Zavoli Alisei momentarily cutting out / switcing to petrol on moderate acceleraton

[danci1973]

Hello,

first, let me introduce myself. My name is Danilo, I'm from Slovenia and this is my 2nd LPG converted car. The first one was also a Subaru, but with a 'factory' conversion with a BRC system and a 2.0 4-cylinder boxer engine. When I bought it (used, of course), the LPG wasn't working and I had to fix it myself and in that process, I gained a bit of basic knowledge about how these systems work (in my case, the gas pressure sensor didn't give a signal, so replacing it made the LPG system switch over, but then the injectors were stuck and I had to help them a bit with a couple of light taps with a hammer).

Now I have a Subaru Legacy 3.0 H6 with a Zavoli Alisei S LPG system, installed in 2017. It has a lube system and I've done about 50k km with LPG since. About a year ago I started noticing that going slightly up-hill, the engine sometimes cuts out, but just for a moment. Didn't think much of it as it was a pretty rare occasion. At the time, I didn't even blame the LPG system.

However, as the frequency of these cut-outs increased, I also noticed that a cut-out is accompanied by an audible click from the LPG tank solenoid and a change of LED status on the Zavoli switch - the green LPG LED turns off while the yellow petrol LED turns on (I do have a video here where it happened a couple of times in a row). Sometimes, the LPG system seems to turn off completely and then comes back as if it was restarted (video here).

I took the car back to the installer, they had a brief look at the computer and there was an error code and based on that, they replaced two coils, but that did not fix the issue.

Although these cut-outs at first appeared random, over the year of living with them I did notice a pattern - (most of the times) it happens with moderate acceleration, at about 1/3 throttle valve opening and only around the 3000 RPM mark (give or take 500). If I accelerate softer or harder, it doesn't happen, it just pulls right through that rev range. It is possible that it's happening under different conditions as well, but for an even shorter amount of time, so that there is no click and I can't have my eyes on the LEDs all the time.

Recently I found a way to pretty much make it happen 'on demand' - the car has a sort-of ECO drive mode, which softens the throttle response and limits the throttle plate opening and in this mode, accelerating from 2000 to 4000 RPM in 3rd, 4th or 5th gear pretty much guarantees one or two cut-outs.

Also, if I try the same on petrol, nothing unusual happens under the same conditions.

When I presented all that to the installer, they seem to be pretty clueless. They thought it could be a bad electrical connection, but I don't believe that (just) a bad connection would be as consistent with regards to the amount of acceleration and RPM range.

Since the Zavoli system is still under warranty, I am inclined to leave it to them, but I'm not really that hopeful and I'd much rather have a working car than a car in their workshop.

I had a few ideas, but ruled them out already:

• LPG tank level - at first it seemed as this only happens if the tank is about 1/2 full - but with the 'on demand' method mentioned above, it happens regardless of the amount of LPG in the tank
• the switch-over button - I mounted that button so that it is flush on the centre console and afixed it with hot glue, so I thought that maybe some glue got somewhere it shouldn't, causing problems - but I have a spare new one, connected that one and nothing changed
• temperature / heat - the Zavoli ECU is mounted in the engine bay, very close to the engine, so I thought maybe the ECU overheats (though this would not explain the acceleration / rev consistancy) - but again, I can make it happen pretty much as soon as the LPG turns on, so can't really be heat soak.

To finish off, here's a list of components used in this conversion (according to the paperwork):

• LPG tank Atiker 650x250 67 L
• multi valve Zavoli / Tomasetto AT-02
• LPG level sensor AEB1050
• reducer Zavoli Zeta S
• injectors Zavoli Pan Jet
• pressure sensor AEB025 (LPG pressure, manifold pressure)
• ECU Zavoli AEB2568

So... Any other ideas?

[LPGC]

When the yellow light flashes while you're running on gas without any beeps from the switch it is because the AEB ECU has noticed that the gas injectors would need to run at (or beyond) full duty cycle to maintain correct mixture according to the calibration map (and compensation for pressure / temperature / etc). Full duty cycle means that the LPG injectors are open constantly, no time to close and open again in time for the next engine induction cycle, thus the LPG system has reached the maximum fuelling it can provide to the engine. Any extra engine load beyond this point would mean the engine would still get the same amount of fuel but would be getting more air, so the engine would run lean if it continued to run on LPG. To protect against lean running at the point of full dc (and also a small window of too rich running at around the point full duty cycle is reached) the LPG ECU forces a temporary switch back to petrol, it will continue to run on petrol until engine load has decreased enough for LPG injectors to be able to provide correct mixture again, during running on petrol due to this condition the yellow light flashes.

This can happen even if pressure is above the level at which the system switches back to petrol due to noting low LPG pressure if the combination of injectors (latency and nozzle size) and pressure doesn't flow enough LPG.

