Build a Backyard Radio Telescope From an Old Satellite Dish

Sep 16, 2025

If you’ve ever wanted to hear the Galaxy breathe, you can. The 21-centimeter hydrogen line is a real, loud-ish astrophysical fingerprint sitting at 1420.405... MHz; it’s the spectral whisper of neutral hydrogen atoms flipping their electron spin, and amateur setups built from cheap SDRs and repurposed dishes can and do detect it. This piece walks you through the physics you need to understand, the hardware that actually works, how to aim and calibrate, and the software/processing tricks that turn a noisy spectrum into a meaningful map. I’ll be prescriptive where it matters: if you want reliable results with the least amount of grief, follow the “dish + tuned feed + LNA + bandpass + SDR + averaging” path I outline below.

Why the 21-centimeter line, and what “detectable” actually means

Neutral hydrogen everywhere in the Galaxy emits (very weakly) at a single frequency: 1420.40575177 MHz (commonly rounded to 1420.4 MHz). Because the natural emission is extremely narrow, what you actually measure from Earth is almost always dominated by Doppler shifts (gas motion) and terrestrial interference; the signal becomes visible only after you collect a fair bit of aperture and average the noise down over time. Amateur builds routinely detect the hydrogen line by pointing a directional antenna towards the Milky Way (where neutral hydrogen is abundant) and integrating for minutes to hours. Think of it like long-exposure astrophotography but in frequency space.

Turn the parabolic dish into a collector, put a feed at the dish’s focal point tuned near 1.42 GHz, amplify/filter before the cable runs, digitize with an SDR that covers ~1.4 GHz, and average the resulting spectra until the hydrogen peak pops above the noise.

What an old satellite dish gives you (and what it doesn’t)

A parabolic dish is just a collecting aperture; its ability to "hear" the hydrogen line is set by physical aperture (diameter), operating wavelength (~0.211 m), and how well you illuminate the mouth of the dish. The gain of a circular parabolic reflector is approximately:

G ≈ η(πD/λ)²

where η is aperture efficiency (typical 0.5–0.7) and the result is in linear units. The dish beamwidth scales roughly as:

HPBW ≈ 70λ/D degrees

So a 1.0 m dish gives you something like 20–25 dBi of gain and a ~15° beam (ballpark), while a 2.4 m TVRO dish is several dB better and has a much tighter beam. These formulas let you predict whether your dish is worth the trouble: a meter-class reflector plus a good front end is enough for a simple hydrogen detection; bigger is always better.

(If you like numbers: using a reasonable efficiency of 0.6 and the 21.1 cm wavelength, a 1.0 m dish yields roughly +21 dBi and a half-power beam of ~15 degrees; a 2.4 m dish gets you near +29 dBi and ~6° HPBW — the point being: even modest dishes concentrate the signal meaningfully.)

Start by measuring your dish. Measure diameter and depth, calculate the focal length from the parabola relation:

f = D²/(16c)

where D is dish diameter and c is the dish depth at the center; this tells you where to place the feed.

The feed: don’t use an off-the-shelf Ku-band LNB and expect miracles

A lot of old dishes are paired with Ku-band LNBs (10.7–12.75 GHz). Those boxes are built to receive microwave signals and downconvert them to L-band for TV tuners; their input circuits are tuned for 10+ GHz and will not sensibly receive a 1.42 GHz sky signal. You need a feed element that actually responds at ~1420 MHz: options that people use with success include simple resonant dipoles or sleeve-dipoles placed at the dish focus, small horn feeds, or purpose-built patch/slot feeds designed for L-band. Horn feeds are historically significant (Ewen and Purcell’s first detection used a horn), but a well-sized dipole or patch matched to 1420 MHz is usually simpler and works fine for hobby work. If someone suggests “just buy any LNB and plug it in”, treat that as a shortcut that will likely fail for the hydrogen line.

