I've successfully sampled the same item from TI more than once, they don't seem to check that you are "resampling".  Done the same with Maxim and ON too.  I would suggest that if 30 days has passed on your last TI sample, get a few more.  Then you can have a few more parts to play with and use to eliminate the chance that the TLC itself is the problem.

OK, I've requested a couple today...let's see what happens.

I should say that I have had lots of problems with TLC5940 chips before and burnt out several of them myself.  For a chip with a "serial" interface it is complex and really you should have control from the uP of seven lines -SIN, SCLK, XLAT, BLANK, GSCLK, and DCPRG and VPRG

Yep, controlling it is quite tricky. The TLC library is quite well done though and takes all the work out of it.

I've done it with just three pins on an ATtiny85: http://arduino.cc/forum/index.php/topic,125212.0.html

Did you receive the samples? How'd it go?

I've successfully sampled the same item from TI more than once, they don't seem to check that you are "resampling".  Done the same with Maxim and ON too.  I would suggest that if 30 days has passed on your last TI sample, get a few more.  Then you can have a few more parts to play with and use to eliminate the chance that the TLC itself is the problem.

I should say that I have had lots of problems with TLC5940 chips before and burnt out several of them myself.  For a chip with a "serial" interface it is complex and really you should have control from the uP of seven lines -SIN, SCLK, XLAT, BLANK, GSCLK, and DCPRG and VPRG as well because it seems to me that you can't load dot correction data or change the source for dot correction data with these just tied to high and low respectively like the example in Arduino Playground.  Arduino Playground says tie DCPRG to high which means take dot correction information from the DC register but tie VPRG low which means you cannot program the DC register.  But the DC register is not guaranteed to be set to any particular value at startup ("The values in the input shift register, DC register and GS register are unknown just after power on.").  How that even works I have no idea.  The MAX7219 is a breeze to use compared to the TLC5940 but no PWM which makes me sad .

This guy controls the extra two lines in his example and uses it to explicitly set GS data:

http://flipmu.com/files/2011/04/Demystifying-the-TLC5940.pdf

If I remember correctly, the DC register/Dot Correction (DC?) is a different method of limiting current, so I should make sure to set that before using the chip? I suppose I will do as you suggested with more samples. I'll stress test one at 5V for a little longer this time as well - I never had problems with 5V until after the chip burned out, but maybe I didn't go long enough. Be nice if there was a MAX7219 for PWM. I have never used it, but I don't know how anything can be as bad of an experience as the TLC5940...

Another possibility is that the chips are getting ESD damage. 

From where? They work fine for a minute. And then spontaneously (or so it seems) stop working. I don't touch the chip or move the bar at all from the time I plug it in to the time that it stops working. And, I could be wrong, but it seems unlikely that a "3rd party" source could knock out two chips.

Where did you buy/get the chips from?

Sampled them from Ti.

Assuming a (low) 3V forward voltage for the LEDs, that leaves a 3V drop on the output of the 5940.  Iref is setting the output to 20mA so that means the chip is dissipating 3V * 20mA = 60mW per channel.  Even with 10 channels you're only at 600mW worst case, which is well below the 3W the DIP package is rated to handle at 25C.

Overall your use of the chip should not result in them failing.  I have never had a TLC5940 fail when used properly.  So let's look for some secondary things that might be caused unexpected problems:

1.  What is supply the 5V to the TLC5940 and are you using decoupling capacitors?
2.  What is the supply used for the LEDs and what is its current rating?  Is it a switching supply?  Does it have a known minimum load requirement?  Have you measured it while the LEDs are operating?
3.  Have you checked with an ohm meter to ensure you picked your resistors for Iref correctly?

1. I'm using a 5V regulator and decoupling capacitors. (I think I remember the regulator getting warm, but for 12 > 5, that shouldn't be too unusual?)
2. It's a 12V switching mode PSU. Rated for up to 600mA. Got it from https://www.sparkfun.com/products/9442
3. The two resistors total 2k, if that is what you mean. (If not, please explain)

Thanks for your time in helping me!

I am getting confusing reports here. Some say that I am within tolerance, and some say I am far out of it.

That is the nature of a free support forum.  You sometimes get people answering questions with only bits of understanding.  Most people do not understand how the TLC5940 works and give (sometimes very) bad advice.

Should I just use a 100R resistor between the TLC and the LEDs?

Absolutely not.  That's what the Iref pin does, it sets the current limit.

The TLC5940 is a constant current source.  It will act like a resistor to limit or allow current to flow, based on the Iref pin.  There is a forumla that is used to determine the current based on an internal voltage reference and the resistor being used.