You implied that the yellow light flashing is more likely to occur at part throttle than at full throttle? This lends me to think that the installer dialled in a 'petrol addition' for high engine loads (because they noticed that their install was incapable of correctly fuelling the engine at correct mixture at combinations of high rpm and high engine load) but didn't set up this petrol addition to occur at low enough engine loads to prevent the full dc / yellow light flashing situation. I.e. put your foot to the floor and petrol addition is applied, enough petrol addition to prevent the gas injectors having to run at full DC, at low loads the engine doesn't need any petrol addition to prevent full DC anyway, but inbetween those situations there can be a situation where no petrol addition is applied but the LPG injectors would still have to run at full DC.

If you hear the tank solenoid click out/in anytime the system is running on LPG but the switch doesn't beep (as it would if you'd run out of gas) it's possible that an oddity of a specific version of firmware causes the LPG solenoids to turn off during the full dc condition. Changing a software setting such as 'tank solenoid with dedicated wire' may prevent the solenoid(s) turning off and back on again. If the switch beeps (as it would if you'd run out of gas) it will be because the system thinks you've run out of gas because gas pressure has fallen to the level at which it thinks you've run out of gas. If the solenoid cutting out occurs only during engine over-run then connecting the brown rpm detection wire to an ignition coil and changing the rpm detection settings in software may prevent it and/or connecting the red/white ignition on detection wire to a different source may prevent it. There's another type of situation that can cause this too but it won't apply to your normally aspirated H6 Subaru - If the car were a highly uprated turbo model that ran full dc pinj at full boost and high rpm it could cause the LPG ECU to believe the engine was on over-run rather than running flat out!

There is no excuse for an installer to fit components that aren't capable of properly fuelling the engine on LPG without resorting to petrol addition. No excuse for them not to be able to diagnose and fix the symptoms you're seeing (and probably those you can't or didn't previously know about such as petrol addition / switching back fully to petrol). This installer may have insights into the problem, in which case they've fobbed you off, or may not have insights, in which case they don't know their job... They don't deserve many brownie points in either case!

Have you checked your LPG system setup and calibration settings? If so, have you noticed how ginj (gas injector pulse duration) compares to pinj (petrol injector pulse duration)?

Simon

[danci1973]

Thank you very much for a quick and informative answer...

The installer I talked to didn't acknowledge that the yellow light does have a specific meaning - when I called him up to tell him that new coils didn't help, he said 'it must be a bad electrical connection then', or something along those lines.

I haven't checked the setup as it's still under warranty and I don't know if they can see that I've been connecting to it - but since I'm starting to doubt their competence, I just might go for it.

If the 'petrol addition' was / is necessary, what could the limiting factor for running on LPG only be? The reducer (Zeta/S - if that's what they really installed) should be capable of 250 kW (which is roughly 40% more than what my engine makes).

Looking at injectors (Zavoli Jet) right now. They seem to have some kind of nozzles / jets with different diameter holes (1.5 to 3.0 mm) - they call them 'gigler' both in Italian and in English:

zavoli_gigler.jpg (10.39 KiB) Viewed 2401 times

Are you familiar with these injectors? Is this something internal to them? How hard is it to take one apart to have a look / measure the current setup?

[danci1973]

Good morning!

I've just been to the garage to check the reducer and injector types:

• reducer is ZETA typ. SUPER (800ZETA/S V2), lot 17.10:
  
  
  
  
  * injectors are Type JET | ZAV, lot 09.2017
  
  

I guess the next step is to check and measure the 'giglers', but I can't find any instructions on how to do that - I only found this drawing, but it's not very helpful:




I found a couple of Zavoli installation instructions for different engines, to try and figure out the 'gigler' sizing:

• 3 cylinder, 998cc, NA, 59kW (19.7 per cylinder) - 1.5 mm
• 4 cylinder, 1984cc, turbo, 195 kW (48.75 per cylinder) - 2.0 mm
• 4 cylinder, 1600cc, NA, 85 kW (21.25 per cylinder) - 1.7 mm
• 4 cylinder, 1200cc, NA, 51 kW (12.75 per cylinder) - 1.7 mm

I though this should be more directly related to the kW / cylinder figure, but apparently not.


Could the actual nozzles, installed in the intake manifold, be too small? Do these come in different sizes?

[LPGC]

The amount of LPG that an injector will flow (when fully open) depends on the diameter of the jet in the 'gigler', gas (reducer) pressure and manifold pressure. Since the reducer keeps gas pressure relative to manifold pressure we don't usually need to be too concerned about gas pressure and manifold pressure individually (especially not on a normally aspirated engine, not unless we're going to get into a really complicated area anyway), we only have to be concerned about gas pressure minus manifold pressure.

Zeta reducers are OK but they don't allow a wide range of pressure adjustment. So for example if an installer fitted nozzles that were a bit small he probably wouldn't be able to fully compensate by increasing pressure, or vice/versa. The Subaru H6 is normally aspirated, will probably be a 3 litre(?) so have 500cc cylinders producing around 40bhp (around 30kw) each? I'd expect to fit nozzles around 2.5mm for such cylinder if fed at 1.4 bar.

