Second build has shipped. Message will appear here when store is ready for third build ordering.

Kiwi 5V power supply observations

I'm posting this in case other Kiwi users experience power up difficulties.

At KPH I have been suffering with one or more of the 7 Kiwis at that site which fail to power up after extended AC power outages.

The DC system consists of an Astron RS-20A 20 amp linear power supply feeding a West Mountain Radio PWRgate battery backup system connected to a 35 Ah battery and to a Powerpole distribution block. To maximize the efficient use of the battery during outages, each Kiwi is fed by its own 12V to 5V/1 Mhz switching converter, and ferrites attached to the SWPSU leads seem to do an excellent job at suppressing the RFI generated by the SWPSU. However when coming back from extended outages, like the 3 day Public Safety Power Shut Off during the October Northern California fires, none of the Kiwis came up until I manually power cycled them.

The Kiwi specification (I think derived from the Beaglebone spec) requires the DC ramp to 5V in 50 mSecs, and I thought that my SWPSUs might be too slow to ramp up. However in measuring the DC supply from the Kiwi barrel connector, I observed about 300 mV of drop on the "5 VE" pin supplying DC to the BB. The BB is spec'd to run as low as 4.3 V, but I had adjusted my SWPSUs with only a DVM attached to them; under any load their output drops 300 mV. So the SWPSUs I adjusted to 5.1V were actually delivering 4.8V to the Kiwi DC barrel connector which resulted in only 4.5V getting to the BB.

So if you are experiencing power up issues you might want to look at the BB supply voltage. I built a test cable which plugs into the USB port which differs by only 30 mV from the "5VE" voltage.

I won't know if there are other power supply issues coming back from outages until the next major winter storm, but with 5.3V on the barrel of the Kiwi they come up reliably from AC power.


  • This is almost certainly a ramp-up time problem (or dynamic voltage problem) and not a static DC voltage level problem (although the later is certainly important). The characteristics of the voltage waveform seen by the Kiwi from simply plugging in a live 5V power cable can be vastly different from powering-up a pre-connected supply with all its caps discharged. And that can be different from just power-cycling a connected power supply such that its output caps (or something else) don't have time to bleed off.

    The only real way to determine the issue is to hook up a scope in single shot mode and view the 5V input waveform present at the Beagle at power-up of a completely discharged supply. All sorts of things could be going wrong. The initial 50 msec ramp-up spec might be met, but then any over/undershoot might be exceeding the Beagle PMIC crowbar thresholds.

    We can speculate forever. But you've really just got to measure it and see what's going on.
  • I measured it with a scope and the ramp up is 1 mSec, so ramp up was not my problem
  • This solution is fairly easy and reliable. I haven't had a restart issue since placing a 90 second timer in line. It gives the router time to boot, and the KiwiSDR gets full power at start and an IP address.

  • edited December 2019
    Thanks for the pointer Ron.
    I have also added an IP controlled DC switch feeding the PSUs, but Yuri's device seems even more failsafe.
Sign In or Register to comment.