you could try to step up the battery voltage to 24V that could sort out the overload protection kicking in. Also, you can check precisely how much power each device draws, that could be problem if you are using 12V to power the RB570UP then you get 10.5V on the output, if some of device is using MAX 8W, then 0.5A @ 11V is not near enough for the device to be powered.
So, best option is to use 24V.
- I'd like to be able to use the RB750UP voltage monitor to measure the battery voltage (a step-up DC/DC converter would make it constant)
- if the problem is too much input capacitance in the radio, higher voltage will make it worse (will take longer to charge with the same current)
- 24V battery really means up to 28V, a bit too much to be safe for some radios (specified "up to 24V"); 18V (3*6V) battery would be perfect but such power supplies are unusual, the 12V ones are cheap
- why so much voltage drop (12V to 10.5V), does the RB750UP switch the power using bipolar transistors and not FETs?
- Bullet2 has no problem to start on a bench power supply, set to 9V and 200mA current limit (draws a little less than 200mA, initially just about 150mA before it fully boots; it could be more under high traffic, but the failing one doesn't even power up for a fraction of a second)
BUT: this was tested in the lab, on a different Bullet2, powering up from a different RB750UP just fine - the devices that fail are at a distant location, difficult to access (need to ask someone in advance for permission to enter there, etc.); it still works fine with the UBNT PoE adapter, it might be partially broken in such a way that it works but draws much more power, but then I'd expect the extra 2-3 watts would burn something inside very quickly, and it has been working fine for a long time. A mystery...
To work around the "fast rise of supply voltage triggering overload protection", I'll try to add some extra capacitance on the output of the power supply (2200uF should be much more than inside the radios, absorbing most of that current spike and slowing down supply voltage changes).