1. Home
  2. Electronics Design
  3. Any ideas for driving an array of discrete LED's without running into heat problems?

Any ideas for driving an array of discrete LED's without running into heat problems?

I have an interesting problem. I have a display that is roughly a

1.5" diameter circle. I am using an array of discrete SMT LED's (104 of them) to create the display. The LED's are grouped into 4 sets of 26 LED's, each of the 4 sets consists of 4 strings of 4 LED's and 3 strings of 3 LED's. These strings are individually controllable and can be turned on/off in various ways to create different symbols on the front of the display (think of it like a big and super bright alphanumeric character display).

The reason for using the discrete LED's is to get the brightness needed... it must be LED and it must be visible in bright sunlight, and discrete LED's provide the necessary qualities.

The first prototype used Allegro Micro power shift registers, plus a linear 1-amp regulator supplying 9 volts. The Vf of the LED's is 2V, and the Allegro chip is a programmable constant current LED driver. I am using 4 of the Allego chips, one per set of 26 LED's (in 7 strings... thereby using 7 of the 8 lines of the Allegro chip).

Running the LED's at 20mA, the whole thing gets super hot. I think the heat is coming from the Allego chips, which very much surprised me. The A6278 chip I am using is supposed to be able to handle 3 watts dissipation at 50C ambient. For the 7 strings of LED's, 3 of them use 3 LED's (2v * 3 = 6v @ 20mA) so the Allegro is dropping 9v @

20mA on those lines or 0.18W, and the other 4 strings are 4 LED's, (2v
  • 4 = 4v @ 20mA) so the chip is dropping 1v @ 20mA on those lines, or .
02W. By my math, that totals 0.62W. Not a tiny amount, but not gigantic. Then again, I do have 4 of the Allego chips on this tiny board, so combined that's around 2.5W.

Then, on the LED board (which is mounted directly on top of the driver board - only about 0.5mm between the top of the Allegro chips and the back of the LED board), I have the 104 LED's in 0603 packages (LiteOn PN LTST-C190KRKT)... those LED's have a power dissipation of 75mW... so 104 of them running full tilt would be 7.8 watts.

7.8 watts from the LED's and 2.5 watts from the Allegros.... just too much heat. The prototype works, but I have to throttle back the LED's to more like 10mA, and even then they get pretty hot... but not overly hot - it's still the Allego chips that get toasty warm.

Of course, I know the problem is too much power disspation, but I am trying to end up with the BRIGHTEST display possible, while keeping heat in check. Any comments? (bear in mind I am not an expert at design - still learning)

1) Are there other LED's that are substantially more efficient and produce less heat in an 0603 package? Others that I have checked are even higher than 75mW!

2) I could multiplex the display, but whenever I have done this, I have always noticed a decrease in brightness (a substantial decrease). Despite the manufacturer saying maybe .1ms pulse at 200mA is OK, at 1/10th duty cycle, I always find that to be substantially dimmer. Have you also found that to be the case?

3) I could mount the top array of LED's on a metal core board and thermally couple that to the housing. I am not sure if I could get away with maybe 5W of dissipation if I were to do that? Any thoughts? On a 1.5" diameter 0.062" PCB?

4) I could change the strings of 3 LED's to strings of 4 LEDs which would result in less power disspiation in the Allegro chips... then reduce the current overall to produce a manageable heat level on the display PCB

Any other suggestions? Any tips and experience with building high- brightness high-density displays made from discrete LED's??? As I said, it has to be visible in daytime full sunlight, so brightness is paramount.

Reply to
ferrari.secret.santa
Loading thread data ...

The A6278 seems to be limited to 45ma on one and only output. You could also use the 90ma limit divided by 16, which gives ~6ma. I would calculate out what the max per output would be using the Ambient temp and the 165c shutoff. That would give you the max dissipation perdevice. You are definitely driving too much current for the device to handle. I would use the ULN2003 perphial drivers a the currents your working with.

Cheers

Reply to
Martin Riddle

The Allegro can handle 90mA on all inputs at the same time, I believe... it's billed as a power shift register... the total current can't exceed 750mA (total ground current) and the dissipation comes into play well before that limit through.

There was an error in my math... each Allegro drives 3 strings of 3 LED's at a 3V drop (9V supply, 2V per LED) and 20mA... that's 0.06W times 3 strings, total 0.18W. Then it drives 4 strings of 4, 1V drop at 0.02mA for a total of 0.08W... so the whole chip is only dissipating 0.26W. That seems low for how hot they are getting.

Reply to
ferrari.secret.santa

The Allegro can handle 90mA on all inputs at the same time, I believe... it's billed as a power shift register... the total current can't exceed 750mA (total ground current) and the dissipation comes into play well before that limit through.

There was an error in my math... each Allegro drives 3 strings of 3 LED's at a 3V drop (9V supply, 2V per LED) and 20mA... that's 0.06W times 3 strings, total 0.18W. Then it drives 4 strings of 4, 1V drop at 0.02mA for a total of 0.08W... so the whole chip is only dissipating 0.26W. That seems low for how hot they are getting.

Your using the Absolute maximum ratings (750ma ground current), which is unwise. If you look at the charts for the A6278, max power is 2.5w at 50c ( dip16 A package )

lets see..

Pdc = DC * (Vce*Io*8)+Vdd*Idd) Pdc = 1.00 * (1.0v*0.090*8)+(5*0.032) = 0.88 w

Which looks ok. You should be within the device ratings, something else is wrong.

