• (Vcc * Icc) + (VOUT X IMAX X (DCn/63) X dPWM X N)   
  (Vcc * Icc) + 2V * 0.12A * ? * 0.9 * 16    (2V * 0.12, Fig.5)
  (Vcc * Icc) + 0.24 * ? * 0.9 * 16             (0.9 assuming high duty)
  (Vcc * Icc) + 0.216 * ? * 16
  (Vcc * Icc) + 3.45 * 1                          (DCn/63) = ? = 1 
  (Vcc * Icc) + 3.45
  Vcc * Icc = ?? + 3.45

So, even w/out (Vcc * Icc) included, 3.45 > 2.45 (dissip. ratings pg.3, at an idealised 25deg_C, and it's worse with heat).

Look, I'm not out to be a d**k, OK, but there's a smoking heap of fried ICs.  For that effect there is a cause, there are no accidents - "Absolute max, Unit" is what they say it is. 
I'm not burning up ICs and some capacitors aren't going to make this go away.

Best of luck.

Do you understand the numbers you plugged in the equation was driving 16 leds with 120mA each?  The discussion was for 20mA each. 

It is understood that all 16 channels cannot be driven at their max.  That is why they include the power dissipation equation.  But at 20mA, the power dissipation is just fine.  Try it.  Let me know what you come up with.  BTW....since dot correction was not stated as being used, DC_n=0.  That would make everything past the '+' equal to zero. 

Does anyone have knowledge of Photodiodes? I plan on using 1 emitter and 1 receiver under each cup to determine whether the light(controlled by tlc) is active or inactive. Could I wire all of the emitters together and have all of the receivers individually relay information to my mega?

Just purchased 50 of these: http://www.fairchildsemi.com/ds/QS%2FQSD2030F.pdf

(10 extra for backup)

Yes, I played with the idea of having a beam across the top and bottom of the staircase for triggering when someone went onto the stairs.  I quickly dismissed the idea for a couple of reasons.  1.  The emitter needs to be pulsed at 38kHz.  This presented a problem to use with the TLC5940 library since the library uses both interrupts.  2.  They draw a large amount of current (rated at 80mA for the ones you bought) and I wanted it to be more energy efficient when idle.

See this for some sample code and a tutorial: http://www.ladyada.net/learn/sensors/ir.html

Perhaps just a light sensor would work for you?  Not sure what exactly you are trying to detect.....light from an LED or presence of a cup.  If it's light from an LED, I can't think of why you wouldn't know if a LED is on or not since your program turns them on and off.  I'm sure I'm missing something though.

Anyone want to jump in and verify that each output has a max of 130 mA? (Of course with the right supply power)

Absolute max of 130mA....recommended max of 120mA.

Along with the total power dissipation being below 1572mW calculated using:
P_D=(V_CC X I_CC) + (V_OUT X I_MAX X (DC_n/63) X d_PWM X N)
You wouldn't even get close to the thermal max if it were limited to just 130mA total output.

The 130mA Absolute Max is for the device in toto, just like the other items called out on page 2 there, it's not "per sink". 
The sum (total) of your outputs, at any time, ought to be less than that absolute max figure, if you go beyond that then stuff starts frying.

[I'm helpful, I'm telling it like it is.]

Tie your reference on page 2 of the spec sheet with the figure directly above it on page 1.  The absolute maximum I_O of 130mA reference is for each of the outputs (0-15) and NOT all outputs combined.  It wouldn't be much of a lighting and display solution for TI if they could only drive 16 LEDS at 7.5mA each, now would it?

Could you maybe make a crude paint drawing of how you jumped from the rails and added capacitors? Any particular areas?

Here's a pic of my bread board setup.....I haven't torn it down yet.  It looks a mess but it works great!

Click for a larger image.  Does this help?

You can't have all 16 outputs on, 20mA each; the Absolute Max output current for the unit is 130mA (PDF, pg. 2)

I've been reading this as the Max for a single OutX!

I would guess either bad connection or voltage issues.  I had voltage issues on some of the rails.  Try running additional jumpers to power the rails, moving where the power is coming into the bread board to a different rail or adding additional/larger capacitors on the rails.  This helped a lot with my issue.  And once I went to my circuit board, everything was solid.

Doing that as we speak, what size capacitors would you recommend?

I agree....that current limit is the absolute max for a single sink.  I'm about to use some TLC5940s to drive some 3w LEDs and plan on tying 8 outputs together to sink 350mA per IC.
P_D=(V_CC X I_CC) + (V_OUT X I_MAX X (DC_n/63) X d_PWM X N)

P_D= (5 * 0.03) + (3.8 * .044 * 8 ) = 1.49W <-- within power dissipation limits of the IC.

