Fun With Science / Globe Deconstruction / Unknown Luminaries · pages 155–157

Stacked, Sharpened — and Beside the Point

What the book gets right about how planetary photographs are made, and why granting all of it costs the argument nothing.

Scope. The premise is conceded: sharpening can invent detail. What is graded is the inference from it, that the planets are therefore unknown. Still to add: the noise-phantom demonstration the page proposes, and a same-date amateur–Hubble overlay, feature by feature.

Three pages of the Unknown Luminaries chapter, pp. 155–157, are given to how planetary photographs are made, under the heading “Image stacking + Deconvolution = Reality?” The two arguments that follow both rest on photographs, one captioned “(image stacked and edited)” in the author’s own hand.

The suspicion of processing: reasonable, and partly correctAs a load-bearing objection: aimed at the decoration
Where this lands

The claim: stacking and deconvolution put the detail into planetary photographs rather than drawing it out, so the pictures cannot stand as evidence. The verdict: the first half is partly right, and conceded below; the second does not follow, because the argument is aimed at detail when the evidence was always motion.

Sharpening really can manufacture detail, and a reader is entitled to want the pipeline explained. But the case for a moving Earth was built on where planets are, and when — measurements that need a dot, not a picture, and most of which predate photography. Grant that every processed planetary image is worthless and Neptune is still where Le Verrier said it would be, found about one degree from his calculated place on 23 September 1846. And on a working amateur’s raw video of Jupiter, stacking 3,215 frames cut the noise power at fine scales by 2,140× while leaving the coarse structure untouched — removing only what lay beyond the 0.389″ a 356 mm telescope can resolve, where nothing real could have been.

The stacking removed only what a 356 mm mirror cannot resolve. Software inventing detail has no way of knowing where that limit is.

An arcsecond (″) is 1/3,600 of a degree; Jupiter is about 40 across. The atmosphere — the seeing — blurs any point of light to between one and four arcseconds. A telescope’s own limit is set by its aperture, θ = 1.22λ/D, and no processing passes it.

What the claim is

p. 156 quotes a working tutorial from the astrophysicist behind the Astraveo channel (the book prints “Astaveo”), whose flat-Earth challenge video the chapter uses throughout: capture around ten thousand frames with FireCapture, stack them in AstroSurface. The book’s reply is “Is he implying that Photoshop is the best method to prove to myself that the Earth is not flat?” p. 157 sets a raw frame beside a processed one — “Real Jupiter” against “Jupiter 4.0?” — over a quotation about a wavelet setting that “brings out a lot of the detail,” captioned “How to manipulate a photo to achieve the desired planet 101.”

Most of this is put as questions, and we grade it as put; the caption is the unhedged form the later pages rely on. The charge is that the detail is put in rather than drawn out, and by p. 161 it has become “Do these deconvoluted and sharpened images even resemble an 88,800-mile-wide ball of gas?”

The concession, and it is not a small one

Sharpening can invent. Deconvolution is an inverse problem, and inverse problems amplify noise into structure unless held back deliberately; push a wavelet slider — a sharpening control that boosts contrast at a chosen scale — far enough and you get ringing at the limb, haloes around moons, and belt edges crisper than anything the atmosphere delivered. Over-processed amateur images are full of exactly this, and two people working the same raw capture routinely produce visibly different planets. How many frames to keep, how hard to sharpen, what to do about colour — the operator’s taste is in the result, and “it came out of image-processing software” is a fair thing to want explained.

The detail was never the evidence

The case for a heliocentric, moving Earth rests on where planets are, and when — and a position needs only a dot.

result	what it was measured from	detail required
Retrograde motion of Mars	naked-eye positions, recorded for millennia	none — no optics at all
Kepler’s laws	Tycho’s naked-eye positional catalogue	none — predates the telescope
Speed of light, Rømer 1676	timings of Jovian moon eclipses	none — a stopwatch on a disappearance
Aberration of starlight, Bradley 1728	stellar positions to tens of arcseconds	none — stars are points
Neptune, Le Verrier 1846	residuals in Uranus’s measured positions	none — see below
Mercury’s anomalous precession	transit timings and positions	none — about 43″ per century

Every one of these predates photography, let alone stacking software.

One thing optics does add to the positional case, and it is fair to say so: spin. Cassini timed a spot across Jupiter’s disc in 1665 and had the rotation at about 9 h 56 min; by the nineteenth century the timings were good enough to show the equatorial belts lapping the rest of the planet by five minutes a rotation, which is the first evidence that the visible surface is weather, not ground. That needs a marking, not fine detail — the Red Spot is a fifth of the disc across — and it needs a clock. The number it produced feeds straight back into the positional physics: rotation plus the measured flattening gives how mass is distributed inside, and the interior rotation timed from its radio bursts since the 1960s agrees with what the clouds showed Cassini to within a few minutes.

