Background: Over the years, I've had various OpenWRT routers which reboot when certain client devices connect. The latest one was an MR 8300 last week. I was able to replicate the problem by running a speed test from a specific older model MacBook. I wondered if the wall wart, a 20 year old linear regulator power supply to which I do not have access, had become flaky. On a hunch, I added a capacitor I had on hand (16V 2200µF electrolytic) inside the junction box where the DC power plug goes to the access point and was pleased that it fixed the problem entirely.
Now I have two questions for any electricians out there:
A. Would it be a bad idea for me to preemptively add capacitors for all the other junction boxes where I do not have access to the wall warts. I intend to use >30V capacitors since I measured the (nominally) 12V DC power at 15V and I like to have extra margin.
B. Is this problem (rebooting with certain client devices) more likely due to
A wall wart that's going bad,
Resistance from the long DC cable (which could be 120 feet, ~37m),
A hardware defect in the AP itself, or
Something else I haven't considered?
I doubt number ① even though the power supply is decades old since it is a linear (not switching) power supply and should be extremely reliable. Number ② seems unlikely since the voltage I'm reading is fine. So, I'm guessing it is number ③. Perhaps a capacitor in the MR 8300 got blown out after a recent storm that knocked out the power?
This sounds like a reasonable stop-gap solution, but I don't know that I would add any capacitors unless and until similar symptoms arise on other APs. Longer term, you should be looking at replacing the wall warts.
Yes, could be.
Yes, could be.
Yes, although less likely given that you solved it via an external capacitor. Nevertheless, the internal power supply circuitry may be beginning to fail.
Yes, but they still age and can drift out of spec. If there are caps inside, they may be failing.
The voltage you read is likely not at peak loading. If there is a low (or no) current, the voltage will measure as expected. But as the current increases on a high resistance wire, the voltage drop becomes more and more significant (V = IR), not to mention that the high resistance means that there is higher power loss through the cable, and thus less power available to the device before the supply circuit reaches its limit.
That's possible, but it's hard to know for sure. If the system experienced a significant over-voltage event (at the router itself), it may damage various components in the power path. However, heat is the real enemy of capacitors, so if the device is warm/hot over time, that's what kills caps.
Going back to the power brick itself -- consider that an old linear supply is at best 50% efficient (likely less). Compare that to a high quality switch mode power supply that can be upwards of 90% efficient. So you could cut the energy consumption of your APs roughly in half just by replacing the wall wart... over time, that can amount to a significant impact on the environment an your energy bills (unless you're solar powered, in which case no bills or environmental impact, but you do regain some capacity).
Can PoE be used? That is the best way forward. You can buy "PoE splitters" which plug in line at the back of a regular router and convert the PoE to 12 volts which plugs into the power input of the router.
There are capacitors at the input inside a router and they can also go bad. "Linear" wall warts are typically just an iron core transformer, diode bridge, and one capacitor on the DC side. The output has a lot of AC ripple which is hard on capacitors downstream. The capacitor is the part most likely to fail and they almost always fail open, so it can be replaced with an external capacitor.
Is PoE considered safe enough for general usage by monkeys? Not to say that there will be any actual monkeys[1] poking at these access points, other than myself, but they are in a public area and maintained by volunteers with no training.
Good point. I had thought of ripples as a switching power supply problem since the frequency is so much higher, but I could imagine an AC ripple stressing a capacitor.
Me too! I used an online calculator to see what the voltage drop was supposed to be and theoretically these shouldn't have been working, and yet they have for nearly 20 years. I'm reading 15V off a power supply that is nominally 12V, with the AP on but, as @psherman pointed out, that's normal for a linear power supply that is not at peak load. At peak, the voltage might be dipping lower than the 12V the MR 8300 is rated for. Thankfully, with the capacitor in parallel, it seems to be passing my stress tests.
Do any not-too-expensive access points have sensors that measure the voltage and/or current? How does one access such sensors from OpenWRT?
Thank you for the good advice. These wall warts are all rated for 12V, but if the OpenWRT One is a harbinger of things to come, I'm wondering if 15V might be a better voltage to standardize on for the future.
Saving the environment is a good enough reason for me, so I'd love to recommend that change! I'm going to need to make a case that whatever I suggest is going to last decades of continuous, maintenance-free use. Correct me if I'm wrong, but aren't switching power supplies less durable — especially the cheap ones that come with consumer-grade access points? I've read that high-frequency ripples are very hard on capacitors and that's why switching power supplies have a finite lifespan. Also, aren't linear power supplies supposed to be more tolerant of folks monkeying about and accidentally shorting them (briefly) since they're designed to handle heat? Again, not saying there will be any monkeys, but the access points and their power plugs are easy for anyone to fiddle with, while the power supplies are inaccessible and would be a PITA to replace if they ever went bad.
'Active' PoE, as standardized in IEEE 802.3af/at/bt, is pretty much idiot-safe. You can safely connect PoE or non-PoE capable devices here, while obviously only the PoE capable ones will power up, you can't damage non-PoE devices this way. There is an active communication between switch and AP, before voltage is injected - and only if both ends support it and agree upon the power draw.
'Passive PoE', where every vendor does its own thing, blindly transferring an unspecified voltage/ polarity over the 'unused' 4 cables is not safe at all. Connect the rj45 without the splitter inbetween and you'll see the magic smoke escape.
No. You need to use power supplies that provide the voltage and current capabilities required by the device. You can safely use supplies with higher current capacity, but the voltage must be the same as is specified for the device. So do not feed a 12V device with 15V... depending on the design, that may fry it (it also may be within the tolerance band, but best not to risk it).
That said, PoE was mentioned earlier, and this is a good strategy. 802.3af/at/bt PoE solutions will negotiate the power needed (at 54V) safely and reliably. If you have a lower power (802.3af or at -- approximately 15W and 30W. respectively) and you need more power, you can upgrade the upstream connection to 802.3bt and get 60W or even 90W of capacity... all over the same CAT cable you're using for data. Obviously this needs PoE based APs, but you can get PoE extractors that are 802.3af/at/bt compatible if you have non-PoE equipment.