• State Resolved
  • Locked Locked
  • Replies 7 replies
  • Subscribers 64 subscribers
  • Views 4961 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.

TPS54560: DC-DC converter frying at startup, large capacitive load

David Bl
Prodigy 90 points
Part Number: TPS54560
Other Parts Discussed in Thread: LM5145, LM76003, LMR14050, LMR16030

Hello,

I have designed a DC-DC converter with the help of WEBENCH, with 48V-input and 24V/3.8A output. I used the components recommended in WEBENCH and followed the layout guidelines given by the datasheet.

The problem is that the converter breaks (short circuits between input and GND, output and GND) at startup.

I tried to start up the circuit with a slowly increasing input voltage, from 0V to 48V, and it worked fine, the output voltage was regulated around 24V when the input voltage went over 24V.

I then tried to start the circuit by applying a 0V-48V step with a lab power supply (current limit set to 2A), and the TPS54560 broke down.

At this moment, the circuit is only loaded by a 2200µF electrolytic capacitor 80mm away. I know that makes a very high inrush current, but the TPS54560 is supposed to have a current-limit mode, right ? And my lab supply was supposed to deliver no more than 2A (which makes 4A at the output), even if a short transient can happen ...

Please find below my schematic and layout :


Thanks for your help,

David

  • 0
    TI__Genius 11085 points

    The biggest problem is the thermal relieves on the PCB layout.

    We understand why thermal relieves are needed for the large pads. But they will cause problem in high power switch mode power supplies. Switch mode power supplies are sensitive to PCB parasitic inductance. The tiny connections increase the parasitic inductance a lot, which will increase ringing, voltage spikes, etc. The high voltage spikes on VIN and/or SW could damage the part, especially 48V is not too far away from 60V Vinmax.

    My suggestion is to remove all the thermal relieves on the pads that belongs to the power delivery path, meaning: input voltage (input capacitors) to SW node to L to output voltage (output caps). Even vias, don't use the tiny connections. Any trace that's in the high current path should be as wide and short as possible. Wide and short traces minimize parasitic resistance and inductance.

    The thermal pad underneath the IC doesn't seem to have top layer copper. The pad has to be soldered nicely to the top layer copper, in addition to the thermal vias to distribute heat to other layers. Heat management is another challenge with such high power design.

    When you say SW short to GND, is the diode still ok? the short circuit is at the TPS54560 or the diode?

    Regards,

    Yang

  • 0
    TI__Genius 11085 points
    The other concern is the 2200µF capacitor. Charging that much of capacitance to 24V need a lot of charge: Q = C*V=I*t
    The TPS54560 internal soft start time is 2.1ms with 500kHz frequency. So the average current needed is I=2200e-6*24/2.1e-3 = 25A.

    I would recommend using another part with soft start pin, so you can prelong the soft start time, such as LM76003, or LM5145.

    Or you can add a load switch with slew rate control between the Vout of the DC-DC and the large capacitors. This way, you can start to charge the caps after start up and control how fast to charge the caps.

    Hope these helps.

    -Yang
  • 0 in reply to YangZhang
    Prodigy 170 points
    hello

    we need dc power supply

    VinMin = 13.0V
    VinMax = 90.0V
    Vout = 12.0V
    Iout = 3.0A

    Design : 1260659/72 LM5116MHX/NOPB

    we count find BSC240N12NS3 G pls tell us equivalent to this part

    Regards

    H.J.Hamsaraj
  • 0 in reply to YangZhang
    Prodigy 90 points

    Hi Yang, thank you for your answers.

    YangZhang said:

    The biggest problem is the thermal relieves on the PCB layout.

    We understand why thermal relieves are needed for the large pads. But they will cause problem in high power switch mode power supplies. Switch mode power supplies are sensitive to PCB parasitic inductance. The tiny connections increase the parasitic inductance a lot, which will increase ringing, voltage spikes, etc. The high voltage spikes on VIN and/or SW could damage the part, especially 48V is not too far away from 60V Vinmax.

    My suggestion is to remove all the thermal relieves on the pads that belongs to the power delivery path, meaning: input voltage (input capacitors) to SW node to L to output voltage (output caps). Even vias, don't use the tiny connections. Any trace that's in the high current path should be as wide and short as possible. Wide and short traces minimize parasitic resistance and inductance.