Do you have series resistors? Vce at 40ma is only 0.7v. What do you mean by 3 volt drop? is this the voltage on the ouput pin? (Vce)

9v and 3 leds@ 2v ea would need a 115 ohm resistor at 0.18watt 3 strings is 3*.18w = .54w. 4 stings of 4 should be 15ohms. And Vdd should be 3 to 5v.

Cheers

Reply to
Martin Riddle

Hi Martin,

I do not have series resistors on the LED strings. I have a 9V linear regulator, which connects to the strings of LED's (strings of 3 and 4) and then each string is connected to one leg of the A6278. The A6278 is a shift register with a constant current LED driver on each leg. So I was using the external resistor of ~1k set the current of ~20mA for each pin, and letting the Allego handle supplying constant current to each pin. Is this not correct? The datasheet says the A6278 is most effective when dropping 0.7 to 3.0V. So, for one of my strings of 3 LED's with a Vf of 2.0V each, that is 6V drop across the LED's, and 3V drop on Vce at the A6278... and then for a string of 4 LED's on another pin, 2V per LED so 8V for the string, and on those the Allegro has Vce of 1V.

So I am figuring that for the 3 legs of the 6278 that have strings of three LED's, the 6278 is dissipating 3V * 0.02A or 0.06W (three of them, 0.18W total). Then on the four strings of four LED's it is dissipating 1V @ 0.02A per leg so 0.08W...totaling 0.26W per A6278 chip.

Am I missing something? Or perhaps not understanding the use of this chip? It is my understanding that the chip can maintain the current on each leg, so even though I have some strings of 4 LED's and other strings of 3 LED's, it should adjust the voltage on each pin seperately to keep them all at 20mA.

And based on that, and given that I calculate 0.26W of power dissipation, I can't understand why the chips are getting very hot since I think I am only at about 10% of their capability.

Then again, obviously something is wrong since it's not doing what I think it should be ;)

thanks!

Reply to
ferrari.secret.santa

...

...

...

...

Based on the data sheet, I agree with your interpretation of the 90mA/ output limit and 750mA ground current limit, and overall power limit.

It does seem likely that the chip is intended to work ok with different string lengths (3 vs 4 LED's), but the data sheet doesn't explicitly say so. It says "designed to operate with driver voltage drops (VCE) of 0.7 to 3V, with an LED forward voltage, VF, of 1.2 to 4.0 V". OTOH, it also says in Absolute Maximum Ratings that Load Supply Voltage Range (VLED) is ?0.5 to 17 V. You should ask Allegro to clarify whether different string lengths are ok.

What A6278 package are you using? If it has an exposed thermal pad, is it attached to a heat sink or a copper plane? Are you clocking the chip at a high rate or rapidly switching outputs off and on?

Perhaps you should add series diodes to the LED cha> Then, on the LED board (which is mounted directly on top of the driver

You miscalculated the LED power dissipation. An LED with 2V Vf at 20mA dissipates at most 40mW, rather than 75, so the LED board dissipation is at most 4.16 watts. This is still a lot of heat and should be dealt with.

--
jiw
Reply to
James Waldby

7
o
N
4

th.

Thanks for the reply jiw... I will contact Allegro and confirm if it holds constant current on a per-leg basis, even with different voltages on each leg. I am using the TP package (eTSSOP with the large exposed thermal pad)... so it should be able to handle 0.25W with no problem.

I still can't figure out why those Allegro chips are getting so hot... I will have to do some more testing and measuring. At first I thought it was the LED board being so close, but the LED board gets "warm", never hot... the Allego chips get hot like they are dissipating multiple watts or something when they shouldn't be. The LED's are being driven correctly, so I guess I can just try to put some series diodes or resistors to reduce dissipation in the 6278 chip.

On the dissipation of the LED - you're right, how stupid of me... I just took 75mW from the datasheet without even realizing that's more than 40mW of total power I am putting through the LED's... duh.

Question... how much of that 40mW that I am putting through the LED's goes into light and how much into heat? i.e. does an LED conver 20% of the incoming energy to light? Or perhaps my understanding is wrong and the LED die will convert all of that 40mW to heat, and the light is just a byproduct of that transaction?

Thanks!

Reply to
ferrari.secret.santa

y
0.7

lso

nt

r

PN

104

mA

s

with.

Generally it is a good idea to keep the distance from the leds to driver chip short to avoid any L di/dt issues.

If you goal is daylight visibility, you can improve it using a filter over the display that matches the color of the LED. This may reduce the power requirement into the LED to achieve your goal.

About all you can do is keep Vled as low as practical. Also, put the same number of LEDs in series so that you can avoid the case where some drivers have to drop an extra Vf of the "missing" LED.

You can somewhat keep chips cooler by putting extra copper on the traces. That is, some of the heat flows through the bond wires. For instance, on the MAX7219, the bond wires are 1.3 mil versus the standard 1mil wire. Heat flow is proportional to the square of the radius, so fat wire does make a big difference. The current source pins are the best target for added extra copper to the PCB.

Reply to
miso

On Thu, 06 Nov 2008 12:24:29 -0800, ferrari.secret.santa wrote: [> James Waldby wrote] ...

...

formatting link
says "high power LED are generating on the average 75% heat versus only

25% light" relative to input power, while a number of other links say that heat output approximately equals power input. So, for design approximation, plan to deal with 3 to 4 W of heat when all lights are on.

The "Light extraction" and "Failure modes" sections of

formatting link
explain a little about how LED's work, and might be useful background. The whole article is interesting reading, except for mistakes.

--
jiw
Reply to
James Waldby

ElectronDepot website is not affiliated with any of the manufacturers or service providers discussed here. All logos and trade names are the property of their respective owners.