I am about to design a constant current LED driver, with the following requirements:

  • The LED string is fed with 24 V
  • The FPGA interface is 3.3 V
  • The system consist of 8 LEDs in a row, with corresponding 8 sensors.
  • The LED current should be defined in 256 steps (in this case with a digital potentiometer => current is defined in 3.3 V / 256 steps)
  • There is a PWM signal that controls whether the LED should be on (when the sensors are measuring) or the LEDs should be off (the PWM is NOT used to determinate the LED intensity, only when to measure)
  • The PWM signal can either be always off, always on, or between 10-40 kHz with 50% duty cycle.
  • The forward voltage of the LED is 1.4 V
  • The max constant current (PWM is always on) is 100 mA and if a pulsed PWM is used the max current is 200 mA.

Any buck/boost solution is not applicable due to the ripple in LED current, which introduces ripple in the emitted light. This is the reason to use a constant current driver, unless someone has an example of a very good buck/boost driver, with less than 1% ripple.

  1. Does anyone have a suggestion of an "out of the box" LED constant current driver chip, that will do the job?

    Because the LEDs have forward voltage of 1.6 V and there are 8 in a row, the voltage drop is 12.8 V, the constant current driver must burn away the rest of the energy and with 0.2 A it means 2.24 W.

  2. Is it a good idea to insert like a 2 ohm, 1 W resistor in series with the LEDs, to take away some heat from the transistor?

    One attractive idea is use op-amp->transistor where the op-amp is fed with the voltage setting the current and the voltage from a shunt resistor below the transistor. The op-amp needs to be fast, to avoid problems with a MOSFET gate.

  3. Is it better to use a MOSFET or BJT (they are more linear)? They have to work in their linear area.

  • \$\begingroup\$ Try using a simulator to compare ideas. \$\endgroup\$ – Andy aka Apr 27 '20 at 11:20
  • 2
    \$\begingroup\$ Welcomet to EE.SE! Please add a schematic of your circuit and your proposals. If you click on edit and the schematic symbol, a schematic editor will open and let you draw one. \$\endgroup\$ – winny Apr 27 '20 at 11:32
  • \$\begingroup\$ Simulation or not, I would still be constrained of my own fantasy and number of ideas, the aim of the post is to get more ideas. \$\endgroup\$ – StefanTh Apr 27 '20 at 12:26
  • \$\begingroup\$ What kind of settling time and flatness do you want on your current pulse? At 40kHz with 50% duty cycle the pulse lasts 12.5µs which is pretty fast so you'd say "I want it to settle to 8 bits in 1µs" for example. Also opamp slew rate might influence turn-on and turn-off time so be careful about that. Also check the settling time on the DAC you're using... or a digipot? How are you using the digipot? A schematic would be useful... \$\endgroup\$ – bobflux Apr 27 '20 at 12:47
  • 1
    \$\begingroup\$ the aim of the post is to get more ideas - basically this means more opinions = off-topic. Questions that solicit opinions \$\endgroup\$ – Andy aka Apr 27 '20 at 14:43

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.