DIY: Cree CXA Arduino Thermal Monitoring & Protection.

Exactly, as stated above, logic level FETs can use lower voltages on the gate to get them conducting and turning on, MOST of the time a logic level part will have an "L" in the part number. Like above, the IRF540, as you can see in the datasheet has a gate threshold voltage of +2-+4V DC, now take a look at the IRL540, which is a very close relative, however the gate threshold is only +1 - +2V DC, which allows a microcontroller using +3.3V, which many, many are using that and lower, down to 1.8V these days; to switch the FET on, and the much higher load connected to the FET without having to use a transistor or some other kind of gate driver circuitry to get it to turn on....

@stardustsailor , you might want to check out one of the part numbers with an L in it, for Logic, if you are going to be driving straight off of the ATMega328p pin, I have seen some instances where they don't fully conduct without it being a true logic level FET.

I am looking over your project now and will continue over the next day or so, have you been able to get your hands on any of the CXA arrays? I know they can't be cheap, I actually saw a demo on them the last time I was at the CREE headquarters, they have an incredible facility. I've been looking for a new project to do in my spare time, looks like I have found it!

Okay, I just browsed over your last thread and took a look at most everything.
I really like your 3d drawings, what are you using to do those? The heatsink design looks good, might possibly be better off with multiple smaller fans, maybe a 2x2 array for each LED, and choose fans that have a high static pressure that will really move the air through those inlets, that is going to be the crucial part.

So are you looking at using three of the CXA's total? I am not 100% sure, I'll have to look back at the datasheet, but correct me if I'm wrong, isn't the forward voltage on those CXA's somewhere in the 72-77V DC range? I can't remember off the top of my head. That is some serious power and some serious heat that we're going to be contended with, I am not sure if using a FET is going to be the best way to do that. What are you going to be pulling your source voltage for the LED's from?

Linear Technology makes a lot of really good LED drivers that I have used in the past, off the top of my head I think the LT3791 may be a contender for the job, it may only have a 60V DC max for the LED's, I will look at finding a part. If we could maybe look at using LED Driver IC's, which have 4000:1 dimming with PWM inputs that we could run from the ATMega328p. Some of the IC's also have built in temp sensing capabilities with an NTC or a PTC, but if you want to get more specific, we could use a digital temp sensors (the DS18B20 from Maxim is good, with a 1-wire serial interface, Analog Devices TMP36 is also a solid linear output temp sensor that is highly accurate. An NTC would do the job, but you just aren't going to have the kind of accuracy and resolution that you could achieve from a discrete temp sensors. I'm sure you have already thought of this, but just putting it out there. And alot of the LED driver IC's can guess a pretty good temp from an NTC or something. So if you are using three of the CXA's, you could use three LED driver IC's, each using two pins from your micro, one each for your PWM input, and another for your ENABLE input to the IC. That won't require any driver circuitry since those are both high impedance inputs. Going this route you will definitely have to design custom PCB's because you are going to have to use surface mount components to get the kind of performance you are after. I'm almost debating rectifying 120VAC from the wall as your supply voltage for the LED's and bucking it down with a driver....

Have to think more and see what you have to say....
Here is the LED driver IC page at Linear, you can put in your parameters and see your choices... Let me know if something catches your eye..
http://www.linear.com/products/LED_Driver_ICs