1. Forum
  2. Main category
  3. Main forum
  4. Pi 5 turns off instantly when 24 V is removed – Multichemistry Watchdog UPS V2

Pi 5 turns off instantly when 24 V is removed – Multichemistry Watchdog UPS V2

  • CP

    Hi,

    We're using the Multichemistry Watchdog UPS V2 (SM-I-033) on a Raspberry Pi 5, powered at 24 V through the HAT's input terminal. A Building Automation HAT is also on the stack, and a touchscreen is powered from the Pi. Both battery holders have Li-ion 18650 cells fitted, and wdt -g vb reads about 4.2 V.

    Problem: when we remove the 24 V, the Pi turns off instantly and its LED goes red. The HAT is still supplying 5 V afterwards. It happens even when we pull the 24 V straight out of the HAT's own input terminal, and it happens with a brand-new HAT.

    What changed: this setup used to ride through power cuts with no problems. At that point, a PCA9685 servo driver board and a relay module were also powered from the Pi's 5 V rail. We then moved those two boards onto their own separate 5 V supply, so the HAT now feeds only the Pi, the stacked HATs and the touchscreen. Since that change, the Pi turns off every time the 24 V is removed. Nothing else changed.

    Our guess is that the Pi's 5 V dips briefly while the HAT switches over to battery. The extra capacitance on those boards was covering the dip, and without it the Pi 5 browns out.

    Questions:

    1. Is a short dip at switchover expected with this board? Has ride-through been tested with a Pi 5 and a touchscreen as the load?
    2. Do you recommend adding bulk capacitance on the 5 V rail, and if so, how much?
    3. Is there a newer revision or a setting that changes how quickly it switches to battery?

    Thanks!
  • alexburcea

    Hi,

    There is a possibility that switching between the main and battery can cause a momentary drop in the Raspberry Pi's 5V rail.  The condition is marginal because is working without problems on some setups.

    A capacitor on the Raspberry Pi 5V rail (placed at the 2x20 pin connector) will solve the problem, and 1000uF should be enough, but bigger is better in this case.

    Because this problem has been reported before, we decided to redesign the card by increasing the capacitor on the 5V rail and by changing the switching mechanism (adding hysteresis and changing the thresholds). We are expecting prototypes in about two weeks.

    Alex

     
Please login to reply to this topic!

Participants

© 2026, Sequent Microsystems | Powered by Xenforum