Sorry to bother you
with the
mandatory cookie drill

Electronic Measurements


Last Modification: April 22, 2013

When doing measurements, interesting phenomena can come across where never been thought of. While measuring the power of a fluorescent lamp during the startup, an inexplicable behavior of the startup sequence led to the discovery that light radioactive material is used in fluorescent tube light starters. Here the story:

Starting behavior of fluorescent lamps

measurement arrangement start-up sequence of a TL
Fig. 1: Test circuit for measuring the behavior of a starting TL.

The initial reason for this measurement was the question how much energy is used during the startup of a fluorescent tube light. For this measurement a 36 W fluorescent lamp set was used. The voltage and current were measured with a digital oscilloscope using a high-voltage differential probe and a DC-current probe. The measurement arrangement is shown in figure 1.

The result of the measurement can been seen in the screen dump below. The scope calculates the power from the instantaneous values of the measured voltage and current. The energy is calculated by integrating the instantaneous power and is represented by the blue line. So at any time stamp it shows the used energy sins t=0.

Diagram startup sequence of a TL
Fig. 2: Startup sequence.

During the startup of a TL the following discrete phases can be distinguished:

  1. This is the inexplicable part. The voltage over the TL circuit is present, but both the current and power are very small.
  2. During this phase, the gas in the starter is warmed up. The current is low, as well as the power. Also observe the slight increase of the energy line.
  3. The gas in the starter has heated up the bi-metal enough to close the contacts. The current flows now through the bi-metallic switch, the ballast and the filaments of the fluorescent tube. Because of the relatively high reactance of the ballast in comparison with the ohmic resistance of the filaments, the power is primarily reactive. Despite that the power amplitude is large, the energy used in this phase is relatively low.
  4. The bi-metal has cooled down and opens. A high induction voltage causes the ignition of the fluorescent tube and light up. The current is now limited by the reactance of the ballast and fluorescent tube. The current and power amplitude are now lower than during the ignition phase "c", but because the circuit has now a more ohmic character, the TL use more energy. Notice that the power curve now is asymmetrical around the zero-line.

The power used by the fluorescent light fixture is proportional to the steepness of the blue energy line. This measurement shows clearly that when the TL has started and works normally the energy line is the steepest and therefore the most power is used. The startup requires a relatively low power.


There is a general prevailing idea is that a fluorescent TL uses a lot more power during the startup than while it's in the normal luminance state. In the MythBusters episode 69 was this assumption examined and confirmed, unfortunately not correct. The MythBusters made an error that revealed clearly where the myth is based on. Only the current was measured during the startup of the tube light. Figure 2 shows that the amplitude of the current during the preheating "c" is greater than when the TL is light up "d". But this tells nothing about the energy that is used per unit time. To determine the true power at a certain time, also the voltage must be measured.
The article Theory en Definitions explains how to measure the power and energy in a correct manner.

Unexplained part start-up sequence

So far this measurement. Remains only the puzzling part "a" during the startup. Repeated measurements of the startup process tell that this current-less period very rarely occurs. Because all components, ballast and the filaments of the fluorescent tube are conductive links, the starter is suspicious.

Measuring the starter

Normally the voltage across a starter is limited to approximately 150 V. But even during the sinus peaks (325 V) and the fairly long duration (3 ms) of a voltage above 150 volts per half period, this is apparently not a guarantee to ignite the starter. It was decided to investigate the behavior of the starter at different voltages. Therefor a step shape voltage is needed that rises quickly from zero to the desired voltage and holds this until the starter ignites.

The test circuit

test circuit
Fig. 3: Test circuit to investigate the response of a starter.

The heart of the circuit is a 1.8 mH coil. During the conductive state of the MOSFET, the linear increasing current though the coil will charge the inductor. At the time the MOSFET is switched off, the energy in the coil will transfer via the diode and 100 Ω resistor into the 10 nf capacitor. This increases the voltage across the capacitor in a very short time and drops afterwards very slowly.

