On my way to bed last night, I noticed that two of the smart devices in my office - the lamp and the shelving lighting - were both offline in Home Assistant. Strangely, the Skyfan wasn’t offline, so it took me a moment to figure out what the issue was: the Unifi access point that covers the rear portion of the house had decided to croak on me, and as they were unable to maintain a decent enough connection they were ushered off by the front one, never to be heard from again.
I spent far too long suspecting my patch cables (very cheap Temu things that I admittedly don’t really have any objective complaints about), I slipped a zip-tie under the AP to get it off the ceiling, and replaced it with a spare I had, which came right up. I adopted it, configured it correctly, and went to bed angry.
Over breakfast this morning I was having a read, and found a few threads mentioning a fairly wide-spread problem some 7 years ago, where Unifi were actually RMAing these devices outside of the warranty period, suggesting something they knew about, which was further irritating to me. Continuing my search, I found a reddit thread mention of devices where they failed to negotiate power-over-ethernet with a switch (exactly how mine were hooked up), so on my lunch break I dug out a POE injector and connected it up and yep - it came to life.
I started thinking about whether it’d be possible, or even worth it, to replace the IC responsible for this when I stumbled upon gold: someone’s not only been through this, but they successfully rescued the device by simply removing (not replacing) a single component!
I’ll replicate, tersely, the instructions here as I don’t trust Ubiquiti not to trash the thread - indeed I couldn’t find the linked instructions to open the fucking thing. But in case the thread disappears, the user TomBK on the Unifi community came to my rescue at least:
- Verify that the UAP does actually power up via a POE injector, you’re wasting your time if it doesn’t.
- Use a plastic spudger to pry it open. There are no screws, there’s five clips and a fuckload of adhesive holding it together. It took more force than you think, but once one clip pops the rest are trivial to undo.
- Remove the four screws to liberate the PCB from the back of the case.
- Flip the board over, and observe near the USB port. With the USB port oriented towards you, looking at the back side of the board, there is a line of four tiny transistors to the right, and then below it are two capacitors, C406 and C2. They may not be present, if they’re not, you’re wasting your time.
- Check the resistance of C2 - if it registers as a short or very low resistance, it’s probably cooked. If it’s open, this is probably not the fault that’s killed your UAP.
- Desolder C2 with a decent iron or soldering gun. It’s tiny, so I just used some flux, flooded the tip of the iron, heated it up and pushed it off. I then used more flux and a clean tip to make sure the pads are not bridged, and then verified no short with a multimeter.
- Plug the device into a switch and see if it powers on. If it doesn’t, this is probably not the issue that killed your UAP.
- Reassemble the device and put it back into service.
I’m not an electronics engineer, so I don’t know if the proposed explanation and solution are sound, but it’s worked for me.
As of the time of writing it’s not skipped a beat for 48 hours, so that’s alright.
At the end of 2024, cutting their self-imposed forecast awfully close (“End of 2024”, we got the letter in the last week of December) we were notified that the fibre outside of our house was lit-up and they were accepting connection upgrades for it. Hurray!
All we had to do was bump our internet plan up to one that doesn’t work on the FTTN (read: VDSL) connection and they’d schedule the upgrade. So I did that, and while I neglected to write down the date the initial site survey was done, it was sooner than I expected. The results of it, however, weren’t that great.
“Please direct me to your outside phone termination point”. Bad news, there isn’t one. The phone cable comes in off the street who knows where, goes inside up the wall, terminates in the ceiling where it fanned out in a star of multiple cables, before we chopped all those off and sent it to one point. That means they’d have to run a trench, and put a new box on the side of the house. We had several options:
-
We put it near where the current line goes in. Aesthetically the worst choice, so I turned it down, but if NBN had communicated clearly and accurately what they needed I would have gone with it. We could have put an outdoor pot plant or something in front of it.
-
We put it near the power ingress. Aesthetically the best choice, being on the side of the house away from everything. However, this meant cutting a trench in our very nice driveway cement and then patching it, not gonna happen.
