Case Study: Auditing Planet's Dove Imaging Constellation

"We image the entire Earth's landmass, every day." — the operating premise of Planet's constellation

Executive Summary

Of all the arguments for the small-satellite revolution, the cleanest is a company that photographs the entire land surface of the Earth every single day using a fleet of shoebox-sized CubeSats. Planet's "Flock" of "Dove" 3U CubeSats — on the order of $150$ operational at a time — turned the searchlight limitation of a single low-orbit satellite (Chapter 9) into a floodlight, not with a better satellite but with many coordinated ones. In this case study we do not design that system; we audit it. Given only the physics of the earlier chapters, we will reconstruct why a Dove sits where it does, what resolution its tiny telescope can achieve, and — the heart of the audit — how many satellites daily global coverage demands, then check our number against the size of the real fleet. We will find that the coverage geometry of §33.5, applied to imaging, predicts a fleet of about the size Planet actually flies.

Skills applied

• Placing a small satellite in its orbit and justifying the altitude (§33.1, §33.3; Chapter 9).
• Estimating imaging resolution from the diffraction limit of a small aperture (§33.3; Chapter 9's "closer is finer").
• Sizing a constellation from a revisit requirement using ground-track spacing and swath width (§33.5).
• Auditing the mass-production / accept-failures philosophy against a replacement-rate calculation (§33.1, §33.6).
• Sanity-checking every number against reality — the habit the whole book insists on.

Background

The system

Planet operates the largest fleet of imaging satellites in history — not a few exquisite spacecraft, but a swarm of nearly identical 3U CubeSats, each about the size and mass of a loaf of bread, mass-produced and revised through many generations. The design philosophy the company calls "agile aerospace" is the threshold concept of §33.5 made into a business: build them cheap, fly a lot, expect some to fail, and iterate the design faster than any traditional satellite program could. The numbers below are approximate, widely reported figures (Tier 2), rounded for legibility.

Why this matters

A single 3U CubeSat in low orbit is a superb little camera and a hopeless monitoring system: it sees a $24\ \text{km}$ strip beneath it for a few minutes and does not return for hours or days. The requirement "image everywhere every day" cannot be met by any one satellite at any altitude — a low satellite cannot see enough at once, and a high satellite cannot resolve fine detail. The only way to satisfy both is a constellation, and the design question is the one from §33.5 in a revisit costume: how many satellites does daily global coverage take? Let us audit whether $\sim 150$ is the right answer.

Phase 1: The satellite and its orbit

A Dove is a 3U CubeSat almost entirely filled by a telescope, with COTS electronics, body- and wing-mounted solar cells generating a handful of watts (§33.3), and — like most CubeSats — little or no propulsion, relying on the low orbit to eventually deorbit it (§33.6). Its sun-synchronous orbit is not an accident: as we saw in Chapter 9, sun-synchronous means every image of a place is taken at the same local solar time, under comparable lighting, so that changes on the ground — a flooded field, a new building, a moving ship — stand out instead of being lost in shifting shadows. For a company selling change detection, constant lighting is the whole product. The choice of $\sim 475$–$520\ \text{km}$ is the §33.6 sustainability logic at work too: low enough that a dead Dove re-enters within a few years rather than becoming long-lived debris.

Phase 2: How sharp is a shoebox telescope?

The resolution of a telescope is capped by diffraction: an aperture of diameter $D$ observing at wavelength $\lambda$ cannot resolve an angle finer than about $\theta \approx 1.22\,\lambda/D$. A 3U CubeSat can fit a telescope aperture of roughly $D = 9\ \text{cm}$ across its $10\ \text{cm}$ face; take visible light, $\lambda = 0.55\ \mu\text{m}$. Then $$\theta \approx \frac{1.22 \times 0.55\times10^{-6}\ \text{m}}{0.09\ \text{m}} = 7.5\times10^{-6}\ \text{rad}.$$ On the ground, from an altitude $H = 475\ \text{km} = 4.75\times10^{5}\ \text{m}$, that angle subtends $$\text{GSD} \approx \theta \, H = 7.5\times10^{-6} \times 4.75\times10^{5}\ \text{m} \approx 3.5\ \text{m}.$$ So the smallest feature a Dove can resolve is a few metres — enough to see a truck, not a licence plate. This matches the reported $\sim 3$–$4\ \text{m}$ resolution almost exactly, and it makes Chapter 9's "closer is finer" quantitative: at GEO, $75\times$ farther away, the same telescope would resolve only $\sim 260\ \text{m}$, useless for this job. Being low is not just cheaper; for an imager it is the difference between a product and a blur.

