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.

UCC28740: Oscillation Stop Behavior of UCC28740

Guru 12065 points
Part Number: UCC28740

Tool/software:

Hi,

We are currently investigating an issue related to oscillation stop behavior in the UCC28740.
Although the exact root cause has not yet been identified, we have observed the following phenomena in the customer’s actual circuit:

  • A large ringing occurs on the bias winding voltage immediately when the FET turns ON.

  • This ringing may cause false detection on the CS voltage, which leads to the DRV signal being turned OFF within approximately 255ns.

  • This behavior is always observed immediately before oscillation stops.

  • The ringing continues superimposed until the auxiliary winding voltage reaches +20V, with a ringing frequency of approximately 12MHz, and takes about 1μs to settle.

  • At this point, we suspect that this ringing is primarily caused by the transformer design.

Based on the above, we kindly request your confirmation and response to the following questions:

Q1.
Please clarify the exact timing of the I_VSL data latch.
For example: Is the CS voltage latched A [ns] after the DRV signal turns ON?

Q2.
On page 16 of the datasheet (CV mode operation), please indicate the switching frequency (fsw) when the peak current Ipp equals 1/4 of the maximum Ipp.

Q3.
In application note SLUAAC5, the following statement is made:

The Vaux ringing in Figure 8-1 is excessive and rings down below ground during tLK_RESET.
This behavior is known to trigger a UVLO fault and shut down the converter.
This is because when the VS pin crosses ground it activates input UVLO fault.

Does this mean that if the bias winding voltage undershoots below GND during the tLK_RESET period (when DRV is OFF), it can immediately trigger a false UVLO fault? Please confirm.

Q4.
Regarding Equation (24) in SLUAAC5, can the described fsw(max) be interpreted as the actual maximum switching frequency in CV mode during operation?
The maximum switching frequency observed in the customer’s circuit is approximately 70kHz.

This matter is time-sensitive. If a complete response is not immediately available, we would greatly appreciate it if you could provide us with an estimated timeframe for your reply.
Thank you very much for your kind support and cooperation.

Best regards,

Conor

  • Q1: leading edge blanking is 230nS nominal. So no CS protection within the first 230nS when the DRV is activated. The delay is negligible.

    Q2: the switching frequency will vary from 170Hz to 30kHz, depending on the load condition.

    Q3: The UVLO fault has to be sampled in three consecutive switching cycles before the flyback controller stop switching

    Q4: Yes

  • Hi Ning,

    Thank you for your reply. I'm not sure if I've conveyed the exact intent of my question, so I'll comment below. I'd appreciate your response.

    Q1: leading edge blanking is 230nS nominal. So no CS protection within the first 230nS when the DRV is activated. The delay is negligible.

    We understand the LEB (Leading Edge Blanking) function of the CS pin. In the current observed behavior, we suspect that immediately after the LEB period ends, the CS voltage reaches the threshold and is falsely detected, causing the FET to turn off immediately. This results in extremely short ON-time pulses, and we believe that some protection function is subsequently triggered, leading to the cessation of switching. Our inquiry is specifically regarding when the IVSL current is sampled. The datasheet defines IVSL(STOP) as 80 µA and IVSL(RUN) as 225 µA, but we would like to know the exact timing during the switching cycle at which this IVSL current is evaluated. Could you kindly provide clarification on this point?

    Q2: the switching frequency will vary from 170Hz to 30kHz, depending on the load condition.

    We are unsure where the 30kHz value originates, but we do recognize that the switching frequency can decrease to as low as 170Hz depending on the load. In our system, false detection on the CS pin tends to occur during startup under light load conditions, where the switching frequency is typically in the range of 25kHz to 30kHz. Our inquiry relates to the diagram on page 16 of the datasheet, which shows that IPP does not fall below 1/4 × IPP_MAX. We would like to know the specific switching frequency (fsw) at which IPP becomes exactly 1/4 of IPP_MAX. From the graph, we can infer that this point lies somewhere between 32kHz and 3kHz, but we would appreciate it if you could provide the exact or typical value.

    Q3: The UVLO fault has to be sampled in three consecutive switching cycles before the flyback controller stop switching

    Does the term “UVLO fault” refer to the condition defined in the datasheet as IVSL(STOP) = 80 µA?
    In SLUAAC5, Figure 8-1 includes the following explanation:

    "The Vaux ringing in Figure 8-1 is excessive and rings down below ground during tLK_RESET. This behavior is known to trigger a UVLO fault and shut down the converter. This is because when the VS pin crosses ground it activates input UVLO fault."

