc.).  I tied them together and forced both triggers on.  If there was no timing difference between cylinders I should not see the timing marks move.  If the scatter was real the marks should move relative to each other.  Well you can guess what I found.  The two marks changed relative position to each other about 3 or 4 degrees.  They walked around a little at different RPMs also.  To prove that both the timing lights were firing at the same time accurately I would put both pickups on the same secondary lead.  Solid as a rock with no relative jitter no matter at what RPM.  Trying different pairs I would see slight differences in the amount of jitter.  There is no way of even knowing that the worse case is even 180 degrees apart.  It may be worse between other cylinders.  There was no easy way to delay one of the timing lights far enough  to try other combinations.  At this point in time it looks very real.  Where it is actually coming from is hard to say.  The distributor bushing were replaced when I setup the EFI.  I pulled the cap off and I can hardly detect any slop in the bearing at all.  Is it in the cam gears ?  I hope not.  The gear drive should be more stable than a chain.  Is it errors in the manufacturing of the shutter wheel ?  Maybe.  More likely in the relationship between the shutter paddle and the hall effect sensor.  If each paddle does not track exactly in the same path as all of the others then the ignition trigger point could change as each paddle crossed the permanent magnet at a different distance.  Could be a combination of all these things.  If it is a tracking issue it would take some precision forming to make each paddle track absolutely perfectly.
    At this point what I am going to do is loose the whole ignition trigger setup in the distributor and install a crank trigger.  You can't get more accurate than a crank trigger in determining where the piston is.  I haven't done this yet.  It will be a winter project.  Looking at different manufactures I see that Crane Cams has a kit that uses a hall effect trigger as the sensor.  This is advertised as accurate between 0 and 12000 RPM.  Static timing adjustment with this unit is appealing.  Also the hall effect trigger output will be easy to interface into the existing hardware.  I'll make another shutter wheel with just one shutter paddle on it for the distributor.  It will be used for nothing more than signaling when the number one firing event occurs.  This is needed to determine the injector phasing.  Another issue is the relationship of the rotor to the cap.  When I had the cap off recently I could see tracks melted into the cap because the rotor is not phase correctly to the cap.  Maybe this is associated with the other errors in the distributor modifications and installation.  Hopefully I can now optimize the relationship between rotor and cap rather than taking whatever I get when the hall effect sensor triggers.

    This plot is a segment from the second run where I collected the current timing period from L003e.  The L003e stash is the value of the timing period that has just occurred.  This segment was cruising down the freeway.  You can clearly see how noisy the data is.  The below graphs expand on this one.

    This plot is a 5 second period cruising around 2400 RPM.  You can easily see the periodic nature of the data.  This is more than random jitter.  Because it is so under sampled it is not possible to determine the exact frequency.  Being the data is aliased its magnitude is probably under stated also.  No way is the motor changing this kind of RPM at these speeds.

    This timing period plot is during the same time period as the above RPM plot.  The 'Y' units is the timer value (L003e) in 8 us units.  So if we pick a big peak-to-peak delta of 47 counts this would translate into 376 us of jitter.  So at 2400 RPM we are turning 40 revolutions a second, which is 14,400 degrees a second.   376 us of time jitter works out to about 5.4 degrees of spark scatter.  Not good for a performance oriented hobby....  This will have to be fixed.


10/20/99 -
    Ordered the crank trigger kit this week.  Decided on using the kit from Crane Cams.  This unit uses a Hall Effect sensor as the pickup element.  This will give me a sharp edge with go signal to noise ratio all the way to the back of the car.  It will easily to interface these levels into the existing hardware.  The other crank trigger units I looked at (MSD, Moroso, Holly) all use a low level magnetic pickup.  It would be more difficult to interface these signals.  Some kind of amp would be needed to raise the levels to something the ECM could use.  Plus it would be more risky running these low levels to the back of the car where the ECM is located.  Another benefit of using a Hall Effect sensor is you can time the motor at zero RPM.
Note (12/01/99): After waiting a month and a half to get the Crane kit they finally informed me it will probably be two to three more months, maybe.  Well, no way guys.  What I decided to do (based on what was available now) is get the Crane Hall Effect sensor and a Moroso crank trigger kit (wheel and mounting hardware).

11/03/99 -
    I plan to modify the existing shutter wheel in the distributor to make it the #1 detector.  I will need to run one more lead into the ECM.  Looking at the wiring harness connector there are no spare contacts in it.  The only contacts in the socket are one currently connected to a wire.  Went to a local pick-n-pull looking for sockets / connectors I could use.  I had read (sometimes between the lines) that this 35-pin connector is used on Fiats, Peugeot, and a early 90's Ford Escort Crossworth.  Well of coarse I couldn't find any of these types in the whole darn yard !  I found one socket from an Audi that looked like it had the same type of contacts in it.  A little more poking around and I scored.  '78-83 BMW's have the exact same AMP socket and hood as out Pro-Flo does.  Most of the ECMs were gone so I am not absolutely sure as to their brand.  One of the Beamers still had the computer in it, it was a Bosch.  Well I hacked every connector I could find and for $20 I came out of the with enough sockets to build five more cars.

