i don't understand...
I guess it is related to "form 109" question:
There are two methods generally used with magnetic ballasts:
1) Ballast input voltage reduction. Work on all ballast without any form of regulation (so HX transformer or series choke reactor), where the actual power delivered to the lamp is dependent on the input voltage. When dimmed, the voltage is gradually in small steps (less then 5%/step, slower then 5%/5min, so the lamp can thermally adopt prior to further voltage reduction) decreased to ~85% (for 60% brightness). This method has many disadvantages (more likely lamp arc extinction, limited dimming range to name the two most important), but one big advantage: The dimmer can be one common for large group of fixtures, so individual fixtures are kept basic (even without PF capacitor, as the compensation is then adaptive in the dimmer box).
2) Ballast output impedance increase. Can be done on all magnetic ballasts, mostly in the form of extra reactive component (inductor on lag ballasts like series choke, HX, mag-reg; capacitor on lead ballasts, like CWA or CWI) serially connected with the lamp, on the full (undimmed) setting bypassed by the contact from the control unit. This extra impedance keep original OCV (only reduce the lamp current), so could be activated at once (having more levels would make the gear even too complex). his method has the big advantage of how deep the lamp could be dimmed (HPS down to 30%), but require the extra circuitry (impedance component, controller,...) to be inside each fixture, making them more complex.
But as it is way simpler then the feed voltage reduction, it is useful with small group of fixtures, where the common, but quite complex dimmer would not pay off.
Other option belonging to this category is assembling the ballast from multiple lower current rated ones, so when connected in parallel yield same characteristics as nominal full power ballast. By disconnecting some sections is then possible to reduce the lamp power. This is useful mainly for serial reactors, as for other types the setup would become too complex...
Electronic ballasts always use lamp power regulation, so dimming is done by altering the target value for the lamp power regulation block. As this mean only altering some resistive divider (or even constant in the firmware, so cheap to implement), many ballasts intended mainly for HPS do offer this feature as standard.
Of course cheap HF selfoscilating ballast designs regulated by ring core saturation can not be dimmed, as the lamp power is given by the feedback transformer core saturation (like in CFL's,...).