    As described, excessive negative ringing on the auxiliary winding (Vaux) during the tLK_RESET period may cause the VS pin to drop below ground and trigger a UVLO fault. Could you please confirm whether this UVLO fault refers to the condition where IVSL falls below 80 µA (IVSL(STOP)) and is detected for three consecutive switching cycles, thereby shutting down the converter?

    Q5: (Additional question)
    We would like to confirm our understanding regarding the parameter "Kvsl = 2.8 typ" listed in the UCC28740 datasheet. Our interpretation is that Kvsl represents the ratio of IVSL(RUN) to IVSL(STOP), serving as a design reference for the hysteresis margin between restart and shutdown thresholds in the UVLO function.
    We believe that Kvsl is not directly used in the internal comparator logic for UVLO detection, but rather is provided as a typical design ratio between the two absolute threshold values. Could you kindly confirm if this understanding is correct?

    Thanks,

    Conor

  • IVSL(STOP) as 80 µA and IVSL(RUN) as 225 µA refers to VS pin for OVP and input UVLO protection. It has nothing to do with CS pin protection. The CS protection threshold is 1.5V and it is not filtered by LEB. (page 13 in datasheet).

    Please read this paragraph for start up sequence.

    Q3: yes

    Q5:yes, it is just a ratio as shown below.

  • Hi Ning,

    I'm sorry that my comment was insufficient. I'd like to understand it accurately, so could you please comment again?

    IVSL(STOP) as 80 µA and IVSL(RUN) as 225 µA refers to VS pin for OVP and input UVLO protection. It has nothing to do with CS pin protection. The CS protection threshold is 1.5V and it is not filtered by LEB. (page 13 in datasheet).

    Please read this paragraph for start up sequence.

    We would like to confirm our understanding regarding the evaluation timing of the IVSL current (from the VS pin), which is used for UVLO and start-up conditions in the UCC28740.
    Our interpretation is as follows:

    • IVSL current flows during the period when the MOSFET is ON (DRV = High), and the auxiliary winding polarity clamps the VS pin to GND

    • During this time, current flows through the external resistor RS1 and is monitored internally as IVSL

    • The IVSL condition is evaluated on a cycle-by-cycle basis, and if the current stays above (for RUN) or below (for STOP) the threshold for 3 consecutive cycles, protection or startup is triggered

    • We interpret this as continuous analog monitoring rather than a single, discrete sampling point per switching cycle

    Is my understanding above correct?

    I would also like to clarify the "timing at which sampling starts" for the IVSL current, so any comments would be greatly appreciated.

    We are unsure where the 30kHz value originates, but we do recognize that the switching frequency can decrease to as low as 170Hz depending on the load. In our system, false detection on the CS pin tends to occur during startup under light load conditions, where the switching frequency is typically in the range of 25kHz to 30kHz. Our inquiry relates to the diagram on page 16 of the datasheet, which shows that IPP does not fall below 1/4 × IPP_MAX. We would like to know the specific switching frequency (fsw) at which IPP becomes exactly 1/4 of IPP_MAX. From the graph, we can infer that this point lies somewhere between 32kHz and 3kHz, but we would appreciate it if you could provide the exact or typical value.

    I haven't made any comments on the question in Q2 above. Can you please answer it?

    I'm very sorry to ask something unreasonable, but we have a meeting with our end customer and it would be very helpful if you could give us your comments by the end of today.

    Thanks,

    Conor

  • 1) IVSL current flows during the period when the MOSFET is ON (DRV = High), and the auxiliary winding polarity clamps the VS pin to GND

    Yes

    2) During this time, current flows through the external resistor RS1 and is monitored internally as IVSL

    Yes

    3) The IVSL condition is evaluated on a cycle-by-cycle basis, and if the current stays above (for RUN) or below (for STOP) the threshold for 3 consecutive cycles, protection or startup is triggered

    Stop needs 3 consecutive cycles below 80uA. start up will also check if current is above 22uA for 3 cycles before true start up

    4) We interpret this as continuous analog monitoring rather than a single, discrete sampling point per switching cycle

    The logic (comparison) will be done at the end of the cycle (when on is end) but the current is continuously compared. 

    Q2: the switching freq is no fixed during start up. Please read the start up section in the datasheet. Under normal operation, you can follow the control law diagram in figure 15.

