* factory ECU, driving ign and the DBW-throttle (and the gasoline-injectors when switched to gasoline)

* piggyback LPG-injector driver

* [VEMS round] for lambda-control feeding voltage signal to the factory ECU's narrowband-lambda input

Car works like a charm on petrol and LPG.

LPG with ~1.15 lambda is perfect (decreases pumping losses)

* no experience yet with higher, say ~1.23 lambda.

----

Trick:

* round's vout (dsub15/pin1) connected to the factory ECU's nernst input, NBO2 sensor disconnected of course

* The round's vout_gndref (dsub15/pin9) had to be connected to system-ground (chasses)

** The installer originally connected vout_gndref to the lambda-sensor ground, which (surprisingly) measured -0.35V (no mistake, minus !) when operating well

** MAT/CLT/TPS sensor malfunction errors came up in OBD-2 tool when round's gndref was connected to this wire instead of chasses

----

vout_conf=04, and the curve set like:

* 0.9V downto 0.76V

* sudden 0.76V downto 0.16V from 1.1 to 1.2 lambda

* 0.16V downto 0

The round experimental (cell.dyndns.ws/f/) 1.5.5 firmware (if the 17 point analog-out curve topmost entry is exactly 0.9V which is internally 46: vout_curve=2E) applies some hysteresys around 0.45V while generating this signal.

* one might see more or less lambda-deviation with or without this hysteresis (0.92V in the topmost entry disables this hysteresis)

----

Room for improvements

* lambda-target is constant at the moment.

** possibly bias lambda-target down (say, by 0.15) at high-power (feed throttle-pedal signal into round ?).