Hopefully I got a bench setup both for the Weber OEM injection IAW ECU and for the Vems.
'''1st tests fw 1.2.12'
GOOD NEWS
On the next figure, the strobes regarding one cylinder are counting from the last: - Ignition, - Vems Injection, - IAW Injection
.
The settings for the injector are the one computed by VT from the engine specs and as we see not too far from IAW
Injection outputs
- 1.2.13 (or newer: and new VemsTune )
- for odd-fire maserati, injection is 30 crankdeg delayed every 2nd injevent (1,3,5)
- We BENCHTESTed 1.2.13 (https://cell.dyndns.ws/f/ May 13).
- inj is also oddfire 90,150,90,150,90,150 see logicdata in http://www.vems.hu/files/maserati/
- (fphil)what was the injector delay set for this run?
- the config (extracted from vemslog) you sent via email
- Adjust the injector delay (in function of RPM) if necessary, to get desired positions. Let us know if you find it's preferred to delay inj 0,2,4 instead (so that adjusted injangle is real as adjusted without cheating)
- Older fw (was evenly-spaced injection events): tested the Vems injection outputs using the sequential injection settings of the config file (fw v1.2.13 from 2013 April)
given by the OddFireSixCyl page.
Here below are the 6 Vems output together with the trigs 
Evenly-spaced. Remember this odd fire engine has an ignition pattern as 90, 150, 90, 150 , ... and one can expect to have this pattern duplicated for the injection one.
(June 2013) 2nd bench testing with fw 1.2.14
As wikied above, the vems support did a May 2nd version of fw 1.2.13 which as been tested good for odd firing at the VEms facilities.
On my side I bench tested good fw 1.2.14. Hereafter the results.
Settings: Injection timing = 1ms, fuel cut = 7200
TEST I
Settings:
1000 rpm
EndInjectionPhase = 370°=360° + 10° (crang trig is 10° before DTC)
-> the yellow cursor marks cyl 2 which is firing, next firing is done by cyl 5 on next trig 90° afterwhile.
The white cursor marks where the injection pulse should have occurs with EndInjectionPhase = 370°
-> injection is odd-pulsing alright
.Test II
Settings:
1000 rpm
EndInjectionPhase = 250° = 370°-120°,
-> The white cursor marks where the injection pulseat 370° before cyl 2 TDC
-> PLEASE DELAY THE OTHER GROUP OF 3 INJECTORS (say nj 0,2,4 ) INSTEAD
- the "other group of 3 injectors are delayed" in [1.2.14] from June
.Test III
Settings: 3000 rpm, 6000 rpm, 7000 rpm
fuel cut = 7200
... Good similar TestII
.Test IV
Settings: 7250 rpm
... good, no injection = fuel cut done.
. Test V
Settings: 7000rpm injection time = 100ms,
time read on scope = 98,165ms = good.
Thanks for the work done to update the firmware
Nest step to cable and test on the car ... (I am really slow !!)
(July 2013) Wiring the injection
Not much work here. I had a spare of an OEM Weber IAW injection ecu, and hopefully, despite the oem injection was semi-sequential every injector was wired down the IAW connector where the shunts were made.
I modified the connector and the spare IAW. This one becomes more or less a connector box for Vems, but also running on its own, feeding the fuel pump, getting the MAT and CLT values to be sent to Vems (I have modified the resistor bridge on the genboard).
So now, simply by swapping boxes, I can run Weber injection or Vems injection.
(August 2013) Running VEMS injection - stage 1-
From the IAW eprom I had some idea about the injection map which could parametrised fw 1.2.15 , so first try = first start but several issues:
Issue 1: The lambda values are jumping up and down with no reasons
By chance G. Lezsak few days before had noticed that this problem was due to a fw bug and the 2013.08.08 built of fw 1.2.15 solved the case.
-> Close
Issue 2: the lambda values from bank A and B are 8-10% apart,
Round values being higher than Vems'
On bank B, I had kept Vems Round since I did not have the components to properly wire the genboard for the 2nd wbo2. I had changed the Round factory values(Nernst target) in order to get 20.9 in open air, although I knew that Round takes the reference resistance set at factory along with the sensor to compute the lambda values and so does need calibration
a) I went back to the bench. Injectors of one banks have to opened after a delay of 30°, hence a bug could have come from here with the change included in 1.2.14 and 1.2.15. I checked this is not the case.
However I noticed a noise say about 2-3 % on the pulse width, with a DC being,per ex. 18.7, 19.3, 18.7, 18.4, 18.3, 18.1% etc,
This is not due to the trigger signals wich comes from the VT because the periodicity of the pulses is rock solid. It may come from the the computation of the angle derivative, possibly on a too straight way.
