Starlink: Outage Data End of May Update

I said at the end of last month that I was looking forward to writing this month’s update. Mid-month, I built a tower and got Dishy up in the air, hoping to cure my obstructions. So how did it go?

In a word: frustratingly.

May 2021 connectivity timeseries. 1sec per pixel, 20min per line. Red=obstruction-downtime, blue=other-downtime, green=no-satellite, dark-grey=no-data (reboot), yellow=uninterrupted connection at least 30min long.

Above is the timeseries chart for the month. Even without knowing what the colored blobs mean, you can see that they change. Change is what I was looking for, after reporting that April had been just like the end of March. Let’s break down the changes.

May 1-13am. Grey bars are May 6 and 7 when Dishy upgraded firmware from 1f86ec34 to fd689710.

The first two weeks look just like April. We were seeing about 38min obstructed, and an additional 38min “other” (formerly known as “beta”) downtime per 12hr through this period. This was starting to get awful. Half the time I was browsing the web, I’d have two tabs open – one to see the webpage I wanted, and one to check whether Starlink was connected. So when the wood for my tower arrived, I wasted no time in starting construction.

May 13-23am. Point A = tower install May 13, B = firmware 68fdc22b May 15, C = support request May 15, D = support response May 18, E = firmware 1752790c May 21.

On May 13, I stowed the dish, unplugged it, mounted it on the tower, and plugged it back in (black bar at point A). All of the obstruction downtime (red) vanished! … but “other” downtime (blue) increased?! I know it takes 24hr to assemble obstruction statistics, so I waited. At 48hr, when I was still seeing 45min of non-obstruction downtime per 12hr with nearly no obstruction downtime, I reached out to Starlink support (point C). Unfortunately, that was Saturday, so even though they called and talked to me, they ultimately told me they’d have to wait for the rest of tech support to check out the details on Monday.

Left: Obstruction view at Dishy’s original mounting position (not on tower), facing north, February on the left, May on the right.

I didn’t hear from Starlink on Monday. I left home mid-afternoon, and spent the night away. When I returned on Tuesday afternoon, an email had just arrived from Starlink (point D). Paraphrased, it read, “Could you check your stats again now?” I opened the app, and found that other downtime had fallen, but obstruction downtime had climbed well over 40min per 12hr! Over the next two days, obstructions only got worse. We topped out around 70min obstructed per 12hr. What happened?!

The lack of obstruction downtime for 96hr after moving Dishy was apparently a fluke. “The stats didn’t reset right,” is basically what support said. That’s not why the stats got worse from Monday onward, though. What made obstructions worse is that that Monday was the second of three days over 70ºF in the middle of spring rains. That’s right, the thing I’ve been talking about racing for months caught up to me: leaves. We went from nothing to nearly full canopy in the space of a few days.

Pre-build tree model from my Tower-building post. 20ft tower would almost clear 55ft trees.

Wasn’t the tower supposed to fix the obstruction problem? Yes, well. Remember when I said I wasn’t sure enough of my measurements to put in a permanent tower? It turns out my measurements were pretty good … but the math I did with those measurements had a major error. I typed “sin” where I should have typed “tan”, and it turns out my main obstruction trees are not 50-60ft tall, but are instead over 80ft.

Corrected tree model. 20ft Dishy tower does not clear 80ft trees.

So, my tower probably saved me from having even worse connectivity. Without being sure how long I’d have to wait to get accurate data, I’m not going to move Dishy off of the tower again to find out. I’d rather not haul the cinder blocks out to restack anyway.

An extra 5ft, to reach 25ft total, does help just a little bit.
May 23-26. Added an extra 5ft to the tower mast at the dark grey line.

In a fit of, “What’s the quickest change I can make to get some improvement?” I added another 5ft of pipe to my tower. Dishy now sits 25ft above the ground. We’re usually under 60min per 12hr of obstruction downtime again. Sometimes when it’s not raining, we’re even back under 50min. Thankfully beta downtime has lowered, to even us out near where we started for total downtime.

Obstruction view near Dishy’s 20ft tower mounting position, facing north. Blur and narrow window are due to my extremely quick tape-my-phone-to-a-pole hack to get this view before a thunderstorm rolled in.

My main northern trees are over 80ft. Trees to the east and west are 60ft. I can’t get far enough from the east and west trees to make anything less than a 40ft tower clear them. I can’t get far enough from the north trees to clear them with any tower much shorter than 60ft. This 25º horizon and 100º field-of-view are kind of a bummer for my location.

“Just get rid of the trees and/or put up a taller tower,” I hear you say. Yes, I know that’s the solution to my current Starlink connection troubles. The problem is that I like these trees, and I’m not particularly wild about spending several thousand dollars on a permanent tower to support a beta service. Can I even rely on the dish getting its stats correct after that move? Once the initial constellation is full, and the service moves out of “better-than-nothing beta”, and we get full details on the potential narrowing of the necessary field of view, maybe I’ll feel better about a tower.

Estimates in forums put the full constellation date between two and three months away. I’ll be continuing to collect these stats, and possibly trying one or two more tower position changes, over that time. In addition, I’m finally collecting official stats on my alternate connection, a locally-operated fixed-wireless ISP. They were here before Starlink, and we might have skipped the beta if that WISP hasn’t had large amounts of unexplained downtime just a couple of weeks before we received our invite. They’ve continued to be the connection we turn to for video calling. Now it’s time to see if they’ve ironed out their issues, and determine if we’d rather just go back to that connection and wait for Starlink to mature. Watch for comparison data next month.