If you find you need the LPG system to be able to supply the engine with around 50% more gas than it can currently supply (and the reducer's flow ability is capable), you could either increase pressure (to keep thing simple, say by 50% from 1.4 bar to 2.1 bar) or you could increase nozzle size (say going from 2mm nozzles to 2.5mm nozzles). Since the reducer isn't capable of supplying 2.1 bar you'd increase nozzle size from 2mm to 2.5mm (2.1 bar wouldn't be a good idea anyway - your type of injectors wouldn't work properly at 2.1 bar)

The amount a nozzle will flow depends mostly on it's cross sectional area (the old pi r squared) and pressure at one side (gas pressure) versus pressure at the other side (manifold pressure).. A 2.5mm nozzle has cross sectional area of 4.91sqmm, a 2.0mm nozzle has 3.14sqmm (will only flow 64% as much as the 2.5), a 1.7mm nozzle has 2.27sqmm (will only flow 46% as much as the 2.5 nozzle), a 1.5mm nozzle has 1.77sqmm (will only flow 36% as much gas as the 2.5mm nozzle). A 3mm nozzle has 7.07sqmm (will flow 44% more than the 2.5mm nozzle).

There are charts for bhp per cylinder versus nozzle size, other charts for cylinder cc versus nozzle size, etc. The charts are OK as a rule of thumb for a layman but they don't consider all factors. Other factors I consider include petrol injector flow rate relative to the power and size of the engine, other aspects of the engine (turbo/supercharged/valvetronic/etc), aspects of the LPG injector I'm selecting the nozzle for (injector opening/closing time, any chamber capacity between it's valve and the metering jet).... But I'd choose the type of injector itself before nozzle size.

Edit - the last paragraph could do with further explanation... Say we've got a normal run of the mil engine that idles with 3ms pinj (petrol injector duration) and max power comes at high rpm with 16ms pinj. On this same engine if we increased petrol pressure by around 20% we'd expect idle pinj to be around 20% lower than 3ms so around 2.4ms and we'd expect peak pinj to be around 20% lower so at around 13ms. The engine hasn't changed, it still needs the same amount of fuel to idle and for maximum power as it did before. In both cases we could fit LPG injectors with a certain nozzle size fed at a certain pressure and in both cases the LPG injectors might need to pulse for 4ms at idle and 16ms at peak power, so really the only thing that would have to change would be LPG calibration... But (especially on some engines) if the petrol system was changed as above it might be better to fit bigger nozzles on the LPG injectors, and/or change LPG pressure, and/or even fit different LPG injectors altogether to suit the engine with the higher petrol pressure and different range of pinj. Too high a multiplier (ginj gas output pulse length versus pinj petrol injector input pulse length) can be bad, it can upset all sorts of aspects from temperature and pressure compensation linearity to getting the dose of fuel into engine intake ports at the correct moment, too low a multiplier is bad (can spoil accuracy of fuelling even during a constant throttle cruise and can mean that with quickly changing throttle position the engine gets too much fuel anyway).. but on the other hand we can't have ginj as low as pinj at near idle if that would mean injectors wouldn't be able to accurately meter fuel because they're being pulsed for a short period of time which is below the minimum pulse period at which they can accurately meter fuel.

Simon

[danci1973]

I finally found the software that works with my ECU (Zavoli AliseiN 6.1.2.0 CAD) and here is what I found the reducer was set to 1.3 bar in the software and the measured pressure at idle was showing about 1.27-1.30 - at hard acceleration & high RPM I saw it drop to 0.8.





I upped the gas pressure so it shows 1.5 bar at idle and changed the 'reducer' setting to 1.5 bar.



The LPG ECU was set up for petrol addition from 3300 to 9000 RPM and with 'Petrol inj. time for petrol addition' parameter set to 12 ms. I don't remember what 'Amount of petrol addition' parameter was set to.





I had to abort the testing as the ECU suddenly stopped communicating with the software and I didn't know what state exactly it was in - back home I pulled the fuse for the LPG ECU for a few seconds and the connection was working again.

If I turn off the 'Petrol addition', I need to be looking at AFR's at high load and high revs, right?

[LPGC]

"If I turn off the 'Petrol addition', I need to be looking at AFR's at high load and high revs, right?"

That's right but there are ways of looking at AFR's without an expensive AFR meter if you don't have one.

Nothing you've said surprises me (settings etc).

Just a quick reply this evening but I'll come back on this thread tomorrow.

Simon

[danci1973]

I can monitor AFR and / or Lambda through an Bluetooth OBD2 adapter and Torque app on my phone.

Looking forward to your insights.

[LPGC]

If you change the software pressure reference you've also got to recalibrate... If you increase the reference pressure and don't adjust the numbers in the map table it will now be calibrated richer on LPG than it was before and vice/versa. Did you recalibrate? Recalibration would have to involve checking mixture / fuel trims at high loads before and after petrol addition kicked in.

