If you’re thinking about putting up a ground station to talk to LEO satellites, let’s get one thing straight: this isn’t a plug-and-play hobby. It’s a race against physics. A typical low Earth orbit bird screams across your sky in 8 to 15 minutes, and during that tiny window you need to lock on, shake hands, pull down telemetry, and maybe push up a command or two. Unlike the lazy arc of a geostationary dish, LEO tracking is a mechanical and RF sprint. I’ve spent years bolting antennas to frozen rooftops and cursing at garbled bitstreams—so here’s what I wish someone had handed me before I poured the first concrete footing.
Start with the Link Budget, Not the Shopping Cart
It’s tempting to order shiny hardware first, but the real work begins with a spreadsheet. The link budget is your master plan. For a typical LEO downlink—say, UHF at 435 MHz or S-band at 2.2 GHz—the numbers are sobering. A satellite screaming overhead at 7.5 km/s, transmitting a watt or two from a tiny patch antenna, will bury its signal deep in the noise. At 500 km range, free-space path loss alone can exceed 150 dB. Add in atmospheric attenuation, polarization mismatch, and a few dB of implementation loss, and you’re staring at a signal that’s 10 to 20 dB below the thermal noise floor. Your job is to claw it back with antenna gain, a cryogenic or near-cryogenic LNA, and smart signal processing.
Define your minimum elevation angle early. Five degrees is a common cutoff, but if you have terrain shielding, you might push to 2 or 3 degrees. Every degree you gain at the horizon adds precious seconds to the pass. Then calculate the required G/T (gain-to-noise-temperature ratio) for your target Eb/No. For a 1 Mbps QPSK link with a 10⁻⁵ bit error rate, you’ll need an Eb/No around 9.6 dB. Work backward from there. If your G/T comes up short, don’t fudge the assumptions. I’ve seen too many builders convince themselves a 0.8-meter dish will work because they “optimistically” estimated cable loss. It won’t. You’ll miss frames, and you’ll miss them when it counts.
Dish or Phased Array? Pick Your Pain
You’ve got two real choices for the antenna: a steerable parabolic dish or a fixed phased array. The dish is the workhorse. A 1.8-meter S-band reflector with a well-tuned feed can give you 28 dBi of gain and a half-power beamwidth around 5 degrees. That’s a pencil beam. At a 500 km slant range, 5 degrees covers about 44 km of sky—plenty for one satellite, but it means your tracking has to be tight. You’ll need a rotator that updates at 10 Hz or faster to keep the beam centered during an overhead pass. Miss by a degree and you’ve lost 3 dB. Miss by two degrees and the pass is toast.
Phased arrays sound elegant—no moving parts, instant beam steering. But they have their own headaches. At low elevation angles, the effective aperture shrinks and gain sags. A planar array that boasts 20 dBi at zenith might deliver only 15 dBi at 10 degrees above the horizon. If your link budget already has a thin margin, that drop hurts. I’ve also watched arrays fall apart in the field. Ice on the radome, a single failed element, or a calibration drift can warp the beam pattern in ways that are hard to diagnose without an anechoic chamber. For most small ground stations, a dish on a solid rotator is still the more honest solution.
Rotator Mechanics: Backlash Is the Enemy
If you go with a dish, the rotator is where you spend real money. A lightweight amateur rotator like the Yaesu G-5500 is fine for VHF Yagis, but hang a 1.8-meter dish on it in a 50 km/h wind and you’ll quickly discover its limits. You need an industrial azimuth-elevation drive with zero backlash. Backlash in the elevation axis is especially cruel. As the satellite passes near zenith, the elevation rate reverses direction. Any slop in the gears becomes a pointing error that can last several seconds—right when the satellite is closest and the signal is strongest. You’ll watch your Eb/No graph take a nosedive. Direct-drive motors or harmonic drive gearboxes are the fix. Encoder resolution should be 0.1 degrees or better. If that’s out of your budget, accept that you’ll lose data near zenith and plan your passes accordingly.
The Receiver Chain: LNA First, Always
The low-noise amplifier is the heart of your station, and it belongs at the antenna feed—not in a warm equipment rack inside the shack. Every meter of coax between the feed and the LNA adds loss, and that loss adds directly to your system noise temperature. For S-band, a GaAs FET LNA with a 0.5 dB noise figure and 30 dB of gain is the standard. At UHF, you can get below 0.3 dB. Use a bias tee to send DC up the coax; it’s simpler than running a separate power cable that can snag on the rotator and yank a connector loose in a cold wind.
After the LNA, you need filtering—and I mean serious filtering. LEO bands are surrounded by terrestrial crud: pager transmitters, DECT phones, airport surveillance radars. A narrowband cavity filter or a SAW filter with steep skirts will keep your receiver from going deaf. If you’re using an SDR, remember that a strong out-of-band signal can clip the ADC even if you don’t see it on the waterfall. A preselector filter before the SDR isn’t a nice-to-have; it’s mandatory.
SDRs and the Doppler Dance
An SDR like the Ettus USRP B210 or LimeSDR gives you the flexibility to capture the whole passband and process multiple signals at once. But LEO Doppler shift is brutal. At S-band, a 7.5 km/s relative velocity produces a shift of roughly ±50 kHz. Your receiver has to track that in real time. Pre-calculate the Doppler curve from the TLE and update the SDR’s center frequency at 20 Hz intervals. Don’t rely on the SDR’s internal AFC—it will lose lock when the rate of change outruns its loop bandwidth. I learned this the hard way at 3 a.m. during a snowstorm, staring at a waterfall that had drifted into static. Hard-code the Doppler profile and sleep better.
