The little LEDs from the first project just weren’t bright enough, so I set about looking for something more substantial. I found them on eBay for just under a pound each. They’re called COB LEDs (Chip on Board Light Emitting Diodes). COB means that there are several (e.g. 9) individual LEDs in the one device. I chose 10 Watt LEDs because they have a similar brightness to an old-style 60 Watt light bulb, which seemed about right.
They are rated as follows:
* Forward Voltage (VF): DC 9-12V
* Forward current (IF): 1050MA
* Output Lumens: 800-900LM
* Colour temperature: 6000-6500K (Cool White) 2500-2700k (Warm White)
* Beam Angel: 140 degrees
* Life span: >50,000 hours
However, those ratings are misleading. You can’t just connect them to as 12 volt battery and expect everything to be ok. It won’t. Well, not for long anyway. There are two (related) things that need to be controlled: current and temperature. Let’s start with current.
The amount of current a device takes depends on the voltage you give it and its own resistance. If the LED were just a resistor then it would be easy. Resistance (R) = Voltage (V) / Current (I). So, R=12/1.05 = 11.4 Ohms. The problem is that applying 12V will give you a higher current than that, and as the LED heats up its resistance decreases, so it takes more current, so it gets hotter, so it takes even more current. The maximum temperature of the chip is 150 degrees centigrade; that’s the point at which it melts. If you’re lucky that means it switches off, if it’s not a good day then it might catch fire.
In fact, what the specification means is that you need to keep adjusting the voltage all the time, between 9 and 12 Volts, and make sure the current never exceeds 1.05 Amps. We’ll come back to how to do this in a while, but the other thing that needs to be controlled is temperature.
An ordinary halogen spotlight will get hot, but it doesn’t need any special cooling because a lot of the heat is radiated as infra-red, along with the light. (If you put your hand a few centimetres in front of it, it feels hot, whereas these LEDs do not.) Although an LED is far more efficient than other forms of lighting, they still convert around 70% of the electricity into heat. If the LED is not going to just get hotter and hotter then that heat has to go somewhere, and that somewhere has to be the air, through a piece of metal to which the LED is attached (i.e. a heatsink).
Choosing a heatsink is, in fact, a non-trivial problem. There are some calculations that will help, but in the end it depends a lot on where the LED/heatsink is to be housed, what the orientation of the heatsink will be, how well the two are attached, how efficient the particular LED is, how hot you’re willing to let it get, and probably other factors as well.
The calculation is based on a value given for the heatsink: Thermal Resistance. This is the number of degrees Centigrade temperature rise to expect for each Watt that’s turned into heat. So, for our 10 Watt LED, if it were to be 30% efficient (quite an assumption!) then it would dissipate 7 Watts as heat.
This is the heatsink I chose. I found it on the internet at RS Components. It’s actually quite small – 50mm x 21 x 19, and it’s rated at 8.5 degrees Centigrade per Watt (8.5C/W or 8.5K/W). So, given a hot day, with a room temperature at 30 degrees, the heatsink will run at 30 + (7 x 8.5) = 90 degrees. That’s hot enough to burn your skin. And the calculation assumes the heatsink is mounted vertically in free air, i.e. everything optimal. Well, it’s not. I tried it at maximum current; the temperature rose slowly, but it reached 90 degrees after half an hour, and it was still rising. Also, remember that I was measuring the temperature on the outside of the LED; the chip itself will have been hotter. (Actually, I turned it off at that point because it started making a strange crackling sound. Not very loud, but a little unnerving.)
So, we have a problem, or rather, a potential problem. Remember that we are fading between five LEDs, so any one LED will be on for 20% of the time. So in theory these heatsinks are fine. The problem would be if for some reason there’s a fault in the software, or the Arduino, or an electrical fault whereby one of the lamps is not turned off. How to be sure it will never overheat? The solution I chose was to run the LEDs at less than full power. 60% should reach a temperature of 65 degrees, or 60 in most conditions, which seemed to be the case when I tried it. That’s still hot, but I think it’s within safe limits.
I mounted an LED on the back of the heatsink – the flat part common to all of the fins. In the final design, the heatsink was screwed to wooden fixings, which reduced the airflow. This problem was created because the electronic design and the frame design were not coordinated, something to improve on next time. In use, the LEDs just get pleasantly warm. It’s very unlikely that one LED will get stuck on full for an extended period – it would probably require a hardware problem with the Arduino or in the wiring. When I was testing the heatsinks their 65 degree maximum only increase a few degrees when they were on their side so I would expect that they will not reach the 90 degrees that seemed problematic.
Sixty percent is just over 600 milliamps (mA). As we saw, the answer to “what voltage do I need for 600mA” is “that depends”. In fact, the answer for my particular LEDs was approx. 9.4 Volts when they’re hot and 9.8 Volts when they’re cold. Now I could just use a fixed 9.4 Volt power supply, but that’s not good practice. Firstly, because the light output would vary with temperature (they could even be seen to “warm up” in some circumstances), and secondly because when I say “hot” I mean after a few minutes. If they were left on for hours with restricted airflow they would need a lower voltage to keep to the same current.
The answer is to use a constant current power supply, that is, it adjusts the voltage until a specific current (in this case 600 mA) is flowing. So, whatever you connect will – within limits – have 600mA pushed through it (whether it likes it or not!). I decided on a simple circuit, where a transistor is used as a variable resistor in series with the LED. The transistor adjusts its resistance until the correct current is flowing. In this instance, the voltage range was from just over zero volts to a little under the supply voltage (because the transistor and current-sensing resistor drop a few volts). So, about 0-9.6V with a 12V power supply.
See the problem? 9.6 is less than 9.8, so the LED will be running slightly dim when cold. Being a perfectionist, this isn’t good enough, so I found a 13.2Volt power adapter that I had floating around and used that (so now it’s 0-10.8V). Here it is running a couple of 5Watt car bulbs in parallel at 600mA. (Note that the digital meter is there to measure the temperature of the transistor – the clothes peg is holding the sensor in place – but I hadn’t set the dial for that when I took the photo, so the number displayed is not temperature.)
This circuit worked just fine, but I decided to change it before building the actual project, but that’s for another post.