The BMB experiment wants to test the lifetime of a top quality CFL under heavy on/off switching conditions. This should mirror a domestic (not professional!) usage.

A 20W OSRAM CFL will be driven in parallel with a 100W OSRAM incandescent lamp, and powered on/off at a frequency of approx. 0.5 min^-1: that is the lamps will be switched on for 1 minute, than switched off for a minute, and so on. Two up-counters will count these switching operations; when a lamp blows, the counter will stop and show the total switches up to failure.

The current political efforts tending to forbid the usage of the classical bulb neglects (among others!) the practical problems of frequent switching. It is assumed that frequent switching will shorten the expected lifetime by up to 85%. In a domestic environment fully equipped with CFL's, many of these will be switched frequently. Think of the lights of the restrooms, the basement, the hallway equipped with motion sensors... Thus the extreme long CFL lifetimes of 8000h or more will probably be unrealistic in practical life; this has been found to be the case by many adopters (the author included; read reader comments on this issue here). As all energy saving arguments rely also heavily on the important difference in lifetime between incandescent bulbs (ca. 1000 - 2000h) and CFL's (8000 - 15000h), they have to be taken witch a pinch of salt.

Starting 04 May 10 with experiment #6, a modified circuit is used.  Jump to paragraph 4.

1. The original experimental setup

To count the on/off switches, a circuit relying on the current flowing through the lamps has been developed:

The small voltage  drop (about 4 VAC) on resistors R1 and R2 caused by the flowing current is rectified and smoothed through a rectifier bridge + capacitor, and will directly drive the coil of miniature relays. The self-powered counter-modules will increment when the coil goes from active to inactive state (i.e. whe current is switched off). If a lamp fails, there will be no voltage available to drive the mini-relay, so the counter state will remain unchanged.

A special hardware structure has been built and installed in a closed windowed cabinet at the LCD (Lycée Classique de Diekirch). A webcam delivers steady pictures of the counters, visible at  http://bmb.lcd.lu  (the webcam connects to the LAN through a Devolo powerline Lan adapter).

The following persons have contributed building the setup:

Marcel THILMANY (+Apr.2009):
Francis MASSEN:    wooden cabinet and miscellaneous hardware

Jean MOOTZ :          PIC circuit to deliver slow 0.5 min^-1 pulses

Francis MASSEN:   design and building of other bmb circuit boards

Here a picture of the setup, state 10 Jan. 2009:

The inclination of the counters can be changed, to avoid annoying reflections; this does not yet give a 100% reflection-free picture.

Two optical fibers are routed from the lamps to the sides of the counters to give a better visual clue on the web screen when lamps are lit (the fibers are the curve dark "cables" facing thelamps)

2. The lamps used

2.1. OSRAM 100 W incandescent lamp

This is a "white-soft" type 100W lamp, with a luminous flow of 1200 lumen. It is made by Osram France, part specification is "Bellalux Soft White, Bella T55 SIL 100, 230V E27/ES. The energy rating is  F  (best would be A, worst G). A scan in pdf format of the box can be found here.

2.2. OSRAM 20W dimmable DULUX EL

This CFL is made in Germany; it has a luminous flow of 1230 lumen and a specified lifetime of 15000 hours (!). It is dimmable, with the restriction that at most 4 lamps can be driven by one 500W rated dimmer. Energy rating is A. Part specification is: Osram Dulux EL dimmable, 20W/827, Lumilux Warm White, 1230 lm, 220-240 VAC, E27, 50/60 Hz. Look here for a scan in pdf format of the packaging.

Both lamps are practical identical in the specified luminous flow; efficiency is 12 lm/W for the incandescent lamp, and 61.5 lm/W for the CFL (about a factor of 5 higher)

3. Start of the experiment

The experiment started the 12 Jan. 2009 at 14:30 UTC (15:30 local time). It was interrupted after a few on/off  cycles, as one of the micro-relays was vibrating. The carbon resistor in the CFL circuit was removed and replaced by an adjustable  wire resistor, which solved the problem. Normal operation resumed at approx. 16:45 local time (15:45 UTC).

The timing of the on/off sequence is practically 58 s (switching frequency is 1/116 = 8.6 mHz ) : the lamps are in the on state for 58s, than switched off for 58s, etc...

4. The modified experimental setup starting 04 May 2010 (Jean Mootz, Guy Schintgen and Francis Massen)

The picture shows the two dimmer switch boards that replace the previous counter driver board. Two optical plastic fibers start from each lamp: the upper goes to the counter support for giving a visual clue when the lamp is ON. The lower fiber goes to the light dependant resistor of the dimmer switch board. When the lamp is ON, the relay of the dimmer switch board closes a contact, and when the lamp goes OFF again, this contact opens and increments the counter by one.

This new circuit solves two problems:

1. the (small) resistor in series with the lamp protected the lamp against sudden power surges during the powering up. This effect does not exist (at least not at this magnitude) in real lamp systems; the resistance of the connecting wires is usually much smaller than the 8 or 40 Ohm used in the original circuit

2. some CFL die ( = remain dark) but continue to draw an appreciable current, as a part of the electronics is not shut off. This current can continue to trigger the counters, even if the lamp is defunct. With the new system, counters remain stuck when the lamp goes definitively dark. Just to be sure, a time-stamped picture from the webcam is stored away every hour, so that inspection of what happened is always possible.

Using fibers to route the light to the back of the cabinet (where the LDR is mounted) guarantees that the dimmer sensor is not triggered by ambient light.

last modified:  07 Nov 2010