0
\$\begingroup\$

I'm making an edge-lit snowflake out of acrylic. It's lit from the center; the LEDs are mounted straddling a circular (40 mm diameter) PCB.

I'm using 24 white LEDs, and since I don't want to work with huge voltages, my plan is to divide them into four 6-LED circuits, driven by a constant current source from 24 V, 20-30 mA, depending on how the thing will look like when fully it.

My original plan was to use through-hole LM317s, but now that I'm laying out the PCB, I've been tempted to use surface mount parts to keep the front-facing side nice and spotless. So far I've been able to find LM317 in SOIC-8 and SO-223, and it's a bit of a tight fit with those. I'm soldering with an iron so I'd like to keep the design as roomy as possible.

Are there smaller adjustable regulators or constant current designs (e.g. SOT-23) that would fit better?

\$\endgroup\$
10
  • 1
    \$\begingroup\$ Is your master 24V source well-regulated? If so, consider using a single series resistor on a 6-LED string for crude current regulation. 1/8 watt resistor might be sufficient \$\endgroup\$ – glen_geek Dec 8 '20 at 18:28
  • \$\begingroup\$ The linear regulator will be dissipating about 300 mW so going too small may be a problem in getting rid of heat. \$\endgroup\$ – Andy aka Dec 8 '20 at 18:34
  • \$\begingroup\$ Oops, left out a detail: I plan to have a regulator per circuit to account for Vf variation and so that power per regulator isn't too great. That's why space is an issue, fitting one regulator would be easy. And no, I don't trust my 24 V power brick that much. \$\endgroup\$ – lrasinen Dec 8 '20 at 18:37
  • \$\begingroup\$ Measure it !!! Report it !!! if 24V +/-2% with No load, then you can trust it \$\endgroup\$ – Tony Stewart EE75 Dec 8 '20 at 18:43
  • 1
    \$\begingroup\$ Probably resistors are fine with about 6V across them especially if it's a switching adapter, but even if it isn't you can measure the voltage and tweak the resistors. Current regulation for something like this smacks of overkill and might make it less reliable. A few 1206 or 2512 resistors seems like a much more robust solution. \$\endgroup\$ – Spehro Pefhany Dec 8 '20 at 18:46
0
\$\begingroup\$

It is best to match the LED String to your regulated Supply or visa versa using any SMPS with <2% error.

  • Each LED is like a 3V Zener typically Vf=3.1 to 3.3 (20 to 30mA)
  • You can use a string of 7, while 6 is less efficient
  • Each 20mA White LED is approximately Vf=2.85[V] + If[A] * 15[Ω] (more good quality modern HB 5mm LEDs, not old ones) i.e Iv> 10,000 mcd
    • high power LEDs have a lower Rs ~ <= 1/Pmax

Thus using only a series Rs to regulate current 24V since both LEDs Vf and 24V are stable.

for V+=24V

  #LEDs If[mA]  Vf[string]  Rs  P(Rs)    
    6    20     18.9V      255   100mW~  (use 1/4W R per string)
    6    30     19.8       140   130mW~
    7    20     22.1       100    40mW~  (1/8W is OK)
    7    30     23.1        30    27mW~   (1/16W is OK to use)
    8       est. 24.05/8=3.00V > - 2.85V = 150mV > / 15 ohms = 10 mA  
            roughly  YMWV With Rs=0
\$\endgroup\$
0
\$\begingroup\$

I am not so sure why you can't include the regulator in the supply. For small currents you can make constant current regulators with a couple of transistors (even two small BJT if your supply voltage is close enough to the total voltage to supply). Two BJTs and the two resistors required to bias and set the current can be soldered together without PCB and sealed in epoxy or silicon in a 'knot' along the cable (it will be smaller than a ferrite bead).
If you wish you can do it with SMD components, it won't take up much space.

If you arrange the LEDs in four rows of six devices you will have something like this

constant current knot

(here I used PNPs because I had them lying around - I adapted an old simulation - but you can change the design to NPNs by substituting the transistors, inverting the polarities of supply and LEDs).

Current and voltage for a single LED, along with the control voltage across the sense resistor:

Current and voltage for the LED

I also stepped the temperature from 15 to 30 degree celsius. The current is reasonably stable (is the star outside, in cold weather? Might want to run a simulation at the operating temperature.)

The simulation shows that if your supply is around 24V, there is very little power dissipated in the circuit (only the supply and the LEDs will have a significant contribute):

power dissipated

If you keep the circuit outside the star, you could also use a MOSFET for QD, but for 80mA, the small drive required by BJT seems the better choice. (Also, 20mA might be overkill - I find high efficiency LEDs extremely fastidious at their rated currents, but that's me).

\$\endgroup\$

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