I can't do the exact lamp but they are all pretty much the same just different scales.
Here is a startup curve with voltage soft-start of an osram 62138 lamp ramping logarithmically up to 12.5V:
(2V/div/3.9A/div/50msec/div)
I calculated the power spike to be 160W at 31A at 50msec.. and 194@ at 24A at 100msec.
now contrast that to a 'current limited' startup like mentioned by JS above..
This time the power numbers mathed out to 80W at 50msec and 136W at 100msec.
notice.. that initially (for the first 20 msec or so) the voltage and current rise at exactly the same rate as with the voltage only soft-start.. but that as soon as the current reaches 20A in this case.. the voltage is clamped and is only allowed to rise to the level that generates 20A.
As the filament heats and becomes higher resistance it takes more voltage to achieve 20A.. and at about 100msec the point is reached where the voltage can resume it's normal logarithmic rise to the set voltage.
To achieve a faster startup and put less heat on the FET, i usually set a startup current about 2 to 2 1/2 x the running current of the lamp.. this is tremendously less power on the filament than 'direct connect'.. and achieves a nice balance of a quick startup (typically 1/4 second or less) and a serious drop on the power spike to the poor filament!
In this same example.. the instantaneous power spike applied to a cold filament of a 62138 lamp wit 12.5V applied will be over 400W easily!
I have measured over 75W spike on a 15W lamp for example.. using only soft start and not current limiting will drop that to 1/6th the power spike. I don't usually bother with current limiting on lamps lower than 50W.. i like the bit of power spike to help heat the filament faster, and it's simpler to program and set the regulator... i typically get power spikes no more than about 50% of running power when i use only voltage ramping soft-start.
-awr