As far as capacitors, I really don't know if there is a great way to select the right one or not.  Perhaps someone more knowledgeable in these things can pipe in.  I used a couple of 1000uF and a couple of 100nF.  The 100nF was a given and the 1000uF was experimentation.

Yes! The flickering is mainly coming from the 3rd and 4th Tlc's. I've tried switching to 3.3v for the chips, but that causes even more flickering!

I tried it out today, low and behold another chip has fried!

I would guess either bad connection or voltage issues.  I had voltage issues on some of the rails.  Try running additional jumpers to power the rails, moving where the power is coming into the bread board to a different rail or adding additional/larger capacitors on the rails.  This helped a lot with my issue.  And once I went to my circuit board, everything was solid.

5v 2A is powering just the 20 RGB led's across the anodes. (60 Led's running at 20mA each)

My arduino is powered by 12v into the Vin. The rating of the 12v source is  5A and 60W. (Also powering my fans)
the second 5v power supply is from the arduino's 5v. This is for the Tlc5940's.

Now I have a better picture of what you were trying to explain. 

From the looks of things, I can see nothing wrong with the schematic.  When I fried the two TLC5940s, it was also the first in line now that I'm thinking about it.  I think mine was a loose/bad connection.  Any flickering of the LEDs?

Nice start for your project.  I see where your priorities are though.   

When I was working on my RGB LED Stairs Illumination project with 6 TLC5940s to drive 30RGB LEDs, I initially had some overheating issues as well (fried two TLC5940s) but was able to overcome them after doing a tear down of my breadboard and setting it back up along with adding some additional jumpers across the power rails.  I'm powering mine with a L7805CV (5v 1.5A) regulator with heatsinks added and a 7.5V 3A wall wart.  When I started with a 12V wall wart, the VR would overheat.  My power with the illumination routines I run peaks at about 1.4A which is too close to the max for my comfort and the VR is only barely getting warm but I'm still going to change it.  I have an external 5V 3A power supply that I've tested on my bread board setup running all 30 RGB LEDs at full brightness (~1.8A) for an hour and it worked wonderfully.  The TLC5940s barely got warm while each was supplying 300mA to the LEDs.

1.9A is too close to the 2A capability of supply power.  Your arduino needs power as well as the TLC5940s to run.  You may be dropping voltage down too much with a struggling VR.  Is it getting hot?  If so, do you have heat sinks on it?  Have you measured the measured the max current of the entire circuit?  Are the fans drawing power from the 5V 2A power as well?  What is the rating of the 12V source you mentioned?

Here is the Fritzing file I created for the project: http://www.mediafire.com/?n4wcun4zxn40oxz

Note that I added a 100nf capacitor across the incoming power since my original design and have reflected that in the file.  I also reduced the current limiting resistors from 82Ω to 68Ω to increase brightness a little.  There was also an issue with one of my traces being routed on the wrong layer requiring me to cut it and use a jumper on the underside.  That was also corrected in the file.

If you build this using my PCB design, you might want to increase the hole size for where the anodes attach to the PCB.  This will make it easier to put attach the cube to the PCB.  I've built 4 of these so far and have managed with needle nose pliers and patience though.

If you want one of my circuit boards, I still have a few and can sell you one for $12 shipped in the US.  Or if you want a parts kit with it , I can do that for $30 shipped in the US.  This would include:

• 130 blue 3mm leds
• 30 transistors (2N3904 or 2N4401 depending on what I have on hand)
• 31-1kΩ 1% resistors
• 5-68Ω 1% resistors
• preprogrammed and tested ATMega328
• 16MHz crystal
• 2-22pF capacitors
• 1-100nF capacitor
• power connector
• two sided solder masked PCB
  

Is it a bounce issue?  You might try using the debounce library if you aren't already...

No, not a debounce issue.  I'm checking for a change in the pressure sensor reading to be below a certain threshold.  It then monitors for it to go above the threshold to tell it my foot has lifted off before counting again.  The counts are very accurate.  It is in the functions that the system completely stops responding for some, as yet, unknown issue.  Don't know if it is some kind of buffer overflow or too many functions called within other functions.  I've been spitting out serial data to try to figure out when/why it happens but haven't found anything relevant yet.

I'd suggest you may want to think about either LM2679 (which is a 5 amp version of the regulator) or splitting the sourcing between two regulators.  Up under a staircase, out of sight, I'd just prefer to think there's not going to be anything that will get hotter than say, 150F.  If you aren't seeing it regularly, don't rely on a fan.  Spread that heat out, and running that close to maximum rating on any component isn't going to be a good thing.

I'd think about adding an lcd and a couple of switches in the closet, maybe one of liudr's keypad shields, whatever-- just a simple menu setup.  From there you could even build out code to be able to build sequences within the interface.. but I'm thinking you would more likely just need a scroll-and-select from a list of pre-built sequences, like you have done.

