---
title: "The galaxy's missing mass, measured in a backyard"
locale: "en"
url: "https://irz.fr/en/articles/detecting-dark-matter-backyard-en"
markdown_url: "https://irz.fr/en/articles/detecting-dark-matter-backyard-en.md"
category: "craft"
tags: ["radio astronomy", "SDR", "dark matter", "hydrogen line", "instrumentation"]
published_at: "2026-08-25T14:00:00.000Z"
author: "Arthur Lacoste"
translation: "https://irz.fr/fr/articles/matiere-noire-depuis-le-jardin-fr.md"
---

# The galaxy's missing mass, measured in a backyard

A horn made of roof flashing, a filtered $45 amplifier, an SDR dongle: a former IEEE Spectrum editor says he detected dark matter from his backyard. Here is what he actually measured, and where the proof stops.

The horn antenna is built from a $25 roll of roof flashing and an emptied paint-thinner can. The amplifier costs $45, the receiver around thirty, the software nothing at all. With that contraption set up in his yard, David Schneider claims he detected dark matter.[1](https://spectrum.ieee.org/dark-matter) The claim deserves to be taken seriously, then taken apart piece by piece: he obviously observed no particle. What he measured are velocities, those of hydrogen clouds drifting through the Milky Way. And those velocities refuse to drop with distance, exactly as predicted when invisible mass dominates the galaxy.

That is the same argument that forced dark matter onto astronomers in the 1970s. It now runs on roughly a hundred dollars of consumer hardware, a laptop and a spreadsheet. So the interesting question is not whether to believe it. It is what is actually being measured, and where the proof stops.

> Illustration: Panorama of the Milky Way: the galactic disk crossed by dark dust bands above an observing site. The target of the measurements: the galactic plane between longitudes 15° and 90°, the only region where the method works. Credit: [Serge Brunier / ESO](https://www.eso.org/public/images/eso0932a/).

## Flashing and a paint can

Schneider is no dabbler: a former IEEE Spectrum editor, he published plans for a smaller horn tuned to the same line back in 2019.[1](https://spectrum.ieee.org/dark-matter) The 2026 version grew for one specific reason: better angular resolution, so he could sweep narrower patches of sky.

Construction remains a weekend job. An online horn calculator provides the dimensions, chosen to fit a 10-foot-by-2-inch roll of aluminum flashing bought for $25 at Ace Hardware and an emptied one-gallon paint-thinner can. Schneider admits that next time he will simply buy the empty F-style can rather than drain one.[1](https://spectrum.ieee.org/dark-matter)

Then came the question of what the thing was actually worth. To measure his beamwidth, he pointed the horn at an Inmarsat geostationary satellite, a bright fixed source: the response came out at about 20 degrees.[1](https://spectrum.ieee.org/dark-matter) Twenty degrees is enormous; an optical telescope divides that angle by tens of thousands. At the galactic center's distance, that beam spans roughly 2,850 parsecs of real space (an IRZ calculation): every single measurement swallows a slice of disk about a third as wide as the Sun's orbital radius. But the method used here does not need sharp imaging. In every direction it looks for the cloud receding fastest. A wide funnel works fine as long as the funnel stays pointed straight.

The receive chain fits in two boxes. First a Nooelec SAWBird+ H1: two cascaded low-noise amplifiers wrapped around a saw-filter centered on 1,420 MHz, guaranteed gain beyond 40 dB, noise figure quoted at 0.8 dB.[2](https://www.nooelec.com/store/sawbird-h1.html) Power travels down the coaxial cable itself through the dongle's bias tee; the module also accepts three external sources.[2](https://www.nooelec.com/store/sawbird-h1.html) Then an RTL-SDR V4 dongle, a direct descendant of the DVB-T receivers repurposed by an entire community, sold for $29.95 alone or $39.95 with antennas by its maker.[3](https://www.rtl-sdr.com/) A very current detail: that model was just declared end-of-life in July 2026, the stockpile of tuner chips exhausted, with a lighter successor already shipping.[3](https://www.rtl-sdr.com/)

On the software side, SDR# runs the receiver,[4](https://airspy.com/download/) the IF Average plug-in stacks incoming spectra minute after minute until the signal surfaces,[5](https://www.astronomy.me.uk/latest-version-of-if-average-plug-in-for-sdr-sharp-software) and Stellarium says where to point. The spreadsheet does the rest.

