I have recurring problems with some Routerboards. The problem is that when there’s something wrong with the power (like momentary power failure, or even unnoticeable fluctuations), they just become “cold”. LEDs don’t blink, the device doesn’t work. Cold, dead, not responding. Only pulling out power supply for at least 20 seconds or more make them work again.
The problem is with 4 or 5 devices in 3 different sites. There are 433, 433AH, 493 and recently 450 boards with the following power supplies, with or without PoE adapter (depends on customer).
Last year I had problem with nearly 10 RB1000s where I had to replace one of the capacitors on the mainboard (because of design bug) to fix the problem.
Does anyone know of the same here? I desperately need help, customers are going crazy when the new, shiny, so-much-advertised devices just behave worse than plain old TP-Link… (Yeah, the most important feature for them is to have router that don’t mysteriously pass away).
Do you think they are simply broken and need to be replaced? Some of them has only 6 months or less…
And what if I already use UPS in most of those places?
One RB433AH is behind 20kVA EATON 9155 looking more like big refrigerator than UPS. And it doesn’t prevent it from exhibiting random shutdowns. And this is the only device in this site which behaves like that (I’ve got there more than 20 servers, some FC arrays, and even 3 more Routerboards - 2xRB1000 plus one other RB433AH).
If you are looking for really good power supplies try these ones: http://zasilaczeonline.pl/
For poorly designed/manufactured power supplies life span of 6 months is possible.
Did you ever considered buffered power supply used mostly in alarm and CCTV systems?
It’s like 2-in-1 solution (UPS + power supply), output voltage pulsation is really small (30 mV p-p)
and switching time is 0 ms
There is link to manufacturer website: http://www.pulsarsc.com.pl/pl/panel.php?lang=PL&m1=1&m2=2
Hope this helps.
Thanks Grzegorz, I have never considered power supply to be a problem here, I’m gonna replace some of them and see what happens. Luckily (or not?) one of the RB433AH hangs almost every other day, so the difference will be observable immediately.
If the power supply fix these problems, I’m gonna blame MT – it’d be their power supply that is unstable and fails
I’ll give you feedback after buying one of the suggested supplies.
I’ll also look into these buffered things – although I suspect I can’t use them in these scenarios.
I forgot one thing which is also important, use surge protectors whenever possible.
I saw a lot of devices destroyed by voltage spikes, especialy when electric engines are used in neigborhood.
Thanks for karma
What might help is to splice in a capacitor on the DC side of the power supply (spliced in-line onto the DC wires of the cable itself). If you are worried about voiding the warranty of your power supplies (understandable), you can buy short male-female 2.1/2.5mm cable adapters and splice in a cap on that. Maybe a minimum of 33,000uf - 47,000uf might be enough (?), but make sure the voltage rating is higher than the power supply’s voltage output.
Any chance you can provide further detail about which cap and what value you replaced? I’ve got a RB1000 that will crash randomly between 7 and 30 days. Does not appear to be a traffic or load issue, and does exactly what you describe - goes cold and needs to be off for 10-20 seconds to come back up…no ip, no console, and of course no supout
Thanks I’ll try that. I’m not skilled in electronics so I’m little woried that this capacitor will explode a second after connecting to the power supply Is this as simple as connecting + of the capacitor to the + wire and the same with - ?
Do you mean 33 THOUSANDS of micro-farads?
It affected only the old versions of RB1000, you can preferrably send affected devices to RMA or fix them yourselves (possibly losing warranty) by changing C309 to 0.1uF 0402 capacitor. Beware this capacitor is really very small in size
Hope this helps.
Yes, + to + wire and - to - wire on the DC side of the power supply. Make sure you don’t get a bipolar/non-polar cap as there is no +/- on these. Regular DC caps have a mark on one side with several “-” to show the negative lead (you’ll see this on the bigger caps of your RB1000 for example). Again, make sure the voltage is higher than the voltage of the power supply. It’s okay to use a 30V cap on a 12V power supply. Depending on the DC voltage you are using, the higher uf rated caps may be harder to find or more expensive. You can add additional caps in parallel (- to -, + to +) to increase the capacitance, so 2 of the 33,000uf caps in parallel would give you 66,000uf total. You could start with smaller caps (10,000-20,000uf) and add additional ones in parallel until you see a difference to determine the total needed capacitance. Once you find the sweet spot, add 10% more to that value to give you some headroom.
The 33,000uf value I suggested is just a guess, but based on some random things I’ve done with caps over the years. Having too much (within reason) shouldn’t be an issue…just don’t install a 1 farad cap
Thank you very much for the info..that thing is tiny! So far I’ve looked at 2 RB1000s I’ve got on hand (one is DOA and the other just came back from RMA which was also shipped DOA) and I’ve noticed that in your pictures there is a cap 301 next to the 309. Neither of my boards have a cap soldered on at 301, so maybe these are the newer boards? Either way, I appreciate the info and will take a look at our problematic RB1000 once I get a backup installed.
It’s not C301, it’s R646 (I understand that resistors are black, and capacitors are bright). You don’t want to solder it in.
I replaced C309 in one such board and it’s ok. This one was particularly problematic before.
By “ok” I mean 10 months without any single issue on all 10 devices up to this day.
