• State TI Thinks Resolved
  • Locked Locked
  • Replies 11 replies
  • Answers 1 answer
  • Subscribers 65 subscribers
  • Views 1960 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.

UCC256404: PFC + LLC power supply problems

Alexander Kharitonov
Prodigy 20 points
Part Number: UCC256404
Other Parts Discussed in Thread: UCC24624

I have recently designed an AC-to-DC power supply (100-250VAC input, 18VDC 8A output) using the UCC28056B as a PFC boost converter and the UCC256404 and UCC24624 as a resonant LLC converter with synchronous rectifiers. See the schematic images below for details.

To design it, I used TI Power Designer and the IC datasheet information + Excel design tools (see the filled out versions posted below) as a baseline, as well as the provided SIMPLIS models and testbenches for tuning, simulation, and compensation loop design.

I have built a prototype of this design, starting with the UCC28056B PFC boost converter alone, which appeared to work correctly on its own (with an external VCC supply).

However, once I tested the complete prototype (with LLC converter, no output load applied), a few seconds after applying the input voltage (240VAC 50Hz), both LLC converter MOSFETs failed simultaneously, in a short-circuit mode (measured after the failure).

I measured the circuit before testing it, to ensure voltage divider values are correct and there were no short circuits or other previously identifiable problems.

I have no immediate clues as to what's wrong with my design to make such a failure happen and I don't know how I can test the circuit or figure out the problem without immediately destroying more components. Can you see anything wrong with my design? Do you have any hints for how I can fix my design to prevent such a failure from happening again?

Thank you very much in advance.

Kind regards,
Alexander Kharitonov

7658.UCC28056X_Design Calculator.xls8080.UCC25640x Design Calculator Rev4.0.xlsx

  • 0
    TI__Genius 16155 points

    Hi Alexander,

    Currently the gate resistance being used is very small. Please use the gate turn on and turn off resistors as shown below:

    I dont see any other issues with your schematic and the design calculator. 

    While debugging, please do following steps: 

    1. Try to debug only LLC portion by applying only DC input voltage directly to the LLC input and you can disable the PFC controller by removing D101. 

    2. In order to start the LLC controller, please remove D203 and Apply VCC voltage of 26V. Once you see RVCC pin voltage, you can reduce the VCC pin voltage to ~15V.  By doing this, use of AC input source is eliminated.

    3. Rrefer section 3 of the app note to know the expected voltages at each pin of the IC.

    Regards

    Manikanta P

  • 0 in reply to Manikanta Pallantla
    Prodigy 20 points

    Hi Manikanta,

    Thank you for your analysis and reply. I have now attempted to debug my modified circuit with your given steps.

    However, I am encountering weird behaviour regarding the shared BLK voltage divider (used for the UCC28056B VOSNS pin, as well as the UCC256404 BLK pin).
    When I attempt to debug the circuit as you described, with the PFC controller unpowered (D101 removed), the BLK voltage only reaches ~0.5V, far below the start voltage. Since the resistor values are correct, I assume the unpowered PFC controller is pulling it down.
    When I replace D101 to power the PFC controller, the BLK voltage now rises to above 1.0V as expected, which allows switching to start - but the PFC controller appears to pull the voltage down regularly (at ~1kHz), seemingly causing the LLC controller to enter an input under-voltage fault state shortly after switching begins. After 1s, the controller resets, and the same pattern repeats.

    I have attached an oscilloscope trace showing this situation.
    Channels are (in order 1-4, top to bottom): RVCC pin voltage, LLC input (bulk) voltage, BLK pin voltage, LO pin voltage.

    What could be causing such behaviour? Is the usage of a shared BLK voltage divider for PFC and LLC not possible here, or is there something else going wrong?

    Regards,
    Alexander Kharitonov

  • 0 in reply to Alexander Kharitonov
    TI__Genius 16155 points

    Hi Alexander,

    Could you depopulate the PFC controller as well and could you check whether LLC is working?

    Regards

    Manikanta P

  • 0 in reply to Manikanta Pallantla
    Prodigy 20 points

    Hi Manikanta,

    I have depopulated the PFC controller and tested the LLC circuit again.

    First, I did a test run with a relatively high input voltage impedance - now, the BLK pin voltage followed the input voltage as expected, though the switching caused the input voltage to quickly fall below the turn-off threshold, stopping the switching again. So this appears to confirm my hypothesis that the PFC controller was interfering with the voltage divider.