-
We put it near the bedroom. It must go on the front of the house, as between the gas meter, the downspout, and the brick wall separating the back yard from the street, there was no clearance to put it on the side of the house, plus they’d have to dig around the gas line. This is what we went with.
For the NTD, this meant they wanted to put the NTD in the bedroom. I knew this would be an argument, but I didn’t realize their suggestion would be this silly… the master bedroom? No thanks.
So we suspended the install, while I waited for our electrician Baz to come out and set up a conduit + pull string for us to put it where we wanted it. He got to that in February, and then we let NBN know, and they scheduled it for Thursday the 6th of March.
A different tech come out, and wow I cannot knock this guy for work ethic. He was here until just after 8pm, trying to find where the conduit came in from the street. He eventually gave up, and did the bits that would at least get him some money: the indoor bit (which went fairly smoothly), and pulling the fiber down the street. Apparently there isn’t a GPON (I think that’s the term?) in every pit, our pit’s actually around the corner, a couple pits away. I got to sticky-beak at this process a bit, it also went fairly smoothly. He ended up leaving as it was getting dark, and said that he’d try get back tomorrow, or maybe NBN would have to make another appointment. No worries, I’m here all day if you’re able to come back, and if not I can’t say you didn’t try!
Anyway, he made it back about midday on Friday, and tried a different approach: instead of ramming a set of empty wires up the conduit and running his tone tracer up those, he unhooked our VDSL and sent the tone back from inside the house. We found the conduit, deeper than expected, sent a pull-rod up the conduit and shook it around a bit to make sure it wasn’t a water pipe, then he cut into it.
So at this point the installation went very smoothly. He teed into the conduit (not 100% sure it’d be easy to pull another one through, but hey, that’s NBN’s problem), ran it up to the PCD he’d put on the wall, pulled the fiber all through, sliced the end off, stuck it in a connector (how about that, I thought they’d splice it to a pigtail, but apparently they get a good enough signal just cleaning up a really nice cut to the fiber and pushing it into a connector). Inside into the air conditioning to run some checks, and then it was active.
AussieBroadband message me a few minutes later, so we swap the cable over into our router from the modem to the NTD. But even though we were on the now-lit fiber, and the IP had changed and everything, I was still only getting the 50/20. Maybe a plan change would fix that, so I teed up moving to their absolute gold-plated plan for giggles. That kicked us off for a bit, and I thought we might be behind CGNAT as well, so I rang up support - they were pretty great, definitely not behind CGNAT, everything looks good their side… they kicked the VDSL connection (we’ll keep it, for free, for about 10 days, in case the other one breaks), and I turned the modem off. Reboot the router, and it came up, everything working correctly, but speedtests still only show the VDSL speeds.
After a bit of thinking, I poked around the UDM-SE’s settings and noted I still had smart queues set up for the 50/20 connection. Set those to appropriate values and bingo, it’s working. Sheepishly hung up with support, and then went off to test things out.
I don’t think we’ll keep the 1000/400 plan - it’s terribly expensive and much greater than what we need. I could use the extra upstream, but the downstream isn’t super useful for us… looking at our stats over the weekend we very rarely reach the point where we’re bursting to even half a gigabit. Steam is capable of saturating the link, but then doing so we have to throttle each client to prevent pings from going up. I might be able to fix this with QoS, but it seems like a bandaid on a bullet wound.
If I put LANCache back on, it’s good for about 300mbps. I’m not sure why that is - it could be the hardware it’s on, or the extra hops through the router. I’ll have a look and see if I can do better than that, but 300mbps is heaps, if we had like a 500mbps downstream that’s plenty of overhead for other things.
As far as upstream, everything I care about has a Linode proxy in front of it, so 100mbps or 400mbps doesn’t really matter that much. Everything else, anything better than the ~12mbps we were getting is gravy.
So I think I’ll drop our plan back a bit next billing cycle and see how we go, to save some money.
We took Shaina, Flip, and family home yesterday, then washed, vacuumed, fueled, and returned the van today, so next on my list is returning all my shit to my office, which was previously used as a temporary bedroom for the visiting family. Starting in almost the hottest part of the day, I went out to the back shed and retrieved the two pieces of my very heavy desk. I then put it together, plopped the desk top on top of if, started setting up my monitors, dug out all the cable-management stuff, then got sidetracked because my desktop PC was positively filthy.
