Crank reference angle vs. timing errors

[socrace]

Some time ago, GM published a chart for the (then new) LT1, supposedly showing the effects of using various crank references per revolution on spark timing error. There's a copy (5th chart down, click to enlarge) at

http://www.gmhightechperformance.com/tech/0310htp_optispark/

This chart shows an error of about 21 deg at 1000 rpm using 4 crank references, assuming 10rpm/ms crankshaft acceleration. When I try to crank the numbers using this data, the error I get is only about 6 deg, assuming acceleration starts the instant the previous period ends, and the error is the result of predicting the acceleration period based on the previous measured constant period.
The only way I can get close to GM's numbers is to assume a crank acceleration of about 100rpm/ms. Is 10rpm/ms or 100rpm/ms a more reasonable max acceleration assumption for an OEM tuned sbc? Or is my math wrong somewhere?
TIA
Bob D

 

[GregLocock]

That's atricky one.

A typical level of torsional vibration measured at the crank nose is 0.25 degrees peak to peak - how does that tie up with your estimates? Why is that not the same as the predicted timing error?

Cheers

Greg Locock

 

[GregLocock]

OK, I've actually read the article now, it is a lot more informative than most of the drivel on the web.

Just thinking about TV levels, at low speed they do tend to rocket up, maybe as high as 3 deg pk-pk, trying to remember a graph from 10 years ago.

So, at 1000 rpm the interval between cylinders firing is 15 ms, so at an acceleration of 10 rpm/ms, or 60000 deg/s that is an error of 1/2*60000*.015^2, 6.75 degrees.

The error is proportional to the acceleration, so you only need to increase it to 15 rpm/ms to get that 10 degree error.







Cheers

Greg Locock

 

[socrace]

Spark angle is calc by the ecu as time (rather than degrees) from a reference point. So if last 90 deg reference period was 15ms, then a 45 degree spark advance would fire 7.5ms in the future. But under constant acceleration, 45 deg is no longer 7.5ms but something shorter, so spark firing will be retarded. GM says this error could be 21 deg, which is huge.

 

[magnograil]

What they are not telling you is where the advance is relative to the reference mark. 1000 RPM = 60ms/rev so 4x reference will give a signal every 15ms. If the acceleration is 10 RPM/ms, then by the next reference mark the engine will be at 1150 RPM, not 1000. At 1150 RPM the next reference is 2ms sooner than the expected 15ms. If the ECU calculated 1000 RPM from the previous two marks, then it will actually fire 6 degrees late as you stated. If they used the two references previous to the last to calculate RPM then they will be off even more. Since they have to calculate RPM then interpolate the tables to calculate the retard angle and then count down a timer for the ignition point, using the 2nd and 3rd previous marks for RPM might be the possible difference. Depends upon how fast their ECU is.
They could keep track of the change in time between reference marks to infer acceleration but that is more computation and cost.

 

[PFM]

OK I follow the logic trail on this just fine, dare we assume ( I do hate to use that word) that this would affect any delay based ignition / EMS system that uses few pulses per engine revolution? I called MSD today and got no solid answer. If you want a real scare, I helped configure one of these "programable ignitions" for a drag race application that had much more timing at 6000 RPM then at 8000 RPM so the timing calculation error would cause the ignition to be too advanced, not retarded. One drops power the other contributes to detonation and engine damage. I see no software solution (without a TPS or learning algorithm to "predict" the RPM rate) that could be used to prevent this. Any ideas on products other than GM????

Thanks,

PFM

 

[magnograil]

You are right, retarding as RPM increases can cause the error to be advanced depending upon the acceleration and retard rate.
FWIW, I am building my own programmable ignition. It also uses optical pickups but unlike GM, mine uses a window which defines the maximum and minimum range of the firing. So although acceleration can cause a delay of the intended timing, the delay is from the a known position closer to the firing point. I am doing away with a distributor completely and using sequentially fired coils. It would be easy enough to keep track of the previous two RPM calculations to get acceleration and correct the RPM before table lookup and interpolation.
Resolution is to 100's RPM and 0.1 degree firing. RPM is stored in 8 bit data giving 25,600 RPM limit. Reducing the limit would give finer RPM resolution. For example, 10,000RPM/256 = 39 degree increments in RPM data.

 

[Azmio]

When it comes to crank speed sensor error, does the number that GM published takes into account factors like: -
1) Manufacturing tolerance for the parts
2) Assembly tolerance
3) Location of the 'trigger wheel' or teeth. If we locate it in front, the middle and at the back of the crankshaft. The number will be slightly different. If the teeth are located at the flywheel, the number will be different too and it is also influenced by the clutch engagement and torque converter dynamic movement.
4) Firing order of the engine
5) Number of cylinders and cylinder configuration (V or inline)
6) Processing speed of the ECU
7) Crank material because the forged, machined and cast cranks also have different material strength and elasticity
8) design and construction of the trigger wheel

 

[socrace]

Thanks to everyone who responded to this.
Unfortunately, it does look like GM's error figures were based on actual measurements of their early hardware, rather than "first principles". At least their error numbers are greater than expected, probably due *mainly* to the latency in their firmware. This looks likely because a "skipped period" error is larger at lower rpm.
To summarize what the errors are, acceleration causes timing to be retarted and dwell to be reduced, using 90 deg crank references the error can exceed 20 deg at 1,000 RPM.