Technical stuff: Issues related to extra injection filtering

Post Reply
Posts: 4578
Joined: Wed Sep 29, 2010 6:01 pm
Location: Yorkshire

Technical stuff: Issues related to extra injection filtering

#1 Post by LPGC »

This is about the way ECU's deal with petrol extra injection.

Firstly for the benefit of those that don't already know what this is...

Some petrol ECU's in certain situations (such as acceleration) pulse petrol injectors for a very short duration between main injection pulses. These 'extra' pulses are so short that they only effect fuelling by a negligble amount but they can cause problems for an LPG installation (which I describe below). The majority of vehicles don't use extra injection, on those that do it may be used to help prevent petrol dropping out of atomisation (instead of a single long pulse from petrol injectors, break the pulse up by having a very small pulse between long pulses). LPG will not drop out of atomisation as it's natural state at normal temperatures is a gas. Extra injection uses the normal port injectors and shouldn't be confused with warm up strategies/functions etc.

Petrol injectors can open and close much quicker than LPG injectors because the moving parts inside only have to move a very short distance between fully closed and fully open positions (because they're injecting a medium pressure liquid), where LPG injectors are injecting a lower pressure gas so the moving parts have to move much further between closed and open positions to allow a similar mass of fuel through.. and while in positions between closed and open will be in a partially open partially flowing position. If we continuously cycle pulsing a petrol injector repeatedly between 1ms open position and 1ms closed position the petrol injector will probably be able to accurately follow this pulsing, so if we cycled it like this for 10ms (1ms on 1ms off) it would probably meter a very similar amount of fuel as it would with a single pulse of 5ms. If we tried the same with an LPG injector the results wouldn't be the same - The injector might stay open the whole time therefore injecting the full 10ms worth of gas, it might stay full closed therefore injecting no gas, or the moving parts might chaotically flutter between the closed and open states so might inject a different amount of gas each time we repeated the same 10ms long experiment (e.g. sometimes 7ms seconds worth, sometimes 3ms seconds worth).

If we were to cut short the above petrol injector experiment after only 5ms and then immediately give the petrol injector an open pulse 10ms long, then the petrol injector would have injected a total of 12.5ms worth of fuelling (because during the first 5 milliseconds it would have been open for only 2.5ms, and then it would be open constantly for another 10 milliseconds). If we did the same with the LPG injector, the chances are it would not inject a total of 12.5ms worth of fuelling - For starters, as described above it isn't very clear how much gas would be injected for the first 5ms section (could be anywhere between 0 and 5ms worth) but the final state of the injector following these 5ms can also effect the later 10ms pulse... If after 5ms the fluttering moving parts in the injector were moving towards the open position then this would give the moving parts a head start in the opening cycle for the later 10ms pulse, so the gas injector might give a little more fuelling than expected for the 10ms section of the pulse / or if the moving parts were already moving towards the closed position then the injector might be delayed in opening so might give a little less fuelling than expected for the 10ms section of the pulse.

The concepts in the above paragraph mean that if nothing is done about extra injection pulses, then regardless of whether a very short pulse between main pulses is reflected accurately by gas injectors in terms of fuelling, an LPG injector even attempting to respond to extra injection pulses can have negative effect on it's ability to respond accurately and inject the proper amount of gas according to the main pulse... Because following attempt to respond to extra injection pulse it may not respond normally to the main pulse or following response to the main pulse it may not be able to respond normally to the extra pulse. As said above, the reason for this is that the LPG injectors take longer to open and close.

So far I have only mentioned the technical problems arising from LPG injectors trying to copy extra injection pulses. Another problem is that LPG ECU's can be confused by extra injection pulses... e.g. instead of seeing main pulse and extra pulse as separate pulses they can see only one pulse which lasts the full duration from the start of one pulse to the end of the other (this example could lead to vastly over-fuelling), or could see only the extra pulse (vastly under-fuelling).

In either case (injector reason / ECU reason) it can be necessary to ignore extra injection pulses (or add extra injection time onto main pulse time) to arrive at a single pulse length to use for LPG injectors per cylinder intake cycle. Of course the LPG injectors cannot do that (the above text describes what can happen if LPG injectors are left to deal with extra injection), so the LPG ECU has to deal with extra injection pulses.

Now I'm getting to the main reasons for this thread..
The first thing I'll say is that filtering can bring about it's own problems due to the way ECUs achieve it. There can be a catch22 involved.. Below I listed a few ways in which extra injection filtering can be detected and filtered...

