Below I pasted a post I just did on this subject on Legacy forum. Can't see the pics due to copying and pasting and the way the other forum works with pics but I'll sort the pics out later. Meanwhile the post on Legacy forum can be seen here http://www.uklegacy.com/forums/index.php/topic/143724-jdm-spec-b-lpg-help/page-6
Haven't had much time to do a write-up since I had Jamie's car, due to having all my usual work to do plus messing with the gearbox on my own car (seems the autobox pump has gone or there's some major internal leak), plus helping my son change the front struts and hubs on his Mondeo and having to repeat because the CV joints pulled apart during the work.
I had Jamie's car for a few days, didn't manage to fix the problem but did identify the underlying issues and am confident I could fix it if we were to change the LPG ECU, possibly also injectors.
There was also a shortfall with the physical install that I did put right - The installer had fitted 'spuds' (gas entry points on the inlet manifold) in too high a position, at least a couple of inches above the petrol injector position. For accurate fuelling and good throttle response, particularly at low rpm, it is necessary to fit spuds as near as possible to the petrol injectors and to have short pipe runs between LPG injectors and spuds. I blanked the existing spud positions and fitted a different type of spud a couple of inches lower down on each manifold port runner right beside each petrol injector and doing so also allowed some of the pipes between LPG injectors and spuds to be shortened, one pipe in-particular to be shortened drastically. Most spuds are just a hollow fitting that screws into a drilled/tapped hole in the manifold, the different type of spuds are similar but the very end of the spud where the gas would usually exit is blanked off, gas instead comes out of the side of the end of the spud and this allows gas to be aimed towards the engine's inlet valves. I regard the shortfall with spud location as a very basic underlying issue, spuds should have been fitted much lower at the time of installation... But failing that, any issue such as hesitation during acceleration at low RPM that Jamie described should have prompted installers to move the position of the spuds to the lower position because such issues are classic symptoms of installing spuds too far from petrol injectors.
However, the main problem with Jamie's setup is that the engine momentarily goes too rich on sudden acceleration at low rpm. It seems the last person to adjust settings on Jamie's car realised this and attempted to compensate/lean fuelling under these conditions by adjusting 'tip-in' settings in the software to a largely negative value. Tip-in is a BRC software handle for compensating fuelling during acceleration, at all times the LPG ECU refers to it's basic calibration 'map' (the map itself can be adjusted by the installer) so tip-in is a fiddle factor that sees the LPG ECU combine the basic map and tip in settings to arrive at pulse time for the LPG injector during quickly increasing engine load... This implies that tip-in can be used as a fudge factor to momentarily correct mixture during acceleration even if the basic underlying map is incorrect... Emphasis here on that last point, because it does seems to me that the basic underlying map is incorrect, I found I could improve that hesitation during acceleration at low rpm by leaning the underlying map, however Jamie's ECU does not allow good control over the underlying map, the control you do have is just a few sliders that can be adjusted and these sliders have global effect on the map. There is a slider for idle mixture control which effects low load/rpm mixture and there is a slider that effects mixture under none low load/low rpm conditions. Some of the various problems with this, particularly in Jamie's case, are that the issues occur at the transition point from one slider to the other, so for starters you've got to adjust both main mixture sliders to overcome the hesitation during acceleration at low rpm issue. Then, considering the sliders have global effect on mixture, if you adjust the sliders in attempt to correct the hesitation you cannot avoid effecting mixture during other driving conditions that you wouldn't want to effect (because during other conditions mixture appears to be correct).. at best this would result in fuel trims being inconsistent (if they were consistent before) and at worst you can end up with MIL light issues due to fuel trims out of range, too lean a mixture during WOT conditions not good for the engine.