    In an attempt to fix the input voltage drop, I supplied the LLC input voltage from a low-impedance regulated supply instead. This, however, quickly led to a similar failure to what I described in my very first post - both LLC MOSFETs failed in a short-circuit mode (gate, drain, source all shorted together).

    Excessive gate voltage should not have been a cause here, since I increased the gate resistances as you recommended (47R charge, 1R discharge through diode), and additionally protected each MOSFET gate with an 18V zener diode (MOSFET used is rated for 30V Vgs peak) - therefore I am unsure what could be causing this failure once again.

    I recorded the event with my oscilloscope, though I wasn't monitoring the ideal set of signals for failure analysis (as the failure was unexpected, of course). Still, maybe they can give some hints for the cause of the failure?
    The recorded channels are RVCC pin voltage, LLC input voltage, BLK pin voltage, and low-side MOSFET gate voltage (driven by LO pin, measured directly at MOSFET). Here is the full operation time-span, as well as the failure event itself:

    Do you have any idea about what could be happening here?

    My own hypothesis is that the MOSFETs failed first, which caused high voltage to be fed to the gate (clamped by the zener diode), feeding back through the controller to the RVCC supply, causing the spike visible there, until the input supply protection turned off the supply.

    However, this still does not explain the cause for the MOSFET failure itself.

    Regards,
    Alexander Kharitonov

  • 0 in reply to Alexander Kharitonov
    TI__Genius 16155 points

    Hi Alexander,

    Could you try the following changes to the board:

    1. When the controller starts switching, I would expect the BLK pin voltage not to go below turn off voltage threshold since there is a ~150uF capacitance at the LLC input voltage. Could you increase the BLK pin cap value to the 1nF from 330pF.

    2. Let's reduce the startup current in the tank by increasing the startup frequency. So, could you increase the LL/SS cap value to 1uF.

    3. The turn off of MOSFETS needs to be faster. Since you connected 47ohm, this would make the turn off of the primary mosfets very slower. As I mentioned before, for faster turn off, please use a separate diode with 1ohm as shown below:

    4. Could you also use primary Mosfets of CFD7 series which has less reverse recovery charge compared to what you are currently using.

    5. I am not sure about the purpose of RV201 in your schematic. Could you depopulate that while testing.

    6. On the secondary side of the LLC, could you depopulate the synchronous mosfets and the controller. And instead, can you use Schottky diodes just for debugging. Once the primary side controller works, you can repopulate them back.

    7. Replace D205 with 15V or 12V Zener. Currently 20V is being used.

    8. Also, could you replace C213 with 1uF cap.

    Regards

    Manikanta P

  • 0 in reply to Manikanta Pallantla
    Prodigy 20 points

    Hi Manikanta,

    Thank you for your suggestions.

    I already had suggestions 3 and 5 implemented at the time of my previous post, so those were likely not problems.
    Since then, I have implemented suggestions 1, 2, 4, and 7, replacing the MOSFETs with IPP65R155CFD7 specifically.

    This appears to have fixed the problem with the LLC controller - it is correctly regulating 18V on the secondary side now. I suspect that the main issue was indeed the MOSFET choice, where the C7's body diodes were too slow to handle the inductive voltage spikes at the switching node.

    However, unfortunately, the issues haven't ended there - I decided to test the full system (PFC + LLC converters) as a whole, and it did seem to work as expected, handling no-load and moderate-load conditions quite well. However, after running for a few minutes at moderate load (~40-50W output power), the circuit suddenly turned off.

    Further investigation and testing after this failure shows that the LLC part of the circuit still works as before, but the PFC controller seems to have failed, as applying VCC voltage to just the PFC controller (using a lab power supply) effectively shorts the VCC rail to ground, just heating the PFC controller chip.

    I have measured the other components of the PFC circuit, and none of them appear to be damaged (shorted or anything), so I am really not sure what could have caused this failure. Based on my schematic of it (see image in original post), do you have any hints for what could cause the PFC controller to suddenly fail in such a fashion, and how I could possibly prevent further failures after replacing the controller chip?

    Regards,
    Alexander Kharitonov

  • 0 in reply to Alexander Kharitonov
    TI__Genius 16155 points

    Hi Alexander,

    Could you first check LLC only upto full load. Once LLC power stage works, you can start connecting the PFC.