Front-end electronics: where you win or lose

Your radio telescope’s sensitivity is dominated by the front end’s noise figure and the amount of unwanted RFI you admit into the chain. The low-regret configuration is: the feed physically at the focal point connects to a low-noise amplifier (LNA) mounted physically close to the feed, followed immediately by a narrow band-pass filter centered on 1420 MHz, then a coax run to your SDR. Keeping the LNA as close to the feed as practical reduces cable loss before amplification; the filter after the LNA protects your SDR from strong out-of-band signals and makes averaging behave. Hobbyists commonly use cheap wideband LNAs such as “LNA4ALL” or small prefiltered products (NooElec, GPIO Labs, Radio Astronomy Supplies make LNA+filter modules aimed at H-line work), and the community has tutorials showing good, low-cost front-end builds using two LNAs and a bandpass. Expect to power the LNA via a bias-tee or local DC feed and pay attention to grounding and RF-shielding to stop mains hash and switching supplies from swamping the sky.

Digitization: Which SDR to choose

You can detect the hydrogen line with a cheap RTL-SDR dongle plus a decent LNA and filter — many hobbyists have done exactly that — but if you want cleaner spectra, higher dynamic range, and easier long integrations, spend more for an SDR with better front-end performance: Airspy and SDRplay units (RSP series) are popular step-ups, and they cover the required 1.42 GHz range reliably. The RTL-SDR Blog V3 dongle is attractive for price and has a built-in bias-tee (handy for powering LNAs), which is why it’s often recommended for first projects. The digitizer’s sampling bandwidth also matters: for mapping the hydrogen line a few MHz of bandwidth is plenty, but you want enough instantaneous bandwidth to capture the local Doppler spread and to place a comfortable baseline for calibration.

Mounts and pointing: how accurate do you need to be?

Because the hydrogen signal is spatially extended (Galactic plane emission), you don’t need arcsecond pointing. For a meter-class dish, rough pointing is enough: point toward the Milky Way or use the Sun as a calibration beacon and then move to the galactic plane. For mapping or interferometry, you’ll want an alt-az rotator or a simple motorized az/el rig; many hobbyists convert an old telescope mount, build a stepper-motor alt/az rotator (Arduino/ESP32 driven), or adopt community rotator designs (the SATNOGS community has low-cost tracker designs you can learn from). If you only want a proof-of-concept hydrogen detection, a fixed dish pointed at the Galactic plane and letting the sky drift by often works.

Software and signal-processing: averaging, baseline removal, calibration

Detecting the hydrogen line is mostly a signal-processing problem. The raw spectrum is dominated by system noise and RFI; the line emerges only after you stack (average) many spectra and correct for baseline ripples and the receiver’s gain shape. SDR# with the IF-Average plugin, GNU Radio, Python (numpy/scipy/matplotlib), and specialized programs like TotalPower (for 21 cm mapping) are commonly used. The usual flow is: capture narrowband I/Q chunks centered on 1420.405 MHz, produce many FFT spectra, average them (time-averaging increases SNR as √time), subtract a smoothed continuum baseline, and then look for the faint, narrow spectral bump or Doppler-broadened profile. Community projects such as Project H-Line provide scripts and pipelines you can reuse instead of inventing everything from scratch.

Radio frequency interference (RFI): the single biggest practical problem

Terrestrial RFI at L-band (cell phones, Wi-Fi harmonics, radar, satellite uplinks, in-house electronics) will often be your limiting factor. Professional observatories maintain radio-quiet zones; you cannot, but you can mitigate. Use a narrow, well-centered bandpass filter at 1420 MHz to throw away strong known interferers, choose shielding and good RF connectors, put the LNA at the feed, and plan long off-source “reference” observations to subtract local contamination. If your neighborhood has heavy L-band activity, that doesn’t mean you can’t detect the hydrogen line — it means you must be disciplined about filtering and averaging, and you may need to move to a quieter site for the cleanest maps.

