I tracked a real satellite over Aberdeen. The signal did something I didn't expect.

Hi all,

Last time I built a link budget tool and swept it across a hypothetical elevation angle — 5 degrees, then 10, then 15, all the way to 90. It told me when a link should close in general. It couldn't tell me when an actual satellite, the kind orbiting overhead right now, would actually be there.

So this week I went and found a real one.

Every tracked object in orbit has a Two-Line Element set, or TLE, published for it — a short block of text describing its orbit at a given moment. Feed that through an orbit-propagation model called SGP4, and for any time you choose, it tells you exactly where the satellite is. I pulled live TLEs for the entire Starlink constellation from Celestrak and asked a simple question: which satellite is going to pass over Aberdeen next, and what does that pass actually look like?

The tool found one: STARLINK-1008, peak elevation 11.3 degrees, 27 samples across the pass. I fed the real, time-varying slant range into the same link budget maths from last week — and the margin curve stopped being a smooth theoretical line and became a real trace of what a ground terminal would actually experience, second by second.

Margin held steady at 14 to 15 dB across the whole pass — comfortably safe. That wasn't the surprising part.

The surprising part was Doppler. As the satellite approaches, the received frequency shifts up — same effect as a car horn rising in pitch as it comes toward you. As it recedes, it flips negative. Across this one pass, the shift swung from about +85 kHz down through zero to −85 kHz — a frequency offset a real receiver has to track and correct for, or the signal simply isn't where the radio expects it to be. It's invisible in a static link budget and unavoidable in a real one.

That gave me a cleaner way to think about the whole problem: margin and Doppler are two separate questions. A satellite can be perfectly safe on power and still demand real engineering effort to keep the frequency locked. Last week answered "will the link close, in general." This week answers "what does the link actually do, right now, across this specific pass."

One honesty note, same as last week: the satellite's orbit and position here are real, live data. The transmit power and antenna gain are still estimated from published figures, not the operator's proprietary specs — a TLE only describes motion, nothing about the radio payload.

Where this goes next: once you can predict a satellite's geometry ahead of time, you can predict its link margin ahead of time too. And once you can do that across multiple satellites, you've got the beginnings of a scheduling decision — which satellite should serve a link, and when. That's a few weeks out. For now, the milestone is simpler: the curve above came from a satellite that's actually up there, not a number I typed into a config file.

Until next time, Shaif