From Inside Electronics
The photos below show the inside of a fluorescent light starter. The starter is used to cause the preheating of the cathodes inside the fluorescent tube before it is started. Once the fluorescent tube has been started the tube conducts electricity and emits light (fluoresces) via the phosphorous coating on the inside the of glass tube. The starter is not required for this to continue and can be removed until the light is turned off.
The starter is essentially two bimetallic strips in a sealed glass capsule, with a small gap between them. The glass capsule contains a conductive gas, allowing sparks to travel across the gap. A capacitor may be connected in parallel to this (as shown in the photos below). This capacitor suppresses the RF noise generated by the device during start up, which would otherwise interfere with radios, televisions and other electrical equipment.
The starter is connected in series with the cathodes and in parallel with the tube itself. A ballast device is connected in series with this pair to control the current once the lamp has been started. This is shown in the diagram below.
When the light is switched on, the tube does not conduct, and current flows through the cathodes to preheat them and also through the starter. This causes sparks between the two bimetallic strips inside the starter, which then heat up. The heat causes the bimetallic strips to bend together, and the gap closes. When the two strips touch a high current flows, but the lack of sparks allows the strips to cool down. The cooling strips then separate and the voltage reappears across the length of the fluorescent tube. By this time the cathodes have been heated sufficiently, and when the voltage appears across the tube it arcs down the length and the tube ignites. If the tube does not start, the starter continues to spark and the cycle repeats.
In some starters the bimetallic strips bend apart as they heat. This also causes the sparks to stop, igniting the tube and allowing the strips to cool down. The starter shown here works on the first method. This was proven by placing the glass capsule on the cooker hob and watching the gap between the strips close.
The ballast used in this circuit is essentially a choke and capacitor in series. The capacitor corrects the phase lag caused by the choke and improves the efficiency. The inductance of the choke causes a voltage spike when the starter switches off. This is due to the sudden change in current flow and can help achieve the initial spark required to start the fluorescent tube. The choke controls the current once the tube has been started, as it has a high impedance at the operating frequency of the tube (about 20kHz). It has a low impedance during start-up when the mains current flows through the cathodes and starter at 50Hz.
Other types of starter are also available, including solid state devices that allow the cathodes to preheat for a fixed length of time and will only try to start the tube a limited number of times. These starters are more expensive of course, but are able to start the tube if the supply voltage is lower or the ambient temperature is too hot or too cold for a conventional starter.
Many other fluorescent lamp circuits exist, but this is the only one to use this type of starter. For more information follow the links below.