Practical walk-through (in words, not bullets)

Start by measuring your dish. Measure diameter and depth, calculate the focal length from the parabola relation f=D216cf = \dfrac{D^{2}}{16c}f=16cD2 where DDD is dish diameter and ccc is the dish depth at the center; this tells you where to place the feed. Build or buy a feed tuned to ~1420 MHz — a short half-wave dipole with a small reflector or a small horn are both proven choices; mount the feed at the calculated focal point with small, non-conductive supports if possible. Attach a low-noise amplifier right at the feed and follow it with a 1420 MHz bandpass filter; keep coax runs to your SDR as short as practical and use high-quality SMA/N connectors. Choose an SDR that covers 1.42 GHz (RTL-SDR V3, Airspy, SDRplay RSP, etc.), enable bias-tee if you need to power the LNA over coax, and capture narrowband I/Q data. Point the dish to the Galactic plane or to the Sun as your first check; you should be able to see a continuum jump on the Sun and, with enough averaging and a clean front end, a hydrogen peak when you point at a dense patch of the Milky Way. For mapping, sample many pointings and build a grid of averaged spectra; software packages and community scripts will help with stacking, baseline removal, and Doppler interpretation. Authoritative community writeups show examples of complete low-cost systems and software pipelines if you want a reference implementation to copy.

Common mistakes and hard lessons

Assume the dish is “good enough” mechanically, but verify the surface for holes, dents, or rust that ruin the effective aperture. Don’t feed the dish with a Ku LNB and expect 1.42 GHz sensitivity. Don’t let long coax come before the LNA. Don’t skip a bandpass filter if you live near transmitters. Power supplies, switching adapters, and LED drivers are stealthy noise sources — bench test the whole chain for spurs before you deploy. If you do eight things well and one poorly, the poor item will dominate your noise budget. Community threads and tutorials are full of people who spent weeks chasing a hum that was a cheap switching wall wart.

Typical parts and rough cost expectations

Costs vary wildly by quality and country, but as a practical benchmark: an old dish can be free; a basic feed you can build from copper wire or buy for tens of dollars; a decent LNA/filter module made for the hydrogen line is often in the tens-to-low-hundreds USD range; an RTL-SDR dongle is roughly $20–40, while better SDRs (Airspy, SDRplay) are in the low hundreds. Add a rotator, cabling, and weatherproofing, and you’re into a few hundred dollars for a robust backyard setup and a few thousand for more ambitious, higher-performance systems. Community “cheap hydrogen line” projects demonstrate the science for sub-$200 front-end + dongle configurations, but the incremental stability and dynamic range of a better SDR and a professionally made LNA/filter will pay dividends fast.

A realistic project plan you can follow this weekend

If you want a low-friction route: find a 0.6–1.2 m dish, build a simple half-wave dipole feed and mount it at the focus, attach a quality LNA (LNA4ALL or a prefiltered module) and a 1420 MHz bandpass filter, plug into an RTL-SDR V3 dongle (bias-tee enabled), and run SDR# with the IF-Average plugin or use an Airspy/SDRplay with their tools. Point to the galactic plane at night, run many 1–5 second captures, average them, and gently smooth the baseline; with luck and a reasonably quiet site, you’ll see the hydrogen bump within an hour. If that succeeds, iterate: improve the feed, shorten the cable, better LNA, or add a sturdier mount. Community repositories and projects (Project H-Line, TotalPower, and the many RTL-SDR walkthroughs) supply step-by-step scripts and examples to copy.

Why you should build this

This is one of the few backyard astronomy projects that lets you directly detect a physical tracer of galactic structure. It teaches real RF engineering, careful measurement practice, and signal processing. You’ll learn about beam shaping, noise budgets, RFI mitigation, and spectral analysis — and you’ll produce maps that are physically meaningful (galactic rotation, Doppler shifts). The hobbyist community has done a lot of the heavy lifting — reproducible guides, pre-filtered LNAs, and ready software exist so you won’t be alone — but the satisfying truth is that the core experiment remains simple enough that a determined, pragmatic person can build it with used parts and a little patience. If you want, I can sketch a concrete parts list tuned to your budget and the specific dish diameter you have, and produce a wiring diagram and a minimal GNU-Radio/SDR# capture script you can run immediately.

(Reality Theory is the official publication for Everything Astronomy and the Universe, which are two groups of the same name that I admin for.

You can find them here:

The main group

The satellite group)

Everything Astronomy and the Universe Newsletter

Discussion about this post