By adjusting the voltage of the power supply or the time the MOSFET is switched on, is the test voltage can be controlled.

The starter under test is connected in parallel across the 10 nF capacitor. The voltage on the starter is measured with a high voltage differential probe.

Test Measurement

impulse voltage across starter
Fig. 4: The impulse voltage across the starter.

Figure 4 shows a typical course of the voltage across the starter. The voltage rises after the MOSFET stops conducting quickly to the desired voltage, 290 V here, and remains this until the starter ignites. The starter responded during this measurement with a delay of 700 ┬Ás. After the igniting the capacitor is discharged by the starter with a typical curve till the voltage reached approximately 150 V. Hereinafter also the starter doesn't conduct anymore and the decrease in voltage across the capacitor determined only by leakage currents.

Multiple measurements show that the reaction of the starter is very random but there is a clear dependency of the test voltage. The higher the voltage the shorter the response time.

For this measurement the protective shell is stripped from the fluorescent starter and the internal capacitor was removed.

Spread in response time

Diagram spread response time daylight
Fig. 5: Spread in response time under daylight conditions.
Diagram spread response time dark
Fig. 6: Spread in response time under dark conditions.

The ignition delay is anything but stable. The first measurements were carried out in the evening where the delay varies from almost 0 to many tens of milliseconds. The next morning the same measurement was performed again, but now it showed a much shorter reaction time than the night before. A very strange situation! After some experiments it becomes clear that the ambient light caused this. Starters are apparently light sensitive. Evidently the photons can speedup the ignition.

Figure 5 shows the spread in the ignition time in the case where daylight entered the room. And the other measurement where the starter was completely shielded from ambient light is shown in figure 6. The oscilloscope images are the result of measuring many ignitions. The difference between these two images reveals clearly the light-sensitivity of the starters. The similarity between the measurements under light and darkened conditions is that in both cases a huge spread in delay time is visible.

The variation in response, especially under dark conditions, could theoretically declare that a long delay could occur before the start sequence of a fluorescent tube light is initiated. This must not be confused with the usually common repeated attempts to ignite fluorescent tubes.

The measurements in figure 6, in which photons are kept outside, shows that the ionization of the gas can occurs almost immediately or after a certain time. The suspicion exists that other energy-rich particles are responsible for ionizing the gas in the starter. This should be further investigated:

External energy-rich particles

histogram ignition delay
Fig. 7: Spread ignition delay of a TL starter with lead shielding plotted in a histogram.

To investigate whether external energy-rich particles play a role in ionizing the gas, the TL-starter is partially shielded with a role of lead. The thickness of the wall is 25 mm and the length of the role is about 100 mm. The starter is placed inside this role, and again also shielded for external light. Despite the starter is now shielded around, the ends are still open. Still a noticeable difference in ignition time should be expected to observe.

There are two comparative response measurements done with and without lead shielding. Every measurement consists of about 5000 sweeps. From both measurements the delay times are plotted in a histogram as shown in figure 7. In these two measurements there is no significant difference whether there is a lead shielding or not. The conclusion is that external energy-rich particles don't have a great influence in ionizing the gas inside TL-starter.


If it doesn't come from outside it must come from the inside. Unfortunately, there is no possibility to screen the interior against itself. No other measurements can be done. A search on the internet shows that the light radioactive Krypton-85 is added to the starters to force a quickly ignition:

radionuclide: Kr-85
Activity criterion: 104 Bq
Activity criterion: 105 Bq/g
activity by product: 1,9.104 Bq

Reverse polarity

Diagram ignition delay reverse polarity
Fig. 8: The ignition delay of the starter with reverse polarity.

There is another peculiarity about starters: They are polarity sensitive. With the same measurement conditions as in figure 7, the two connections on the starter are swapped. Figure 8 shows that the ignition delay is now a lot shorter. A real explanation for this phenomenon is not immediately ascertainable.

Please enable Javascript if you want to respond.

Responding to articles is temporarily disabled.