Forum

[tribonnie]

Hello, the system uses the MAP sensor to manage the injector timing. Without a cam sensor, at engine startup the first 2 engine revolutions are used to determine when the crankshaft is on the ignition stroke. With this data I doubt whether the ignition is a wasted spark but I don't know. Surely no one would use "wasted injection" these days, it is such a crude system with regards to fuel efficiency and emmision control. Let's not forget that the EFI engine management system is quite crude based on a small number of crank sensor pickups and limited number of sensors. To be fair it does not to be much more than this on such simple pushrod, low powered engines. Richard

[jefrs]

Suggest a simple method to dfetermine whether the ingjector squirts once or twice might be to monitor the injector pulse rate vs the crank rpm. It will either be at the same rate ot at half of it. I don't have the equipment for that nor is it something the Dynojet software (laptop) reports, which is a pity, although it does report Pulse_Width and Duty% as well as rpm etc.



It seems to me, considering the way it uses an O2 sensor that it should only inject the once. Where in the dydle that occurs is a matter of timing.

[Thack]

OK, Rattlebattle, I have some definitive data for you.



The fuel injection sub-system in my "black box flight recorder" now works like this:



1/ It monitors the ignition primary and waits until the moment it drops from 12V to 0V (i.e. when the coil is switched "on" to charge its magnetic field).



2/ At that point it starts sampling the fuel injector electrical feed. Each 100 microseconds (0.1ms) it samples the feed wire to see if it is 12V (injector not actuated) or 0V (injector actuated). (Both ignition and fuel injection drive pulses are active low.)



Every time it encounters an "active" signal on the fuel injector, it adds 100 microseconds to an internal "time" counter. Thus the time counter contains the cumulative "active" time on the injector, to a resolution of 0.1ms. Also, every time the injector changes from inactive to active, it records the start of a new injection pulse on a separate "pulses" counter.



3/ While doing this, it's also monitoring the ignition line, and as soon as the next ignition pulse commences it stops monitoring the fuel injection.



Thus it monitors the injection system for an entire four-stroke cycle (spark to spark) and then reports the result back for (eventual) graphical analysis.



These are the results:



1/ At idle and on the over-run it almost always uses 1 pulse per cycle (I mean per four-stroke cycle) with a total injection duration of about 3.7ms.



2/ Under more load it uses two or three pulses per cycle, three typically being associated with large throttle openings. The maximum injection duration I've measured so far is 16ms, but it's early days yet and I haven't given the bike a good thrashing.



In the next post I'll show you a screenshot and talk you through it.

[Thack]

You will see five parameters, with total fuel injection duration at the top, down to road speed at the bottom. On the middle graph you will see four letters: A, B, C and D. These represent points in the data to which I'll refer.



Point A: Idling at a junction. The FI system is using just one pulse per cycle, and the injection duration is sitting steady at about 3.7ms per cycle. The throttle is closed, the engine speed is about 800rpm, and the road speed is 0mph.



Point B: Observe the throttle position (orange) first. You will see three clear peaks as I accelerate through the gears, and beneath that the engine rpm rises to a little over 3000rpm in each gear. You can see the rate of speed increase getting shallower with each higher gear, as you'd expect. At the bottom you can see the road speed steadily rising to about 32mph.



Now look at the FI duration graph. As soon as I crack the throttle the FI duration jumps to about 14ms before dropping off a little (acceleration enrichment). Again you can see the peaks which align with the throttle movements. As you can see, my throttle hand wasn't completely steady (there was a LOT of traffic), and this is even clearer in the FI duration line.



Interestingly, at those peaks of injection duration, the number of injection pulses goes up to 3 per cycle, and back down to 1 per cycle each time I snap the throttle shut to change gear.



Point C: Travelling along at a vaguely constant speed, with a slight hill which I crest, and then gently begin to decelerate (towards the next traffic lights). Most of the time its injecting for around 7ms per cycle using two pulses. There is a little bump towards the end of section C which I can't explain, except that it's probably a minute throttle position change.



Point D: It looks like the traffic clears slightly because - still in fourth gear - I gently open the throttle a little which causes the small increase in rpm and road speed you can see in the bottom two graphs. The FI duration graph shows the injection period going up from about 2.5ms to about 7ms in response, and it briefly uses 3 pulses per cycle before reverting to 2 pulses.



There's masses more testing to do now, and it's too early to put anything more than 90% confidence in those results, but I thought you might be interested to see them anyway.

[Thack]

Oh, I forgot to mention that the engine does not use a wasted spark (except possibly at start up), so one cycle is always two revolutions of the engine.

[Boxerman]

Interesting! I'd certainly like to know more, when you have the data.

Thanks!
Frank

[jefrs]

Thanks Thack. Very interesting project you have going there!



The PCV does display but not record injector pulse duration as well as throttle position as a percentage with an option to display delta-throttle (rate of throttle change). It also displays ignition advance.



I shall check but I don't think the PCV injector pulse duration matches what you have reported.

I'm more than a little used to counting high speed pulses in the lab so I do understand what you're doing. I'll check my data and get back here with something. Which won't be tonight.



The question for me is - because the timing pulse is on the alternator running on the crank hence at engine speed, how does the ignition know which revolution of the two to give a spark on? What other input determines the compression/power stroke?

Is that the MAP? Is that only the MAP or is something else sending data?

[Thack]

Everything I have at my disposal (ie my data and my university lecturers) says it's the MAP.



My uni tutor said that such engines will often run wasted spark and double-pulsed injection for the first few strokes until the engine picks up and it can get a reliable enough "suck" signal from the MAP, at which point it reverts to the non-wasted spark and timed injection.



I want to do some more comprehensive testing to make sure I can fully believe the data I've recorded, but it definitely points to something more complicated than the traditional one-pulse-per-cycle.

[Thack]

On the subject of agreeing with the PCV: I'm pretty confident that my injector pulse durations are pretty close because I've measured them in three different ways (but only at idle) and they all agree.



It was because my measurements started to go loopy whenever I opened the throttle that I started to investigate further, and found these multiple pulses.



By the way, the PCV might be reporting the pulse time minus the injector dead time. An injector typically takes about a millisecond to open and maybe 0.25ms to close again, and the ECU takes these into account when driving the injector. In other words, the electrical pulse is deliberately a bit longer than the actual required "open time" of the injector in order to compensate for the dead time.



My system measures the electrical pulse duration. Maybe the PCV is reporting the "open time" of the injector (having factored out the dead time) as that is what is actually the most important when it comes to controlling the mixture.



For my purposes I don't need to worry about the dead time because it is (almost) constant and isn't relevant to my research.

[jefrs]

Thack in order to verify data it is considered normal for another lab to be able to reproduce the experiment, usually by other means. We have visitors approaching so wife has me cleaning the house, not been able to get down the shed, yet ...

Otoh I have retired and no longer have access to lab equipment.



On the subject of dead time I well remember listening on the phone for good 3 hours to a colleague at NPL explaining his PhD thesis on dead time (re event pulse counting), extending, non-extending, rise times, pulse integration, pulse shaping ad naus. Actually quite interesting but way too heavy for here. When printed a PhD thesis should be thick and heavy enough to hold the office door open, he sent me a nice door stop

This "dead time" applies to the counting system being 'down' whilst processing the last count when the next pulse arrives, and so is incapable of registering the next count; the lost counting time is usually added back on to the gross count time at the end, if one is able to calculate the 'dead time', which is not always possible hence imposing an artificial dead time that is known to be longer than the real one can be.