Starlink: Outage Data End of July Update

Published Sunday, August 1, 2021 by Bryan

July marks my sixth month of Starlink Better-Than-Nothing Beta service. Many of these months have been rough, and even July brought its own oddities, but as I look through the data I've collected, it's obvious that the network is rapidly improving right now.

July 1-14, 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 connectivity

The month started with new firmware - 4990ce8d on the morning of the first, and 26d12814 on the morning of the second. Whether it was changes in those firmwares alone, or their installations coincided with other constellation updates, the reduction in the number of interruptions for the following two weeks was noticeable. The improvements we had been expecting in the lead-up to the finished initial constellation seemed to be confirmed.

July 15-24, 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 connectivity

Then came very strange behavior on July 15. Suddenly, statistics were logging a 1 second beta downtime nearly every 15 seconds. You can see this in the image above - the single blue dots vertically aligned in an even pattern. It was a strong pattern for a full ten days. It seemed to be a widespread problem in the Starlink subreddit. If the previous two weeks hadn't been such an improvement, we might not have noticed the slight increase lagging connections. In fact, looking at the previous figure, it seems like we did miss this pattern appear briefly on July 9. I haven't yet seen a good explanation for what may have caused it. Its appearance didn't coincide with a firmware change (we got 2bc83694 on July 11). Outages at exactly a 15 second period means it almost had to be associated with the calculation of the satellite connection schedule. But whether it was a problem, or Starlink reserving each of those seconds to run an experiment instead of serve other traffic, is unknown.

July 25-29, 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 connectivity

The 1/15 second pattern dissappeared almost completely on July 25. That's the same day that my dish upgraded to firmware 6390f89c … for the first time. Strangely, sometime early on the morning of the 27th, our dish rebooted itself, and reverted to firmware 2bc83694. When I saw the 1/15 second pattern return, I decided it was time to take the Starlink subreddit's over-recommended advice and power-cycle the dish manually. When the dish came back up, it had returned to firmware 6390f89c. The pattern seemed to continue for the rest of that day, but mostly dropped away again afterward.

July 30-31, 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 connectivity

As I write this, we've been on the latest firmware, e825ab2c, for two days, at it is nice. It's clear just from the color of the chart - literally. We haven't seen connectivity this clear since at least early March. It may be our clearest ever. Stats say we're down to a combined 20min to 24min/12hr downtime.

The best part about the improvements this month is that we haven't changed anything about our installation. The dish is still on top of our tower, in the same place at the same height, with the same trees growing around it. In email this past week, SpaceX confirmed what I suspected my data showed last month: the new satellite-switching feature has been enabled for everyone. With the additional satellites getting into place offering more choices to switch to, we've seen a big difference.

Categories: Starlink