Hacking A Non-Dimmable LED Fixture

For most of us, the solution to having a non-dimmable LED light bulb but needing a dimmable one is a simple as a drive to the store to get the right kind of bulb. But that seems downright boring, not to mention wasteful, so when [Leo Fernekes] was faced with this problem, he looked for a way to make a non-dimmable bulb dimmable.

To be fair, there was a financial aspect to this hack, too. [Leo] had a bunch of cheap non-dimmable light fixtures he wanted to put to use. He started with a teardown and reverse-engineering of a light strip, which contains little more than LEDs and a small buck converter. His analysis of the circuit led him to a solution for dimming the light: inserting a MOSFET as a shunt around the LEDs. That and the addition of a diode to isolate the LEDs from the current regulator would allow for simple PWM-control of the lights via a microcontroller.

As is typical with these things, there were complications. [Leo] found that a timing problem resulted in flickering LEDs; the fix came from adding a sync circuit that cleverly leveraged a flip-flop inside the PIC16 microcontroller he chose for the circuit. His prototype incorporates these modifications, plus an interface that supports the DALI protocol for architectural lighting control. As always, [Leo] is quick to point out that mixing line voltage into your projects is not without risks, which he takes pains to mitigate. And as is also typical for his projects, [Leo] gives just the right amount of detail to understand the theory behind his design.

22 thoughts on “Hacking A Non-Dimmable LED Fixture

  1. I did not watch the video yet, the SO is watching tee vee, but it seems to me if you put a shunt across the LED’s you are just turning the power you are shunting into heat and bypassing the whole reason using LED’s in the first place?

    1. It’s not making the setup use any more electricity than it originally did (excepting that of the additional control circuitry). However, it’s not making it use any less when it is dimmed.

      I kind of wonder if it might be easier to make a whole new LED supply instead of having to reverse engineer and hack into each unit. The main variables across different LED light fixtures is how many LEDs are in series or parallel.

    2. It’s a constant-current driver, so the power it’s delivering drops as the load voltage drops. Drop the load voltage to nearly zero by shunting with a low-resistance mosfet, and the power drops quite a lot.

      1. It doesn’t work that way. When you drop the load voltage, the voltage across the constant current regulator increases and the power is dissipated in the regulator instead. This is because the supply voltage is constant.

        What the circuit appears to be doing, it shunts the bridge rectifier filter capacitor to ground and basically starves the regulator so it can’t pass as much current, which causes a huge ripple in the DC suppy and will probably kill the capacitors in a short order.

        1. The schematics on the website are a little different from the connection shown in the video, but the same point applies. It bypasses the whole capacitor and LED circuit by drawing current straight from the low-pass filter through the regulator.

          This has the effect of dumping the charge in the smaller 2 uF capacitor through the MOSFET instead of the LEDs. Now since the larger tank capacitor is no longer in the loop, the filter capacitor is seeing a whole lot more ripple, and as I said before, will probably burn out after a while.

        2. > When you drop the load voltage, the voltage across the constant current regulator increases and the power is dissipated in the regulator instead. This is because the supply voltage is constant.

          That’s how a linear constant-current regulator works, but this thing claims to be a buck converter and has a big ol’ inductor to back that claim up. I’ll allow that I can’t find more than a cover page for a datasheet and that I’m never going to slog through 20 minutes of video blather for the sake of what should be one minute of reading, but I can’t see much point in a buck converter if it just ends up acting like a linear regulator.

          1. At full power it’s behaving just like a normal buck converter, with its associated efficiency.
            Shunt fet dimming is surprisingly common for dimmable LED lighting, although with modern LED drivers it’s finally beginning to get eclipsed by having dedicated dimming circuitry in the driver itself.
            Shunt FET dimming is capable of very high frequency dimming, which many LED drivers can’t handle: they have a startup time measured in dozens or hundreds of milliseconds to do things like charge the gate driver charge cap, for instance, and a lot of them didn’t used to have separate enable circuitry that was capable of handling KHz dimming, which a shunt fet can do with ease.
            Shunt FET dimming is also capable of a much greater dynamic range of dimming, which is a consequence of the high switching speed it can handle. A lot of older drivers might only have a 20:1 dimming range by driving their enable pins with a PWM. Shunt dimming can manage 1000:1 without a problem.
            Newer drivers do cycle by cycle adjustment without any problem so they have good dynamic range via enable pins.
            Here’s a pretty good explanation of LED dimming strategies. (Written by my coworker/ex-manager, I guess I should note.)

  2. BigClivedotcom did a couple videos (youtube) on the Dubai lamps, LED bulbs that are underdriven to reduce heat, increase lifespan and reduce consumption. He used small ceramic capacitors in series with the input voltage. It worked quite well for a number of bulbs. I now have some setup in my house this way.

        1. Take a clear yellow LED and hook it up to oscilloscope leads, then point the LED at the lamp at close range. It acts as a solar cell and produces a voltage pulse proportional to the brightness of the lamp.

          You can’t see the flickering head on because of persistence of vision effects (stroboscope), but it creates a sort of jittering effect around it when the flickering bulb is the only source of light in the room, or when all the lights are flickering at the same rate. It’s most noticeable in the peripheral vision.

  3. I found a power brick for 42v regulated and thought it might come in handy. Then I stripped a flat screen of it’s cracked LCD and found it’s LED’s to be 4 separate strings with 4 sinks in a chip I looked up, constant at 350mA per string. About 67 volts. So the bench supply up to 15v and that 42 volt in series all 4 strings tied together. The brick plus the 1.25 to 15v variable turns out to go from moon light panel to about 2/3 of original current which is a thin panel under the overhead shelf above the workbench. Way better than the 15 watt florescent glass transmitter over the place before. 1 cm thick and light photographers lust for, shadowless even light.

    A pot for the thrown together wall wart 12v dimmer works over a range without much heat waste, the worst comes from the 317 regulator as the dimmer. DC to daylight no RF flicker or triac buzz or power phase transients.

    Since I run it low most of the time I figure it ought to last, the monitor it once was fell when new. The aluminum strip is in air not a case and runs warm at most, cold when matching lumin in that normally darker area than well lit around in front.

    1. Want to see something really impressive: look at the technology it takes to draw tungsten wire 0.04mm in diameter, blow a glass bulb, and seal the wire through the bulb so there aren’t any leaks.

  4. That seems like a lot of work, but is always fun to transform something not inherently amenable to adaptation to do what one wants. That is a cool demonstration and analysis of the problem of needing to prevent flicker. The last place I worked had expensive commercial LED fixtures that had flicker when dimmed. This goes the pros one better.

    Having said that, I did a quick look on the web and found a single channel LED driver module with a DALI control input for $9.98 US. Of course, who knows who much quality is in that ten Dollar module,

Leave a Reply

Please be kind and respectful to help make the comments section excellent. (Comment Policy)

This site uses Akismet to reduce spam. Learn how your comment data is processed.