May be it is off topic, I did some tests about the power consumption when driving a CFL by help of a fly back transformer:

Attached see a circuit which is in essence a MOSFET with a pulse width modulated signal at its gate. The MOSFET drives the primary of a fly back transformer.

By programming a TI-launch pad

http://processors.wiki.ti.com/index.php/MSP430_LaunchPad_(MSP-EXP430G2)?DCMP=launchpad&HQS=Other+OT+launchpadwiki

I can create whatever pulse signal I want.

The best result is a 300 Hz signal with a 0.9% duty cycle. At 2 Volts it lights up a 15 Watt CFL with about 3 mA.  This 6 Milliwatt operation is the lowest I could get. The CFL is nicely visible in the dark (like a night light) but does not have any use.

To get useful light I need 10.000 Hz with a duty cycle of 30%. This means 150 mA at 2 Volt, i.e. 0.3 Watt.

If one is modest, 3000 Hz with a duty cycle of 8% needs about 25 mA at 2 Volt, i.e. 50 Milliwatt, and produces a modest light which allows you to pass through a room without hitting anything.

Never mind the crude contraption, it is only to test many MOSFETs and fly back transformers. I will build a nicer lamp, but the program still needs some more work. The microprocessor needs less than a Milliwatt an the whole Launch Pad costs less than 5 Euros (including the USB-cable), so it can stay in the finished lamp.

The MOSFET used is crutial, because many MOSFETS do not operate well below 3 Volt.

The best I found: VISHAY SILICONIX - SI2377EDS-T1-GE3 - MOSFET,P KANAL,20V,4.4A,DIODE,ESD,SOT23
Farnell Order Code: 1858946  (0.48 Euro)
0.1 Ohm at 1.8 Volt
(this MOSFET is terribly small, a pain to solder and to use, P-channel, 1 = OFF, 0 = ON)

This experiment serves to show what sort of low power operation of a CFL can be done with "conventional means".

But LaserSaber's circuit and his famous transformer seem to light a CFL much more efficiently, although it seems to need quite a high Voltage (up to 80 Volt) to start it up and to work for some minutes on an electrolytic capacitor. (The initial power input might well be in the Watts.)

This experiment also shows that it is indeed necessary to build "something unconventional" to have it "ring" for some minutes with only an electrolytic capacitor.

Greetings, Conrad

Therefore a question: what is the original core in the fuji transformer? Is it also ferrite or something else?

Thanks. I might just grab an inferior one and give it a try first. With almost half Al value, the current consumption will double...

Another question: what is the core material used in the bifiler coil? I looked into a book, and it seems even a bifiler should have an inductance, maybe around 50 uH, depending on the core material -- in fact, everything has an inductance.