Mercury vapour lights use a gas such as neon to strike a
running current limited arc, the lamp starts to heat up,
vaporizing the mercury (which at low temps may be solid -
mercury theremometers can freeze you know), the
temperature and pressure in the bulb rises as the
mercury vapour carries the current. If the pressure rises
too high for the mains supply to sustain the arc, the
arc extinguishes and the lamp has to cool before
it will re-start. Then it does the above cycle again.
This repeating cycle usually means a defective or
worn out lamp. High pressure lighting mercury
vapour lamps (in the UK 240v supply) run off the
mains through a current limiting choke, regulating
the current to 2A or so. Not sure about the lower
voltage US mains, they may need some special
arrangement to get the arc started.


Mercury lamps use argon as the fill gas, and for 120v operation they use an
autotransformer ballast just like large (700-1000W) require to run on 240v
mains.



Low pressure mercury vapour tubes (not used
for lighting) strike at 1KV or so, and run at
typically 100mA from a current limiting transformer
called a leak transformer. Leak because some
of the primary flux is leaked away through a
shunt so as to limit the current. Making it
a constant (well near enough) current, high
voltage transformer.

SOX (low pressure sodium) lamps use a similar type of transformer, this is
why while the lamp efficiency is very high, the system efficiency is rather
pathetic for the smaller sizes (15W ballast loss for an 18W lamp!) . Some
modern units use electronic ballasts which are more efficient, I'm not sure
how much more though.