tlc5940 power dissipation

Forum 2005-2010 (read only) Hardware Interfacing
1

hello,
I have still some problems with the TLC5940 - actually with the power dissipation, because I burnt one TLC (my problem is that I don't understand the pd-formula in the datasheet, so I can not calculate it.... :frowning:

In detail: I drive 25 LED sheets at 12 V. Everey sheet needs 0,057A so IRef is 680 Ohm. I use the TLC library from metalab. I daisychained two TLC's; the first drives 16 LEDsheets, and the other the last nine. I feed the TLCs with 5 V, even If I could use just 3,3. (I have read the post of nima (http://www.arduino.cc/cgi-bin/yabb2/YaBB.pl?num=1204292903), but I dont believe that the 5 V is really the problem...?
-->Datasheet: 0-60mA; Vcc< 3,6 V and 0-120mA, Vcc > 3,6V))

All the setup worked perfectly for two weeks, while trying out different little programs. But when I started to let run all the programms for a longer time, at the programm which dimms all 25 Ledsheets at the same time on to off and vice versa, some of the Leds started to flicker or they had like an offset in brightness. When I excluded the "dimmall_programm" everything worked again. From time to time I held my finger to the TLCs and they where not hot! But nevertheless, after two days, the TLC who has to drive the 16 LED sheets was dead (of course I tried again the "dimmall_programm") - and the other who has to drive just 9 is still alive.
So my suspicion is that it consumes too much power when all the LEDs are dimmed (0%-100%-0%) at the same time......Unfortunately the power dissipation formula You can find in the TLC datasheet is not really clear for me.

PD = (Vcc x Icc) + (Vout x Imax x (DCn / 63) x Dpwm x N)
where:
VCC: device supply voltage (in my case: 5)
ICC: device supply current (in my case 30???? I looked at the datasheet under Icc...?)
VOUT: TLC5940 OUTn voltage when driving LED current (in my case 12 ?)
IMAX: LED current adjusted by R(IREF) Resistor (0,057)
DCn: maximum dot correction value for OUTn (63, I dident have done Dotcorrection, the DC Pin goes directly to 5 V, so Dot correction is 63...???)
N: number of OUTn driving LED at the same time (16)
dPWM: duty cycle defined by BLANK pin or GS PWM value (????4096???? here I have actually no idear what is exactly meant :-/

The formula with my values gives the result of 44976,624....mW?? or what? this result is of course far away from the max. rating of 2456mW for the DIL TLC...
so I have no clue if my calculation is totally strange (probably and hopefully yes) or if it is in fact not possible to dimm all (16) LEDs at the same time (what I can't believe).

Had anybody similar problems with dimming all channels at the same time?

Or had anybody problems with the capacitor between Vcc and GND? (if the capacitor is dead there is a short on the chip...this happend to me twice!)

For any help, hints and suggestions Im very gratefull!

Thanx a lot
lr

2

(Note In the following numbers like 2.34 should be read as two point three four.)

Most of the maths is correct. The two things wrong is the ICC which is 30mA typically but 60mA maximum so it is always better to use the maximum here. This gives 0.3W for the chip.

The Dpwm is the duty cycle and will range from 1 if it's on all the time to 0 if it's off all the time. So maximum power is when it's on.

That gives a total power of 10.944 + 0.3 = 11.244 Watts.

Now the package is only rated at 2.456 Watts so you can see you are over the top by several times. This is way over what a heat sink would give you (the 2.456 Watts assumes an infinite heat sink). You will kill the device if all the LEDs are on at the same time. You can look at the formula to see what you could cut down to bring things into line. Like the number of LEDs on at once, or how bright they are (duty cycle) or the drive voltage or the LED current.

You have to decide what you want to cut back on. You will in practice get away with over dissipation, if it is for a short time only, as the thermal capacity of the chip will prevent damage but it is not reliable in the long term as you have found.

Or had anybody problems with the capacitor between Vcc and GND? (if the capacitor is dead there is a short on the chip...this happend to me twice!)

This should not happen and suggests that there is too much ripple current for the type of capacitors used. Use one with a low ESR (effective serial resistance) of the types used in switch mode power supplies.

3

Thank you very much for Your fast answer!
At least its not that much I calculated;)
As well thanx for the tip for the right capacitor.

As it is important to dimm all LEDs together, I will try to limit the LED current. With an amplifier...
(In the thread http://www.arduino.cc/cgi-bin/yabb2/YaBB.pl?num=1204292903, somebody had tried this with an ULN2803)

lr

4

Regardless of any other capacitor, the first one should be something like 0.1uf (100nf) ceramic with short leads. Because the dimming can impose some pretty significant current on the supply you should also have a moderate size low-ESR electrolytic or tantalum capacitor in parallel. Remember, not all tantalums are low ESR.

Which package are you using? If you are using the 28 pin DIP then you will get a lot of improvement if you glue a heatsink on top of the device.

The Geek

5

thanx geek,
but my main problem is how to limit the LEDs current to a point, where I'm able to dimm all at the same time from 0 to 100%....I don't think I will get that with an "infinite " heatsink;)

Has anybody an idear how to do this? (instead of 0,057 A just 0,005 A per LED, means somehow to amplifie the output of the TLC wich is limitetd to the 5mA....)

I'm wondering that nobody else had this problem, because to dimm all at the same time seems to me a very obvious wish;)
Still for any suggestions very gratefull!!
lr

6

No you can't use an amplifier to boost the signal. This is because the chip provides a constant current source to control the brightness and you can't amplify that.

The problem is that at 57mA per LED (I assume that's what you mean by sheet) it is much more than normal users require. Also why do you have to drive them with 12V, get that down to reduce the dissipation.

you will get a lot of improvement if you glue a heatsink on top of the device.

Yes it will improve things but will NEVER be better than the 2.5 W given in the data sheet, these figures are derived assuming an infinite heat sink. That means the surface temperature never gets above 25C.

you should also have a moderate size low-ESR electrolytic or tantalum capacitor in parallel

I would never use a tantalum capacitor as a power decoupling. This is because the failure mode for these is a short circuit. One place I worked at it was almost a sackable offense to do so.

7

Thanks again for your good explanations!
I can not use a lower voltage because its a LEDstripe (with three LEDs, configured to run with 12 V) who is fixed in a LEDsheet (Acrylglassplate).
I'm just building a model, the final project will have another electronic controll - so as I had worked last year with the TLC, it was for me the simplest/fastest solution. And I was naivly blended by the datasheet talking about 16 channels which can provide each up to 120 mA, (without making the power dissipation calculation :wink:
I will try it with four or five TLC's and reducing the LEDcurrent by a higher REF-resistor. For the model, it will work without 100% brightness.

mercie,
lr