I'm looking to power 50 3W red leds as cheaply as I can.

I have a 100W LED COB that I powered with an old laptop power supply ([email protected]), connected to a 150W DC-DC step up. This is a constant current setup I believe, as I control the voltage from the DC-DC board. Also, I have a 1 Ohm, 10W resistor attached to the LED. This setup works well, and I only drive the LED at 60%, max.

Laptop PSU>150W DC-DC>Resistor>100W LED

Can I do something similar to power these 3W red LEDs? How would I modify it? The whole setup cost me $12 USD. I know fairly little about electronics, so I want to make sure I'm doing things correctly. Also, the "data sheet" on my LEDS says "Forward Voltage:2.2V~2.8V Forward Current: 700mA", and even without doing any math, this doesn't add up to 3W, it's more like 1.68W. After checking all the LED sellers, they all seem to do this. What gives? I've heard of Red LED voltage drop, but don't understand how it works.

  • \$\begingroup\$ Related: Why are these LEDs listed as 3W when the forward voltage is 2.4V, and forward current is 700mA? \$\endgroup\$
    – JYelton
    Sep 26, 2014 at 4:41
  • \$\begingroup\$ this is industry trying to make more cash on the weak minded... \$\endgroup\$
    – Gilad
    Sep 26, 2014 at 7:51
  • \$\begingroup\$ No, read the datasheet. It is 3W, usually with heat sinking, and the 700mA number is just a "test current" to get the measured forward voltage and listed luminous output. More than likely the typical rated current IS 1000mA at 3V forward drop, giving you a nice round 3W. Just because their listed parameters at the test current of 700mA doesnt mean it is not a 3W LED. \$\endgroup\$
    – KyranF
    Sep 26, 2014 at 8:43
  • \$\begingroup\$ As the current through an LED increases, the forward voltage consumed in the process increases rapidly, but it also works the other way too - if you adjust the voltage a small amount below the required forward voltage, you will see huge variations in current going through it. The best way to drive an LED is to use constant current rather than voltage control. I think you could search the internet for simple 1A constant current circuits and make a chain of these LEDs up to the expected input voltage from the 3V drop per LED. E.g if 15V available, do 4 LEDs and leave room for current ctrl \$\endgroup\$
    – KyranF
    Sep 26, 2014 at 8:49

2 Answers 2


In addition to Russell's answer, I'd like to provide an alternative which I used for my 3kW Strobe light I designed, which uses 24V and does 120A during the pulse, over 48 parallel 12W white LEDs.

In your case you would use a string of LEDs in series as Russell describes, but instead of the LM317 you may use a low-side N channel MOSFET and a "hardware" constant current feedback loop. The feedback consists of a "shunt" resistor and a NPN transistor and another resistor or two.

I found the picture I used as reference for my design, which worked fine. I had my limit at 2.3A per LED though, yours will be much less crazy. Lower current is better for heat issues anyway.

constant current LED driving using MOSFET

It is very simple, you merely calculate the voltage built up over R3 in the picture so that by Ohms law, your required current will generate 650mV-700mV, enough to turn on the base of the NPN transistor and therefore 'turn off' the MOSFET's gate.

R1 in the picture should be something useful, like 2.2-10K as a pull-up to charge the Gate.

Obviously make sure R3 is rated to handle the current going through it. For 1 Amp and to create 0.7V drop you can use a resistance of 0.7 Ohms, and it will need to be rated 1.5x above the expected dissipation just for good measure, so I^2 * R = ~1.5W-2W rated axial or chip resistor will work fine. I did all of my stuff surface mount, but in my huge strobe light I never went above 5-10% duty cycle in reality, which means I could have down-rated my components significantly, but I kept them as high power packages just in case.

  • \$\begingroup\$ That works well enough in this application. Minimum headroom is better than LM317 = 0.6V-0.7V range for Q1 whereas LM317 is 1.2v resistor drop plus 1 to 1.5V extra for LM317 itself. The above circuit would allow 7 LEDs per string from 19.5V. | LM317 is less exciting when things go wrong as Imax = 1A - 2A range. If things go wrong the above circuit can deliver whatever the power supply can send, for a while. BUT a very useful cct in this application. \$\endgroup\$
    – Russell McMahon
    Sep 26, 2014 at 13:41
  • \$\begingroup\$ Your LEDs are being 'slightly' overdriven ! :-) . But, you know that :-). What LEDs are you using. What is the application? Photography? \$\endgroup\$
    – Russell McMahon
    Sep 26, 2014 at 13:42
  • \$\begingroup\$ @RussellMcMahon My strobe light was my honours project in Mechatronics, designing the light with the intent to assist with robot vision (stereo vision) for day/nighttime (diurnal) navigation. A picture showing operation is here: wiki.qut.edu.au/download/attachments/157229075/… The guy sitting in the farm robot's seat is the vision guy Patrick Ross. My strobe flashes at 10Hz, so overdriving my LEDs with 5ms pulse was below the rating. Cree XML-EasyWhite 12V modules are what I used, rated at 1A. It is okay to pulse at up to 4A! \$\endgroup\$
    – KyranF
    Sep 26, 2014 at 14:45
  • \$\begingroup\$ So yes, I guess you could say it's for photography. Robotic photography :) The strobe trigger is a TTL signal, falling/rising edge trigger so it can be interfaced with all sorts of things. \$\endgroup\$
    – KyranF
    Sep 26, 2014 at 15:05
  • \$\begingroup\$ I won't go looking up XMLs now - 4:15am and need to be up earlier than nice. But 4:1 peak:rated max would be unusual with modern phosphor LEDs at any duration. I'm certainly not saying that what you are saying is incorrect - just that it does not match anything I recall seeing for this type of LED. My recall and my completeness of survey may well both be wanting :-). \$\endgroup\$
    – Russell McMahon
    Sep 26, 2014 at 17:13

Here is a constant current drive solution using the PSU only.
Strings of 6 LEDs in series. MAYBE 7.
19.5V - 6 x 2.8V = 2.7V headroom = marginal.
If VLED is more like 2.2V then MAYBE 7 LEDs OK - try and see.

LM317 per string.
Use TO220 version or equivalent thermally.
Use 19.5V supply directly.
V+ to LM317 Vin

LM317 will act as CC (constant current source with R across Vout- Vref and output from Vref. Icc ~= 1.2V/R.

For 700 mA
Icc from LM317 = 1.2V/Icc = 1.2 / 0.7 = 1.71 Ohm-> use 1.8 Ohm. Change R for different Icc

1.8 Ohm resistor from LM317 Vout to Vref
R = at least 2 Watt rated - preferably 5 Watt LED drive from Vref (NOT Vout).

6 x 8 = 48 so two short. Could use 4 x 5 LEDS + 5 x 6 LEDS. For 5 LED strings add series R = V/I = 2.5V/.7 =~ 3.3 Ohm resistor 5 Watt rated.

LM317s will dissipate =~4W so need good heatsinks.

Anyone is welcome to add a diagram.
Time vanishes ...


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.