1. Delay opening an LPG injector until it's respective petrol injector pulse length has been longer than the maximum length of an extra injection pulse - the downside of this way is that the LPG injector follows the petrol injector pulses but with that delay, which can lead to problems if the delay means not all gas injected makes it past the engine inlet valve into the cylinder during this induction cycle (in which case this induction cycle will be lean, the fuel that doesn't get into the cylinder on this induction cycle will go in on the next cycle so the next induction cycle will be rich). A similar negative situation is possible on an Atkinson cycle engine, where the delay could mean gas is still being injected from the LPG injector when air and gas are being forced back out of the cylinder past the inlet valve back into the manifold.
2. This involves the LPG ECU beginning the pulse to open respective LPG injectors at exactly the same time petrol injector begin pulse but then noting the first instance of extra injection pulses and then switching to method 1, remaining in method 1 while-ever extra pulses are still occurring. This would mean that most of the time (with no extra injection) pulses would not be delayed, but then as with 1 while extra injection is detected fuelling would still be delayed. Also, extra injection would not be filtered in the first instance.
3. This would involve the LPG ECU noting usual injection order of the engine and by this would know if the start of any pulse were at expected moment for a main pulse or if this were an extra pulse, if an extra pulse then don't turn on the LPG injector. Downsides of this: It could be difficult to determine such correct moment on engines that switch fuelling strategies (e.g some Mazda which switch between sequential injection and group injection, but these engines bring certain other issues to get around anyway) / where RPM wire is not connected how would the LPG ECU if first few pulses following injector cutoff (e.g. from over-run) were main or extra pulses? Even with RPM wire connected, on engines with enough cylinders the difference between advanced/retarded ignition could be enough to have the ECU expect one cylinder to be having fuel injected when it sees another cylinder having fuel injected. Some extra pulses are closer to main pulse than others, some may be too close for 3 to be an accurate method of detection.
4. One I just thought of but a very simple one... Similar to method 3 but in this case the ECU doesn't need to anticipate the pulsing of a petrol injector relative to other petrol injectors pulsing, it just needs to anticipate the petrol injector in question pulsing (and to note how long ago the same injector finished pulsing).. If another pulse is seen to occur much earlier or later than anticipated it might be interpreted as an extra pulse.

The most likely way of avoiding the catch22 out of those listed might be 3 but then again this method might not be a catch-all regards filtering extra injection on all engines under all circumstances, for that 1 might be better, or 2 as a compromise...

Different manufacturers seem to use different ways of extra injection filtering but none of them seem to publish detailed explanations on how their ECU's do it.. Advice seems to be over-simplified to the extent 'with this enabled injection pulses below X milliseconds will be ignored, this can be necessary on some engines to prevent...' and just go on to explain the issues that can arise from extra injection (those that I described nearer top of page) with no explanation of methodology. Injecting gas at exactly the correct moment is critical on some engines, on other engines the way in which extra injection filtering is done is more critical, on others both are important, on others neither is of importance. We can see how the 3 methods above would each have different effects on timing of gas pulses in relation to petrol injection pulses. The effect of having delayed gas injector pulse can be similar to the effect of having injector pipes too long. Installers (at least those whom have thought about this) are left to read between the lines and draw their own conclusions on many technical points (extra injection filtering mentioned - a similar catch22 must be involved on ECU's capable of petrol addition), and then to apply that to anticipating which ECUs give the best results on which applications but never having their thoughts/insights confirmed by manufacturers. Regards extra injection filtering, the list of methods won't be a concise list (there can be different ways and also combinations of ways) but even in this list there is a methodology (3) which would need to work slightly differently (in terms of timing) depending on the number of engine cylinders. Using method 3 it would be possible that an ECU might work properly (including filtering of extra injection pulses) on an engine with a particular number of cylinders (let's say 8 because we might expect the manufacturer of an ECU capable of use on an 8 cylinder install to do most testing on 8 cylinder engines) but for the same ECU not to work properly on an engine with a different number of engine cylinders (let's say 6, because quite a lot of 8 cylinder ECUs are fitted on 6 cylinder engines).