I saw RPM noisy as well on the bench without reason.
b) Back to the car, swap the wbo2 connectors, round and vems calibration -> round still higher.
Finally I put back the original Nernst target value, round does not gives 20.9 (18.8) but around 1 the 2 lambdas are quite the same!!
Probably some kind of non linear correction for large lambda values are not done with Round as it is done with Vems. Better to have known that before.
-> Close
- "Safe mode"
I should have enable an "IButton" at some time and I now had no ign. or inj. (test mode good).
To get rid of this "nice safety feature", I finally wrote the "prohibit" word with (190,239). Thereafter the red safe mode indicator went off but still no inj or ign, until, by chance, I upload a fw + config on the genboard which was actually running, and I miraculously recovered inj + ign.
- yes "VemsTune sets correct value automatically after upload of full (.vemscfg) config (to leave "safe-mode").
- maybe you can recommend a place elsewhere in the help, where one looks first and it's more intuitive to find ?
But for now on to get ign. + inj on that genboard, I had to fed it with the triggers signals before switch on?
- no, with uploaded config (correct config values burnt, that disable safe-mode) can be powered down/up and fed trigger signals afterwards. (it's more usual to start cranking after ignition power is on; of course it'll also sync on a higher RPM signal)
Issue 3: genboard disconnects more than often
I thought it was because of the noise of my new injection, or fw version etc .. since before, the connection was steady good. Afters many tests I recalled I changed recently the motherboard of my netbook and I noticed that the prolific cable was shielded on the side of the computer not on the RS32 connector side.
I soldered a wire from this usb metallic shield to the car ground, et voilà rock steady connection with prolific.
I thought it was because of the noise of my new injection, fw etc .
Very good hint, thanks.
-> Close
Issue 4: Idle IAE
I rapidly got reasonable VE and pulse width injection values for lambda 93 at 20° CLT or 90° CLT which compare withe the values I had scoped from the oem Weber ecu. However when the IAC DC for weber was about 30% @ 90°, it needs to be 80% with Vems.
This is something I do not understand since I have checked on bench the similitude of the Weber and Vems signals (frequency, polarity, ...)
- Maybe the weber has a low-voltage flyback for IAC output ?
- low-voltage flyback is not the standard for v3 because 30V flyback gives better injector-control; but also possible to order v3 with low-voltage flyback on 1 or more injector-outputs.
- if using a diode (any diode >=1A eg. 1n4004, 1n4007 or BY399) in parallel with the IAC; anode on controlled IAC pin; cathode on +12V IAC pin.
- but 30% .. 80% is a bit too much to be caused by the switchoff-effect (flyback voltage)
- Just in case... Does IAC regulate in the right direction, btw ? Higher PWM duty % => opens the IAC and feeds more air ?
-> Open
Issue 5: Idle control PID vs Achille's heel
I did no succeed to set the idle as good as it is with oem ecu (cpu 6804)
After some tests, I have finally put PID=0,0,0 and set the IAC DC curve as good as I can. So the engine idles as it likes about 950-1200rpm depending on the warm up enrichment curve and VE
Before I gave a look to the PID.
- First surprise at the right of the set point there is a moving IAC RPM Target (for the P, I and D).
Would the Paris'arrow reach Achille heel in a finite time?
- yes, if the PID tune is anywhere close, target will converge to final_target RPM (from above!) rather quickly (RPM dropping below is much more annoying than to temporarily have a slightly higher RPM).
Theorically this non linear control gives a 1rst order response(exponential), but in practice there is always a significant final error which can only be reduced by a huge value of P which in turn makes the control on the left side of the target instable. So to recover the stability there ones put a large value for D which, because of the noise on the velocity etc, makes a wrong action.
- Another surprise, there is no interpolation for the IAC DC between the end of cranking (say 450) and the MAP entry at the first column (say @800rmp) of the VE table. The P term does the job for IAC DC. Hence one may get IACDC=100% with the VE value for 800rpm.
Probably when the start is well tune, the engines rpm jump across that gap [450rpm,800rpm] easily and the tuning in thisband does not matter.
- I have not understood what the option "Asymmetric PID conf" does.
- if enabled, bigger (twice) values to feed more air (push RPM up) than to remove air
- Hopefully the I action is well described. I did not use it yet. Certainly this is the way to go along with the idle ign control.
-> open
For the present time I need to continue tuning for the steady states, particularly above 1.4 bar. Also I have to set the MAT/TPS enrichment map since actually when MAT=40° idle goes with lambda 0.93, and, when MAT=60°, lambda 1.02.