> The horn's bill
> **Under $120 without cables**
> - roof flashing, 10 ft × 2 in: $25
> - SAWBird+ H1, filtered 1,420 MHz LNA: $45
> - RTL-SDR V4 dongle: $30–40
> Free software: SDR#, IF Average, Stellarium. Sources: IEEE Spectrum, Nooelec, RTL-SDR Blog.

For scale: in 1950, Edward Purcell obtained $500 from the Rumford Fund to bankroll his student Harold Ewen's horn, roughly twelve times more in today's purchasing power (an approximate IRZ conversion using US inflation). The project consumed a year of weekends.[7](https://www.gb.nrao.edu/fgdocs/HI21cm/ephorn.html) Hardware stopped being the bottleneck long ago. What really changed is that the complete chain now fits in one shopping basket.

## The hum of 1951

What the horn aims at cannot literally be heard: an emission line at 1,420.405751 MHz, produced by neutral hydrogen when its spin component flips. The wavelength, 21 centimeters, has a precious property: it passes through the dust clouds that hide the center of the galaxy from optical telescopes. Where the eye fails, hydrogen talks.

The idea was born during the war, in Leiden. Jan Oort, convinced that radio could map the Galaxy, asked his student Hendrik van de Hulst to find a usable line. Van de Hulst identified hydrogen's hyperfine transition in 1944 but judged detection improbable: the transition is so slow the emission seemed doomed to stay invisible.[6](https://www.nrao.edu/archives/Ewen/ewen_HI.shtml) Oort pushed on and funded the attempt anyway. That stubbornness started the race.

It ended on Easter Sunday 1951. Around three in the morning, Ewen, with his horn jutting out of a fourth-floor window of Harvard's Lyman Lab, got a signal he immediately attributed to Doppler shift. His receiver used a trick new to astronomy, frequency switching, which cancels background noise by comparing two adjacent channels. The project had cost $500 and a year of weekends; rain had once flooded the lab through the clogged horn, and students used it for snowball practice.[6](https://www.nrao.edu/archives/Ewen/ewen_HI.shtml)[7](https://www.gb.nrao.edu/fgdocs/HI21cm/ephorn.html)

The Dutch confirmed on May 11 after adopting frequency switching, the Australians on July 12. All three teams published in the same issue of Nature, September 1, 1951. And the Dutch paper's title says everything: Muller and Oort announced the line "and an Estimate of Galactic Rotation".[7](https://www.gb.nrao.edu/fgdocs/HI21cm/ephorn.html)[8](https://www.nature.com/articles/168356a0) Within months the line had become the instrument that measures how the Galaxy turns. Purcell himself shared the 1952 Nobel Prize for his work on nuclear magnetic resonance.[7](https://www.gb.nrao.edu/fgdocs/HI21cm/ephorn.html)

> Illustration: Harold Ewen and Edward Purcell standing next to the horn used for the first detection of interstellar hydrogen. Ewen and Purcell in 1956 beside the 1951 horn, now displayed outside Jansky Lab at Green Bank. Credit: [NRAO/AUI Archives](https://www.nrao.edu/archives/items/show/21383).

So the backyard experiment replays the founding measurement of radio astronomy. Not a commemorative copy: the same line, the same kind of horn, the same Doppler physics, with seventy-five years of cheap components added.

