The test data requirements defined the necessary test setup.
- Turn-on time: Requires luminosity (lux) and time (milliseconds) measurements.
- Warm-up time: Requires luminosity (lux) and time (seconds) measurements.
- Dimming performance: Requires luminosity (lux) and input power (watts) measurements.
To measure light output I used the TSL2561 luminosity sensor produced by TAOS (Texas Advanced Optoelectronic Solutions). I bought the light sensor from Adafruit with the surface-mount light sensor on a breakout board for easy interfacing.
The reason for using this light sensor instead of an ordinary photodiode is that you can obtain measurements of the usable light (for humans) that make up the lux measurement. A photodiode usually measures a broader spectrum than is usable for visual illumination by humans and does not allow computation of the spectrum profile needed for lux measurements. The TSL2561 uses both a wide spectrum photodiode and a narrow spectrum infrared photodiode to allow computation of the true lux measurement (see figure below).
The difference is significant when comparing the light from an incandescent lamp where the IR radiation is about 3x the visible light against an LED or CFL lamp where the IR radiation is around 25% of the visible light.
Adafruit provides the circuit and software information you need to make your own luminosity measurements using an Arduino. For my tests I used a USB Boarduino (also available from Adafruit) which is functionally nearly the same as an Arduino but is easy to use with a solderless breadboard. I also wrote additional software to provide a menu-driven interface for simplify making the measurements and recording them.
To accurately measure light output versus time requires a triggering system to start the timing measurement when power is applied to the lamp circuit. For the trigger circuit I used an AC to 9VDC linear wall adapter with the resistor divider trigger circuit shown below. The wall adapter is plugged into the same power strip used to switch power to the lamp being tested. When you switch on the lamp the circuit generates a trigger. The signal from the trigger circuit is used as an input to digital pin 2 of the arduino or USB boarduino. Other input pins could be used but pin 2 is convenient for using a hardware interrupt for triggering if desired. Because my circuit does nothing while waiting for the trigger, I currently just poll the signal level every msec until the signal reaches the trigger level.
If you use this circuit, be sure to measure the output of the wall adapter first to verify the output voltage. Unregulated dc power supplies will put out much higher voltage than they specify on the label when driving a light load (the trigger circuit). Adjust the resistor divider to limit the maximum voltage to be at or below the digital logic supply on the arduino or boarduino.
I tested the linear wall adapter and resistor divider circuit with a digital oscilloscope and found the signal reaches the trigger level within 10 to 15 msec after switching on AC power. A switching power supply (wall adaptor) I tested took about 40 msec.
Power measurements for the dimming tests were made using a P3 Kill A Watt meter shown above.
Although inexpensive and convenient, these meters only provide single watt resolution. Because many of the lamps I test are in the 10-20 watt range the signal digit resolution is too imprecise. This is especially true when dimming down to one or two watts.
I also did not have a way to feed the power measurements electronically into my other measurement data. I had to record the data manually and then add it to my other data in the spreadsheet. I plan to develop a method both for improved resolutions and electronic reading of the power measurements in future tests.
For these tests I used a cardboard box to eliminate other sources of light from the lamp tests. The cardboard box isn’t very classy but will enable accurate and repeatable measurements provided the lamp being tested and the luminosity sensor don’t change position. To hold the lamp securely in position, a ceramic lamp based was mounted on an 8 inch diameter sheet metal end cap you can see in the above photo. This had the added benefit of providing heat sinking when testing the higher power incandescent lamps. During testing the box is closed.
The lux measurement, which stands for luminous flux, is a measure of lumens/ square meter. Because the luminous flux depends on the distance from the lamp, the angle of the light (altitude and azimuth), and the characteristics of the lamp design, a single lux measurement tells nothing about the total lumen output of a particular lamp or how it compares with another lamp. The single lux measurement should be proportional to the total lumen output for the particular light as long as the light, sensor and surroundings remain unchanged. So a reading of 50% of the full power lux measured when dimming means the lamp output is 50% of the full power lumens.
Turn-on time test
For the short period turn-on measurement, I set the light sensor integration period to 13 msec, record readings every 25 msec until the data shows the initial turn-on time. The fast integration interval results in some loss of resolution for luminosity measurements due to quantization errors, but this is not important for the turn-on time. The triggering circuit initiates the timing measurements when power was applied.
For the long-period warm-up measurement, I set the light sensor integration period to 101 msec, and record data at 1 second intervals until the light output has stabilized. The trigger circuit was used to initiate timing measurements.
Dimming measurements are made with the lamp fully warmed up. For these measurements I use a 101 msec integration time and manually control the data sampling when I have the dimming power level stable. I always take two readings at each power level to watch for drift. This only happens if the light has not fully warmed up.