What does reducer pressure fall to now you've increased pressure to 1.5 bar (it fell to 0.8bar before) ? If it still falls to 0.8 bar it could be because you've also inadvertently placed more demand on the reducer (because you didn't reclalibrate which would have involved lowering the numbers in the table for the higher reference pressure) or it's because of a design constraint of the reducer... such design constraint (and some design reducers do have this constraint) can mean that you can happily adjust low load pressure from a reducer but regardless of where you set low load pressure high load pressure will generally fall to the design constraint... in which case you had might as well set the reducer back to the original lower pressure because the ECU won't have as wide a range of pressure compensation to deal with if pressure is between 0.8 and 1.2bar compared to if pressure is between 0.8 and 1.5 bar and lower range of pressure compensation can lead to more accurate fuelling. Also because on this system you can't manually set (in software) the pressure at which the system switches back to petrol (switch beeps like you've run out of gas) like you can on other systems (including some other AEB systems), the switchback pressure will be roughly half of whatever the reference pressure is set to.. Changing reference pressure from 1.2bar to 1.5 bar will have raised switchback pressure from 0.6 bar to 0.75bar which is very close to the 0.8bar the reducer was supplying under load... But leave pressure at 1.5 bar for now.

Can you post a picture of your map table (interested in the figures which are the 'multiplier' for pinj to arrive at ginj - if these are too high it will definitely account for the yellow light flashing) and a pic of your 'modify carb' settings (interested in extra injection filtering settings, incorrect setting of which can lead these ECUs to lose rpm detection) ? The problem with the modify carb settings in this version of AEB software (Zavoli and this release of software) are that there's no extra injection ID time shown, it's usually incorrect setting of ID time that can make the system more likely to lose rpm detection.

12ms Is the default pinj for start of petrol addition (when petrol addition is enabled). Depending on the numbers in your map table (and whether or not they are correct) and depending on what pressure falls to now, this 12ms figure may need to be decreased to 9ms and the amount of petrol addition may need to be increased from 2ms up to (say) 5ms. Most engines have peak pinj of around 16ms, most Subarus have peak pinj of around 12ms... The 12ms default figure would represent about 3/4 throttle for most engines but would represent full throttle on most Subaru engines (which in conjunction with low reducer pressure and/or small injector nozzles would explain why it cuts back to petrol at part load - petrol addition not activated but not at full load - petrol addition activated). Iirc the H6's have peak pinj closer to 16ms than other Subarus but even if the petrol system has peak pinj of 14ms if petrol addition only kicks in at 12ms it's not far off from kicking in only at very near full throttle.

Simon

[danci1973]

I did not recalibrate, haven't really thought about it.

Anyway, I went to the garage with the intention to take some additional screen shots with the engine running on LPG. As I haven't driven it today, it was cold and while it was warming up, I noticed, that the reducer temperature (T.reduc) was rising very slow in comparison to the engine coolant - coolant reached 80°C while the reducer was just about hitting 32°C.

Another problem I noticed - when the engine bay got hot, the Zavoli ECU stopped communicating with the software - it simply dropped the connection. At first I was able to re-connect if I switched over to petrol, but a couple of minutes later that didn't work either. Instead, I had to find the fuse for the Zavoli and pull that, reconnect it and communication was back for a couple of minutes (I actually had to do that for almost each screen shot I took).

So, here they are:







The 'Modify Carb' was the last one I took and look at the T.reducer - 44°C - I'm pretty sure the cooling fans have cycled at least once by that time, meaning the coolant was well over 90°C.



I thought about turning petrol addition off just to see if it would work - like this:

[danci1973]

Today I had a bit of time after work, so I went ahead and changed the 'Gas/petrol' so that it doesn't use 'petrol addition'.

While doing that, I discovered, that the communication problem occur only if it's switched to LPG - in petrol mode, the communication works for any length of time, after switching to LPG the communication stops within 10-20 seconds...

So with that in mind I went for a ride and used petrol to get the engine to working temperature. Shortly before reaching my 'test range', I started 'data acquisition' in the Zavoli software - still running on petrol. When I reached the clear road, I switched to LPG, waited for a couple of seconds for it to settle and then floored it.

Here are my findings:

• gas pressure
  • while I was driving on petrol, it was fluctuating between ~1.6 and 2.1 bar
  • after switching to LPG, it was dropping slightly to 1.36 bar (I was accelerating lightly)
  • after flooring it, the pressure was dropping still, hitting minimum at 1.02 bar - that was at about 6200 revs - after that, the pressure rose slightly to 1.1 bar.
• injector timing
  • Pinj at max. acceleration from 4000 - 7000 RPM was 13.16 - 14.04 ms
  • Ginj at 4000 was 17.66 ms, it rose slightly to 18.5 and hovered around there up until 5000 RPM, then it climbed to 20 ms at 5500 RPM, but the started dropping back to 15.14 at 7000 RPM.
• Lambdas, as recorded by Zavoli, are weird and probably wrong - while I was driving around on petrol, lambda1 was 3.29. When I switched to LPG and started accelerating, it started dropping and at 7000 RPM, it reached 3.09. Lambda2 was 0.98 for 1st half of the log and then jumped to 1.0 for the 2nd half.
  • there are two other parameters (OX and OX2) that have values between 0 and 1.015 / 0.995 - not sure what these are...
• Gas and reducer temperatures were 38-42°C
• slow corrections (LTFT) were between -3 and 7, while fast corrections (STFT) were between -18 and 11 (STFT) / -16.4 and 26.78 (STFT2) - however, this was all while I was still using petrol. After switching to LPG and flooring it, it went 'open-loop' and both STFTs were 0