Observations and reading between the lines - 1. Like some other BRC systems this system is calibrated using a manifold pressure sensor that is only attached during the calibration stage, during which time it learns to anticipate what manifold pressure will be during all the varying driving conditions (after the map sensor is removed). Like most LPG systems it needs to know what manifold pressure is in order to know the difference between gas pressure and manifold pressure (like other BRC systems this system calls difference between gas pressure and manifold pressure delta pressure). Gas pressure on all vapour type sequential LPG systems (including this system) is directly measured, most systems directly measure manifold pressure too (above I described how this system doesn't directly measure it but anticipates it). Most LPG systems, including this system, need to know delta pressure because for any given LPG injector pulse length that injector will flow more gas if delta pressure is higher, so systems need to compensate LPG injector pulse length for delta pressure, even for slight/momentary fluctuations to delta pressure. The potential problem here is that if the system does not correctly anticipate manifold pressure it cannot properly compensate LPG injector pulse length for delta pressure. Now suppose a system that has learned to anticipate manifold pressure seems to have generally done a good job of learning, this is verifiable because if we trust the reducer to keep manifold pressure constant above manifold pressure we can check this by just watching delta pressure, if delta pressure behaves like we expect during various driving conditions (i.e. remains the same) then it is likely that manifold pressure has been learned properly... However, this may still leave question marks about accuracy of anticipation of manifold pressure during quickly changing manifold pressure and in fact monitoring pressures (directly read, ECU anticipated and calculated delta) in the scope facility built into the LPG software does seem to reveal some pressures that I wouldn't expect during quickly changing pressure situations. 2. The system seems to be designed to be very plug n play but it doesn't seem to offer the type of control over fuelling that is possible using other systems (particularly some systems). Seems to learn the underlying map during it's 'autocal' stage and then present the installer with only the few slider handles to effect control over the underlying map that I mentioned in the paragraph above. Any system that relies on autocal to get the basic shape of an underlying map correct will struggle on certain vehicles, most particularly on vehicles that have very short petrol injector pulse lengths, mostly because LPG injectors take far longer to open and close than petrol injectors.. Most engines idle with pinj (petrol injector pulse length) of around 3ms (milliseconds) at idle with a warmed engine, if an LPG injector takes 1ms longer to open and flow fuel than a petrol injector then if the LPG injector (at it's provided delta pressure) flows the same as the petrol injector while open then for same flow as the 3ms pinj, ginj would need to be 4ms, so the underlying map at the 3ms pinj point would need to multiply the 3 by 4/3 = 1.33. For this same average engine with same spec injectors and same delta pressure, for 10ms pinj, ginj would only need to be 11ms which would mean a multiplier at the 10ms point of only 1.1... From this we can imagine a graph of the necessary multiplier going all the way from say 2ms (engine very light load on over-run) to around 16ms (engine torque peak). The graph would be curved, it would start at a low multiplier figure for very low pinj's (because the LPG injectors, once starting to open, also take time to close), would peak somewhere maybe at around 3.5ms, then would fall off at a diminishing rate, so we end up with a multiplier graph that looks a bit like a stretched out sine wave. This multiplier graph is what I am terming in this post as the shape of the underlying LPG map.. Installers normally have direct control over the shape of this map but do not have the same control over the shape of this map on this system, instead we just have the sliders which in simple terms lower or raise entire sections of the map. So suppose Jamie's engine idles at 1.8ms pinj, so if we use the same 1ms slower LPG injector spec that we used in the example for normal engine above, we might expect Jamie's ginj to need to be 2.8ms and multiplier of 2.8/1.8 = 1.56 but the multiplier would still need to be 11/10 = 1.1 at the 10ms pinj point of the map and 17/16 = 1.06 at torque peak. On overrun Jamie's engine might go down to pinj of 1.4ms and due to open/close time of LPG injectors at 1.4ms multiplier may need to be 2 or may need to be less than 1, regardless of which the shape of the map will be very different to what it would be on a normal engine. It may be that the autocal would get the basic shape of the map just about right on a normal engine but get it very wrong on Jamie's engine. Without getting too technical, the point to draw/conclude from this is that engines which use very short pinj's are much more demanding in terms of LPG components spec, much more demanding to calibrate properly (sliders to control fuelling not enough control, especially if autocal anticipated incorrect shape of the underlying map), some injectors (those that are not quick enough to open and close) would have little chance of doing a decent job at all because they would make it necessary to compromise drive-ability on LPG at low loads at one extreme against ability to run at high engine loads on LPG at the other extreme (consider what shape of map would have to be if LPG injectors took 3ms to open/close and couldn't accurately meter fuel below 3ms pulse time.. 4.4/1.4 at idle = multiplier of 3.14, 19/16 at full load = 1.19 but this would restrict ability to provide rich enough mixture at high rpm flat out because 19ms is longer that it takes the engine to complete a 4 stroke cycle at high rpm, so for high rpm full load we would have to increase pressure, but increasing pressure would mean that at idle LPG injectors would need to pulse below the minimum duration at which they could accurately meter fuel - a problem here would be the type of LPG injectors regardless of ECU). BRC injectors aren't bad in any of these respects but are not what I would have selected for this install, one of their drawbacks is that they are on rails of 2 so even if the rail could be sited between the 2 port runners on one side of the engine pipes between injectors and spuds will be longer than would be possible using single injectors that are not on rails but could be mounted almost directly onto the spud. I reckon that limitations with the ECU is the aspect that is preventing all the issues being cured.
Original arrangement on driver side with inlet air ducting and water header removed.
new nozzle position and pipe length cut down
original type spud
spud type I fitted lower down
I think I took a screen shot showing pressure response, if I find it I may post it here