Current Mirror for 3 or more Strings?

[LED Boatguy]

I've replaced a couple 4' fluorescent fixtures in the kitchen with a motley collection of warm and neutral white LEDs. This batch is being run off 2 40W Xitanium drivers—each running 3 strings of 6 LEDs. I've voltage-matched the strings at operating temperature (3.07V-3.74V LED voltages), and they're well balanced as is, but I like the idea of having a current mirror just in case—and to turn off the string if one LED burns out.

I've used the current mirror shown in figure three http://www.ledsmagazine.com/features/6/2/2 to good success with two strings but when I add a third or fourth string, the circuit no longer shuts down when one LED opens. I've tried putting the highest vF string first, last, random—doesn't work. Works fine with two strings.

Any ideas or other circuits come to mind?

Thanks

PS The lights look great, BTW.

 

[SemiMan]

Oh brother! ... they just don't make engineers like they used to....and not magazine editors either I see. OK off my soap box.

So if one of the strings goes into current limiting....what will happen.... well if the main LED driver is a constant current source, say of 700mA, then it will keep increasing the output voltage until it can get to 700mA. As per the article, the circuit limits current to 455mA. However, the LED driver is putting out maximum voltage. The voltage has to be somewhere, so it is going to be on the left transistor...which is going to be hot and need a significant heat sink. That is something to keep in mind.

There is another way of doing the current limit that quite simple though perhaps not as flexible. All you need to do is put two forward biased diodes between the base of each transistor in the current mirror and ground. These are pennies each. This will limit the maximum voltage on the base of the transistor to approximately 1.4V. You could play with a standard diode and a schottky to get to 1.2V, but 1.4V works well.

Assuming 0.7V for Vbe of the current mirror transistor, then the maximum current in any chain is limited to (1.4-0.7)/1.5 = 470mA. 1.5 is the value of the resistor. Practically the limit is going to be lower than 470mA. You can extend this to as many chains as you want.

Note that the minimum hFe for the BD139 at 350mA is likely on the order of 30-35 (not typical, but what you may actually encounter). That means for 350mA, you are going to need on the order of 10mA into the base. With the 680ohm transistor between the last LED and the base, you will need a 680ohm * 0.010Amps = 6.8 volts extra! That means worst case you need 6.8V across this resistor, 0.7volts Vbe, and 0.5V or so across the 1.5ohm resistor or about 8 volts extra!!!

I would pass on their current limit scheme, get rid of the 680ohm resistors (but keep the 1.5 it is essential for current matching), and then use the two series connected diodes in the base to limit current.

Next I would consider closely reviewing what I have written and see if I am right or I have lost my mind (I could have made a mistake in my math). Assuming I have not, may want to consider a note to LEDS Magazine suggesting better peer review.

Semiman

I've replaced a couple 4' fluorescent fixtures in the kitchen with a motley collection of warm and neutral white LEDs. This batch is being run off 2 40W Xitanium drivers—each running 3 strings of 6 LEDs. I've voltage-matched the strings at operating temperature (3.07V-3.74V LED voltages), and they're well balanced as is, but I like the idea of having a current mirror just in case—and to turn off the string if one LED burns out.

I've used the current mirror shown in figure three http://www.ledsmagazine.com/features/6/2/2 to good success with two strings but when I add a third or fourth string, the circuit no longer shuts down when one LED opens. I've tried putting the highest vF string first, last, random—doesn't work. Works fine with two strings.

Any ideas or other circuits come to mind?

Thanks

PS The lights look great, BTW.

 

[Steve K]

Just jumping in here.... I'll agree that on the surface, the article looks sensible, but then you have to ask why someone is trying to run two strings off of a single current source? My instinct would be to use a buck converter (or boost?) to get the voltage a few volts above the max forward voltage of the string, and then use a simple linear current regulator in each string.

If you wanted an excuse to design a circuit, you might set it up so that there is a voltage regulator feeding each string through it's own ballast resistor. The current from each string is sensed, and then the regulator uses the highest current sensed to control the voltage. The other strings would be lower current, and therefore dimmer, but it would achieve the dubious virtue of having the strings track together, sort of...

I would certainly recommend a separate current regulator for each string. Of course, I would also recommend sticking with the original 4' fluorescent tubes, since they are probably just as efficient, and probably $100 cheaper. (?) I'd be more inclined to use the LEDs to replace incandescents in dimming applications, or maybe some CFL's that get turned on and off a lot. (but that's just me)

in any case, I hope this provides a little technical guidance.

regards,
Steve K.

 

[LED Boatguy]

I took a peek in CPF since it's been a few years and wanted to see what's new. Looks like the Cree XM-L2 has dethroned the MC-E as top dog in the flashlight world--at least for Cree. BTW: Is anyone making D M@g heatsinks for 3-up XM-L2s?

I also thought I'd lookup old threads that I had started. This one was about replacing 4' fluorescents with vf-matched series/parallel strings and a current mirror. Not sure if the current mirror is working, because nearly 6 years and a couple thousand hours later, not one LED in the two units has gone Tango Uniform. So far, so good.

 

[Steve K]

wow.. a 5 year old thread comes back to life... amazing...

I'm glad that the circuit has been successful, but a few thousand hours shouldn't be exceptional if the LEDs aren't being stressed. At this rate, you ought to get at least 10 times this amount of life out of it, so figure it is good for another 60 years or so??