This thread has been locked.

If you have a related question, please click the "Ask a related question" button in the top right corner. The newly created question will be automatically linked to this question.

  • Resolved

tlc5940 damaged

Prodigy 30 points

Replies: 9

Views: 9052

Hi,

 I am designing a led display with the TLC5940 drivers. I use 7 of them connected by the serial outputs to the next serial input, so cascaded as the datasheet described. The dispay is multiplexed 10 ways with 1.5msec timeframes, so the grayscale clock is defined to do 4096 cycles in this time. I don't care about the dot correction or the error output, so the mode is selected always to use GS register (DCPRG and VPRG is wired to GND). I drive them with 5 signals (GSCLK, SIN, SCLK, BLANK, XLAT). VCC is stable 5.2V. See the schematic here.

My problem is that, TLC ICs are randomly get damaged and I can't find the reason behind it, because I use normal VCC and only 20mA/output (2k resistors). Some of them were damaged immediately during the first power time, some of them a few minutes later, and some of them are still working. The ones which were damaged are getting very-very hot when I apply the VCC.

 I guess my design and the controlling algorithn should be correct, because there are pieces which are still working, and drives the LEDs fine in the same socket, where the previous IC died.

Have you got any idea, or tip about the conditions when this type of driver can be damaged? Please help, as I have 10 dead pieces now :-(

Best regards,

 Huszty Gergo

 

  • dear Huszty Gergo,

    Please check the ESD diodes integrity on the damaged parts.

    Do you know how to perform that?

    Could you also please provide what and how is connected to JPx connectors?

    Have you read our SLVA267A?

    V.

  • In reply to Vincenzo Pizzolante:

    Hi Vincenzo,

     Thanks for your answer!

     Sorry, I don't know how to check the ESD diodes, or where are they at all :), is it described in the datasheet?

    JP1-JP5 : directly connected to LED cathodes. One pin is connected to 10 LED's cathode, but one of them are powered once, because of the 10 way multiplexing.

    JP6: directly connected to LED anodes. One pin is for 100 LEDs. Those FETs are performs the multiplexing.

    JP7: not used yet

    JP9: GND and +5V pins (1,2,11,12) are not used, the other 7 pins are coming directly from an ATmega32 uC, which is connected to the same supply (5.2V / 4A).

    I looked into the SLVA267A document. The only difference I found in schematic is that, there is a separated VCC made for the LEDs, does it matter? I've seen several applications on the internet, using common VCC like mine.

    Can this issue caused by any bad algorithm (for example turning off the LEDs anode, while the TLC5940 output is sinking, etc.) ?

    Hint: previously this application was designed with TLC5947s, and it was working fine. Later I wanted more then 30mA / channel, thats why I moved to TLC5940, however they are still running on 20mA. I want to increase the current to 40-50mA after all this kind of issues are fixed.

    Br,
     Gergo

  • In reply to Gergo Huszty:

    Ok, there are a few marginalities in your system.

    First of all, IRLIZ24 are power mosfet that you are driving with logic gates’ output that are not strong enough to switch them on/off. You can replace them with the smaller, cheaper and better performing (in this case) with 2N7002 or even smaller signal mosfet; if power / current rating of 2N7002 is not enough for your purpose then a mos driver is needed.

    Second, you are using these mos as pass-gates on the VCC / VIN power rail: normally p-channel mos type are used for driving high-side pass-gate, while the n-mos are more suitable for low-side pass-gates, but you can live with this (even if this is not the canonical way).

    Third, I didn’t get why are you overriding the switches already embedded in TLC with the IRLIZ24N… refer to application example (figure 22 on page 21) of TLC datasheet.

    Last but not least, the absolute maximum input voltage of the TLC is 6V, max 5.5V is recommended: you are using a power supply 5.2V + tols; depending on layout quality the disconnection of the load could cause load dump (overshoot on VCC that could reach 6V and destroy the devices). This, joined to the fact that the big IRLIZ24N don’t switch on/off properly, could be the cause of the failure you observed.

    V.

  • In reply to Vincenzo Pizzolante:

    Hi,

     Maybe I didn't tell you the details. So I use the FETs because I want to drive 1000 pieces of LEDs in one matrix, thats why I need to somehow multiplex it. I reached this by spliting the display into 10 layers, and driving one of them at once. Therefore I need TLCs for only 100 LEDs, not for 1000 (not to say the power consumption will be also suitable).

    At the beginning, I measured the 4017 outputs, which controls the FET gates, and the active outputs were above 4V. According to the IRLIZ24N's datasheet, at VGS=4.0V, the ID can be even 10A.

    I have to use such big FETs because of the 100 LEDs current is at least 2A (and if my target is 40mA/LED, it can be 4A at all), so I can not use signal MOSFETs. The easy way was to use TTL input gate capable ones, that's why it is IRLIZ24N. I agree with you in that, I should use p channel FETs, but this can not cause the failure, as you said.

    I didn't notice any false light from the layers, that are currently switched off, if you were poinging to this with the switch on/off problem.

