Design Current-Based Power Supply for LEDs from 120Vac Grid

[KingNothing]

I'm a little stumped trying to design a current-based power supply for some power LEDs. Essentially I want them to ramp up from 0mA current to 350mA inverse logarithmically - like a graph of current through an inductor after a switch has been closed. I want the time constant to be approximately one second.

All this needs to feed power from 120Vac, 60Hz main power grid. So, I'll likely have a bridge rectifier with a capacitor in series with a resistor, possibly with another cap in parallel with that branch. That can provide me with a nice, smooth ramping voltage across the capacitor.

Now I say to myself, how in the world do I translate that into a current? The obvious solution is a resistor, but I really, really want to avoid dissipating heat through a resistor when the circuit is fully on.

Any suggestions?

 

[berkeman]

What is the overall voltage of the LED stack? Are you using matched LEDs to get the same brightness at the same current?

I don't think you can just use a rectified version of the AC Mains, especially if you want to control the current. The usual way you do it is to put a low-value sensing resistor at the bottom of the LED string, and use the voltage across that in your feedback circuit. The traditional way is to use a high-side Buck DC-DC converter, with feedback taken from the low-side sense resistor (like at 1.25V). Alternately, you can to a linear regulator on the high-side, again using the low-side sense voltage to control the voltage drop across the linear regulator's pass element.

 

[KingNothing]

What is the overall voltage of the LED stack? Are you using matched LEDs to get the same brightness at the same current?

I don't think you can just use a rectified version of the AC Mains, especially if you want to control the current. The usual way you do it is to put a low-value sensing resistor at the bottom of the LED string, and use the voltage across that in your feedback circuit. The traditional way is to use a high-side Buck DC-DC converter, with feedback taken from the low-side sense resistor (like at 1.25V). Alternately, you can to a linear regulator on the high-side, again using the low-side sense voltage to control the voltage drop across the linear regulator's pass element.

Yes, the LED's are matched. There will be anywhere from 0 to 8 LEDs in series, and the power supply I use should accommodate 0 to 8. There should be a voltage of about 32V available - each will need at most about 3.9V, ever.

I know I can't only use a rectified version of the AC mains. That's just where the power has to originally come from.

I apologize for the dumb questions, I haven't ever dabbled with regulating a current. I think I understand enough how to set the current with a BJT and MOSFET, and a sensing resistor. However, I don't understand how to get the current to 'ramp up', other than making the resistor 'ramp up'

 

[berkeman]

For the ramp-up, you should be able to just use an RC signal someplace in the feedback circuit. Maybe post a block diagram or initial circuit approach, and we can suggest a place to add the delay.

 

[KingNothing]

I've attached the circuit I am trying to simulate. The top part is just a bridge rectifier, probably not really of interest yet. The lower part is the current-controlling part, but I'm having a lot of trouble just getting that to work right. The current is by no means constant through the diodes, which I was hoping for.

 

[KingNothing]

Just wanted to give a little update. I have been trying to investigate other ways of accomplishing this - possibly by using pre-made LED driver circuits such as the 03021-A-E-350 BuckPuck: http://ledsupply.com/03021-a-e-350.php

I've attached a diagram that comes with their datasheet, showing the dimming control circuit, which is a 5K pot. When that 5K pot has 5K across it, the LEDs are fully on. When it has zero resistance, the LEDs are fully off.

I am guessing that internally, the thing measures current through the CTRL and Reference terminals...so, could I not just place a capacitor there in place of the pot? Shouldn't that "look" to the IC similar to a resistor with current climbing from zero to infinity?

 

[berkeman]

It looks like the Buck Puck might work for you. It would be best for you to understand exactly how that control works, so that you can implement an RC or whatever the best way to give you your dimming.

You can also look at the ON Semiconductor LED drive ICs, like the NCP3065:

http://www.onsemi.com/PowerSolutions/content.do?id=15102

BTW, another way to control brightness (the way it is often done with LEDs) is to PWM the drive current off and on at some frequency above the flicker fusion frequency of the eye. A couple hundred Hz is usually fast enough. So you could consider turning on and off the buck circuit's enable at a duty cycle that follows the brightness curve that you want to achieve at turn-on.

 

[Bob S]

Here is a LED constant current regulator circuit taken from doug3460's post in the thread
http://forum.allaboutcircuits.com/showthread.php?t=24128&highlight=led+current+regulator.
See the attached thumbnail from doug3460. Although it is designed for potentiometer control, the concept can be adapted to the requirement discussed here. See more discussion on circuit improvements, and other proposed circuits in the same thread on Allaboutcircuits.com.