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.

LMR36506-Q1: Fixed voltage & frequency model oscillates too fast and deivers too low VOUT

Part Number: LMR36506-Q1
Other Parts Discussed in Thread: LMR36506

I have a problem with the 3.3v buck regulator not getting up to correct voltage but appears to deliver 1v.

Chip is LMR36506MSC3RPERQ1 which has fixed 2.2 MHz switching frequency and 3.3v fixed output.

3.3v line (cyan in picture) reaches 1v (which happens to be Vref on adjustable models) when applying 8v to Vin (V+).

SW pin (Yellow in picture) switches with low amplitude at high frequency (>7MHz) and it pulls 102mA and heats up to 52°c. Unfortunately I cannot disconnect load which is supposed to be 2 mcu's.

Is this a output overload issue or is there an error somewhere?

  • Hello Kyrre,

    This is a very strange behavior. Here are two things I would like to start off with:

    1. Can you double-check the device being used?
      The schematic does have the correct part number but I noticed pin 8 is labeled as "FB" instead of "VOUT/BIAS". There might be a possibility of an incorrect orderable part number.
    2. Regardless of the device, your waveform is not making sense. The device is operating about 3x above the fsw range of the device and VIN is only getting up to 2.5V yet able to produce an output of 1.0V. Maybe you are being limited by your input supply? Double-check your input supply.

    Regards,
    Alejandro

  • Yes I believed it may be wrong as well so I did check the part before posting.
    It is marked MCCQ which is correct according to datasheet.

    Full marking:
    MCCQ
    0AJ
    P5DC

    I also loaded the system with an additional 1Ohm power-resistor to see what would happen and it increases current draw from 102mA to 180mA while output drops to 400mv. It gets hotter and the waveform looks cleaner and around 5 MHz.

    Yellow in graph is Vin probed on input capacitor.
    Cyan is output 3.3v rail which is not...

    I am wondering if there may be an overload as the power of 800mW in should make things really hot but my thermal camera only sees the LMR heating up.
    At 76% eff the input power would equate to the output limit of 0.6 amps and 1.04v (could be it's way of limiting?).
    This doesn't match the datasheet though.

  • As a sidenote, I did leave it running for a while, monitoring with thermal cam, and it did not creep. It stayed at 1.04-1.06v and heat was confined to the LMR chip.
    The board is designed to be good thermally as it is very small and will be in a confined space. It is only 35mm diameter circular-ish.
    There was no other hotspot of note and heat from LMR spread quickly in board as planned. (I have soldered wires directly to board in the test as I am waiting for the right connectors for the backside...)

  • I managed to remove my output coil and disconnect FB/VOUT/BIAS pin from load.

    I tried to confirm if this was indeed a 1v vref type device with need of a feedback network on pin 8.
    Using a 10µF capacitor and a 1k/1k network to pin 8 I still get 1v output... Switching frequency still in the 7MHz range.

    This is very strange.
    I do not know where these devices were sourced but I will find out.
    Meanwhile I will order some devices from reputable source and try swapping it.

  • Kyrre,

    Strange, looks like VIN is reaching 8V but SW node is only reaching 2.5V

    Understood; please provide an update when you do swap out the device.

    Thanks,
    Alejandro

  • I received my ev al board (LMR36506MSCEVM) today and will test more tomorrow, but inspection so far reveals chip is not marked according to datasheets.
    This makes me very unhappy.

    Datasheet says it is supposed to have a LMR36506MSCQRPETQ1 which from what I can gather is LMR36506MSCQRPERQ1 in tape format.
    However, device marking does not match anything in datasheet. It is clearly labelled PC1Q and I wonder if it is preliminary.

    I also received legitimate parts from DigiKey which are labelled identically to my current problem-parts.

  • Hi Kyrre,

    The LMR3605MSCEVM features the LMR36506MSCQRPERQ1 (adjustable output) and not the LMR36506MSC3RPERQ1 (3.3-V fixed).

    I checked internally with the team and the top-marking of the device are correct. They are indeed the LMR36506MSC3RPERQ1 (3.3-V fixed) device. All three device codes you have provided have been verified as correct. This verification eliminates the probability of an incorrect device.

    Please provide information after you have swapped out the misbehaving device.

    Regards,
    Alejandro

  • I have swapped chip on a new board, Result is no heat and no switching.

    Next I will try a dead-bug assembly with just the required inductor, capacitors and resistors for activating mode.

  • I tried a dead bug assembly and it works.
    I then reverted to the board with a tighter layout and I can find no fault but it still fails with new chip.