12/06/99 -
    After modifying the ECM code to send the timing period (L003E) every ignition trigger event I was able to make a couple more runs.  I wanted to get real time timing data before installing the crank trigger.  It takes about 1 ms to get a byte out the serial port so I was able to get real time information up to about 4000 RPM.  More than fast enough to see what is happening.  The below plot is quite busy, about 15 minutes of run time.  The time legend below it probably won't help you much but it keeps it straight for me.
 
   Xgraph
   Xgraph - markers

Legend:
A - Initial start.
B - Data start school Stop sign.
C - Shopping center signal.
D - Stop sign at vets.
E - Onto freeway.
F - 65 MPH 5th gear.
G - Off freeway.  Through several signals to get back on south bound.
H - Off freeway.  Through several signals to get back on south bound.
I - 65 MPH 4th gear.
J - 65 MPH 5th gear.
K - Off the freeway to gas station.
L - Restart at gas station
M - Back home.
N - Couple hot re-starts

    Below is a couple of expanded plots.  There is no way that the motor could possibly accelerate as fast as the timing numbers suggest.  Can't really determine a period here even though the noise has some rhythm to it.  Well when I get the crank trigger installed we shall see......

12/12/99 -
    This weekend I got the crank trigger wheel mounted and centered.  I had to mount this pup on the outside of the crank pulleys because I didn't have the room to move the other pulleys outward to keep the belts aligned (I didn't want to either).  This meant that there was nothing to align the trigger wheel to.  Had to set up the dial indicator to adjust the centering of the wheel.  It wasn't too hard to get it centered.  The worst part was finding enough room to mount the dial indicator.  Had another scare also.  For a time it looked like the Hall Effect sensor and the lower radiator hose were going to occupy the same space at the same time.  After a little tweaking it looks like it should clear.  Maybe a 1/16" to 3/32".  Miles of clearance compared to other things on this car.  If the wheel was mounted where it was supposed to on the damper there would have been no problem at all.  I had to space the mounting hardware out 2" to get the trigger centered over the wheel.  Also had to degree in the flywheel.  There is no way to turn the motor over from the top side of the car anymore.

12/18/99 -
    Got the modifications done to the ECM this week.  I had to change the input topology of the old distributor input conditioner circuit.  The output of the hall effect sensor is not a well behaved square wave as it should have been.  This unit has two outputs, one they call the 'Point' is the square wave and the other is a output that mimic a magnetic trigger.  I think what is going on internally is a hall sensor driving an op-amp, which is the square wave output, and an AC coupled output used for magnetic trigger installations.  Their output amplifier does a fair job of sourcing current to about 50% of the supply voltage.  From there it really slows down (takes 2.5ms to get to the rail !).  Obviously I want to detect the trigger on the fastest portion of the waveform. The original not used magnetic trigger input circuit off of J1-4 I changed to sense the CAM position signal coming from the distributor.  Basically modified it to look like the original distributor trigger input circuit.  Both circuits contained a 4.7k resistor between the output of the LM2904D op-amp and the input of the 74HC14 schmitt trigger input.  Most operation amplifiers do not like to drive a capacitive load.  You end up with a phase shift which cause the op-amp to start oscillating.  A common way to prevent this is to install a small resistance between the output and the capacitor (in this case the input of the 74HC14).  This resistor is generally in the hundred(s) ohm range.  Why they used a 4.7k is unclear.  What this excessively high value does is real slow down the edge which adds some more uncertainty to when the inverter senses a change of state.  I reduced this value to 562 ohms which still gives me the needed phase margin and speeds up the edge.

12/28/99 -
    Well got all the hardware installed and the firmware done today and fired it up for the first time with the crank trigger.  Worked first time !  Timing was easy.  My initial guess was off by 1 degree.  Under the car and move the pickup one tic mark and the timing was dead nuts on.  Remember how much jitter there was when the 'Base Timing' was on trying to time it at 10 deg BTDC ?  Well it is all gone.  All I had to do was increase the speed maybe 50 to 100 rpm to smooth the cam a little and it was as solid as a rock.  Changing the RPM further made no difference at all anymore.  All that jumping around before was the direct result of the ECM mis-computing when the next firing event was supposed to happen.  It was really obvious while I was checking to make sure that my new shutter wheel in the distributor had enough leading/trailing guard band, how much jitter is introduced into the system by the original distributor pickup.  The timing jitter between the leading edge of the distributor pickup and the crank trigger's was quite dramatic.  I wish you could see the noise, you'd run out and install a real crank sensor.  I haven't taken it out on the road yet put did take some timing period data while its still on the jack stands.  As you can see by the graphs below there is a remarkable difference.  Where before I was getting variations in the hundreds of RPM now it is closer to two to five RPM.  Well worth the effort.
 
 
  Xgraph
  Xgraph - markers
Look how much smoother the RPM and timing period data is !
5 to 10 RPM worth of jitter only !
Look at the units, one count per !

The below link will take you to some wave forms taken in the car to verify performance.  One of them you can clearly see the jitter and noise between then distributor pickup and the crank trigger's edges.  Also some observations on the original Edelbrock shutter wheel and Rotor-to-Cap relationships.