I recently found these 8 pin RGBW LEDs. It looks like a good deal. I may get them soon to use in a DIY flashlight. It's surprising that you can get them without a common gnd. So how necessary is a heat sink when they are being operated at full power, and would a random steel washer or bolt suffice? I really don't want to spend too much. Also, what is the cheapest way to drive them reliably? Is there any way to make a current source that can supply 300mA for less than the cost of an LM317 or other linear regulator? Would a BJT work, or is the gain too unreliable?

 

Definitely have to do a heatsink design. A bolt or washer would be good for a few seconds before
burn out.

An LM317 is a 10 - 50 cent kind of part. But cannot handle 300 mA x 4. An external BJT +
R + heatsink + Zener can do basic current regulation. A more efficient way is to use PWM
if you are going to control brightness.

You running all 4 leds simultaneously ? Want to do brightness control, whole array
or individual color ?

Regards, Dana.

 

Definitely have to do a heatsink design. A bolt or washer would be good for a few seconds before
burn out.

An LM317 is a 10 - 50 cent kind of part. But cannot handle 300 mA x 4. An external BJT +
R + heatsink + Zener can do basic current regulation. A more efficient way is to use PWM
if you are going to control brightness.

You running all 4 leds simultaneously ? Want to do brightness control, whole array
or individual color ?

Regards, Dana.

The heat sinks actually look pretty cheap, so buying them is not too big a deal. But is thermal paste really required, or do I just need to ensure there is little gap between them?

I plan to PWM each color at a low frequency, independently, from a microcontroller (arduino nano). They would each have their own current source, so it would only need to be able to handle 350mA. But then there would be a mosfet. But having the mosfet and the current source seems unnecessary, and I do not want to waste so much power. So could I PWM the LED and put an inductor in series? Would I be able to use the inductor as a current shunt in some way? Could use an analog pin or op amp to measure the voltage, and adjust the duty cycle or frequency accordingly?

 

The heat sinks actually look pretty cheap, so buying them is not too big a deal. But is thermal paste really required, or do I just need to ensure there is little gap between them?

Paste is not so much electrical isolation as improving heat flow from device to sink.

I plan to PWM each color at a low frequency, independently, from a microcontroller (arduino nano). They would each have their own current source, so it would only need to be able to handle 350mA. But then there would be a mosfet. But having the mosfet and the current source seems unnecessary, and I do not want to waste so much power. So could I PWM the LED and put an inductor in series? Would I be able to use the inductor as a current shunt in some way? Could use an analog pin or op amp to measure the voltage, and adjust the duty cycle or frequency accordingly?

Normally PWM approach drives an led with a series limiting R to set max current.
Its not precise, but neither is the human eye. And just a power mosfet to do the
switching. So no L needed. That will accomplish wide range avg current control.
One could use a sensor, cal it for its response curve, and then use A/D to measure
sensor and modify the duty cycle being sent to each PWM channel.

Regards, Dana.

 

Paste is not so much electrical isolation as improving heat flow from device to sink.



Normally PWM approach drives an led with a series limiting R to set max current.
Its not precise, but neither is the human eye. And just a power mosfet to do the
switching. So no L needed. That will accomplish wide range avg current control.
One could use a sensor, cal it for its response curve, and then use A/D to measure
sensor and modify the duty cycle being sent to each PWM channel.

Regards, Dana.

I was just trying to think outside the box, with an inductor instead of resistor. So what you're saying is that CC is not really too important, as long as there is adequate heat dissipation? And is the thermal paste absolutely necessary?

 

I was just trying to think outside the box, with an inductor instead of resistor. So what you're saying is that CC is not really too important, as long as there is adequate heat dissipation? And is the thermal paste absolutely necessary?

What do you have against thermal paste? Why increase your risk of failure? Just use it like you're supposed to and be done with it!

 

What do you have against thermal paste? Why increase your risk of failure? Just use it like you're supposed to and be done with it!

I just don't want to spend too much. Could you get it cheaply?

 

I just don't want to spend too much. Could you get it cheaply?

Fair enough. Not sure where best to get it in small quantities at reasonable prices.

We use giant tubs of it at work (more than you'd want for hobby use) and we also get free mini tubes of it sometimes in with orders for parts that would need it. I've been lucky and not had to source it myself.

 

I just don't want to spend too much. Could you get it cheaply?

Anyplace that sells computer parts. Since I don't know where you live hard to say but there are dozens of online places like NewEgg just as an example. Locally to me I would just go a few miles to Micro Center. A less than $5 syringe last a long, long time. You are looking for "thermal compound" as a common name.

