TLC5940 LED driver - overheating with 2 in series

Discussion in 'The Projects Forum' started by bongo_g, Sep 13, 2012.

  1. bongo_g

    Thread Starter New Member

    Sep 13, 2012
    7
    0
    I am working on a project to use 4 Texas Instruments TLC5940 16 channel LED drivers to control a total of 64 one-meter RGB LED strips. The TLC5940 sinks a constant current when an output channel is on, and this current is set by reference resistor R(IREF). I am using R(IREF)=2K ohms for an ouput channel current sink I(REF) of 20mA.

    Here is a drawing of my output stage:
    [​IMG]
    V(LED) is 12V and R(PU) is 470 ohms. When the TLC5940 pulls 20mA across the 470 ohm resistor, 12V drops by 0.02*470=9.4V to 2.6V. This switches the p-channel MOSFET on, providing power to the LED strip.

    Here is an Eagle schematic of my circuit:
    [​IMG]
    This circuit works perfectly when just a single such circuit is used to control 1-16 LED strips. When I connect the TLC OUT-1 from the first such circuit to the TLC in of a second such circuit (as specified in the TLC5940 datasheet , the second TLC5940 begins to overheat and malfunction. If I use this second circuit by itself it operates well.

    I suspect that I am somehow sinking current from the first circuit thru the TLC5940 in the second circuit, and somehow exceeding the power handling capacity of this second TLC5940.

    Any ideas about what might be wrong with my design, or how I might further test to identify the source of the problem?

    Thank you in advance for any assistance!!
     
  2. bongo_g

    Thread Starter New Member

    Sep 13, 2012
    7
    0
    Thank you @bretm for such a quick reply. In my original post I was referring to labels on my schematic. If you look at the schematic, you will see that SOUT is connected to the 4th pin of the connector OUT-1. This will connect to the 4th pin on the connector IN on the next board, which is SIN.
     
  3. #12

    Expert

    Nov 30, 2010
    16,252
    6,751
    You don't need to use a 470 ohm resistor to pull up the gate of a MOSFET. Change to 10,000 ohms and see if the heating stops.
     
  4. bongo_g

    Thread Starter New Member

    Sep 13, 2012
    7
    0
    #12, what you suggest seems like it would limit the current flowing into the TLC5940. However, I think it will cause an unacceptable delay in turning the MOSFET off. The TLC5940 is using PWM to rapidly turn the MOSFET on/off to affect brightness of the LED, so on/off response is critical. The gate of the MOSFET needs to be pulled up to V(LED) quickly and I do not think a 10K ohm R(PU) pull up resistor will do this quickly enough.
     
  5. #12

    Expert

    Nov 30, 2010
    16,252
    6,751
    Yes, a larger resistance of the gate resistor will cause less current to flow into the TLC5940.

    How fast do you want the LEDs to turn off?
    What part number are you using for the MOSFETs?

    Which pins are you using to connect the output of one TLC5940 chip to the input of another TLC5940 chip?
    I see Sout to Sin on the datasheet. Should not cause problems, but that is whay you asked.
     
    Last edited: Sep 13, 2012
  6. bongo_g

    Thread Starter New Member

    Sep 13, 2012
    7
    0
    I am using a GSCLK (greyscale clock) signal for the TLC5940 of 1MHz, or 1 microsecond per tick. The TLC5940 can set a brightness (greyscale) value for each output anywhere between 0-4095, which it accomplishes by turning the LED on for 0-4095 ticks of the GSCLK. In practice, I am only using 256 variations, so the shortest on or off pulse to the LED is 4096/256 = 16 ticks or 16 microseconds. Thus I want turn on/off time to be some reasonable fraction of this... maybe 4 microseconds? Did I do this math right?

    Also, have you had a chance to look at my circuit schematic? Is there something I have done to cause current to drain from one board through the TLC5940 on the second board?

    Thanks again for your help #12!
     