## Noise and integration

What arrives at the end of the cable looks nothing like a signal. The hydrogen line drowns in the receiver's thermal noise and in the radio sky itself. No setting makes it pop out at once: it gets built up. The IF Average plug-in adds spectra together over several minutes; useful signal grows with the square root of integration time while noise lags far behind, and eventually a bump emerges.[5](https://www.astronomy.me.uk/latest-version-of-if-average-plug-in-for-sdr-sharp-software) A few minutes of averaging per direction were enough for Schneider to get a usable spectrum; patient amateurs stretch integrations over hours for the last few decibels.

A bump, rarely a spike. A 20-degree beam swallows myriads of clouds at different distances at once, each with its own velocity along the line of sight. The displayed spectrum looks like one broad crest, often resolvable into several overlapping ones.[1](https://spectrum.ieee.org/dark-matter) Measurements published by PhysicsOpenLab with a similar setup show the mechanism plainly: near longitude 12° their spectrum is dominated by the nearly motionless peak of local clouds, the ones rotating along with us; the receding component grows toward 45°, then recedes again by 90°.[9](https://physicsopenlab.org/2020/09/08/measurement-of-the-milky-way-rotation/) The galaxy's spiral structure reads directly in those bumps appearing and sliding.

The order of magnitude explains why such modest hardware suffices. Do the math: the rest frequency sits at 1,420.405751 MHz, and a Doppler shift moves frequency proportionally to speed. One km/s corresponds to a slide of about 4,737 Hz; 200 km/s slides the peak by roughly 950 kHz (IRZ conversions). The entire galactic choreography observable from Earth therefore fits inside a window a few megahertz wide. A tiny window, comfortably within reach of an 8-bit USB dongle once the SAW filter in front of it keeps strong transmitters from blinding the input.

> The Doppler scale
> **A galaxy inside two megahertz**
> - MHz at rest, the neutral hydrogen line: 1420.405751
> - wavelength that crosses the dust: 21 cm
> - of shift per km/s (IRZ calculation): ≈ 4.7 kHz
> Δf = f · v / c. At 200 km/s the peak slides by about 950 kHz.

Schneider models each spectrum as a sum of bell-shaped curves, one per cloud, directly in Excel. From that decomposition he keeps a single number per direction: the largest redshift, belonging to the fastest-receding cloud.[1](https://spectrum.ieee.org/dark-matter) The project's entire scientific processing lives in that sorting step. The counterintuitive part is that this kind of astronomy gains nothing from amplifying harder. It gains from averaging longer.

## The tangent point

What remains is turning apparent velocities into a rotation map. The method dates back to radio astronomy's beginnings and assumes circular orbits. From the Sun, sitting R₀ from the galactic center, you observe along the galactic plane between longitudes 0° (straight toward the center) and 90°. Each sightline crosses clouds on many different orbits, but the one approaching Earth fastest is the one whose orbit grazes the sightline at the tangent point. That maximum is the measurement.

Two formulas do all the work:

V(R) = V_obs + V₀ · sin l
R = R₀ · sin l

> **Tangent-point geometry**
> Diagram showing the Sun, a circular orbit around the galactic center, and a sightline tangent to the orbit, with the formulas V(R) = Vobs + V0 sin l and R = R0 sin l.
> - Galactic center
> - Sun
> - line of sight
> - rotation
> - V(R) = Vobs + V₀·sin l
> - R = R₀·sin l
> - tangent-point method
> - l
> From the Sun, the fastest cloud along the sightline is the one grazing its orbit at the tangent point. Each galactic longitude l becomes a velocity-distance pair.

The constants deserve a look. Amateur guides often use 220 km/s and 7.6 kiloparsecs for the Sun's orbital speed and distance from the center.[9](https://physicsopenlab.org/2020/09/08/measurement-of-the-milky-way-rotation/) The most recent professional reference, published by Sofue and colleagues in 2025, adopts 235.1 km/s and 8.178 kiloparsecs.[10](https://www.ioa.s.u-tokyo.ac.jp/~sofue/papers/sofcomp/2025-pasj-Inner-Rot-Curve-MilkyWay.pdf) These values have moved by several percent across decades, and every amateur result inherits that drift: the method converts longitudes using whatever constants its era trusts.