[LPGC]

Difficult to keep up with changes made but I believe you've:
Upped reference pressure from 1.3 to 1.5 bar and increased pressure at the reducer?
Disabled petrol addition?
At some stage (not shown in screenshots) set software to expect lambda probes being connected (or at least you mentioned lambda readings from data acquisition?

With your gas temp reading being so close to reducer temp reading it seems your reducer temp is reading low. When the engine is sat idling, engine coolant is up to full temperature and reducer temp reading is only around 45C, if you touch the reducer does it feel like it's really only 45C or does it feel hotter?

Before making changes to reducer pressure and reference pressure the gas pressure reading dropped to 0.8bar, now it only drops as far as 1bar, and this is with petrol addition disabled (more demand on the reducer than previously, so without increasing reducer pressure we'd expect pressure to drop below 0.8bar)? Seems this was a good move, the only exception to it not being a good move would be if your injectors were now pulsing below the minimum duration they could accurately meter gas at low loads (idle etc). At idle your gas injectors are now only pulsing for around 2.3ms, is idle still smooth with the engine cool (on gas but warming up) and hot? Are fuel trims relatively steady at idle or are they moving more than they move on petrol at idle?

The reason for the Zavoli lambda voltage readings may be that the lambda wires are not connected or may be that they're connected but the car has wide band lambda probes (which you cannot get a reading from except via OBD). The OX and OX2 readings look like wide band lambda probe equivalence values where 1 would be chemically correct (14.7 parts air to 1 part petrol), 0.95 would be 5% rich, 1.05 would be 5% lean. See if on petrol when you floor it these values come down to about 0.8 (which would confirm this)? If the car does have wide band probes there may be a 'command lambda' reading in OBD2.

Better to disable the 'adaptivity' checkbox in 'modify carb', this will disable the LPG system from adjusting ginj according to fuel trim readings ('corr gas' reading in the map screen). Probably better to leave that disabled all the time but you definitely want it disabled during calibration.

"Ginj at 4000 was 17.66 ms, it rose slightly to 18.5 and hovered around there up until 5000 RPM, then it climbed to 20 ms at 5500 RPM, but the started dropping back to 15.14 at 7000 RPM." Could be unusual readings. Are we definitely talking ginj and not pinj? If pinj had these readings I'd expect the reason to be momentary switching back to petrol due to full dc of ginj around 5500rpm, I'd also expect LPG calibration to be too lean. If we're talking ginj (I expect we are because you've also said peak pinj is 13.16ms>14.14ms, which is what I'd expect on a Subaru) I'd expect it was because petrol addition was still set starting from 5500rpm, which could also partly account for the 1 bar gas pressure (opposed to 0.8bar previously). Could even be that the reducer temperature reading is accurate, in which gas the system may have switched back to petrol at high loads because your config has 'temporary petrol mode when Tgas min' enabled in 'gas/petrol'.

Bare in mind that you should be aiming for fuel trims on gas to be close to fuel trims on petrol. If fuel trims on petrol are going negative 10% it could be because fuel trims have been learned for running on LPG which is 10% more positive than petrol because LPG calibration is 10% leaner than it should be. When we're talking about calibration and fuel trims we're talking about fuel trims at specific loads (of combination of manifold pressure/airflow and rpm), so we're talking at specific points in the map, the map has to be correct at every load point.

Simon

[danci1973]

Difficult to keep up with changes made but I believe you've:
Upped reference pressure from 1.3 to 1.5 bar and increased pressure at the reducer?

Correct.

Disabled petrol addition?

Correct.

At some stage (not shown in screenshots) set software to expect lambda probes being connected (or at least you mentioned lambda readings from data acquisition?

Incorrect - I only mentioned lambda1 and lambda2 because they are in the data acquisition file.

With your gas temp reading being so close to reducer temp reading it seems your reducer temp is reading low. When the engine is sat idling, engine coolant is up to full temperature and reducer temp reading is only around 45C, if you touch the reducer does it feel like it's really only 45C or does it feel hotter?

I haven't tried that - will try later this afternoon.

Before making changes to reducer pressure and reference pressure the gas pressure reading dropped to 0.8bar, now it only drops as far as 1bar, and this is with petrol addition disabled (more demand on the reducer than previously, so without increasing reducer pressure we'd expect pressure to drop below 0.8bar)? Seems this was a good move, the only exception to it not being a good move would be if your injectors were now pulsing below the minimum duration they could accurately meter gas at low loads (idle etc). At idle your gas injectors are now only pulsing for around 2.3ms, is idle still smooth with the engine cool (on gas but warming up) and hot? Are fuel trims relatively steady at idle or are they moving more than they move on petrol at idle?

