• State Resolved
  • Locked Locked
  • Replies 5 replies
  • Answers 1 answer
  • Subscribers 39 subscribers
  • Views 1024 views
  • Users 0 members are here
Support feedback
Options
Options
Related

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.

CCS/TIDA-01420: LMG3410RWHT driver forms very short fault impulses

Max Lashk
Prodigy 180 points

Part Number: TIDA-01455

Tool/software: Code Composer Studio

Hey all,

We are trying to run TIDA-01455 which was produced by given reference design. We have connected it to LAUNCHXL-F28379D by twisted wires (photo below).

It works in common, but at the 200-250 V on the Udc bus, driver LMG3410RWHT starts sending fault signals to the controller.

Our first guess was interference and noise to fault wire but the oscilloscope has displayed this:

Cyan - muxed fault signal from all legs;

Green - fault signal phase U

Red - fault signal phase V

Yellow - fault signal phase W

PWM frequency is 100kHz, dead time 200us, current 0.5A, temperature 40-50 degrees Celsius.

According to the LMG3410RWHT datasheet, the fault reset time is 350us. Also, there is description: "LMG3410R070 provides 1) latched OCP option, by which the FET is shut off and held off until the fault is reset by either holding the IN pin low for more than 350 microseconds or removing power from VDD."

So, we have understood it in such a way that fault signal _shorter_ than 350 us is impossible. Is it correct?

But our tests showed that one leg generates 2us fault width, another 0.2us. Why? Also, we have connected the oscilloscope to fault signals of the driver of one leg directly (FLT_T and FLT_B pins at the output of the optocoupler) and have got the same picture - both switches (top and bottom) generate 0.2us faults simultaneously.

It looks like it is not noise or interference on our messy long interconnection wires - fault signal is clean and reproducible. 

Why fault happens is an interesting question too, but the first question - is it a real fault? Why is signal so short? How to investigate the reason?

What we did:

1) Checked power supply on drivers during fault - it is ok;

2) Reproduced the problem several times - fault time and leg changes, but approximately it is 0.2-2us on average;

3) Connected more ground wires; twist PWM wires with the ground; twist fault wire with the ground;

  • 0
    TI__Expert 3755 points

    Hi Max,

    The board does not look like the TIDA-01420. Did you refer the TIDA-01455?

    Regards,

    Pawan

  • 0 in reply to Pawan Nayak
    Prodigy 180 points
    Yes, you are right, my fault. It is TIDA-01455. I will edit my post and tags.
  • 0 in reply to Max Lashk
    TI__Prodigy 640 points

    Hi Max,

    Thank you for the detailed information on the fault. LMG3410 generates a fault during one of the following three conditions - Overcurrent, Overtemperature and VDD UVLO. If Fault is generated due to Overcurrent, the fault reset time is 350us as you had mentioned - since the fault duration is less than 2us, we can rule out Overcurrent. Since the temperature is below 50 deg. C, the fault is not due to Overtemperature also. So, I guess the fault signal is generated due to VDD UVLO and the UVLO related faults clear as soon as the UVLO condition is cleared - there is no reset time for UVLO as with Overcurrent. I think the UVLO fault is generated due to noise on the VDD rail. I understand that the VDD supply looks good but can you please verify this - did you capture the VDD waveform on full bandwidth or did you apply any filtering during scope capture. Given that the fault appears for short duration(< 2us) and seems to aligned with a switching event, I strongly suspect the noise on the VDD supply generating the UVLO fault signal. 

    • Please share the RDRV value(connected to pin 30 of LMG3410) that you are using - a lower RDRV results in higher dv/dt and this can be the cause of noise. Please try using a higher RDRV value and check if the problem goes away.
    • Please share the layout changes from TIDA-01455 compared to your board - I see there is a connector change and a daughter card(seems to be an iso. DC/DC for LMG3410 driver power supply) that is not on TIDA-01455. 

    Regards

    Siva

  • 0 in reply to Siva Mohan
    Prodigy 180 points

    We have got it.

    The problem was in the PWM_EN signal. It looks like the mistake in the PCB. Look at the picture:

    According to optocoupler datasheet this EN1 signal (pin 7) influence to the operation of the FLT_T_U signal (pin 5), but not the PWMs (pins 3,4). So, to make it work, we have to short PWM_EN at 3.3V. And we did it. But - there is no capacitor at the input of EN1 signal, so the long wire can produce some noise on it, which leads to FLT drop for a short time. We have added a small capacitor in parallel to R20 and it fixes the problem.

    To be sure (before adding this capacitor) we also have checked that there is no real fault (signal drop) from the driver at pin 12 of the optocoupler, but there is at the optocoupler output (pin 5).

    So, there are two problems in the PCB:

    1) PWM_EN does not work as expected

    2) To work with the board as is, capacitors must be added for each optocoupler to eliminate noise on EN1 input.

    PS:

    By the way, it will be nice to add information about the fault length of the LMG3410RWHT for different fault types to the datasheet. It isn't clear from the datasheet that times are different, and this is important to correctly receive faults.

  • 0 in reply to Max Lashk
    TI__Prodigy 640 points
    Max,

    Thank you for the valuable feedback about the problem solution.

    Regards
    Siva