Site Selection: Find a Quiet Spot
You can build the finest receiver chain on the planet, but if your site is next to a 4G tower, you’re dead. LEO ground stations need radio quiet. Spend at least 48 hours with a spectrum analyzer—not just an SDR with a rubber duck—surveying the site. Look for intermittent sources: a toll collection system on a nearby highway, a harbor radar, a weather station’s 400 MHz radiosonde. These can wipe out a pass at exactly the wrong moment.
Terrain shielding is your best free upgrade. A valley or a hill to the north (in the northern hemisphere) can block terrestrial interference while leaving the sky open. Most LEO passes for polar-orbiting satellites come from the north or south. If you can put a ridge between your antenna and the nearest city, you’ll drop the noise floor by 10–20 dB. That’s like doubling your dish diameter without spending a cent.
Grounding and Lightning: Don’t Skip This
Ground stations are lightning magnets. A 2-meter dish on a 5-meter mast is a very attractive path to ground. Install a proper grounding system: multiple ground rods bonded together, a heavy copper strap from the mast to the ground ring, and gas-discharge surge suppressors on every cable entering the building. Coax surge suppressors must be rated for the frequency band—a 1/4-wave stub that works at 144 MHz will look like an open circuit at 2.2 GHz and do nothing. Polyphaser makes band-specific units. Use them. And unplug everything when you’re not actively tracking. No protector is perfect.
Tracking Software and Orbit Propagation
Gpredict is the workhorse for amateur and small professional stations. It handles multiple satellites, computes Doppler shift, and outputs rotator control commands via Hamlib. But TLEs age quickly. A TLE more than 3 days old can produce pointing errors of several degrees for a LEO satellite in a low-drag orbit. Automate TLE fetching from Space-Track or CelesTrak at least twice daily. If you’re tracking a satellite that performs orbital maneuvers, you need operator-in-the-loop updates. A 1-degree pointing error on a 1.8-meter dish at S-band can cost you 3 dB of gain. That’s the difference between a clean decode and a frame full of CRC errors.
For critical passes, consider adding conical scanning. By dithering the antenna position and measuring the received signal strength, you can peak the beam on the actual satellite position rather than relying solely on the propagated TLE. This compensates for orbital decay, TLE staleness, and mechanical misalignments. It adds complexity but buys you back those lost decibels.
Data Processing and Storage
Once the bits are demodulated, you need a pipeline. For a single satellite, a Python script that ingests UDP packets from the SDR software, runs frame synchronization, and dumps CCSDS frames to disk is sufficient. For a constellation, you need a proper telemetry server. Look at NASA’s OpenMCT framework for visualization and command. Store raw IQ recordings of every pass. Disk is cheap; lost data is not. A 10-minute S-band pass at 2 Msps complex samples consumes about 4.8 GB. A 20 TB RAID array will hold a year’s worth of passes. When the satellite anomalies start, you’ll be glad you have the raw captures to reprocess with different demodulator settings.
Automation and Remote Operation
Ground stations should run unattended. Write a scheduler that reads TLEs, computes pass times, and queues tasks: power on the preamplifier 5 minutes before AOS, start the SDR recording, command the rotator, and shut down after LOS. Use a watchdog timer to kill the transmitter if the software crashes. A stuck transmitter can interfere with other ground stations and potentially damage your own receiver on the next pass. I’ve seen a station’s LNA get fried because the scheduler hung and the transmitter kept blasting 50 W into a terminated load that slowly cooked itself. Redundancy in control paths is not paranoia—it’s engineering.
Frequently Asked Questions
What is the minimum dish size for LEO S-band reception?
For a typical 2.2 GHz downlink with a satellite EIRP of 3 dBW and a required Eb/No of 9.6 dB for BPSK at 1 Mbps, a 1.2-meter dish with a 0.5 dB NF LNA provides roughly 3 dB of margin at 10 degrees elevation. A 0.8-meter dish works at zenith but loses lock below 30 degrees elevation. I recommend 1.2 meters as the practical minimum for reliable data capture across the full pass.
How do I handle multiple satellites passing simultaneously?
If the satellites are in the same frequency band, you need either multiple antennas pointed independently or a phased array with multi-beam capability. A single dish can only track one satellite at a time. For constellations like Planet Labs’ Doves, which use different downlink frequencies, you can use a wideband antenna and multiple SDR receivers, each tuned to a specific satellite’s Doppler-corrected frequency. The antenna must have sufficient beamwidth to cover the angular separation of the satellites, which typically requires sacrificing gain for a wider pattern.
What is the biggest mistake first-time ground station builders make?
Underestimating local RF interference. A spectrum analyzer survey is tedious but essential. The second biggest mistake is poor cable management: using lossy coax, skipping the LNA at the feed, or allowing water ingress into connectors. Water in a connector creates a diode effect that generates intermodulation products from nearby transmitters, raising your noise floor by 10–20 dB intermittently. Seal every outdoor connection with self-amalgamating tape and a weatherproof boot.