    I didn't know thermal reliefs could cause enough inductance to break circuits ... I will take this in account and discuss it within my company, thank you.

    YangZhang said:

    The thermal pad underneath the IC doesn't seem to have top layer copper. The pad has to be soldered nicely to the top layer copper, in addition to the thermal vias to distribute heat to other layers. Heat management is another challenge with such high power design.

    Yes it does, it has copper on the top layer and on the bottom layer, it's just not displayed in that image.

    YangZhang said:

    When you say SW short to GND, is the diode still ok? the short circuit is at the TPS54560 or the diode?

    When I removed the IC, the short circuit disappeared, so the diode is still allright.

    YangZhang said:

    The other concern is the 2200µF capacitor. Charging that much of capacitance to 24V need a lot of charge: Q = C*V=I*t
    The TPS54560 internal soft start time is 2.1ms with 500kHz frequency. So the average current needed is I=2200e-6*24/2.1e-3 = 25A.

    I would recommend using another part with soft start pin, so you can prelong the soft start time, such as LM76003, or LM5145.

    Or you can add a load switch with slew rate control between the Vout of the DC-DC and the large capacitors. This way, you can start to charge the caps after start up and control how fast to charge the caps.

    I think this is the actual problem, we are trying to sink way too much current from that circuit ...

    I will have a look at pin-to-pin compatible circuits with soft start (if such a thing exists), or find a way to reduce the inrush current (resistor + parallel switch).

    Thanks again,

    David

  • 0 in reply to David Bl
    TI__Genius 11085 points

    Actually there is a pin-to-pin compatible IC with soft start exist: LMR14050. www.ti.com/.../lmr14050.pdf

    This part has internal compensation. The COMP pin of the TPS54560 is replaced by a soft start pin.

     -Yang

    Modified:

    I just realized that you need 48V input voltage. The LMR14050 wouldn't work. The LMR16030 should work for you. But it is a 3A rated part instead. We don't have a 60V 5A part pin2pin with the TPS54560 now.

  • 0 in reply to YangZhang
    Prodigy 90 points

    Hello again,

    I have added a 5.6 Ohm resistor in series between the filtering capacitors and the large capacitor that was causing a huge inrush current. That seemed to solve the inrush current issue, as the converter now breaks less easily ... But it still does !

    Here is a global schematic of my application :

    This time, the problem seems different :

    • If I slowly rise the input voltage, the converter seems to work fine : its output stays at 24V when the input is over 24V
    • When I suddenly turn my input supply on, which makes the voltage rise in approximately 300ms, the output voltage goes up to 30V, and I have to wait a long time or load the output to make it stay at 24V.

    Is this a normal behavior ? Should I put a minimum load at the output ? Is there any input dv/dt requirements ? (I couldn't find that in the datasheet)

    Here is a scope trace of the output voltage (dark blue) during ramp-up : (no input voltage trace, sorry ...)

    The converter broke again when I tried to apply a load at the input ... I don't know if it is related with the previous issue or not. That loads sinks a huge 3.8A inrush current, but it should be covered by a large capacitor right in parallel. In fact, the TPS54560 should only charge that capacitor and provide the 70mA steady state current for that load ...

  • 0 in reply to David Bl
    TI__Genius 11085 points
    I think it still relate to too much in rush current and too noisy board.

    When you ramp the Vin slowly, Vout follows Vin before Vin reaches 24V, so the large cap at the output was also charged slowly. When Vin goes above 24V, Vout is regulated at 24V. Since large cap has been charged to 24V already, there's no large current needed, so everything is fine.

    If Vin ramps to 24V fast and Vout cannot follow, it will supply as much as possible current to charge the large cap. If load is added before the cap is fully charged, the large current will be there longer, since charge on the large cap is drained out at the same time it is charged.

    After large cap is charged, if load is added, any load current will still come from the regulator, since any charge taken by the load needs to be refilled by the regulator.

    I'd suggest you probe the inductor current waveform to see what's happening when it operates ok and when it was damaged. You can lift the inductor from the board, solder the SW side on with the inductor stand vertically on the pad. Then solder a piece of thick wire from the other pad of the inductor to the VOUT pad of inductor footprint on the PCB. This way, you can use a current probe on the wire to observe the current behavior during start up and loading conditions.

    -Yang