Excellent suggestions.  Thanks.  I am also a bit concerned about using the regulator long term.  At this point with the animations I have, the current may reach 1.3A for a few seconds and that is it.  The regulator barely gets warm with the animations I have set up.  But I have ordered some LM1084IT-5.0 from China.  These will allow me to get more headroom and do an easy replacement of the one I have now without a redesign of the PCB i have on there now.  I am going to also look at getting at least one LM2679, LM2676 or LM2670 to play with.  These are more complicated to set up but appear to generate very little heat.

Currently, I'm working on using a pressure pad tapping routine to select different modes.  It is very simple and elegant.

• Ambient light level has to be high to enable the tap counting (at night, turning on the stair light is adequate)
• When pressure pad is activated, count subsequent taps for 2 seconds
• If there are at least 3 taps, flash the lights on the stairs indicating the selection mode has been selected
• Start cycling through brief visual representations of each animation (about 2 seconds each)
• After each animation, pause 1 second and check to see if the pressure pad is still pressed.  If it is, go to the next animation.  If it is not, select this animation as the one to use.

I have it all programmed out and it functionally looks great, but it is currently unstable going into and out of the tapping routines multiple times and becomes completely unresponsive.  Trying to figure that out.    If I can't, I'll have to come up with something else. 

I just ordered 2 - 1 m strips of the LPD8806 controlled RGB LEDs from Adafruit so will be playing with these as well.

I also ordered 200 - LPD8806 ICs and 400 - 5050 RGB LEDs from a China wholesaler to build my own lighting strips.  It is much more cost effective doing it yourself.  For about $45 (inc s/h) you get 32 RGB LEDs that have 16 LPD8806s on it.  For $57, I got 200 RGB LEDs and 100 LPD8806s.  That's enough for 6 meters worth at the same spacing.

Now I need to get enough SMD resistors, SMD capacitors, and design some PCBs to allow stringing these together as needed.  Only thing lost is flexibility of the strips. 

Oh that makes sense.

So do you upload a new program each time or do you have like 1 microcontroller with 1 program and then another with a different one and then a main switch with on-off-1-2-3?
or how do you change from light show to light show?

Also did you need a carpet foot to get under the carpet?

I upload a new program each time.  Except when I want to demonstrate it, the program will be the basic light up and light down routines.  I am going to code in a double tap on the pressure sensor to put it in demo mode so I don't have to upload a new program each time.  Holidays will require uploading a new program to use special lighting patterns as well.  But again, if I want to make it totally computer free, I could use a double tap to go into a mode selection menu that will alternate between patterns with tapping of a pressure pad.

Right now, since I'm changing the program fairly frequently, I have a laptop plugged into the board in the closet under the stairs and I log into the laptop from my upstairs office (Remote Desktop) to do the programming. 

I have not placed the top pressure plate under the carpet yet.  Downstairs, it is wood flooring and there was already a runner at the base of the stairs so the plate went below the runner.  Upstairs, it is on the carpet and I threw a mat over it.  Once I am confident enough that this solution will work long term, I will put the upstairs pressure plate under the carpet.  I don't know what the 'carpet foot' you mentioned is though.

Okay.
So i'm also curious, what kind of pressure sensor did you use?  Like, say i stepped on the left side and then the sensor was in the center.  How does the sensor cover the entire step?

100# flexforce pressure sensor.


To ensure that it is activated, I sandwiched the sensor between two 1/16" stainless steel plates.  The pressure sensor has a dab of hot glue on each side of it and the plates have dabs of hot glue between each corner.  This ensures that even if you step on the outer edge of the plate, it will still transfer pressure to the sensor.

In the breadboard setup, I also experimented with:

• IR LED emitter and IR detector - this was the most complex solution to set up since the emitter has to blink at 38kHz, be aligned with the detector, and require much more power.  It would also be hard to eliminate dog triggering without mounting it high and then it would be hard to make it blend in.
• IR proximity detector - this brought it down to one package to install but it requires a lot of depth to mount making it hard to blend in.  The range is only about 36" on it but on a staircase, the width is only 30" so technically would work well.
• PIR sensor - this also would require a lot of depth to mount (~2") although not as much as the IR emitter/detector above.  Blending in issue is still there using this.
• Ultrasonic sensor - this could be rigged up two ways for versatility.  Either one at the top and bottom shooting across to the other wall or two in the middle of the facing staircase 180° from each other.  The advantage of the middle is that I could do an installation of one box housing both of them and could track people on the staircase as well as where they were on the staircase.  Theoretically, I could have lit the stair they were on a different color.

I decided on the pressure sensor for simplicity.  It was easy to set up and easy to blend in.  If after a period of time I discover these don't work as well as I would like, I still have the hardware to implement one of the options mentioned above.