• State Resolved
  • Locked Locked
  • Replies 5 replies
  • Subscribers 62 subscribers
  • Views 602 views
  • Users 0 members are here
Support feedback
Options
Options
Similar topics

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.

LM3433: Driver burns with cable between driver and led

Bruno Borges de Oliveira
Prodigy 110 points

Part Number: LM3433

Good afternoon.

I'm having a problem with the TI's driver LM3433 described on link below:

www.farnell.com/.../1640920.pdf

It was assembled with a 56 x 2.5 mm cable between driver and the led, and I'm using a 2kHz PWM pulse to control the led brightness with potentiometer setting to minimal current.

This system worked well for a while, but suddenly the driver smoked and stop working.

I think that the problem is the inductance of cable that probably is generating spikes and burning some components on the driver. The board has elements to add a snubber circuit to reduce this effects, but this elements are not specified on project file. So I would like to ask two questions:

1 - How could project the snubber circuit or find out the components values?

2 - How could I detect the burneds components on the board to replace them?

I apreciate your help.

Best Regards.

Bruno Borges

  • 0
    TI__Mastermind 25500 points

    Hello Bruno,

    That board does some serious current, so we can dig deeper if need be. But your line of thinking is likely correct. With that amount of current any inductance between the PWM shunt FET and the load can cause serious voltage spikes. The end result is usually degrading the shunt FET by over-voltage over time and that is the component that usually fails. Unfortunately it is really hard to tell if something else failed because the spikes can also be on common ground which could affect the IC or either switching FETs. Usually if you ohm each out and there are no shorts they are ok. The shunt FET is usually the only failure in this case though (no guarantee in your case).

    As for a snubber that could for sure solve the damage problem and improve PWM dimming performance. There are plenty of app notes on the web that could likely give you more detail. But the general experimental way is to look at the frequency of the ringing caused by the voltage spikes. Then add output capacitance until the frequency is cut in half. Then you can calculate the corner frequency resistance and add it in series with the capacitor. Often you will need to tweak it a little to filter out the ringing by better than 3dB. Let me know if you have further questions.

    Best regards,

    Clint

  • 0 in reply to Clinton Jensen
    Prodigy 110 points

    Thanks for your reply Clint.

    Yesterday I did able to fix the board replacing LM3433, LM2937 and all three LM5111. Now it is back to the functional state.

    But now, to avoid the problem again, I reduced the cable lenght to 5cm and I will add the snubber. I understood your steps about project the snubber, and found a TI note about this (link below):

    www.ti.com/.../slva255.pdf

    But I still have some doubts:

    1- In the schematic of the board (page 4 of the link at first question) there are three elements to compose snubber circuit (C17, R32 and D4). The steps that you suggests allows me to dimension C17 and R32. How can I dimension D4?

    2- Based on TI note slva255, to look at the frequency of the ringing caused by the voltage spikes I need to put an osciloscope in switch node, wich means the pin 2 of J4, is that correct?

    3- To look at this frequency should the PWM pulse be switched on or I could do the tests only with the led on (without PWM)? I do not want to take the chance to burn the board again ...

    Best regards,

    Bruno

  • 0 in reply to Bruno Borges de Oliveira
    TI__Mastermind 25500 points

    Hello Bruno,

    To answer your questions:

    1. I have found the diode isn't needed. A placeholder was put there just in case we needed to avoid a negative voltage across the LED. But we found it really wasn't required if the RC is selected correctly. If one was needed for some reason a simple 1A schottky would be sufficient, but it shouldn't be needed.

    2. The voltage ringing you need to snub is across LED+/LED-, or more correctly across the shunt FET. You need to snub the high frequency spikes at the rising and falling edges of the voltage across the FET.

    3. You need to be PWM dimming to see the spikes on the edges of the shunt FET turn on/off. But if you want to reduce the likelihood of damage you can reduce the PWM dimming frequency for the test. You only need to see the edges and it sounds like it is the repetitive spiking over time that caused the failure from your description. 

    Regards,

    Clint

  • 0 in reply to Clinton Jensen
    Prodigy 110 points

    Hello Clint.

    Thanks again for your explanation, it's being very useful. I already can see the spikes in the osciloscope like shows the SLVA255 TI Application Report.

    Now I'm trying to following your steps to dimension the snubber to avoid this spikes. When I finish here I tell you about the results.

    Thanks a lot.

    Best Regards.

    Bruno

  • 0 in reply to Clinton Jensen
    Prodigy 110 points
    Hello Clint.

    I finished the snubber project, and the results is a 10uF capacitor with a 70m ohms resistor (wich I replaced by a short). The result is 40% reduction in spikes. I'm using 1kHz PWM pulse, and the spike's frequency is now 222kHz (before it was 649kHz).

    Thanks a lot.

    Best regards.

    Bruno.