🔧 Engineering Reality: Diffraction is a ceiling, not a guarantee — real resolution is also limited by the detector's pixel size, pointing stability, and the atmosphere. But the diffraction estimate tells you what is possible with a given aperture, and it explains why you cannot cheat physics with a small satellite: a sharper image needs a bigger aperture, and a bigger aperture needs a bigger satellite. The $10\ \text{cm}$ face sets a hard limit on how sharp a CubeSat imager can ever be.

Phase 3: How many satellites for daily coverage?

Now the core audit. A Dove images a $w = 24\ \text{km}$ swath directly beneath its ground track. As it orbits, the Earth turns underneath, so each successive orbit lays down a fresh strip to the west of the last. Daily global coverage means those strips must tile the equator with no gaps in one day.

Strategy first. Find the orbital period, hence the number of orbits per day; that fixes how far apart (in longitude) successive ground tracks fall. Then ask how many satellites, evenly spaced around the orbit, are needed so their interleaved tracks close the gap to within one swath width.

The orbit radius is $a = R_E + H = 6{,}371 + 475 = 6{,}846\ \text{km}$. From Kepler's third law (Chapter 8), with Earth's $\mu = 3.986\times10^{5}\ \text{km}^3/\text{s}^2$: $$T = 2\pi\sqrt{\frac{a^3}{\mu}} = 2\pi\sqrt{\frac{(6{,}846)^3}{3.986\times10^{5}}} = 2\pi\sqrt{8.05\times10^{5}\ \text{s}^2} = 2\pi\,(897\ \text{s}) = 5{,}638\ \text{s} = 94.0\ \text{min}.$$ So the satellite completes $1{,}440 / 94.0 = 15.3$ orbits per day. During each orbit the Earth rotates $360^\circ / 15.3 = 23.5^\circ$, so successive ground tracks are spaced $23.5^\circ$ of longitude apart. At the equator, where one degree of longitude is $40{,}075/360 = 111.3\ \text{km}$, that gap is $$23.5^\circ \times 111.3\ \text{km}/^\circ = 2{,}610\ \text{km}.$$ A single satellite's $24\ \text{km}$ swath cannot fill a $2{,}610\ \text{km}$ gap in a day. Place $N$ satellites evenly around the same orbital plane, though, and their tracks interleave, each falling $23.5^\circ/N$ apart. To close the gap to one swath width: $$\frac{23.5^\circ}{N} \le \frac{w}{111.3\ \text{km}/^\circ} = \frac{24}{111.3} = 0.216^\circ \quad\Longrightarrow\quad N \ge \frac{23.5}{0.216} \approx 109.$$ About 109 satellites in the plane give daily coverage of the equator — and higher latitudes are covered more often, because the ground tracks converge toward the poles. Planet flies on the order of $150$.

Phase 4: Auditing the philosophy — mass production and replacement

Why $150$ and not exactly $109$? Because a mega-fleet audit is not only geometry; it is reliability at scale (§33.6, and Chapter 32). The extra satellites are margin: for guaranteed no-gap coverage with imperfect tiling, for cloud cover (an optical satellite sees nothing through clouds, so multiple looks help), and — crucially — for the satellites that fail. CubeSats built from COTS parts on short lifetimes do die; the constellation is engineered to shrug it off.

