I have started a project for my home lighting, I'm trying to build a DC LED module which would change the color temperature from 3000K to 1800K as dimmed with reducing current. I'd like to do it mostly passively without any current sense or threshold circuitry. I also want it to run on my current TRIAC dimmable constant current driver.

The most sensible way I could find to do it is to use two different temperatures of LEDs, route all the current to warm array in the beginning of dimming range, later ramp up the current of the cool array for the rest of the dimming range. Maybe, preferably, completely turn off the warm array at the end of the dimming range.

Firstly, I'd like to share my current "progress" with you, if you can call adding a resistor as such:

I was able to come up with a very simple circuit to do it, I kept the number of warm LEDs in series low, thus the total forward voltage, put a series resistor to them, then parallelly connected the cool array with higher number of LEDs in series. The LTSpice schematic and DC current source sweep analysis results given below:

LTSpice schematic of the resistor limited dim to warm circuit

DC Sweep analysis results

The red curve is for the cool array, green for the warm array.

However as you might have quickly realized, the series resistor dissipates up to 800 mW of power at the maximum brightness, and also the warm LEDs stay on at the end of the cycle(not really a big issue though). For a total of 4.8 Watts power consumed in the circuit, 0.8 W is directly wasted on the resistor.

I've seen some other Warm-to-Dim passive DC LED modules online, some add a transistor, some just use the resistor approach. Here is an example of such arrangement with transistors, but i wasn't able to figure out the circuitry from the pictures:

Commercial DC Dim-to-Warm product

I can just speculate that the transistor assisted circuitry is much more efficient since the transistor or MOSFET just switch current instead of dissipating the unwanted amount.

I failed on designing some transistor or MOSFET circuits, I tried to sense the current on the warm LED branch, then feed it to the transistor so the cool branch could turn on after some set current.

Maybe me being an amateur hobby electronics guy, originally a mechanical engineer, couldn't see the obvious thing they've done there, or couldn't come up with a very simple transistor arrangement to achieve it.

Can you recommend me a circuit that:

  • Use minimum number of components even at the expense of performance
  • Doesn't waste as much current as the resistor circuit I was able to build
  • Have it's crossover current adjustable with the value of a resistor
  • And possibly have the warm LED branch completely turned of at the maximum current level?
  • \$\begingroup\$ I have done this with an array of Blue, Red, Orange , White LEDs with an adjustable DC supply and string voltages from 8 to 20V with currents calculated with passive R controlled strings and strings of 3W White LEDs. It requires careful calculated V,I, power levels and not a simple solution. Your best bet is to buy the CREE smart color Lights. \$\endgroup\$ Commented Jun 10, 2020 at 22:48
  • \$\begingroup\$ e.g. DC voltage dimmer (using 65W laptop source and custom LEDs not 12V strings.) [1]: i.sstatic.net/gojJL.png [2]: i.sstatic.net/H2VMa.jpg \$\endgroup\$ Commented Jun 10, 2020 at 22:56
  • \$\begingroup\$ @TonyStewartSunnyskyguyEE75 i plan to use custom LEDs, not 12v strips. The strip was an example to the passive Dim-to-Warm approach with transistors. I'm okay with carefully measuring the forward voltage of the LEDs and adjust the resistor value according to it. The commercial products that passively dim to warm on DC current adjustment are not available where i live and i want to do it myself. Your build looks nice by the way. \$\endgroup\$ Commented Jun 10, 2020 at 23:38
  • \$\begingroup\$ Ha! Yours truly actually have a patent for the very circuit you are proposing using a linear ramped current source. I hope you are not planning to put this to market. \$\endgroup\$
    – winny
    Commented Jun 11, 2020 at 8:35
  • \$\begingroup\$ @winny definitely not, it's for personal and hobby uses. I even looked up for patents to copy if such a thing exist. If you know such a patent you can share it with me! \$\endgroup\$ Commented Jun 11, 2020 at 9:33

2 Answers 2


1) Start by defining your ideal White light.