I will have to check that - because of the communication loss issue I have warmed up the engine using petrol.

The reason for the Zavoli lambda voltage readings may be that the lambda wires are not connected or may be that they're connected but the car has wide band lambda probes (which you cannot get a reading from except via OBD). The OX and OX2 readings look like wide band lambda probe equivalence values where 1 would be chemically correct (14.7 parts air to 1 part petrol), 0.95 would be 5% rich, 1.05 would be 5% lean. See if on petrol when you floor it these values come down to about 0.8 (which would confirm this)? If the car does have wide band probes there may be a 'command lambda' reading in OBD2.

I shall try this later.

Better to disable the 'adaptivity' checkbox in 'modify carb', this will disable the LPG system from adjusting ginj according to fuel trim readings ('corr gas' reading in the map screen). Probably better to leave that disabled all the time but you definitely want it disabled during calibration.

Ok, will do so.

Ginj at 4000 was 17.66 ms, it rose slightly to 18.5 and hovered around there up until 5000 RPM, then it climbed to 20 ms at 5500 RPM, but the started dropping back to 15.14 at 7000 RPM.

Could be unusual readings. Are we definitely talking ginj and not pinj? If pinj had these readings I'd expect the reason to be momentary switching back to petrol due to full dc of ginj around 5500rpm, I'd also expect LPG calibration to be too lean. If we're talking ginj (I expect we are because you've also said peak pinj is 13.16ms>14.14ms, which is what I'd expect on a Subaru) I'd expect it was because petrol addition was still set starting from 5500rpm, which could also partly account for the 1 bar gas pressure (opposed to 0.8bar previously). Could even be that the reducer temperature reading is accurate, in which gas the system may have switched back to petrol at high loads because your config has 'temporary petrol mode when Tgas min' enabled in 'gas/petrol'.

Shouldn't the reducer temperature be much closer to coolant temperature?

Bare in mind that you should be aiming for fuel trims on gas to be close to fuel trims on petrol. If fuel trims on petrol are going negative 10% it could be because fuel trims have been learned for running on LPG which is 10% more positive than petrol because LPG calibration is 10% leaner than it should be. When we're talking about calibration and fuel trims we're talking about fuel trims at specific loads (of combination of manifold pressure/airflow and rpm), so we're talking at specific points in the map, the map has to be correct at every load point.

That is how I set-up my previous car (also Subaru) with BRC system - I made sure that fuel trims were pretty much the same on petrol and on LPG. That made sure that the transition was smooth and it worked well overall.

I'm worried about the communication loss when I switch to LPG, because that way it will be impossible to calibrate anything. I noticed that two wires, connected to the battery + and ground with a 15Amp fuse, are rather thin - is all the power for the ECU & injectors drawn from there?

[danci1973]

I may have found (at least a part of) the issue...

I went out to warm up the engine and the feel and measure the temperature of the reducer - as the coolant was well into 90's (°C), the reducer temperature was still shown as 23°C. That had to be wrong, as I have normal level of coolant, heating works and all that. It also meant, that I wasn't even able to switch to LPG.

When I stopped and touched the reducer, I almost burned my fingers - it was properly hot. I then measured the temperature with an IR thermometer - the reducer casing was at 68-70°C. So I stuck my head deeper and saw a loose wire, right next to the reducer. I pulled it out and, of course, it was the temperature sensor, which was pulled out of the brass fitting. I loosely shoved the sensor back into the fitting and the rest of the drive, it was showing much more 'normal' temperatures - still lower than what I had measured with the IR, but at least it was all working.

I also checked the Pinj time at idle (with hot engine, obviously) - it was about 2.4-2.6 ms. Zavoli claims that these injectors can work with 1.9 ms.

I then did a couple of full throttle accelerations and AFR went down to 11:1, which I'd say is OK?

As you predicted, the short term fuel trims are negative and pretty high on LPG now, I think I saw -25 - but I didn't do 'acquisition' this time as I wanted to get home to fix this sensor and glue it back into the fitting.

So more on (re)calibration in the next days.

BTW, the communication loss - didn't happen today, even after switching to LPG, I was able to monitor the data through Zavoli software. This may have to do with the fact, that previously, I used the Torque software for monitoring AFR, which (also) uses OBD2 protocol, whereas today, I used the 'BTSSM', which uses the SSM protocol.

[LPGC]

You'll want to sort the reducer temperature sensor issue because it not only affects when the system switches to LPG it also affects fuelling on LPG, the system compensates fuelling for reducer temperature. 68C Is more like it!

AFR of 11:1 is probably near where we'd expect for a full throttle run, it's not far from the 0.8 equivalence ratio I mentioned. But compare this AFR on LPG to AFR doing the same run on petrol, you don't want to go any richer on LPG than petrol.

