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.

AC brown-out supervisor

Other Parts Discussed in Thread: TPS3850, TPS3851, TPS3430, TPS3803

Hi team,

Do we have AC brown out supervisor, such as IN1M101?

It seems no for us.

Thanks.

Robin

  • Hi Robin,

    We do not have that specific device but I believe we can use our devices to accomplish the same thing. Can you speak more about the application and how the device is being used? What are the key requirements?

    If any information is sensitive, feel free to email me directly at michaeldesando@ti.com

    -Michael
  • Hi Michael,

    After confirmation with customer, they need the NDFPO and FPO output signal, and delay setting function.

    You could refer to the figure in competitor's datasheet as below. As for the threshold, no detailed requirement from customer.

    Do we have low cost solution to achieve the function?

    Thanks.

    Robin

  • Robin,

    After investigating the requirements, I believe we have a single IC solution that can replicate the competitor functionality and more because it includes fully programmable "CD delay" which we refer to as "CWD watchdog timeout" and programmable rising edge delay which we call "CRST delay" which may be useful for phase shifting / glitch immunity. This solution can also detect if the IN1 singal is overvoltage in addition to low voltage / brown out.

    Using the TPS3850 www.ti.com/.../sbvs301a.pdf
    You can directly connect IN1 to "SENSE" if using the 1.2V variant and the brown out falling threshold will be set to 1.152V and the overvoltage threshold will be set to 1.248V if using the 4% option. If using the 7% option, the UV will be set to 1.116V and the OV will be set to 1.284V. Using the adjustable version, you can use a resistor divider to set the UV and OV threshold so this is no problem and can be adjusted.

    When IN1 rises above the adjusted threshold, the RESET output of TPS3850 will rise up just like NDFPO does above. When IN1 falls below the adjusted threshold, the RESET output will fall just as NDFPO does above. This RESET output is then connected to WDI in order to monitor when the pulses are missing. The duration of time labeled CD Delay above is set with CWD capacitor on the CWD pin of TPS3850. When this delay expires and no falling edge arrives on SENSE input like IN1 above, the /WDO will trigger low just like FPO does above. The key to replicating the functionality is to set the rising edge delay on TPS3850 to the minimum of 700us using 100pF on CRST pin of TPS3850.

    We are assuming the frequency of the IN1 signal in the above timing diagram is constant and defined such as 50Hz or 60Hz for example. In this case, the Window Watchdog functionality of the TPS3850 will not cause any issues as there will be no chance to cause an early fault.

    Please let me know if this makes sense or if there are any questions you might have.

    -Michael
  • Hi Michael,

    Really appreciate your very detailed explanation.

    Still some details need to confirm with you.

    (1)For NDPFO generating, only TPS3850H01 could be used, because for this part, only undervoltage is monitoring, instead of functioning as a window comparator.

    If window comparator is used, we must make sure the VIT+(OV)>peak value of AC voltage after divider, otherwise the nRESET will turn low when this happens, which is not same with IN1M101. 

    If so, the VHYST of undervoltage is about 0.5% (0.2~0.8%), which is fixed for the device. We could not set the rising edge and falling edge for the monitored voltage.

    (2) The watch dog is in window mode. Only 1:8, 3:4, 1:2 could be selected.

    For AC input, the normal NDPFO should be 20ms intervals for 50Hz as an example. To avoid early fault, tWDL(max) must less than 20ms(even smaller). If long delay time for FPO is needed, we should choose 1:8 ratio, then the longest time we could achieve is about 20ms*8. It is relatively limited.

    Please feel free to correct me if my understanding is wrong.

    Again, great thanks for your fantastic suggestion using our existing sockets.

    Robin

  • Robin,

    1) You bring up a good point. Ideally we would like to use TPS3851 which is only undervoltage monitor, but the issue with TPS3851 is the rising edge delay is fixed to 200ms so the /RESET output will be delayed by 200ms upon rising. For the TPS3850, the rising edge delay is programmable so that's how we can fix that issue by setting CRST to 100pF to set rising edge delay to small value of 700us. For the TPS3850, in order to make the window voltage monitoring work, we can add a hysteresis resistor from /RESET output to SENSE input to widen the window.

    2) Again, ideally we would like to use TPS3851 since this device has a standard watchdog, instead of the window watchdog, but we can't because of the reset delay. I think by setting the TPS3850 window ratio to 1/8 and using a constant frequency signal, we can use the fixed window configuration with CWD = no connect, SET0 = 0, and SET1 = 1 to set the lower boundary to 1.85ms typical and the upper boundary to 27.5 ms.

    3) Another issue that may arise is the timing between /RESET transitioning from the SENSE input transitioning, and how this will impact the WDI functionality. After /RESET goes low, the watchdog will ignore the WDI input (which may only be a problem when going back high since the /RESET must be in logic high state before the WDI receives a transition low, otherwise the WDI will ignore the transition and /WDO output will remain logic high)

    4) I think a more robust solution is to separate the voltage sensing and watchdog functionality into 2 separate devices for a few reasons:
    a) If we use a voltage detector such as TPS3803, this device only has undervoltage monitoring so this fixes the issue of overvoltage accidentally causing a fault. TPS3803 also has 5us rising edge delay compared to TPS3850 which has minimum of 700us making the timing much more real time. Then using a standalone watchdog like TPS3430, the watchdog timeout can be programmed with lower boundary of 1.85ms and upper boundary of 27.5ms just like TPS3850. I feel better about the 2-device approach for this specific application but I don't know is 2-device approach will satisfy the customer. Please provide feedback.

    -Michael
  • Hi Michael,

    I agree with you that 2 chips will be more robust because watchdog is disabled when nRESET is asserted.

    But for two chips, the price may be not that competitive with competitor's. But I will have a try.

    Great thanks for your very strong support.

    Robin