So I elected to clean it up before plugging it in, otherwise I wouldn’t. While there, I inverted the motherboard again, because I’ll put it on the other side of my desk. While I did that, I pulled everything out and cleaned it top to bottom, instead of just blowing the dust out and calling it good.
At this point I started to get irritated by the fan noise - I removed the EK Furious Vardar that’s generally the primary culprit, but it’s still noisy… what’s going on? After quite a bit of experimentation and peeking, I learned that I’d made two mistakes at some point in the past. First, I plugged the case PWM splitter (which controls the exhaust and one intake fan) into the pump header, which always runs at 100%. Oops.
My second mistake was when I setup Linux on this machine, I did not configure power management on my GPU. So it’s running at full noise, even when idle, pumping out an additional ~80W of heat any time the machine is on at all. Oops.
Fixing both of these, it’s still not dead-silent (the fan I replaced the Vardar with is only marginally more quiet, but at least it doesn’t have the god awful spin-up noise the aggressive blades of the Vardar make). It’ll do for now.
As we were unable to book a tour of the Aradale Asylum, and due to the fact we couldn’t go to the Grampians (Gariwerd) due to a tremendous bushfire, we headed the other direction and went to Mount Arapiles instead. We accidentally drove to the top, due to most of the parking spots being full, but near the top we found a nice shaded spot to leave the van.
We started down one path before deciding it was a bit difficult, and headed up to the other lookout. We found a sign that showed a fairly short loop, so we opted to set out on that, following the trail down past Pharos Rock, not realizing that this trail is actually quite difficult. We also failed to take into account that driving to the top of the mountain, hiking down, and then hiking back up is positively the worst way to see a mountain, but anyway.
Down past Pharos rock we passed a small Rock Wallaby, with Joey, hanging out just outside a tiny crevice. We passed it so close, without noticing, that I was the last one on the trail and spotted it completely by accident, saying “hey, did you guys see this?”. It didn’t seem overly bothered by us, but we definitely weren’t going to get off the trail to get any closer.
At the bottom of a fairly difficult descent, we wandered around, spooked another Kangaroo which equally spooked some members of our group, then started up another ascent to finish the loop - ironically ascending on the same path we decided was too difficult, but in for a penny in for a pound we needed to finish this thing and get back to the car. Saw a nice shingleback type lizard just hanging around on the path - Shaina actually thought at first that someone had taken a shit on the trail until it blinked at her.
Back to the car, exhausted, we had a lovely drive home then stopped off for ice cream, which should have happened in Halls Gap, but alas, they’re evacuated right now.
Back to Ararat again today, this time to vist J Ward, an old prison and later asylum. Sabriena, Duncan, and I have visited this before, but Sabriena and Shaina in particular are really interested in creepy stuff so we thought it’d be fun to do. We looked into it too late to be able to do the Aradale Asylum the same day, so we figured we’d go back there tomorrow for that.
The tour is entirely volunteer run, and they actually update it as the researchers find new information, and as a result the tour has improved significantly since the last time we were there nearly 9 years ago (though it was by no means bad then). Our tour guide told us that they were apparently lucky enough to have at least a couple of former warders come through, who filled in many blanks with their own stories, and we in turn got the benefit of them. There’s still a ton of speculation, but it was a load of fun.
We also got to climb the rebuilt watchtower - these were apparently torn down when a new doctor came in to take over operations of the asylum, but the volunteers rebuilt one as the stone was left on-site and they apparently had all the information needed to do so. I’m not sure if this tower wasn’t rebuilt yet when we last visited, or whether it just wasn’t part of the tour, but we definitely didn’t get to do that last time.
Off to find some lunch, then back home to rest for the afternoon. Sabriena tried to book a tour of Aradale, but it turned out the tour was too full to accomodate us all, so we scrapped the idea, in favour of going to Mt Arapiles instead.