  • For me, it might be 4500'K white 150k lux 40W to simulate bright sunlight Perhaps you can get this using an array of Philips dimmable 8W PAR30 White bulbs.

2) Then define or simulate the power or lux level and colour spectrum of a sunset using a mix of orange and/or red and dim white and maybe a blue horizon.

3) Then consider a neutral white (4500~5000'K) for your base with the constant sunset lights to warm up the blended light.

  • Then experiment with dimming the white only to capture the effect of sunset, such that the blue red orange gradients only tint the white warmer.

    • This only works well with blended or indirect light sources such as what I tried with ceiling egg grating painted flat black.

    • It could be perimeter ceiling light with a valence. If normally illuminate an area with 300W Halogen indirect light (e.g. floor lamp) , consider 25% of this for LED power. You can get more or less better efficacy, but you are going after special effects.

This was the least technical answer I could make for a non-EE.

i.stack.imgur.com/gojJL.png enter image description here

I modified a Universal Laptop charger with a 4 wire connector and a few parts with a pot to adjust the voltage 3 to 35V with LED indicator in a small box lower left instead of a fixed 19.5V 65W.

But I am suggesting "dimmable" PAR (parabolic reflector) Edison-base AC bulbs to match your desire using a Triac control. Dimmable LED bulbs are pretty common not but none of them match, so I get bulk qty in a box in case of failures.

For the Orange, Red emitters Blue, you might be better off using RGB remote strings avail on Banggood , Ebay etc.

  • \$\begingroup\$ p.s. I do own a Lux meter and also have CIE filtered Panasonic Light Intensity sensors AMS302T (? obsolete or Mouser or grey market) but did not need for my app. If it were a commercial design, then I would use testable specs. mouser.com/catalog/specsheets/panasonic_napica-catalog.pdf These are great inexpensive eye-corrected spectrum light intensity sensors that run off an L7805 reg and gain resistor. \$\endgroup\$ Commented Jun 11, 2020 at 1:00
  • \$\begingroup\$ Thanks for your detailed workflow, as you said, i defined my working temperatures. I want to simply crossfade between a group of 1800K LEDs and another group of 3000K LEDs. Your simulation of sunset looks dope, but i really don't need that much of a precision. I'm just wondering how do people achieve this dim-to-warm effect on LED modules(COBs, strips, bars, etc.) that only have two leads. They are dimmed to warm using only a single channel variable current LED driver, with only two wires. The current balancing between the warm and cool leds are done by the passive components on board. \$\endgroup\$ Commented Jun 11, 2020 at 1:02
  • \$\begingroup\$ As i stated in the question, i reached my previous goal of dim-to-warm build, but it's inefficient. Maybe as you said i'm just better off with accepting that inefficiencies. I can somehow manage to adjust the circuit to achieve acceptable dimming temperature changes. \$\endgroup\$ Commented Jun 11, 2020 at 1:06
  • \$\begingroup\$ low efficiency is OK . but overtemp is not OK. Normally % PWM is used to dim more efficiently. Without measurable specs, I can only do handwaving suggestions \$\endgroup\$ Commented Jun 11, 2020 at 1:16

Here is how I did it in the end,

I had to give up on using only passive circuitry, also had to use a current sense resistor. Put a LM317 regulator in series with the warm array, in current regulating configuration. I modeled the actual LEDs I wanted to use according to the I-V curves given in their datasheet, used an LED modeling guide for SPICE from this answer. I've chosen the current sense resistor as 15 ohms.

Here is the final schematic i used: enter image description here

Here is the power delivered to warm array(green line) and cooler array(red line) with respect to the total current, the current sense resistor power dissipation isn't included in the graph but it maxes out at 110 mW: enter image description here

And this is how I implemented it physically. Excuse my unaesthetic solder work because I had to put everything on the same board, I did some scraping to create new pads: enter image description here

It works just fine, the temperatures of the components are fine, nothing is too hot to touch for less than 10 seconds. The LM317 is well heatsinked by the aluminum PCB, and its maximum power dissipation in this circuit is 1W.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.