In the above list of methods, 3 (and edit 4) are the only method that begins the pulse of LPG injectors at the same time as petrol injectors whilst still facilitating extra injection filtering on the first occurrence but this method would not be possible on ECUs that do not base individual LPG injector pulses on individual petrol injector pulses, so such (usually older) ECU's would need to use another method, perhaps method 1 or 2 or 4. Newer ECUs which are truly sequential may use method 3 but if a manufacturer previously made ECU's which were not truly sequential (so couldn't use method 3) then updated ECU design to truly sequential and at the same time updated extra injection filtering from method 1, 2 or 4 to method 3, then this would need to involve very different firmware, not only for the change to truly sequential but also for the change in extra injection filtering. Just because extra injection filtering, rpm interpretation, etc, worked OK on said manufacturer's old version of the ECU (on engines with any number of cylinders) this would not necessarily mean the same manufacturer's new ECU would work on engines with any number of cylinders. If the new ECU worked OK in all these respects on engines with a particular number of cylinders but not on engines with a different number of cylinders, we might expect something like case 3 to be used and for a problem to exist in firmware. I have noted problems related to extra injection filtering and rpm detection on some manufacturers newer ECUs when fitted on vehicles with a particular number of engine cylinders, where the same manufacturer's old version of the same ECU would have worked as expected. This is one of the reasons that led to me think up method 3, as this is the only method that would require a different approach / timing in firmware for engines with different numbers of cylinders... Otherwise I may not have thought any ECU advanced enough to use method 3, which just led me to think of method 4.

My daily driver is an old mid 90's Chrysler Grand Voyager V6 with a very basic EFI port petrol injection system, converted to LPG (obviously!) with a particular make LPG ECU (which I'm not going to mention here). When I fit the old version of the manufacturer's LPG ECU it doesn't seem to matter at all where I set controls regards extra injection filtering, the 'ball' in software which shows petrol injector duration never flicks to the top of screen (to the very low load part of the map) during acceleration and no adverse effects occur from any combination of extra injection filtering settings. When I fit the same manufacturer's new version of same ECU to the same car the ball on screen does flick to the top during acceleration and this LPG ECU seems to be very fussy regards how I set extra injection filtering controls to overcome the otherwise poor drive-ability (which will be due to the ECU reading incorrect petrol injector pulse durations during acceleration). On my car the RPM wire is connected (because I sometimes fit customer old ECU's on my car which require rpm connection to determine whether a customer has a bad ECU or their fault is due to some other issue). Not tried it yet but I dare say if I did not have RPM wire connected I might expect some strange RPM readings during acceleration with the new LPG ECU on my car, because I have seen all the same so many times on other 6 cylinder engines with the new ECU (including similarly basic engines such as mine, but this does seem to tend to be worse with more advanced engines). On an 8 cylinder engine install, I wouldn't expect any such issues with this ECU manufacturer's old or new ECU.. So I will continue to install what I rate as a great ECU for 8 cylinder engines on 8 cylinder engines and continue to avoid fitting this same make ECU on 6 cylinder engines.

Of course much of the above is speculation but something very similar to the concepts discussed in 1234 must be true to facilitate extra injection filtering, while the potential problems I mention related to timing in case 3 also seem to agree with results in real world applications (with certain ECUs).
I would like manufacturers to issue more advanced technical information (including on the subject of extra injection filtering) so that it might be of benefit to more advanced users, or at least be of interest. Advanced info needn't be as detailed as letting us see coding of firmware etc (which would be difficult to get to grips with for most of us), it could be in the form of flow charts which outline the purpose of individual ECU internal processes and when those processes are called upon by the ECU, such as if/when the various processes call each other and if/when hardware interrupts (perhaps created by an event such as a petrol injector opening/closing) call upon them. But they could start with a single character answer to the question on extra injection filtering such as replying with just 1,2 or 3 (referring to the methodolgies I listed above), or maybe suppliers could ask their manufacturures on our behalf and then explain to us how extra injection filtering is handled in case of each type of ECU, just 123 would do but a bit of a write up would be even better... I don't think anyone really knows this sort of stuff other than manufacturers, it is stuff that would be of interest (at least to me!) even if it were unlikely there'd be anything to pick up on from flow charts etc. I have already come to a few conclusions myself, I have a cheap oscilloscope but reckon would need a dual trace scope to compare petrol and LPG injector pulses at the same time to delve much deeper.. while those LPG software on screen graphs which supposedly can show comparisons between petrol and LPG injector pulses don't always seem to accurately reflect actual signal input/output. In fact while on subject, some of that advanced info from LPG ECU manufacturers could include dual trace scope images comparing LPG injectors pulses to petrol injectors pulses using different settings in extra injection filtering and on engines with various numbers of cylinders... I reckon if they did that they might learn something themselves and this could lead to them improving their ECUs.

Full time LPG installer
Servicing / Diagnostics / Repairs to all systems / DIY conversion kits supplied with thorough tech support
Mid Yorkshire
2 miles A1, 8 miles M62,

Post Reply