    Also, could you let me know how are you connecting the lab power supply to the PFC controller? Its already connected to RVCC pin of the LLC controller right through a diode. What's the voltage you are applying from the lab power supply?

    Could you power the PFC controller from the RVCC pin. That way startup will be smooth in the combined LLC PFC power system.

    Regards

    Manikanta P

  • 0 in reply to Manikanta Pallantla
    Prodigy 20 points

    Hi Manikanta,

    The LLC power stage works as expected at full load.

    As for the PFC controller, during my initial testing of the combined system (PFC + LLC), it was powered from the RVCC pin through the diode D101, as you described. In this scenario, the PFC+LLC system worked for a while, but then failed.

    After the failure, I noticed that RVCC was failing to start up once power was applied (specifically, it was power-cycling once every second), so I removed D101, which allowed the LLC stage to work by itself again.

    Then, I tried to power the PFC controller alone, using a lab power supply set to 13V at 0.1A max (which, in the past, was able to power the PFC stage normally). However, this caused the lab power supply to be current-limited at ~3V 0.1A, indicating a short circuit failure of the PFC controller.

    Regards,
    Alexander Kharitonov

  • 0 in reply to Alexander Kharitonov
    TI__Genius 16155 points

    Hi Alexander,

    I am not sure what caused the short. So, Let's debug only PFC section for now. Could you replace the old PFC controller with a new controller. Also, remove the LLC controller for now. 

    Could you increase the VCC cap (C107) value to 10uF. Let's make sure PFC is working up to full load.

    Regards

    Manikanta P

  • 0 in reply to Manikanta Pallantla
    Prodigy 20 points

    Hi Manikanta,

    I have replaced the PFC controller and started debugging the PFC section.

    At lower load, in DCM operation, the circuit appears to work without problems - here are some DCM waveforms.
    Channels are, in order 1-4, top to bottom: Output voltage, MOSFET VDS, MOSFET VGS, ZCD/CS Pin Voltage.
    Note that the VDS trace is not accurate due to impedance in my test setup - so the VDS rise/fall times appear much slower than they actually are. The ZCD/CS trace is much more accurate.

    However, at higher load, when the circuit is operating in CRM, something goes wrong. In particular, it operates as intended for multiple seconds, delivering regulated output power - but after a while, the audible noise of the circuit starts changing (a kind of hiss with rising frequency), and a few seconds later, the circuit fails. This time, it caused the controller and the MOSFET to fail (Drain-source short circuit), also destroying the current sense resistor.

    I tried to measure the behaviour of the circuit, both in the first few seconds of CRM operation, as well as once the audible noise starts changing. I don't really see anything wrong with this CRM operation (comparing with the waveforms given in the application note). Here are some traces, both an overview, and detail views at different AC input wave positions:

    So, in summary, the PFC circuit fails destructively in CRM operation, and I'm not sure what could be causing this. Do you have any ideas for what I could try in order to fix this?

    Regards,
    Alexander Kharitonov

  • 0 in reply to Alexander Kharitonov
    TI__Mastermind 40070 points

    Hello Alexander, 

    In my experience, when a circuit appears to work "for a while" then begins to make noises and fails, it is usually (not always, but usually) a sign of some thermal runaway.  This happens often where a converter works fine for a long time or indefinitely at lighter loads, but then fails after a short time when the load is increased.  

    In the case of your PFC, the three main components I suspect are the MOSFET Q101, output diode D103, or boost inductor L102. 
    Please examine the heat dissipation capability of your design as regards these components.  D103 is an SMD and its pads may be too small to dissipate its losses.  Q101 is TO-220 and may need a suitable heatsink system.  L102 may be saturating if is is not properly designed for the application. 

    If you have a thermal camera available, you can run your prototype at low power and see which components start to get hot. 
    Gradual increase in load power can reveal the component which is getting hotter faster than the load is going up, and you can shut down the test before it runs away to destruction.   

    I suggest that, until the PFC prototype is proven to operate reliably from 0 to 120% load over min to max input line voltages and frequencies, there is little point to attaching the LLC circuit and running them simultaneously.  Each converter should be debugged independently to verify expected operation before connecting them together.   Where signals may have to interconnect (such as RVVC and BLK), the PFC section can have test circuits which emulate how the LLC would drive those signals, and the LLC section can have test circuits which emulate how the PFC would load those signals.  Accurate emulation should include conditions where the PFC or LLC sections would be powered and unpowered. 

    Regards,
    Ulrich