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.

UCC28880: UCC28880: 220VAC -> 5VDC (30mA) - UCC blows up now and then

Part Number: UCC28880

Hi All,

We have built a small form factor, UCC28880 based 85-265 VAC /50Hz to 5 VDC (30mA) supply which in turns feeds into an LDO to provide 3V @ 30 mA 

We have tested it rigorously, but out of every 500 or so, the UCC will blow up without notice. 

We are unable to understand the exact cause, I am detailing the circuit and conditions below, community support would be much appreciated.

BOM: 

U1 -> UCC28880

L1 -> 1mH, 100mA

L2 -> 1.2mH, 100mA

D1, D2, D3 -> IN4007

Conditions of Use:

The circuit is connected in parallel to large load (2 KW power tools (inductive loads)), usually when the 220 VAC is being generated by a generator supply. We have tested the circuit extensively under all possible stress conditions of load, short, surge. But mysteriously every now and then 1 blows up, without much explanation. The UCC usually either blows up when it is powered on for the 1st time or even it works perfectly for the first time and blows upon powering it up again.

We are absolutely clueless about the solution and where to look.

Doubts:

Could there be a design flaw on our part or some measure we forgot to incorporate? Or maybe an issue with the inductor(s) [L1 and/ or L2] which causes it to stay charged and burst current through the UCC28880D?

What could be the possible diagnosis of this issue?


Much appreciate the support in advance!

Thanks,
Rohit

  • Dear Sarkar,

    Thanks for using TI's products in your design.
    In your design, you use 1N4007 for D3 as a freewheeling component.
    In our application note use ultrafast diode STTH1R06A.
    I think long Trr might cause issue during transition, maybe you can change D3 to a suitable one.
    Hope you can find the bug on your circuit

    Best Regards
    Kevin
  • Dear Kevin,

    Thanks for the suggestion, we have identified the same, we shall test it tomorrow and update further.
    But the larger issue is we are not being able to make the circuit fail in controlled conditions, so that we can debug root cause.

    Thanks,
    Rohit
  • Dear Rohit,

    Let's see what happen after you change the diode.
    I am not sure what are you mention here. Is that you want to capture the waveform at the fail time to find out the root cause?

    Best Regards
  • Dear Kevin,

    Exactly, would like to capture the waveform during breakdown in order to find out root cause.

    Would you suggest any method or other suggestions to detect root cause?

    Thanks,

    Rohit

  • Dear Rohit,

    Ok, I think the reason is base on the Trr. When the switch conduct and diode turn off, will have a short period of short circuit which cause by the diode's reverse recovery. After a period of time, the diode's temperature get higher, so the Trr time getting worse which cause higher current spike when the switch conduct. I think the current spike is the main reason to cause your IC broken. So, if you want to get the waveform, you can put a current probe on the input of switch(IC) and adjust the trigger on the scope to get the waveform you want. But it take time.
    I hope this suggestion will help you to find the bug.

    Best Regards
    Kevin
  • Dear Kevin,

    We have tried it by changing the diode, but we are realizing that the problem may be elsewhere. 

    Following are the situations where the UCC is burning out.

    1) If we remove C2 (1uF, 450V) - My assumption is, this is happening because in oscillation the inductor L1 (1mH) is not getting a path to discharge, hence this burning out of the UCC happens the moment the power is applied and not in running condition.

    2) If we remove C3 (1uF, 450V) - The circuit is functioning properly

    3) If we replace L1 with a 0 ohm resistor, the circuit is still functioning properly

    4) If we remove L1, C2 & C3 the circuit is still functioning with certain noise in the line

    I am unable to understand that what situation is causing an obstruction of the path for discharge of the L1, which is probably leading to the backflow into the UCC and causing it to fail (Short circuit temporarily and then burn out to be an open circuit).

    Would request your observations.

    Much Thanks,

    Rohit

  • Hello Rohit,

    You have found some important clues to solving this problem. It seems to me that oscillation in the C2-L1-C3 tank might be either over-voltaging the UCC28880 HVIN/DRAIN inputs or driving them to below GND. Please probe the voltage at HVIN and DRAIN (with isolation probe or isolate the AC input using a 1:1 transformer) to see the stress on these inputs during normal operation.

    There are additional circuit considerations. I used the UCC28880 Design Calculator (available at this link: www.ti.com/.../sluc625 ) and tried to estimate your application specifications. For “Specification” parameters, I used 85Vac to 265Vac input, 4.9V to 5.1V output, 100mV ripple and 30mA load (Iout). In the “Calculation” tab results, row 29 indicates DCM operation and row 28 indicates up to 0.3A peak current in the output inductor (L2 in your design).