Before comparing anything, the raw measurement needs cleaning. Earth spins, orbits the Sun, and the Sun drifts relative to nearby stars: three motions contaminating every measured velocity. Earth's orbit alone imprints a seasonal swing on the order of 30 km/s onto each speed. Radio astronomers remove the whole lot by converting to the Local Standard of Rest, a standard correction documented step by step in serious amateur projects.[9](https://physicsopenlab.org/2020/09/08/measurement-of-the-milky-way-rotation/)

The method, incidentally, only lives on one quarter of the sky. Past longitude 90°, no sightline cuts any orbit head-on anymore: no tangent point, no direct conversion. That valid domain is why Schneider sweeps between 15° and 90°.

He picked six longitudes, from 15° to 90° spaced roughly fifteen degrees apart, far enough that each measurement stays largely independent, with Stellarium serving as star chart.[1](https://spectrum.ieee.org/dark-matter) His six velocity-distance pairs match the inner rotation curve published by Sofue's team in 2025 reasonably well. The two innermost points came out low; a second round of curve fitting brings them closer without landing them exactly on trend.[1](https://spectrum.ieee.org/dark-matter) That local anomaly is nothing to be ashamed of: the galactic center is precisely where the tangent-point method meets its worst limits, as we'll see.

The essential point fits in one sentence: the estimated orbital velocities do not diminish with distance from the center, quite the opposite.[1](https://spectrum.ieee.org/dark-matter) If the galaxy's mass were concentrated in the middle the way the Sun's mass rules the planets, Mercury sprinting at 47.4 km/s against Neptune's 5.4, the decline would be unmistakable.[1](https://spectrum.ieee.org/dark-matter) There is none. Something keeps tugging on those clouds where nothing shines.

That is exactly the argument that, applied to Andromeda in 1970 by Vera Rubin and Kent Ford, gave the file its optical centerpiece: their spectrograms covering 3 to 24 kiloparsecs showed total mass growing nearly linearly out to about 14 kpc, then more slowly, well past the visible disk.[11](https://ui.adsabs.harvard.edu/abs/1970ApJ...159..379R/abstract) Schneider does nothing else, in our own galaxy, with a hundred dollars of electronics.

## A debt that kept piling up

The anomaly did not fall out of the sky overnight. Andromeda was already spinning on photographic plates in 1914: Max Wolf and Vesto Slipher noticed spectral lines tilted along the major axis, a sign of rotation, and Francis Pease measured it in 1917 after seventy-nine hours of exposure on the Mount Wilson 60-inch.[13](https://ned.ipac.caltech.edu/level5/Sept16/Bertone/Bertone4.html) In 1939, graduate student Horace Babcock pushed M31's rotation curve out to some twenty kiloparsecs and found velocities too high for a disk made of stars alone; he cautiously blamed light absorption while conceding that new dynamical considerations might be needed.[13](https://ned.ipac.caltech.edu/level5/Sept16/Bertone/Bertone4.html)

In 1959, Franz Kahn and Lodewijk Woltjer observed Andromeda hurtling toward the Milky Way at 125 km/s. Assuming the two galaxies form a bound system, they derived a mass lower bound far beyond what visible matter could supply; in retrospect, one of the first clean hints of dark halos.[13](https://ned.ipac.caltech.edu/level5/Sept16/Bertone/Bertone4.html)