Trims at -25% mean the system is trying to lean the mixture which points to LPG calibration too rich... But compare to readings on petrol.

Some of the later AEB ECU's do seem to lose communication with software at high loads. Sometimes it seems due to voltage at the ECU being pulled down due to pulsing injectors for longer, it seems to happen less often when using an AEB branded cable, but sometimes there doesn't seem to be any reason for it. The LPG software doesn't communicate using OBD protocol...

[danci1973]

You'll want to sort the reducer temperature sensor issue because it not only affects when the system switches to LPG it also affects fuelling on LPG, the system compensates fuelling for reducer temperature. 68C Is more like it!

I already fixed the sensor - I took the brass fitting out, removed the old glue and cleaned it well, then I glued the sensor back in to the fitting with 2 component glue (Mannol 2K-PUR) that should tolerate temperatures up to 150°C.

AFR of 11:1 is probably near where we'd expect for a full throttle run, it's not far from the 0.8 equivalence ratio I mentioned. But compare this AFR on LPG to AFR doing the same run on petrol, you don't want to go any richer on LPG than petrol.

Trims at -25% mean the system is trying to lean the mixture which points to LPG calibration too rich... But compare to readings on petrol.

I don't remember exactly what values fuel trims had, but they were positive. Maybe petrol injectors need cleaning...

Some of the later AEB ECU's do seem to lose communication with software at high loads. Sometimes it seems due to voltage at the ECU being pulled down due to pulsing injectors for longer, it seems to happen less often when using an AEB branded cable, but sometimes there doesn't seem to be any reason for it.

I couldn't find the actual pin-out of the LPG ECU, but I really don't fancy unwrapping all of the wiring just to check for voltage drop - but if problems persist, I can replace the power supply wires with thicker ones.

The LPG software doesn't communicate using OBD protocol...

That's not what I meant.

I know the LPG ECU uses OBD protocol to get fuel trims and other parameters and according to the Zavoli installation guide, it's connected to one or two of the pins on the vehicle's OBD connector (depending on the protocol).

However, I also have a Bluetooth OBD interface which I was using to monitor AFRs on the Subaru ECU. It is possible, that this interface was also using the same protocol and the same pins as the LPG ECU and maybe the LPG ECU gets too confused...

Yesterday, instead of OBD app (Torque Pro) I chose to use a different app (BTSSM), which uses the Subaru SSM protocol along with different pins on the OBD connector.

[LPGC]

Petrol injectors may need cleaning but it's more likely they're fine, just that fuel trims are a bit positive. Just calibrate the LPG system so trims on LPG are the same as trims on petrol.

Here's the AEB2568 pinout (your Zavoli ECU is a AEB2568). viewtopic.php?f=32&t=15853

Correct - if a code reader and LPG ECU both try to communicate with a vehicle via OBD at the same time, one of them probably won't be able to connect properly.

[danci1973]

Took the car out for a test drive and it switched to LPG much sooner than usual - I guess the sensor was 'loose' for a while and has fallen out completely recently, when I was fiddling with the pressure adjustment. The idle felt pretty normal all the time, from cold to hot.

At full acceleration and high RPM, the equivalence ratio (lambda) goes down to 0.88 on petrol (12.9:1). LPG runs quite a bit richer,as lambda goes down to 0.75 (11.6:1) - considering the stoichiometry is about 15.5:1, so that will definitively need some adjusting.

While I was hoping the loose temperature sensor might've been the cause of the 'cutting out' issue, but unfortunately, this is still happening. So I was thinking about what you said regarding the RPM signal and how some settings in 'Modify carb' can cause the LPG ECU to loose the RPM signal - and looking at 'live data', I could see that it is occasionally loosing the RPM signal. I even managed to record a short clip:

https://www.youtube.com/embed/hOeyqMTARoA

So I went to the 'Modify carb.' screen and the only option that seemed relevant was the 'Extra-injection cutting', which was checked. If I un-check this one, two additional settings appear:



If the 'Extrainj. sensitivity' is checked, one can adjust the slider below, but I found that the only setting that somewhat works is with the slider to the full left. However, it seems that it needs the 'Extra-injection cutting' option turned on, as without that, the throttle response is very bad.

However - even though the two extra setting disappear when I check the 'Extra-injection cutting', it seems that 'Extrainj. sensitivity' has to be enabled, otherwise the throttle response is very bad...

Unfortunately, after I fiddled with these settings, I couldn't get it to 'cut out' any more - even if I returned all the settings back to initial state.

What's with the brown wire you mentioned? What should it be connected to (the engine has 6 individual coils on plug)?

[LPGC]

When stochiometric ratio is correct lambda equivalence ratio is 1 but lambda equivalence ratio is not the same as lambda signal voltage.
Stochiometric ratio and equivalence ratio = 1 is 14.7:1 for petrol and 15.4:1 for LPG.
Sensors read oxygen content in exhaust gasses (so they directly read equivalence ratios but do not directly read mixture unless they are calibrated for the fuel being used). Since OBD data is calibrated for petrol correct mixture on LPG will still be with a mixture ratio reading of 14.7:1.