  • Hi Ning,

    Thank you for your reply. I have five additional questions based on your answer. Could you please answer them?

    2) During this time, current flows through the external resistor RS1 and is monitored internally as IVSL

    Yes

    1). Is it correct to understand that the IVSL monitor is always monitored in an analog manner when the FET is ON (DRV = HIGH)? For example, if the FET is ON for 4 usec, it is always monitored for 4 usec, and if the IVSL falls below 80uA at this time, is it correct to understand that 'IVSL (stop)' is detected once (cycle)? Also, does this analog monitor have a filter time constant for detection?

    2). Does the monitoring mentioned in 1) include the Leading Edge Blanking (LEB) period? Or is the IVSL not monitored during the LEB period?

    The logic (comparison) will be done at the end of the cycle (when on is end) but the current is continuously compared. 

    3). At what moment does one cycle end? Please tell me the specific timing, such as the moment the FET turns OFF, the moment the FET turns ON, etc.

    4) In our testing, the oscillation stopped when the CS pin was falsely detected and the FET was turned off immediately after the LEB period. This behavior is unexpected for an IC, but can the IVSL monitor the correct value in this case?

    5.) The following is an excerpt from SLUAAC5.

    'The Vaux ringing in Figure 8-1 is excessive and rings down below ground during tLK_RESET. This behavior is known to trigger a UVLO faults and shut down the converter. This is because when the VS pin crosses ground it activates input UVLO fault.'

    The above "UVLO faults" are IVSL (STOP) = 80uA, but doesn't IVSL monitor when the FET is ON? Isn't the FET off in the above state?

    Thanks,

    Conor

  • Hello,

     

    Your inquiry has been received and it will be reviewed in the order it was received.

     

    Regards,

  • Hi Ning and Mike,

    Thank you. I'm very sorry, but due to the development schedule, it would be very helpful if you could answer by today.

    Thanks,

    Conor

  • Hello,

     

    Your inquiry has been received and will be answered in the order it was received.

     

    Regards,

  • Hello Connor,

     

    Ning has asked me to look into this for you.  I have gone back to the original post and think I can help from there.  Please see my comments below.

    ------------------------------------------------------------------

    A large ringing occurs on the bias winding voltage immediately when the FET turns ON.

    • This ringing may cause an OVP trip.

    This ringing may cause false detection on the CS voltage, which leads to the DRV signal being turned OFF within approximately 255ns.

    • The device has 235 ns of leading-edge blanking which will not protect against this ringing
    • You can filter this ringing out by adding a 220 pF capacitor frim CS to ground. This will form a low pass filter with RLC to filter out the noise.

     

    This behavior is always observed immediately before oscillation stops.

    The ringing continues superimposed until the auxiliary winding voltage reaches +20V, with a ringing frequency of approximately 12MHz, and takes about 1μs to settle.

     

    At this point, we suspect that this ringing is primarily caused by the transformer design.

    • If you have excessive ringing on the aux winding it will trip OVP.
    • You can snub this with an RC snubber across the output rectifier even the aux winding diode.
    • https://www.ti.com/lit/an/sluaac5/sluaac5.pd describes how to setup the snubber to dampen the riming.

     

    Based on the above, we kindly request your confirmation and response to the following questions:

    Q1.
    Please clarify the exact timing of the I_VSL data latch.
    For example: Is the CS voltage latched A [ns] after the DRV signal turns ON?

    • I don’t think this is part of your issue.
    • If the current coming out of the VS pin is less than IVSL(run) during FET on at startup the design will go into under voltage lockout mode. (IVSL(run) = 225 uA. Please note excessive ringing during this time could cause this to shut down to UVLO.  This also can be removed with snubbing.

    Q2.
    On page 16 of the datasheet (CV mode operation), please indicate the switching frequency (fsw) when the peak current Ipp equals 1/4 of the maximum Ipp.

    • When the controller is operating below 32 kHz the peak current will be controlled to ¼ the maximum.

    Q3.
    In application note SLUAAC5, the following statement is made:

    The Vaux ringing in Figure 8-1 is excessive and rings down below ground during tLK_RESET.
    This behavior is known to trigger a UVLO fault and shut down the converter.
    This is because when the VS pin crosses ground it activates input UVLO fault.

    Does this mean that if the bias winding voltage undershoots below GND during the tLK_RESET period (when DRV is OFF), it can immediately trigger a false UVLO fault? Please confirm.

