I just received an almost identical lamp from Amazon.com.
Globe Electric 5240101 Fluorescent Desk Lamp, Black
The lamp is identical to the one I demonstrated in the video, except that it is black. This lamp body is ideal for the LED conversion and it costs under $15 plus shipping. I like its compact footprint and the fact that it can be folded or rotated to shine light in any direction. I bought the last one they have in stock, but Amazon says that more are on the way.
I have a friend who is a member of a model railroading club. He wants to replace hundreds of small energy-hogging incandescent bulbs with LEDs. He is currently using triac-based dimmers. As long as he uses an isolation transformer or completely insulates all wiring, my circuit will be a safe and simple solution to his problem. I'm going to help him with that.
This circuit is very clever, and I enjoyed deciphering how it works (although I have not built it). However, I wonder about the driver for the second string of LED's. As you mention, different color LED's have different voltages. So if the "second string" voltage is greater than the primary string, then Q3 will be saturated and you basically will have two strings of LED's in parallel, and your current regulating resistors will also be in parallel (110 ohms total). On the other hand, if the secondary string has significantly less voltage either due to the rating of the LED's or to a lower number of LED's in the string, then you will have a significant collector-base voltage across Q3. It's only rated to 40 volts, and you could easily exceed the rated power dissipation if the C-E voltage reached around 30 volts. There's well over 100 volts available from the power supply, so I think this is a problem. Perhaps a 2SA821STPQ-ND would be a better choice (210 volt VCE, 200mA capability).
You're absolutely right about the transistor, but you need to balance the voltages in the strings to keep a low voltage across Q3. If you do that, then the specified transistor is fine. Also, if you don't do it, Q3 will unnecessarily waste power. The primary string must always have equal or greater voltage than the secondary strings. Also, the 220 ohm current limiting resistors will never be in parallel, as only the resistor in the primary string is used to set the primary LED current. The secondary string will have its own seperate current sense resistor. The circuit in figure 3 of the article was used as the breadboard in the video.
The LED driver circuit in this article is an adaptation of a voltage regulator circuit that I created years ago to fix an overheating problem in a cheap Chinese-made stereo set. If interested, see the link below.
Much more recently, I found similar but different versions of the concept in three "EDN Design Ideas" articles, by different authors. I have PDF copies, but no links, if you're interested (downloaded from the site when it was free). Two of them were used to power LEDs, but neither circuit automatically adapts to the number of LEDs used, nor would they accommodate precision dimming. Mine is the only version that takes the output off the source instead of the drain of the MOSFET. This allows switching the high side of the load, with the more powerful N-channel MOSFET (i.e. half the on-resistance of a similar sized P-channel MOSFET). However, this requires a bootstrap circuit (R2, D6, D7 and C2) to provide the turn-on bias voltage for the MOSFET. It is the high-side switching that allows the precise dimming and the automatic adaptation to the number of LEDs used, plus it does not require dropping high voltage across an energy-wasteful resistor.
armorris, What a cool project. I find it to be quite simple in design and its very efficient (5W). I've worked on a lot of LED lighting fixtures and they seem to be over complicated with microcontrollers and switching mode power supplies. I'm really amazed at the PMOSFETs with no heatsinks. Its amazing what you can do with PMOSFETs. Very nice project!!!
I'm going to have to build one (or more) of these.
You state a minimmum of around 7 LEDs, so I figure around 21V min.
I have some reels of 12V prewired series/parallel string surface mount LEDs (3 chips and a resistor, every 50mm). If I cut some into sections, and wire 2 or more 12V sections in series, I'm guessing they should work with some triming of your circuit values. Other than the extra losses in the built-in series resistros, do you see any problems with this?
The final showdown is under way in our first-ever Gadget Freak of the Year contest. Who will win an all-expenses-paid trip to the Pacific Design & Manufacturing Show? It's up to you, dear readers, to tell us.
For industrial control applications, or even a simple assembly line, that machine can go almost 24/7 without a break. But what happens when the task is a little more complex? That’s where the “smart” machine would come in. The smart machine is one that has some simple (or complex in some cases) processing capability to be able to adapt to changing conditions. Such machines are suited for a host of applications, including automotive, aerospace, defense, medical, computers and electronics, telecommunications, consumer goods, and so on. This discussion will examine what’s possible with smart machines, and what tradeoffs need to be made to implement such a solution.