www.irf.com

- 4 -

RD0803

One strategy that has been used is to omit the preheating phase and steer the

oscillator frequency to resonance during ignition using feedback from the output circuit.

This ensures that at switch-on the highest possible ignition voltage will be applied to the

lamp. In this way the lamp will ignite in whichever position the 3 way switch is set.

Such a scheme could reliably ignite the lamp when the DC bus is at 300V, however

without correct preheating the ignition voltage of the lamp and consequently the peak

current in the MOSFET half bridge during ignition will be higher. Also the life of the

lamp is substantially reduced when there is no preheat due to far greater stress

occurring on the cathodes at the point of ignition.

Ignition when the DC bus voltage is at 150V is very difficult. Tests indicated that

sweeping the frequency down through resonance sometimes failed to produce

sufficient ignition voltage leaving the ballast in open circuit running mode. The

conclusion from this is that the ballast needs to oscillate at resonance for an extended

period of time in order for the lamp to ignite at 150V considering that the output inductor

and capacitor have been designed to produce 100% lamp power at 300VDC bus when

the frequency is 40-45kHz.

Many CFL ballast designs do not incorporate a current sense and shutdown function to

protect the circuit in the case of ignition failure and so the ballast would eventually fail if

left switched on due to the high MOSFET switching losses causing thermal destruction.

This would not matter with and integrated ballast / lamp type product when the lamp

has failed.

It has also been observed that hard switching occurs at the MOSFET half bridge when

the DC bus voltage is low in position 1 since when the ballast is running it will be close

to resonance, bearing in mind that the resonant frequency shifts downwards in run

mode. Hard switching is very undesirable because of the high peak currents that occur

when each MOSFET switches on. This has been shown to result in a higher rate of

field failures in ballasts due to MOSFET failure.

The conclusion is that the approach to design described above is unable to provide a

reliable ballast.

The dimming level can also be controlled by simply changing the frequency. By

changing the frequency between 3 defined settings, however, it was found to be

extremely difficult to set a point where the dim level is 50%. The problem with this is

that the lamp current against ballast frequency characteristic of the system exhibits a

very sharp knee such that as the frequency increases the lamp current is gradually

reduced up to a point at which a small increase of frequency will result in a very large

reduction in the lamp current.