    I notice that VCC from pin 7 does not come up to 3.3v on most of the power-on cycles.
    A few times it does power up and generates too high switching frequency on SW to inductor.

    Can this be some sort of decoupling/racecondition problem from coupling in the layout layers?
    I have not managed to trigger the behaviour on my dead-bug setup but it has much less capacitance.
    The 8v input rises over 20ms in board where it fails but I can not find any timing constraints on Vin in datasheets.

  • Kyrre,

    I'm a bit confused with some of the naming conventions. I would appreciate if you clarified some confusions on my end to make sure I am understanding correctly.

    1. "I have swapped chip on a new board": The "faulty chip" was placed on a new 35mm circular board? or the new TI EVM board?
    2. "I tried a dead bug assembly and it works": The dead-bug assembly was performed on the "faulty chip", correct?
    3. "I then reverted to the board with tighter layout and I can find no fault but it still fails with new chip": A "new ordered chip" was placed on the circular board and it did not work, correct?

    IF I understood correctly (before clarification), this looks like a board/layout problem and not a device malfunction. Test results:

    • "faulty chip" tested on circular board = does not work
    • "faulty chip" tested in a dead-bug assembly = works
    • "new ordered chip" tested on circular board = does not work

    Not once was any chip tested on the TI EVM board, correct?

    Thanks,
    Alejandro

  • Sorry Alejandro, I wrote this after staring at things for 6 hours straight.
    I did not do a good job of explaining.

    I tested the EVM board as-is, and since it has an adjustable chip on it and I could not see any special changes in the layout I decided to go ahead and swap the non working chip on my round board with a new one from DigiKey.
    When this did not work, I mounted a chip from DigiKey upside down on a piece of protoboard and wired out the pins as pr their use.
    Attached pic is before I added the 3.3v feedback etc, but I used a 4.7µF and 100nF capacitor in parallell both for input and output as well as 100nF for bootstrap and a 1µF for VCC. I let VCC provide pullup to PGOOD via 10k and I used a good 10µH inductor from coilcraft. (I used a dual one often used for flyback stuff but it was better to solder than the ones on my board.)

    This starts up every time with VCC rising, then the switching begins at 2.2 MHz and output rises steadily to 3.3v as expected.

    The only difference is the load, and PGOOD that goes to a microcontrollers nRESET line, yet my board does not start.

    Here is my layout. I have tried removing the PGOOD pullup as well as changing the UVLO input to connect directly to VIN via 0R.
    There is a 3.3v plane under this layer connected to the output via microvias (thus invisible) and a groundplane below that. Power comes in from other side of board. There are 2x22µF and a 1µF (not 100nF) on Vin, and 4.7µF and 100nF on output. Feedback comes directly from the output via 3.3v plane.

    Only real difference here is the lacking 100nF capacitor on Vin?

  • I added 100nF capacitor to Vin but that did not help. Then I changed supply configuration to speed up risetime of Vin and this cleared up the latching issue!
    The oscillation is now correctly 2.2 MHz! Vout still only rises to 1v and chip gets hot so there may be a short? I can't find it with the DMM but will continue to look.

    Turns out at least that the internal oscillator in the chip does not like slow rise on Vin!

  • I swapped the bootstrap capacitor assuming something was odd with the low oscillation amplitudes.
    Now it works! But, if I supply it with the slow rising Vin (25ms 0-8v) it latches up again. Not sure why this is but I see VCC does not reach 3.3v.

  • Kyrre,

    That's good news. What voltage are you reading on VCC? The VCC rail typical measures 3.15 V.

    Regards,
    Alejandro

  • Slowing down the Vin risetime makes it fail now and then, and the vcc only reaches 600mv.
    Trying a few more times I can get slow rise to work now and then: (Fast rise works every time.)
    Yellow is VCC. Cyan is Vout (my 3.3v rail).

    Looking at the Vin (Yellow) vs the VCC (Cyan) and using a bit higher voltage (12v vs 8v) to get it to start:

    I cannot find this documented anywhere so something is still fishy. I may just have stumbled on an issue. I will try to start the eval board with slow rising Vin to confirm.

  • Kyrre,

    Any updates on the test ran on the eval board (TI EVM)?

  • I had a day-trip to our manufacturing facilities yesterday so could not update.

    Today I did not manage to latch up the eval board which uses adjustable chip.
    I have tried the dead-bug as well and only managed it twice but did not get a scope shot saved. It appears to be a narrow window of opportunity.

  • Understood; please provide an update when able to.

    Thanks,
    Alejandro