Kiwi keeps crashing
Hi all, my Kiwi keeps randomly crashing - or more accurately looks like its powered down.
If I power-cycle it comes right back up, and might last a day, two, three days maybe.
The power supply is good, its on solar power with a large battery, according to the solar controller the battery output is never below 12v, so the 5v converter the Kiwi runs off must always be OK. Power cable only about 200mm.
Any ideas please as to why this might be happening? Its annoying as the Kiwi is remote from my home and need to drive there to power-cycle!
thanks
EI2HH
Comments
I would not “assume” the 5V is always good, I’d test it.
Is the 12V-5V is on the “output” side of the charge controller not directly wired to the battery?
If so what is the cut off voltage set to on the output?
What actual voltage does the 5V convertor output and what is it’s current capability?
By large battery, how large and how much solar panel, sky cover etc?
73 Stu
The charge controller doesn't have a separate load output, you connect the load across the battery. Its a pretty basic one - I tried an expensive one which did, but it was very RF noisy!
The 5v converter is 5.2v output and is rated at 5A.
It's difficult to test the 5v on an ongoing basis, I'd need to be metering it when the Kiwi actually failed I think - but anytime i've been to the site there's at least 12v on its input and 5v out.
Thanks for the thoughts.
A
Are you using the correctly sized 5V plug at the Kiwi power connector? If not, and the 5V cable is subject to being moved slightly by the wind, then this would explain the problem. We had someone complain about this exact issue a while back and it is documented here (second paragraph): http://kiwisdr.com/quickstart/index.html#id-power
Not to mention the small potential contact areas anyway.
No its definitely the correct one, good and tight.
Tin foil hat comment I assume you have a good admin password.
It can be hard to assess the tightness of the center pin as the outer barrel tightness dominates
Its a pity there's no watchdog...
thanks
A
Next time this happens, note the state of the single blue LED to the "right" of the Ethernet RJ45 (other side compared to the four status LEDs). If it's OFF then there has been a power quality event such that the Beagle's power management IC (PMIC) has shutdown (e.g. momentary dip below 4.75V due to bad connections, cable brownout on current spike etc). It is not the result of a software crash.
A software watchdog does not help in this case. A power watchdog perhaps. But the Beagle does not implement that as it is always expecting to get clean power. Just like your PC motherboard!
I suspect this will turn out to be something really obvious but I like the challenge.
Things you have left us guessing are actual battery and solar capacity, how you know it never goes below 12V, if you have tried rotating the barrel plug?
Barrels are funny things and unless they have inner spring contact they can disconnect in some rotations (even in the warm), as Jim said that inner pin is hidden and it does not take much to disconnect while the outer is still tight, in fact too tight could mean the plug is held central in the outer shell, that is not good as it needs to be offset slightly by the outer spring contact to bear on the inner pin.
JKS I will check that LED today and let you know.
PN- The modem is on the 12v side of the supply. It is of course cold and possibly humidity is high - although the system is in a waterproof box it is not an IP-rated box. Maybe it should be.
The modem is using an external antenna mounted about 1.5m away.
The battery is 70AH, the panel is 100w @12v, usually in sunlight it runs at around 20v, in gloom around 14v.
I wonder if it would be best to solder wires to the board to bypass the barrel. The company I work for have a legacy product which internally used a barrel for the dc connection to the motherboard (an audio processor) and this caused problems, we advise soldering it now.
A
these work well, CCTV people use them
I'd not solder the connector I'd review the single 100W panel.
Will do some calculations at lunch but to me that is cutting it close in winter. I have a 110AH battery and 2x100W panels and that is good but can get surprisingly low over a couple of dull days (my panels are flatter than optimum for my latitude).
Those CCTV things are very variable, I cut one apart and it had a single strand of thin wire poorly soldered to the outer....
--Later--
Ok some guesstimates, Kiwi and Router, allowing some overhead and DC-DC inefficiency about 17W - roughly 400W per day. Winter maybe 8-9 hours daylight so the panel needs to average 44W per hour, through the winter day(light), I'm pretty sure it wont do that on many days. Given the charger does not have correction for cold temperatures the battery is not fully charged either so perhaps 40AH usable without damaging the battery. That does not even consider runs of dull days which are common at this latitude, snow, condensation, shadow etc.
In summer it should work but in winter there is just not enough guaranteed sunlight to service a single 100W panel with that load. I'm happy to be corrected, no expert.
Thats all possible. I'm heading over to it this evening so I'll measure some stuff.
Thanks for those calcs, makes sense.
A
Just for the record (and to see how far out I was on the guesses) I measured the solar output of 100W panels (a few minutes ago, full cloud cover). Roughly 3-6W per panel, 3PM while the sky looked like this.
The battery was charged OK but not float voltage (despite 3x100W connected at the time). Load is just a Kiwi (BBG) and a low power network switch with Fibre module.
I think I was too pessimistic on the power used (without knowing the router) but optimistic on the solar output. I'd read on the net cloudy days gave 10-25% panel efficiency, I was not expecting 3-6%.
If you have tracked it down John might want to append the real cause rather than "crashes" it more correctly should read "Kiwi bravely endures constant power cuts".
73
Stu