In fact, there was a study done that shows the rapid-start T8 magnetic ballast actually demonstrates shorter lamp life than an instant-start electronic ballast in frequently-switched applications!

I have also seen rapid-start electronic ballasts - these I think have temperature-controlled (PTC) preheat, and a discharge (dim glow) forms along 1/3 to the whole length the tube at the same time, or slightly after the preheating starts, for usually about a second, and then the tube ignites fully. When the lamp is hot, though, it starts instantly.

These are the Programmed start: First they just heat up the filament and only after some programmed time they increase the voltage for ignition.
The partial glow during the preheat is caused by the voltage across the lamp: Although the aim is just heat up the electrodes, it does not mean there can not be any voltage across the lamp, it just has to be low enough, the lamp does not ignite.
From practical implementation, there are two options how to approach the preheat/ignition/run:
- First uses a kind of switch, which initially bypass the tube, so the arc can not form, but let the current flow just through the filament. With this, the inverter is in fact loaded by just an L and small R, so operates as during normal lamp run.
After some time the switch opens and let the inverter operate into a series LC, forming the required high voltage for the ignition. At this stage the operating frequency is steered by the LC resonance. When the lamp ignites, the arc effectively damp the LC, so the inverter operate into a LR load, so the inverter frequency become more-less controlled by the current.
This concept mean nearly no voltage across the lamp during preheat, but it's only practical implementation mean using a PTC as that switch: Initially it is cold, so low ohmic, so bypasses the resonant capacitor. The preheat current heat this PTC too, till it's resistance increase and allow the LC to generate the high voltage. This is usually employed in the very simple selfoscillating ballasts and very frequently in CFL's, because the whole circuit is just a power halfbridge oscillator: So simple, but not able to do much.

- The second way is to keep the resonant output stage all the time the same, but sweep the frequency: Initially operate above the resonance, so there is not enough voltage across the capacitor (so lamp) to ignite it, but the current is sufficient (~1.5..2x the rated lamp arc current) for the electrode warmup. Or, in case of voltage heating, the voltage across the auxiliary winding on the inductor generate quite sufficient heating voltage. But during this time, the voltage across the lamp is a bit less than 150V (a figure for F36T8; the F32T8 would be about the same), way too little for the lamp to ignite at any temperature.
After the timer expires, the frequency starts to sweep down towards the operating frequency, driving the LC through it's resonance, so generate high voltage. This high voltage (at least 1kV for F36T8) is then sufficient to ignite the lamp at any temperature and even surface leakage (humidity, dust,...) condition. After the frequency reaches the normal operating level, the lamp is assumed lit. With these, either the circuit is designed so, it may stay in the "operation frequency" state forever (it does not stress any component and the voltage on the output is way below 600V - the limit for a normal "600V OCV" rated wiring), or it detect the fault and shut down.
This control requires quite sophisticated circuit, so practically it mean an integrated circuit, where all the functionality, except the power transistors, is implemented on-chip, driving a pair of power MOSFET's. The circuit has around just few passive components to program the required operating characteristics (frequencies, preheat time,...) and it's supply, so the circuit on the PCB is not that complex at all (as all the complexity is integrated on the controller chip).

In one of my more complex and expensive ballasts - this "Energy Savings" dimming ballast - only in complete darkness can you see the cathodes preheat (and sometimes one or two will form a slight arc across the cathodes) but otherwise the tubes do not glow during preheat - and it just appears that there is a one-second delay before the lamps light. I have read on some ballast manufacturer's web site that this is the design of a "proper" programmed-start ballast, such that the voltage across the lamp is kept as low as possible during preheat. Even if I turn the power off for just a second, the preheat delay is always there.

This Lights of America ballast was very cheap and runs the lamps in parallel, but does have end-of-life protection (which, when tripped, cuts off both lamps and the power has to be shut off for at least 5 seconds to reset it). The first lamp usually glows at one end and travels down about 2/3 the length of the tube, then ignites, and then the second lamp ignites. When warm, the lamps start instantly.

Indeed, the glow during preheat is an unwanted effect, it can be tolerated when the electrodes are already warm (so not immediately after power ON, but with about 1s delay, just for short time before ignition). Otherwise an immediate glow after power ON means the voltage is indeed too high for the preheat.
The fact the ballast does not glow could actually mean one of two things: The voltage is really low (the good scenario), or the preheat time is too short and/or preheat current too low (the lamp glow only when the electrodes really have reached the emission; and this scenario is the bad one). And more frequently I met the second one.
To distinguish, you would have to really monitor the filament temperature. But that is very difficult, as the incandescence of the correctly heated filaments is very weak.

Quote from: RyanF40T12
Do they make rapid start T-8 ballasts?
>Not anymore, but yes they did... startup is just like the F40T12 version. I have a few of them myself Smiley

They actually still make magnetic ballasts for F32T8, but they are hard to find and are intended for locations where the interference from electronic ballasts is unacceptable.
Programmed Start will give the longest lamp life, and I would recommend it for most uses.

@Medved:
The ptc is not the only switch method used for preheating(altought it's used in 95% of cases).
Osram uses for some of these ballasts a circuit very similar to this one. I've also seen a rele used as preheating switch (in one 230v electronic ballast and in some 12v ballasts).

Yes, there are other methods, but such ballast concept has so many other design flaws, the only it's advantage and reason why it is still used is it's simplicity. So anything more complex mean the ballast loose it's only advantage over the use of frequency controlled designs utilizing one of the many ballast control chips, whose circuits are only slightly more complex than the basic selfoscillating circuit, but most of the issues are solved there (correct programmed start operation, EOL protection, supply brown-out handling, with some chips closed loop constant current regulation over wide supply voltage range,...).
So even when it is sometimes used, as you wrote, vast majority are just a PTC.

And for Osram, I have disassembled few Quicktronic ballasts and there was nothing more than just a PTC, only not directly across the complete lamp, but bypassing just one of two capacitors connected in series to form the resonant capacitance. And their inverter circuit is usually a bit different: They are using MOSFET's and a LC resonant circuit in the gates, probably to keep the frequency independent on what the lamp does (or just to be able to use MOSFET's instead of bipolars).

Yes Osram use to put the PTC in parallel of one of the two series capacitors. I've used too this method in my instant start conversion, by selecting the value of the bigger capacitor (the one not in parallel with the PTC) you can adjust the preheating current:
Bigger capacitor lower the preheating current (and the voltage seen by the lamp) and smaller one rise the current (and also the voltage seen by the lamp, so care must be taken to not cause the strike of the lamp).

I have some osram quicktronic ballast for 18W lamps of the same model, in some the preheating is made by a PTC and in others (I think the newer) with the transistor switched system. The ones with the PTC resistor doesn't preheat well the cathodes, they start more like a rapid start ballast(probably the voltage on the lamp is already too high during the preheat phase).