The purpose of these settings is to filter 'extra injection' pulses. Extra injection pulses are very short duration pulses of petrol injectors that occur between the normal petrol injector pulses (normal pulses occur once per cylinder intake cycle). The reason for filtering these short pulses is to avoid pulsing gas injectors more than once during any cylinder's intake cycle. Gas injectors take longer to open and close than petrol injectors and if gas injectors try to pulse more than once during an intake cycle it can lead to overlapping gas injector pulses which has the effect of vastly enriching the mixture. For the very small amount of extra fuel that extra injection pulses provide it can make more sense to totally filter out (disregard) extra injection pulses or at least avoid pulsing gas injectors more than once per intake cycle by adding a bit of extra length to the single gas injector pulse per intake cycle.

There are various explanations in documentation that comes with various versions of software on what/how the extra injection settings do but none of the explanations seem to agree exactly with what the settings do. Some AEB2568 software that looks very similar to Zavoli software features an extra box called 'extra injection ID time' which (as the label suggests) is the minimum pulse duration above which a petrol injector pulse will be considered a main (normal) pulse and below which a petrol injector pulse will be considered an extra injection pulse. In AEB software like yours the slider position is usually considered the setting that sets ID time. Extra injection cutting (box) enabled will just completely filter out / disregard extra injection pulses, when extra injection cutting (box) is disabled the sensitivity slider may try to compensate LPG fuelling for the petrol extra injection pulse by adding extra duration to the LPG injector pulse. In other versions of AEB2568 software there's another box which shows the number of milliseconds the slider position points to, full left is 0ms, full right is 3.8ms. Particularly on engines that pulse petrol injectors for shorter than average duration at low loads, if you set the slider too far to the right it will cause problems because the system will start to filter out normal injection pulses in belief they are short enough to be extra injection pulses... if your engine idles with pinj of 2.6ms and you set ID time at more than 2.6ms the system will regard normal petrol injector pulses at idle as extra injection so no wonder the engine runs rough if you set the slider to the right!

I've noticed in some cases on AEB2568 systems these that extra injection filtering settings can affect rpm detection.

I don't think Subaru H6's feature extra injection. This may not appear to make full sense but try enabling the Extrainj.sensitivity box, move the slider full left, then disable the Extrainj.sensitivity box. Try this with the ExtrainjCutting box disabled (how does it run?) then enabled if it seems better with this enabled.

On some similar AEB2568 systems/software there are 4 indicators in the software screen next to the image of the changeover switch (bottom left corner of screen) to denote occurrence of petrol extra injection, extra injection filtering occurrence, enhancement filter and petrol contribution. Petrol contribution is usually petrol addition as set up in the 'gas/petrol' screen but I've noticed it also seems to occur when rpm detection seems to be playing up, which in turn can be due to incorrect extra injection pulse identification, which can be due to incorrect settings in the extra injection filter section... but now always due to incorrect settings in the extra injection filter section.

The system can detect rpm from frequency of petrol injection pulses, or you can wire in the brown wire to the pulsing wire on an ignition coil (and change rpm detection settings in software 'changeover' screen so the system knows to expect rpm signal from the brown wire by enabling 'type of revolution system' box). I've even found that rpm detection settings that are greyed out when the brown wire isn't used can affect pulsing of the 'petrol contribution' indicator and rpm detection. Before connecting the brown wire you may want to experiment with temporarily enabling 'type of revolution signal' so that you can change settings that were greyed out, then disabling 'type of revolution signal' again. I'd advise trying setting the drop box at the side of 'type of revolution signal' to 'weak' and ignition type to 'single coil'. If you do connect the rpm wire you'll need to leave the 'type of revolution signal' box ticked anyway.

It doesn't surprise me that the connection didn't drop out after you fiddled with extra injection filtering settings.. As said above, these settings can affect more than just extra injection filtering, they can also affect rpm detection and I've also noticed they can affect coms with the laptop. But other settings such as rpm detection settings also don't seem to do exactly what they're supposed to. Especially when these (later) AEB2568's are fitted on 6 cylinder engines... If your ECU was an earlier AEB2568 you wouldn't be having problems with rpm detection or with extra injection filtering, but older AEB2568s have their own drawbacks compared to more recent AEB2568's on modern engines. The first engines I stopped advising AEB2568's for were 6 cylinder engines.

Simon

[danci1973]

The 'Type of revolution signal' was already checked and it was set to 'standard'. That means the brown wire is connected somewhere and I tried to find out where, but it seems it was done somewhere in the bulk of the engine wiring loom and I didn't fancy unwrapping it all (not today, at least).

So I decided to chance it and change that setting to 'Weak'. I also to lowered the gas pressure to 1.4 bar. After that I drove around all afternoon without getting a single LPG 'hiccup'.