    Maybe there are really some glitches in the supply voltage, which I could not measure and above 6V.

    My questions are:
    - What do you think, if I try to fix this "supply over 6V" problem with some fast overvoltage protecion, can it solve my problem completely?
    - What do you think, how could the TLC5947 s survive the same conditions (the modification was only to add GSCLK line to the ICs in the TLC5940 version) as it also has 6V absolute maximum power supply voltage.

    Please tell me if I misunderstood your findings.

    Have a nice day,
     Gergo

     

  • In reply to Gergo Huszty:

    Ok, let me say that I was a led display designer....

    in big panel solution / screens for pubblicity we use a TL592x family based.... plenty of them connected in daisy chain...

    because the size of the solution (we are talking about pubblicity screeens, something like 6x4m) the problem you have to face with is the power distribution...

    normally a minimum of 12V to supply the led was used, while a 5V was locally obtained for the ICs...

    the point is that VLED will drop along the panels because the R of the PCB copper, because the connections...

    in your system, you are using the 5V to supply both the led and the IC: very bad idea for big size panels (for little solution it could work because little also means low power consumption)...

    if you are going to bias 1000 led @ 40mA each you need 40A, that will cause the problem I mentioned in this and the previous email...

    you don't need an High-Side switch, I can't see the reason for that...

    furthermore, if you use IRLIZ24N turned on / off by logic gates that are little current capable you can turn on/ off them for sure, but the time needed for this will be unacceptable for the purpose...

    always because the size, the 2nd biggest problem you will face with will be, for sure, the synchronizazion: even in daysi chain connection (also called snake) the data transmission is not a problem because DATA_IN / DATA_OUT are available for the pipeline, while the clock / sinchronization signals aren't... this means you have to use a buffer to drive the clocks at the source and you need to terminate the line (long pcb tracks act as transmission lines at high frequency, causing reflecting waves that lead to oscillation / overshoot / failure in communication....

    V.

  • In reply to Vincenzo Pizzolante:

    Sorry to revive this thread but I've found it helpful. 

    I am attempting something close to what the original poster is except I've been driving tri color leds.  The common cathode leds are each connected to one pin of a tlc59401 and the anodes are connected to 3 different mosfets, one for each color.  This way Each color can be displayed in rapid sequence to display a 36-bit color value while only using 1/3 the number of drivers.  The disadvantage of course is code complexity and brightness.  I'm only using common 20ma leds and the project requirements do not require high brightness.  

    I also am experiencing power supply over/undershoot problems which I should overcome with a better supply/filter.  

    The question I have is this:  Is there any problem with driving LEDs in this manor?  I've read a ton of literature on the led drivers and nothing sticks out at me but I thought that this technique would be more common that I've seen so far. 

    Thanks

    -Jay

  • In reply to Jason Sherwood:

    Hello Jason,

    Just from physics I do not see any reason why you should not be able to drive the LEDs this way. There are some practical issues that I can see, like keeping the brightness when the color is changed (e.g. when turning off the blue in a RGB LED and the common cathode is connected to a 20mA sink, now the red and green will have more current going through them) or keeping the color when the temperature changes (e.g. normally red, green and blue will vary differently with different temperatures and therefore when the temperature shifts, it might be that now the green one takes more current than the red one).

    But as long as it is not extremely important to keep the color exactly the same (as it is in a big LED screen for example), this attempt should work.

    Best regards,
    Brigitte

    Best regards,
    Brigitte

    For more information on buck-boost devices have a look at www.ti.com/buckboost

  • In reply to Brigitte:

    Thanks for your reply Brigitte.  Only one color will be on at any given time.  I quickly cycle through each color  in a manor similar  to a DLP television or projector.   I'm not worried enough with color accuracy that temperature induced color shifting will be a concern.  The problem I've been having recently is excessive noise on the LED drive rail.  I'm driving the tlc59401 at 3.3v but the LED voltage is 5v.  the 3.3v rail stays pretty stable however the 5v rail is all over the place, even with just two led drivers enabled. Do you happen to have any pointers on how to cope with the drastic change in current draw between the time one greyscale cycle is over and the next starts?

    Thanks again for your reply

    -Jason

  • In reply to Jason Sherwood:

    Hello Jason,

    The switching action of the external FETs you have in the anode might generate this issue. It should be possible to stabilize the 5V by adding small capacitors close to the switches. Do you maybe have some current limit in the 5V rail. This could be the power supply you are using, or the width of the lines you are using to connect to the 5V. It could also be helpful to use a plane for the 5V to stabilize it.

    When you count all the external FETs and the additional software generation and evaluation, is it still less expensive to use common cathode parts instead of common anode ones with one current sink on each of the colors?

    Best regards,
    Brigitte

    Best regards,
    Brigitte

    For more information on buck-boost devices have a look at www.ti.com/buckboost

This thread has been locked.

If you have a related question, please click the "Ask a related question" button in the top right corner. The newly created question will be automatically linked to this question.