The 21 cm line turned a hint into a case file. After the first radio rotation curve of M31, published in 1957 from Dwingeloo, Kenneth Freeman wrote in 1970, in the appendix of a paper on exponential disks, that if the data were right, these galaxies must contain additional matter undetected either optically or at 21 cm, and at least as massive as everything already seen.[13](https://ned.ipac.caltech.edu/level5/Sept16/Bertone/Bertone4.html) In 1972, Roberts and Whitehurst pushed M31's radio measurement out to 120 arcminutes from the center, far past the last visible stellar disk. Roberts and Rots extended the evidence to three spirals in 1973: their curves declined slowly or not at all at large radii. The following year, two influential papers independently concluded that galaxy masses had been underestimated by a factor of ten. Bosma stacked twenty-five flat curves in his 1978 thesis; Rubin, Ford and Thonnard published their optical curves of spiral galaxies soon after.[13](https://ned.ipac.caltech.edu/level5/Sept16/Bertone/Bertone4.html) The invisible halo became the standard interpretation.

The backyard experiment plugs into a fifty-year-old case file. It reproduces exactly one piece of it: our galaxy, seen from the inside.

## The honest limit

Should we therefore write "dark matter detected at home" with no asterisk? No, and the reasons are instructive.

Start with the instrument. A 20-degree beam is no imaging instrument; it is a funnel blending whole swaths of the galactic disk. Professional surveys like HI4PI chart the entire sky at fractions-of-a-degree resolution, complemented by CO surveys from Nobeyama, Mopra and CfA probing the dense gas the HI line struggles to see, especially toward the center. The 2025 reference curve additionally folds in ultra-precise maser positions from very-long-baseline interferometry plus Gaia data.[10](https://www.ioa.s.u-tokyo.ac.jp/~sofue/papers/sofcomp/2025-pasj-Inner-Rot-Curve-MilkyWay.pdf) Against all that: six points, hand-pointed, fitted in Excel, with no absolute flux calibration.

Then the method. The tangent-point method assumes circular orbits and an axisymmetric galaxy. Sofue's unified curve itself reads in three acts: a brutal climb through the bulge, from roughly 100 km/s at 20 parsecs from the center past 250 km/s by half a kiloparsec; a long plateau around the solar circle; then a monotonic decline out to 25 kpc.[10](https://www.ioa.s.u-tokyo.ac.jp/~sofue/papers/sofcomp/2025-pasj-Inner-Rot-Curve-MilkyWay.pdf) The model reproducing it combines a central black hole of 4 million solar masses, a bulge, a disk and an NFW halo, the shape cold-dark-matter simulations expect; that halo alone yields a local density of 0.107 GeV/cm³, about the mass of one hydrogen atom inside a two-centimeter cube (an IRZ conversion), explicitly given as a lower bound since the outer decline pulls mass away from the fitted halo.[10](https://www.ioa.s.u-tokyo.ac.jp/~sofue/papers/sofcomp/2025-pasj-Inner-Rot-Curve-MilkyWay.pdf) And the center stays unruly: Sofue shows in 2025 that eastern and western curves diverge near the nucleus, the probable signature of a weak bar inside 4 kpc perturbing velocities in exactly the zone where the amateur points stray off trend.[10](https://www.ioa.s.u-tokyo.ac.jp/~sofue/papers/sofcomp/2025-pasj-Inner-Rot-Curve-MilkyWay.pdf) Schneider's defect there is not a crude mistake; it is the method's predictable weakness, in the place where it is least clean.

And then there is the comment posted under the article by Paolo Salucci, one of the specialists on how dark matter distributes inside galaxies.[1](https://spectrum.ieee.org/dark-matter)[12](https://arxiv.org/abs/1811.08843) Two corrections. First, the presence of a massive, non-luminous, non-baryonic component stands independently of catching any beyond-the-Standard-Model particle: nobody has seen the particle, and that was never the experiment. Second, rotation curves are NOT flat, he insists; it is their precise shape, central rise, plateau, slow decline, that encodes the amount and nature of the halo.[1](https://spectrum.ieee.org/dark-matter) Sofue's unified curve indeed declines monotonically beyond the solar circle out to 25 kpc, and the local halo density he derives from it, 0.107 GeV/cm³, explicitly counts as a lower bound.[10](https://www.ioa.s.u-tokyo.ac.jp/~sofue/papers/sofcomp/2025-pasj-Inner-Rot-Curve-MilkyWay.pdf) The shorthand "flat means dark matter", repeated everywhere, is a caricature professional data moved past long ago.

