Configuration for backlight leds on PCB

[yeah_indeed]

Hello,

I'm working on a project with some old display tubes (Numitron, uses around 5 V with filament instead of high voltage gas discharge with Nixies) for which I'd like to have backlighting via some smd leds. The power input for the board is 12 V at the moment, with a 7805 to step it down and regulate 5 V for the uC and peripherals.
Since I'd like to have pwm control over the brightness of the backlight, my plan was to connect 6 leds (one for each tube) in series with a n-mos at the bottom and control its gate via the uC that way. The forward voltage of the leds is 2V (they're orange/amber) in the datasheet and in my experiments. On my breadboard this worked quite well without any current limiting due to the low resistance needed anyway, and utilizing the ds-resistance of the n-mos as such current limiting resistance.

I was wondering if that's the best solution to my problem or if anyone has some suggestion on how to improve this. I'm using current sink drivers for the tubes but all outputs are taken up, so unfortunately I cannot wire them that way.

In short: I have a 12 V and 5 V supply and need to pwm control 6 leds with Vf = 2 V each. What's a good way for doing that? I have some board space to work with so that's not a huge issue.

[Buriedcode]

Well if it works, then it works, but you're relying on the 12V being accurate and stable - always nice to have some headroom for voltage.

With 6 LED's you can have a string of 6, two strings of 3, three strings of 2..
As you're using PWM I dont' think it would hurt to drive them hard at 20mA (assuming they're 20mA LED's!) which means for 3 strings of 2 in series.. (5V - (2*2V))/0.02 = 50 ohm resistor for each string. Maybe drop this to 33ohm depending on the saturation voltage of the transistor (or the on resistance on a MOSFET).  At 20mA small signal transistors are fine, like a BC337 or any general purpose NPN.

If you really want constant current, then a simple NPN current sink like this:

http://www.seventransistorlabs.com/tmoranwms/Circuits_2010/Current_Sinks.png

It has a larger voltage drop (~1.4V), so you'll need to use the 12V line, but that means you could have two of these for two strings of 3 LED's.. 3*Vf = 6V, plus the drop across sense resistor = 6.7V.. the transistor will have 12-6.7 = 5.3V across it. At 100% duty, it'll dissipate 5.3*0.02 = ~106mW, a TO92 will get warm. PWM signal is just used on the transistors base.

I guess the way to decide would be to do a mock up, and fiddle with the driving current of the LED's. If they appear to look good at 10mA, then measure the Vf.  If they need to be brighter, run them at 20mA, and check the max Vf.  All depends on whether the Vf hovers, or goes above 2V or if they all consistently are below it, if they are then you can get away with driving them all in series with a very low value resistor just to make sure.

[yeah_indeed]

Well if it works, then it works, but you're relying on the 12V being accurate and stable - always nice to have some headroom for voltage.

With 6 LED's you can have a string of 6, two strings of 3, three strings of 2..
As you're using PWM I dont' think it would hurt to drive them hard at 20mA (assuming they're 20mA LED's!) which means for 3 strings of 2 in series.. (5V - (2*2V))/0.02 = 50 ohm resistor for each string. Maybe drop this to 33ohm depending on the saturation voltage of the transistor (or the on resistance on a MOSFET).  At 20mA small signal transistors are fine, like a BC337 or any general purpose NPN.

If you really want constant current, then a simple NPN current sink like this:

http://www.seventransistorlabs.com/tmoranwms/Circuits_2010/Current_Sinks.png

It has a larger voltage drop (~1.4V), so you'll need to use the 12V line, but that means you could have two of these for two strings of 3 LED's.. 3*Vf = 6V, plus the drop across sense resistor = 6.7V.. the transistor will have 12-6.7 = 5.3V across it. At 100% duty, it'll dissipate 5.3*0.02 = ~106mW, a TO92 will get warm. PWM signal is just used on the transistors base.

I guess the way to decide would be to do a mock up, and fiddle with the driving current of the LED's. If they appear to look good at 10mA, then measure the Vf.  If they need to be brighter, run them at 20mA, and check the max Vf.  All depends on whether the Vf hovers, or goes above 2V or if they all consistently are below it, if they are then you can get away with driving them all in series with a very low value resistor just to make sure.

Thank you for the elaborate response. Yeah the no headroom on the 12 V and absence of a proper current limiting resistance halted me a bit. They should be 20 mA as I remember but I'll double check that. It doesn't have to be constant or accurate for that matter, I'll tweak the pwm to make it look nice in three settings or something.
Do you mean to have 3 strings or 2 in series and then have one transistor to control the current? Or would that need a transistor for each string, as the forward voltages of the leds won't be accurate with respect to each other (temperature drift, etc)?

I appreciate the list of options, I will surely test out different configurations but have stayed away from parallel setups due to the imperfections of the leds.

[Buriedcode]

Apologies for the late response! it didn't appear on the 'show new replies to your posts' thing.

I meant one constant current sink per string.  For your 12v supply that would mean two strings of three, which I appreciate can take up space (5 passives each).  I wouldn't worry too much about constant current though, a simple NPN for each string, and a current limiting resistor.  The two transistors can be controlled by the same line, so that reduces it to two resistors and a transistor per string.

Although constant current is used in many situations, for just backlighting, any difference in current between the strings with a series resistor (caused by slight variations in Vf of the LED's in that string) will be imperceptible. Paralleling LED's can cause headaches as some *might* be much dimmer than others, but as long as theres current limiting somewhere (a resistor) then I don't think you can kill them.

Again the tests would be best, 6 in a string, with a small current limiting resistor (just in case the 12V creeps up to 12.5V) would probably look fine.  Reduced Vf is only really a concern in white LED's where the change in colour temperature can be noticed.

[Zero999]

The switching transistor can double as the current sink.

The 10R resistors are just for current sharing. The 100R resistor sets the current through both strings in parallel. I = (V_PWM - V_BE)*R_E = (5 - 0.7)/100 = 0.043 = 43mA.