  7. #12

    Expert

    Nov 30, 2010
    16,252
    6,751
    The only thing I see wrong about the board is that P$1 is not labeled so I can only hope you are giving the chip less than 5.5 volts.

    I could also say that humans can not see things that change faster than about 50 times per second. If the LEDs are for people to see, you do not need to use Mhz speeds. However, I do not see how to slow this chip speed.

    I still need the part number of the MOSFETs to calculate a turn off time.
     
  8. bongo_g

    Thread Starter New Member

    Sep 13, 2012
    7
    0
    P$1 is connected to 5V supply from an Arduino board.

    I am actually controlling RGB LED strips, and the LEDs are pulsing on and off as fast as they are to affect the brightness of each color. Color changes may only be perceptible 50-100 times per second, but that means I need to control each color (R-G-B) 3x that fast, and the TLC5940 is setting an on pulse as some fraction of 4096, so it uses very fine time divisions.

    MOSFET is International Rectifier IRFU5505PbF
     
  9. #12

    Expert

    Nov 30, 2010
    16,252
    6,751
    I did the math and see that you have done well for the time budget. I see 244 cycles per second. Agree?

    I calculate 1.2k ohms maximum to shut off in 4 ticks, 470 ohms will shut off in 1.5 ticks, but I am not the best at this calculation.

    I see no problems where I first looked.
     
  10. #12

    Expert

    Nov 30, 2010
    16,252
    6,751
    Other helpers, please. I am failing here.

    ps, here's the datasheets
     
  11. bongo_g

    Thread Starter New Member

    Sep 13, 2012
    7
    0
    Thanks #12. Let's see if others have ideas?
     
  12. nickelflipper

    Active Member

    Jun 2, 2010
    280
    35
    I had a tlc59401 heat up because of crossed wires on the parallel chip to chip connections. Double check those or for solder bridges. Also the data sheet shows separate sense resistors for each chip, and that's how I did it with no problems.

    The output stage diagram seems lacking. All the led strips I have seen are parallel configuration with their own voltage reducing resistors. So how many parallel leds per strip color channel, or total amperage each FET is trying to switch?
     
  13. bongo_g

    Thread Starter New Member

    Sep 13, 2012
    7
    0
    @nickelflipper - I have checked my chip to chip connections and they look OK. Also, I do have a separate R(IREF) current setting resistor for each TLC5940, if that's what you were talking about.

    Sorry, the output diagram is not completely accurate - it was something drawn by a TI engineer months ago to suggest a circuit for me to try. I am indeed using LED strips with LED wired in parallel and do contain current regulating resistors. I am using RGB strips, but only one color is lit at a time. Each p-channel mosfet is powering 1 meter of LED strip (60 LEDs) @ 12V * 400 mA = 4.8W
     
  14. nickelflipper

    Active Member

    Jun 2, 2010
    280
    35
    @bongo
    60 rgb leds x 20ma = 1200ma per color channel. I can only hope that the schematic is incorrect then on the Iref resistors? as that shows it as being connected to the next tlc.

    It sure seems like the low resistance pullup is competing with the pfet gate for current. I'm am just not familiar with that configuration. I just used a CMOS shift register in a multiplexed pfet row driver configuration and it worked great.
     
  15. ricbordenave

    New Member

    Nov 19, 2012
    1
    0
    Hi! Sorry for my poor english. Exactly the same happens to me, but the differences are:
    12 burned out TLC
    using: p-channel IRF5940
    pullup resistor of 1K
    added an extra resistor between OUT`s and mosfet`s Gate (like a current buffer-spring)
    The mine does not overheat, until they are dead, 5-15 minutes later. after the defunction, they are very useful for cooking bacon...
    I have 5 TLC daisy chained an for THIS strange reason, every single board works well, but when I daisy chained them, TLC on boards 3,4 and 5 burn out. shuld I mabe use ULN`s (for ex ULN2804) and after that N-channel mosfets? Any suggestions?
    Thanks in advance.
     
Loading...