> **What 'detect' means here**
> - orbital velocities on the 21 cm line.: Measured
> - a halo shaping the rotation curve.: Inferred
> - the dark matter particle itself.: Not observed
> After Paolo Salucci's comment on IEEE Spectrum (August 15, 2026).

So what is left of the backyard experiment? A demonstration, and a real one. Reproducing at home, with a metal funnel, the full causal chain, line, Doppler, geometry, mass anomaly, is the difference between reading a result and having executed it. To me this is where the project becomes interesting for any maker: a foundational measurement of twentieth-century science now runs on hobbyist hardware, with no permit, no lab, no budget.

Anyone wanting to dig further has concrete options: more than six longitudes, longer integrations, a commercial feed like the Discovery Dish IEEE Spectrum featured last year, or moving the processing into GNU Radio to control every stage.[1](https://spectrum.ieee.org/dark-matter) Those two lingering inner points await another session. The galaxy, for its part, keeps transmitting, continuously, on 21 cm, license-free and unencrypted. The last remaining obstacle fits in a shopping basket.

## References

1. [David Schneider, 'Detect Dark Matter's Mark From Your Backyard', IEEE Spectrum, July 30, 2026](https://spectrum.ieee.org/dark-matter)
2. [Nooelec, SAWbird+ H1 dual LNA with 1,420 MHz SAW filter (product page)](https://www.nooelec.com/store/sawbird-h1.html)
3. [RTL-SDR Blog, V4 dongle announcement and end-of-life follow-up (rtl-sdr.com)](https://www.rtl-sdr.com/)
4. [Airspy, SDR# download](https://airspy.com/download/)
5. [Lichfield Radio Astronomy Observatory, IF Average plug-in for SDR#](https://www.astronomy.me.uk/latest-version-of-if-average-plug-in-for-sdr-sharp-software)
6. [Harold 'Doc' Ewen, 'Detecting the Interstellar Hydrogen Line, 1951', NRAO Archives](https://www.nrao.edu/archives/Ewen/ewen_HI.shtml)
7. [NRAO Green Bank, 'The Ewen-Purcell Horn' (detection history)](https://www.gb.nrao.edu/fgdocs/HI21cm/ephorn.html)
8. [H.I. Ewen & E.M. Purcell, 'Radiation from Galactic Hydrogen at 1,420 Mc./sec.', Nature 168, 356 (1951)](https://www.nature.com/articles/168356a0)
9. [PhysicsOpenLab, 'Measurement of the Milky Way Rotation' (tangent-point method)](https://physicsopenlab.org/2020/09/08/measurement-of-the-milky-way-rotation/)
10. [Y. Sofue et al., 'The Inner Rotation Curve of the Milky Way', PASJ (2025)](https://www.ioa.s.u-tokyo.ac.jp/~sofue/papers/sofcomp/2025-pasj-Inner-Rot-Curve-MilkyWay.pdf)
11. [V.C. Rubin & W.K. Ford Jr., 'Rotation of the Andromeda Nebula from a Spectroscopic Survey of Emission Regions', ApJ 159, 379 (1970)](https://ui.adsabs.harvard.edu/abs/1970ApJ...159..379R/abstract)
12. [Paolo Salucci, 'The distribution of dark matter in galaxies', The Astronomy and Astrophysics Review 27, 2 (2019)](https://arxiv.org/abs/1811.08843)
13. [G. Bertone & D. Hooper, 'A History of Dark Matter', chapter IV: Galactic Rotation Curves (NASA/NED)](https://ned.ipac.caltech.edu/level5/Sept16/Bertone/Bertone4.html)