The 5940 does NOT provide a voltage on its output pins.  So your voltage supply must be large enough to drop the forward voltage of all of the LEDs connected to a channel AND leave some small voltage drop for the output of the 5940.  Otherwise, it will not be able to limit current correctly.

I am not sure I understand how you are wiring up your LEDs.  Could you please provide a diagram of how your LEDs are connected to the TLC5940 and what resistor value you are using for Iref?

Here.
http://i.imgur.com/PuIJf.jpg

You do not need to get the transistors, if you are driving the light bars at 12V at 20mA, you should be fine as long as the light bars are made up of 4 or 5 leds, your voltage drop should be enough for the tlc5940 to handle. 

So basically you are saying that 2V per channel is too much? That is not enough to drive a 4th LED. Should I just use a 100R resistor between the TLC and the LEDs?

If the chip is not getting hot then what about the power supply you are driving it from. It could be that this is shutting down. Make sure the ground of this supply is connected to th ground of the TLC chip and that that is connected to the arduino ground.

I think I can rule that out. I have taken the chips and put them back in a breadboard design (that worked before) and they exhibit the same behavior.

I am getting confusing reports here. Some say that I am within tolerance, and some say I am far out of it. What is the case here? Should I just invest in some PNP transistors and resistors to drive the LEDs?

Again, sorry for the belated post. *adds direct bookmark*

I am only running the LED clusters at 12V @ 20mA.

How are you setting the current and what voltage are the LED clusters dropping, off that 12V?

It's a TLC5940, you set the current via a resistor on the TLC5940 itself and hook the cathodes of the displays to the TLC5940.  If you are using a matrix you have to control the top side of the matrix as well using PNP/P-channel transistors or MIC5891 or whatever.  If you are just controlling 16 individual LEDs you just connect the anode to +V.

"I am only running the LED clusters at 12V @ 20mA"

The resistor sets channel maximum current, not "cluster" current.  There are 16 channels.  16 channels @ 20mA @ 100% PWM = 320mA.  Free air maximum output current for the package is 130mA, supposedly.  I am not good at reading these parts of the specs sheets, can someone take a second look at this?

http://www.ti.com/lit/ds/symlink/tlc5940.pdf

But you say it is not getting hot.  Mine were definitely  getting hot while I abused them.

By 'cluster' I mean 3 LEDs in sequence. So +12V -> LED -> LED -> LED -> TLC5940. This is essentially the same as one LED, right?

Yes, quite enough.  The absolute maximum settings is 6v.  There is a table that says "Absolute Maximum Settings".

That's for VCC and not for the chip's outputs.  The outputs are rated to 18V.

Ok, that's what I thought. This is strange. Is it possible I just got a bad sample?

Sorry I haven't followed my own thread here. I expect to get email updates... Anyway, each LED is 3.2-3.6v, so together that is 9.6-10.8v. Is that enough for the DIP version of the chip to be overheating? Also, if I remember correctly, I don't remember the chip being hot. Nor do I remember running more than one cluster at a time when the chips stopped working. However, I am probably wrong. I guess it's time to break out the transistors and external resistors again...

Ethan

P.S. Thanks for the link. I will have to look at it more in depth when I get some time.

For a while now I have been trying to create an LED light bar using the Arduino and a TLC5940. However, the TLC's keep going belly up. They work just fine for a few minutes and then the LEDs become very dim and (when a nightrider style light program is running) turn on and off at times when the shouldn't. After a few seconds of this, the LEDs fade completely to the point that I can't see them at all. I do not get what is happening. The chip isn't getting hot, I have deca-checked my design, and I am only running the LED clusters at 12V @ 20mA. Has anyone ever heard of this happening before? Does anyone know hot to correct this? And are my two 'dead' chips recoverable?

Ethan

It worked. Thanks again. I feel like an Idiot

Oh, Yes I did I will try it again Thanks!

For a little while now I have been trying to Interface Arduino and Microsoft Visual BASIC .NET 2008. However all my attempts Have failed. I was wonder if someone could tell me how to Interface them. I dont understand Firmata so I am using Serial, But in VB.Net it tells me that Access to COM15 (My arduino COM) is Denied. Can anyone help me? I am fairly new to both Arduino and VB.Net so if someone could take it down to Level 1, I would appreciate it.

Thanks!

Yes, I did follow the link, but I don't quite get it. If I tell the LCD440 Library to autoscroll, why does it go onto a different line? Shouldn't it stay on one line? And why does it get super mixed up when I tell it to scroll with multiple lines? Characters end up where they aren't supposed to be.