Ron

 

Would something like this work for 10s of heat sinks?
https://www.amazon.com/Thermal-Ther...&qid=1535495924&sr=8-7&keywords=thermal+paste

 

Yes. That will work just fine.

Ron

 

Ok. Could I reliably operate my mosfet in the ohmic region to eliminate the need for a resistor? I'm just trying to think of ways to reduce the components needed.

 

Ok. Could I reliably operate my mosfet in the ohmic region to eliminate the need for a resistor? I'm just trying to think of ways to reduce the components needed.

So how would you control the gate voltage? You are looking at a very small region where the mosfet can act like a resistor. The linear portion is really small. I don't see it working well but maybe others have some thoughts to share. With your case I see the mosfet working as a switch and not a resistor.

Ron

 

So how would you control the gate voltage? You are looking at a very small region where the mosfet can act like a resistor. The linear portion is really small. I don't see it working well but maybe others have some thoughts to share. With your case I see the mosfet working as a switch and not a resistor.

Ron

I would maybe have some sort of voltage divider that steps the μC GPIO voltage down, and feeds it to the mosfet. This would allow for current limiting and switching.

 

Ok. Could I reliably operate my mosfet in the ohmic region to eliminate the need for a resistor?

Yes if you did a control loop with processor.

No if no control loop to take out device to device variations, LED,.MOSFET, Vdd,
T.....

Regards, Dana.

 

Yes if you did a control loop with processor.

No if no control loop to take out device to device variations, LED,.MOSFET, Vdd,
T.....

Regards, Dana.

I think I'll try that BJT circuit you suggested. If I used BJTs like this, how would I heat sink them? Also, what values should I use for the resistors and zener diode in it if I want a current of 300ma?

 

A lot more detail is needed to do this right -

1) What is power supply V ?
a. Max V
b. Min V
c. Battery C
d. Battery technology

2) Description of functionality of design from user point of view

3) Accuracy of mixing, spectral response

4) Physical volume available for heat sink(s)

5) Operating environmental T range

To name a few.......

Regards, Dana.

 

A lot more detail is needed to do this right -

1) What is power supply V ?
a. Max V
b. Min V
c. Battery C
d. Battery technology

2) Description of functionality of design from user point of view

3) Accuracy of mixing, spectral response

4) Physical volume available for heat sink(s)

5) Operating environmental T range

To name a few.......

Regards, Dana.

Sorry if I didn't give enough details. Well, I did the math just now, and I realized that just a resistor might not be too bad. I don't think I'll go with a linear regulator or BJT circuit. They say the LEDs can handle up to 350mA, but 300mA is recommended. I plan to have one thing powered by a 5V 2A USB port. If I use a 5.5 ohm resistor, if there is a fV of 3.4V, there is a current of 290mA. If there is a fV of 3V, there is a current of 363mA, which shouldn't destroy it.

I also want to make an insanely bright RGBW flashlight using 10 of the LEDs in series. Since I need decent efficiency, the only option I see is SMPS. If it consumes up to 30 or 40 watts, it will be hard to have a decent battery life if so much power is wasted. I already have some power mosfets and current shunts. So I am thinking it might make sense to make my own CC SMPS for that. The μC would control the current and handle regulation. The one thing is, I do not really know what the best way to amplify the shunt voltage is. Would a simple LM358 IC be able to amplify it for the μC's ADC without issue, and with reasonable accuracy (withing a few mA)? I will probably use a one ohm shunt, meaning the op amp would have to be within a few mV. I know about input offset voltage, noise, and all that stuff, but do not really know too much about how op amps behave in real-world situations.

Also, it's getting kind of late for me, so I have to get going. Unfortunately I have some real world obligations, like sleep.

 

It looks like all the Amazon parts you linked to have no detailed spec's and no datasheets. No-Name-Brand cheap Chinese stuff?
How do you know the thermal resistance of the LEDs for you to select a suitable heatsink??

 

If you use a micro with a differential input A/D that can handle the shunt current measurement.

If it has adequate resolution and appropriate Vref range (low).

And in general if doing smps with a micro you want speed in the control loop,
to permit high speed and low L values.

PSOC 5LP might be a decent choice. Some parts have 2 12 bit SARs onboard,
each capable of 1 MSPS. But these are not low cost parts, but do have PWM,
OpAmps, Vref, diff inputs, ARM core.

If you used separate SMPS there are parts out in industry that are low cost, can
do > 1 MHz, and use a PSOC 4 with 1 SAR to do PWM, Current Sense, ARM core,
DAC, etc. The low cost SMPS are Micrel, bought out by Microchip.

See attached.

The PSOC SAR has a Vref range 1.024V, 12 bits, so LSB = .25 mV

Regards, Dana.