    Your original posting indicates that L2 is 1.2mH and rated for 100mA operation. This suggests to me that L2 may be close to saturation at the 30-mA (DC average) load level, and the peak inductor current may be higher than the 300mA peak that the calculator predicts. Please check the L2 current with a current probe to see if the current slope becomes non-linear and curves upward, which indicates saturation. This may not happen immediately but over time as the inductor heats up with extended operation time.

    I suspect that excess peak current can excite the input C-L-C stage and cause it to resonate. The calculator also predicts that the typical switching frequency could be around 6kHz. The C-L-C tank resonance works out to be about 7kHz, which could be excited by the switching frequency to resonate with higher amplitude than normal. Furthermore, L1 is also described as rated for 100mA, and excess peak current from saturating L2 can deplete C3 quickly and build up higher currents in L1.

    Even furthermore, the design calculator recommends to use output inductance of 1.32mH when the maximum input requirement is set for 265Vac. If I reduce this spec to 240Vac max, the inductance recommendation drops to 1.2mH. You describe your input as 220Vac, but that does not account for line variations nor for possible application at input voltages higher than 220Vac (nominal). Please use the calculator with your min and max specifications to determine the appropriate output inductance and peak current rating for it.

    I tis not clear that the line-filter inductor (L1) requires the same peak current rating since C3 should absorb most of the ripple current into the UCC28880, however I suggest that 1mH may be oversized for the load level (5V x 30mA = 150mW), and the tank resonance frequency is too close to the typical switching frequency. I suggest to reduce the value of L1 sufficiently to eliminate any resonance that may be excited by the converter, and to check the peak current in it to make sure its rating is sufficient to avoid saturation. Of course, you’ll need to re-verify EMI compliance after making any changes.

    Regards,
    Ulrich
  • Dear Ulrich,

    Much thanks for the detailed observation.

    We have been using this design with an output consumption of 10mA @3V via the LDO for quite some time. In some of our recent field deployments, it started to blow-up but very rarely. From our today's testing, I think we have been able to pin point the condition under which it is blowing up, request your expert opinion on a probable solution.

    1) tested it from 10-280 VAC/50hz using a variac, very intensively with a 1KW inductive load in parallel. - passed for all tested

    2) tested with 2KV surge - passed

    Failure condition ( 1 out of a hundred or more)

    1) when connected in a line with over 3KW or more of load and suddenly the circuit breakers is switched

    Had to test with many units over the day to make 2 of them fail.

    2) when connected in line with 5kw of load and suddenly the main linear switch connecting the unit was turned on

    I am suspecting a Switching Surge Phenomena by Circuit‐Breakers and Line Switches. But I am using a MOV in the circuit (10D471k), which I am doubting maybe insufficient.

    Would be very helpful to receive your expert opinion from the above observations.

    Thanks and Regards,

    Rohit

  • Hello Rohit,

    I agree with you that it appears to be very likely related to the input line conditions in a real-world environment. The controlled test conditions, even with 1kW inductive parallel load, do not appear to generate sufficient stress to lead to failure.

    However, when tested in an industrial environment with 3kW and 5kW parallel loads, I believe that switching of the AC line results in larger surges and spikes than the test conditions. The AC line itself can be quite inductive and a turn-on current surge or turn-off current interruption can generate large spikes to the UCC28880 input (even with the C-L-C filter).
    I checked the Bourns 10D471K spec and it may not begin to clamp voltage until up to 517Vpk for only 1mA of current. If it has to clamp several amps, the MOV voltage can be in the 600 to 700Vpk range.

    I suggest to try a lower-rated MOV (maybe 10D431K, or even 10D391K for test purposes) to see if the failure rate diminishes or goes away completely. I also still think it would be a good idea to observe the voltage at HVIN/DRAIN, if you have a probe that can handle the peak voltage. And a current probe around the inductors to verify the peak currents in them under normal and abnormal conditions.
    Even thought your typical load seems to be 10mA, the design rating is for 30mA so I believe it should be tested there to verify compliance to the design requirements, not just the typical load.

    Finally, if a lower MOV works, you may need to add some series input-line resistance to limit the current into the MOV to allow the MOV to clamp the voltage lower during a line-spike condition. Also, you may need to consider whether the 10mm disk MOV has sufficient surge absorption capability for a lifetime of service in a harsh ac-line environment. Maybe a 14mm or larger size is necessary to avoid premature burnout. (But that is putting the cart before the horse. First let's see if a lower voltage clamp solves the problem in the first place.)

    As an aside, I also think that D3 should be an ultrafast diode as Kevin recommended earlier. Too much reverse recovery current in the slow 1N4007 diode can also be a possible source of trouble with higher shoot-through current when the UCC28880 turns on and input voltage is at an abnormal peak.

    Regards,
    Ulrich