A Simple Linear Constant Current LED Driver Circuit: Some Concerns

[agwdm2]

Hello everyone,

I've a few questions regarding the following circuit floating around several forums. The pros: low part count, only 0.6V of overhead.
(More details at http://laserpointerforums.com/f51/simple-linear-mosfet-dimming-circuit-pwm-70812.html)
 


As I'm about to begin an expensive color mixing project with many luxeon rebel LEDs, I have the following concerns:


1. Forget the PWM control parts (shown in grey) for now. Apart from that, how good is the regulation? I don't have a scope, I would like to know if there are any ripples, and if so, how to reduce that.


2. I've read some app notes by Avago and Osram-opto on driving high power LEDs. For linear circuits, they recommend the plain old LM317 or LDO-regulators like LM2941CT, both of which need extra input/output caps for stabilizing.




3. Now suppose I complicate things by feeding a PWM signal . Is there any restriction on the frequency/duty cycle of the PWM for the current output to be stable? Asking this since many of the commercially available PWM dimming drivers are limited to 200-1000Hz frequency and 10-100% duty cycle. I want to know the answers for both the circuits.


UPDATE: I settled on a design using Onsemi's CAT4101 for the driver and PIC24FJ128GA106 for the PWM generator.

[SeanB]

As Q2 heats up the current limit does drop at 2mV per degree, as this is using the Vbe as a reference. It is somewhat non linear with supply voltage, but is good enough to drive a LED load to within about 2% if you use a 1% resistor as a shunt/sense resistor. Q3 should be another NPN so that you are sure to saturate the switch and turn off the mosfet fast, and it will be better to reduce the value of R1 so that you have around 1mA flowing to provide the pullup so that you can get fast turn on of the current source. Keep lead lengths short around Q1, Q2 and the sense resistors, and keep the gate leads short as well. A 100-220R series resistor on the gate will help damp oscillations on the mosfet during turn on as well, along with having a large value capacitor on the supply rails to decouple them.

[agwdm2]

Thanks for your reply, my concern was essentially the MOSFET switching delay. However I still don't understand why an NPN Q3 will act faster than a PNP.

Also, the PWM will be supplied from TLC5940. I guess I've to change R4 to a pullup resistor in that case, since its outputs are open collector type?

[oPossum]

The MOSFET turn on will be rather slow with the R1 being 1M. I would change R1 to 1k and R4 to 4.7k to speed it up.

[agwdm2]

I came across a very helpful TI appnote explaining high speed MOSFET switching: http://www.ti.com/lit/ml/slup169/slup169.pdf (I didn't know that MOSFETs come with stray capacitance/inductances, like capacitors have ESL/ESR. I've a lot to learn!)

Another site http://neuroelec.com/2011/03/high-pwm-frequency-for-led-good-or-bad/ discusses the issues with PWM speed/duty ranges. It seems that high switching speed(<2uS) is necessary for my application since TLC5940 has 12bit output and I might need very low duty cycle  (say 4/4096, i.e. 10bit resolution) without observing any flicker. Don't know if its possible though. I am planning a PWM frequency of 200Hz.

@SeanB,
Can you please explain more why Q3 should be NPN?

[westfw]

high speed MOSFET switching

WHAT "high-speed MOSFET switching"?  It's a linear circuit, isn't it?  Any PWM can be relatively slow and need not be turning the MOSFET hard on or off anyway, right?

That said, it also means that you'll need a MOSFET capable of significant power dissipation.  A lot of the "modern" transistors are aimed at ... switching, and while they have a high current rating, they're counting on either being full ON or OFF (and driver circuits being careful to push them quickly), and can have pretty moderate ratings when it comes to power dissipation.  And you'll need a real heatsink as well.

[agwdm2]

WHAT "high-speed MOSFET switching"?

Read my previous post. When I'm operating at a PWM freq of 200Hz and a very low duty cycle, say 4/4096, the MOSFET needs to turn (soft) on for ~5uS (very briefly) and then (hard) off again for ~5mS. Similarly it needs to (hard) off very briefly for duty cycle very close to unity. (I picked 4/4096 as I must need at least 10bit of PWM resolution so that the color change is fluid enough.)

Any PWM can be relatively slow and need not be turning the MOSFET hard on or off anyway, right?

Sure it won't be hard on ever since it'll be operating in the ohmic region (so its not like conventional switching), but it should be hard off during the off time of the PWM? At least that's what I'm trying to achieve by reducing the delays.

Also I'm well aware of the heatsinking requirements and calculating the heat dissipation in the linear mode, its just I*(V_supply - V_f _of_LED_at_I - 0.6). The average dissipation will be much less since I'll be cycling through 9 different LED strings/MOSFETs for a color changing rainbow effect.

[SeanB]

NPN will give a lower saturation voltage to pull the mosfet off, so a logic low will be better defined. As well it will not have any effect on the high voltage when off, with the PNP the logic high voltage needs to be higher than the regulation setpoint ( depends on the mosfet so will be somewhere between the threshold voltage and the supply rail) otherwise it will pull the voltage to the high point and your current will be lower.