    • Yes

    Q4.
    Regarding Equation (24) in SLUAAC5, can the described fsw(max) be interpreted as the actual maximum switching frequency in CV mode during operation?

    • Yes

    In regards to troubleshooting this design should look at VDD, the voltage across the aux winding, current sense resistor and Vout with an oscilloscope.  Trigger on when Vout or VDD drops out and look at the last three switching cycles before switching stopped.  This will tell you whether it is OVP or CS that is causing the device to shut down.  From there you can use application note SLUAAC5 to remove the fault by RC snubbing and filtering.

    Regards,

  • Hi Mike,

    Your response is very helpful. I would like to know what information customers want to know, specifically the 5 questions I replied to Ning 3 days ago.

    Some of these may overlap with the questions in my first post, but I would appreciate it if you could answer these questions.

    Thanks,

    Conor

  • Hello Connor,

     

    Thankyou for pointing out there were further questions that needed to be answered.  I will take this one at a time.

     

    1). Is it correct to understand that the IVSL monitor is always monitored in an analog manner when the FET is ON (DRV = HIGH)?

    It is looking for IVS to be greater than IVSL (run) for 3 consecutive switching cycles while the FET is on to turn on.

    It is looking for IVS to less than IVSL(stop) for 3 consecutive switching cycles while the FET is on.

    Please note that if the VS signal is extremally noise the controller can determine the ringing at switching cycles and could turn off the controller.

     

    For example, if the FET is ON for 4 usec, it is always monitored for 4 usec, and if the IVSL falls below 80uA at this time, is it correct to understand that 'IVSL (stop)' is detected once (cycle)? Also, does this analog monitor have a filter time constant for detection?

    • If the fet is on for 4 uses and there is not ringing and IVS drops below IVSL(stop) for the consecutive switching cycles the convert will turn-off
    • If while the FET is on there is excessive ringing during one switching cycle and the IVS drops IVS(stop) 3 times this will also trigger a fault.
    • To avoid issues make sure your aux winding voltage is clean.

     

    I will at the other questions later.

     

    Regards,

  • Hi Mike,

    Thank you for your reply, I appreciate your support. Can you answer any questions other than Q1?

    Also, my customer has obtained a waveform based on your advice, so I would like to share it with you along with the circuit diagram information. Since it contains confidential information, is it possible to send it by private message? A friend request has been requested. Or if you prefer email, please let me know your email address.

    Thanks,

    Conor

  • Hello,

     

    This is a public forum and technical support is given on the forum.  If you need confidential support you will have to work with your local Texas Instruments Field application engineer.

     

    I believe you should be able to resolve your issues with recommendations already made.

     

    Moving on to your other questions.

     

    2). Does the monitoring mentioned in 1) include the Leading Edge Blanking (LEB) period?

    This is for over current protection fault and not LEB.

    Or is the IVSL not monitored during the LEB period?

    Ning Tan said:

    The logic (comparison) will be done at the end of the cycle (when on is end) but the current is continuously compared.

    CS Leading edge blanking as the beginning of the on time.

    3). At what moment does one cycle end? Please tell me the specific timing, such as the moment the FET turns OFF, the moment the FET turns ON, etc.

    The UCC28740 will keep the FET on until the peak current controlled at the CS pin between Vcst(min) and Vcst(max) to control the output voltage.  Once that current is met the FET turns off.

    The FET turns on at zero current or on the valley of the resonant ringing.  This is depended on how much dead time is required to control the duty cycle.

    4) In our testing, the oscillation stopped when the CS pin was falsely detected and the FET was turned off immediately after the LEB period. This behavior is unexpected for an IC, but can the IVSL monitor the correct value in this case?

    It takes 3 faults to turn-off the IC not just one.  Application note SLUAAC5 will give you guidance on how to resolve these issues.  I also made recommendations above.

    5.) The following is an excerpt from SLUAAC5.

    'The Vaux ringing in Figure 8-1 is excessive and rings down below ground during tLK_RESET. This behavior is known to trigger a UVLO faults and shut down the converter. This is because when the VS pin crosses ground it activates input UVLO fault.'

    The above "UVLO faults" are IVSL (STOP) = 80uA, but doesn't IVSL monitor when the FET is ON?

    It does not matter if the FET is on or off.  If the aux winding signal rings below ground when it should not the controller will misbehave and shut down.

     Isn't the FET off in the above state?

    The FET could be off or turning off.

     

    If you need further support please repost in the e2e starting a new thread.

     

    Regards,