Put a number on it. Suppose the fleet loses $\sim 15\%$ of its satellites per year to failures and natural decay. To hold $\sim 150$ on orbit, the operator must launch $$0.15 \times 150 \approx 22\ \text{satellites per year}.$$ Twenty-odd 3U CubeSats a year is a handful of rideshare slots (§33.4) — utterly routine, and cheap enough that each replacement is a better, newer design than the one it replaces. This is the whole business model in one calculation: a fleet you continuously refresh, where losing a satellite is a scheduled cost, not a catastrophe. Contrast a flagship: lose your one billion-dollar imaging satellite and the mission is simply over. The audit confirms the threshold concept of §33.5 — reliability has moved from the satellite to the system.

Phase 5: Sanity checks

Three checks, because a number you have not sanity-checked is a number you do not yet believe:

1. Resolution vs altitude. Our $3.5\ \text{m}$ GSD at $475\ \text{km}$ matches the reported $3$–$4\ \text{m}$, and scales correctly (a higher orbit would blur it) — consistent with Chapter 9.
2. Fleet size vs geometry. Our $\sim 109$ satellites for daily equatorial coverage sits just below the $\sim 150$ actually flown, and the difference is exactly the margin (overlap, clouds, failures) that Phase 4 predicts. The order of magnitude is right, which for a back-of-the-envelope audit is the win.
3. Direction of every knob. Wider swath → fewer satellites; higher orbit → wider swath but coarser resolution; more failures → more replacements. Every dependency points the physically sensible way.

Our reconstruction, fed nothing but Kepler's third law, the diffraction limit, and the coverage geometry of this chapter, predicts a fleet of about the size Planet flies to do a thing — image all of Earth's land daily at a few metres — that no single satellite of any size could do. That is the small-satellite revolution audited from first principles.

Discussion Questions

1. Planet chose sun-synchronous orbits rather than a random inclination. Using Chapter 9, give the two distinct advantages (lighting and coverage) this buys an imaging constellation.
2. Our estimate assumed a single orbital plane. If Planet spread the same $150$ satellites across three planes instead of one, what would improve, and what would get worse, for daily global coverage?
3. The diffraction limit ties resolution to aperture, and aperture to satellite size. Explain why this means a CubeSat imager can never match a flagship's resolution — and why daily revisit is the advantage it offers instead.
4. Phase 4 treated a $15\%$ annual loss rate as a feature, not a failure. Under what cost conditions does "expect losses and replace them" beat "make each satellite reliable"? When would it not?

Your Turn: Extensions

• Option A (analysis). Recompute the fleet size for a coarser swath of $40\ \text{km}$ (a wider-field, lower-resolution sensor). How many satellites now give daily equatorial coverage, and what did you trade away to need fewer of them?
• Option B (computation). Write a Python function sats_for_daily_coverage(alt_km, swath_km) that computes the orbital period, the inter-track longitude gap, and the number of satellites for daily equatorial coverage. Reproduce the $\sim 109$ result for $(475, 24)$. (Do not run it; hand-trace and add # Expected output:.)
• Option C (design). Planet images in optical light and is blind through clouds. Sketch how you would size a radar imaging constellation instead (radar sees through clouds and at night) — would you need more or fewer satellites for the same daily coverage, and what new subsystem constraint (Chapter 25) dominates?

Key Takeaways

1. A constellation turns "see everywhere, all the time" from impossible into routine. No single satellite can image all of Earth daily; $\sim 150$ coordinated CubeSats can.
2. The fleet size follows from geometry. Orbital period sets the ground-track spacing; swath width and satellite count close the gap. Daily equatorial coverage needs $\sim 109$ satellites — close to the $\sim 150$ flown, the difference being margin.
3. Resolution is set by physics, not budget. The diffraction limit ties sharpness to aperture, so a CubeSat trades resolution for revisit — a few metres, but every day.
4. Reliability lives in the system. A $15\%$ annual loss rate means $\sim 22$ cheap replacement launches a year — a scheduled cost, not a catastrophe. The satellite is disposable; the constellation is the asset.