And the moons the chapter itself points to. The four Galilean satellites are the best positional laboratory a small telescope has: Galileo watched them change places night by night in 1610; their eclipses in Jupiter’s shadow, their transits, and the shadows they throw on the disc — the very shadows the chapter argues about at pp. 160–161 — are published years ahead to the minute and timed by amateurs with a stopwatch; and their orbits are what weigh the planet. A moon a second of arc across is a dot in any telescope. What carries the argument is when the dot goes out, and a dot cannot be sharpened into anything.

Neptune is the case that settles it. It was calculated from the discrepancy between where Uranus was predicted to be and where it was measured to be. Le Verrier published a position; Galle pointed a telescope at it on the evening of 23 September 1846 and found the planet about one degree away. Nobody could see anything on it: no feature at all was recorded on Neptune until methane-band CCD imaging from the ground in 1979, and its weather was first resolved by Voyager 2 in 1989. The standard objection is that Le Verrier was lucky — his predicted orbit was too large and his mass too great, and the planet sat near the place he named partly because of where it happened to be in its orbit that decade. Fair, about the orbital elements. But what the Uranus residuals constrained well was the direction of the unseen mass at that epoch, and direction is what was predicted and found. A wrong orbit that still puts a telescope within a degree of an undiscovered planet is evidence that the pull was real, and says nothing about pictures, because there were none.

Apparent sizes on the evening of Dury’s eight-planet photograph, 22 February 2025:

planet	distance	apparent disc
Venus	0.38 AU (Earth–Sun distances)	44.3″
Jupiter	4.88 AU	40.4″
Saturn	10.56 AU	15.7″
Mars	0.81 AU	11.5″
Mercury	1.25 AU	5.4″
Uranus	19.71 AU	3.6″
Neptune	30.79 AU	2.2″

Uranus and Neptune are at or below the one-to-four-arcsecond blur the sky itself imposes — barely more than points from the ground, through any telescope, however processed.

The two planets in that photograph whose images a sharpening pipeline could most easily have invented are the two whose orbits are known best relative to anything anyone can see of them. The chapter attacks the quantity that was never carrying the weight.

Position is also a much weaker thing to ask of an image than detail: resolving a feature means separating two close points, which seeing destroys; locating an object means finding the centre of its light, which averaging helps. And an astrometric frame — one exposed to measure positions — carries its own calibration — the field stars in the same exposure have known catalogue positions and fix the scale, orientation and distortion of that image. Nothing of the sort exists for a wavelet-sharpened disc, which is why the objection at pp. 155–157 has force there and none here. The book’s own two tests run on positions, not pictures: the eight-planet challenge at pp. 162–163 asks whether a simulation reproduces the placement of the planets, and the Jupiter argument at pp. 158–161 turns on where a shadow sits relative to a moon and when it moves. The imaging critique and the tests it is meant to support are not measuring the same thing.

They are not only visible. They are pulling on us.

There is a second channel the book’s objection cannot touch, because it does not run through a camera: these bodies exert forces we can measure, and some of the measurements are not made by looking at the sky.

Mercury’s orbit turns, and mostly the other planets turn it. Mercury’s perihelion (the point of its orbit nearest the Sun) advances by about 575″ per century against a fixed frame. The famous Einstein residual is 43″; the other 532″, some 92 per cent of the whole effect, is simply the other planets pulling on Mercury, computed from their masses and positions and agreeing with what is observed. Venus and Jupiter do most of it.

Venus and Jupiter are written into rock. The largest-amplitude term in the variation of Earth’s orbital eccentricity is the g₂ − g₅ term — the difference between the rates at which the orbits of Venus and Jupiter precess. Its period is about 405,000 years, it drives climate through insolation (the sunlight a latitude receives), and it is recorded in sediment: the lake beds of the Newark and Hartford basins, the Chinle Formation at Petrified Forest, Japanese pelagic ribbon cherts, the Early Jurassic marine sequences of the Bristol Channel. Shown stable from roughly 215 to 202 million years ago in the Newark–Hartford cores, and further back by the Chinle correlation, it is now a calibration clock for geological time — a “geological orrery” that constrains the planets’ past motions from strata.

The gravitational pull of Venus and Jupiter is legible in mudstone, and geologists count the layers to tell the time. There is no telescope anywhere in that measurement, no sensor, no stacking, no wavelet, no Photoshop.

Jupiter weighed with a neutron star. Millisecond pulsars are clocks; timing one means converting arrival times to the solar system’s centre of mass, and a wrong planetary mass shows up as a wobble in the residuals at that planet’s orbital period. From four pulsars the Jovian system mass comes out at 9.547921(2)×10⁻⁴ solar masses — about four times better than the Pioneer and Voyager flyby values, and within a factor of twenty of what Galileo achieved in orbit.

Three limits. These effects are secular, accumulating over centuries and hundreds of millennia; nothing here says Jupiter influences a Tuesday. They are inferences inside a gravitational model, not proofs of it; what they add is channels that share no instrument or pipeline. And the sedimentary link runs through climate — orbit to insolation to lake level to lithology — each link with its own uncertainty; the 405,000-year signal is robust enough to serve as a metronome, but it is not a direct measurement of Jupiter.

What we know about Jupiter that no picture shows

The chapter’s question is what the planets are, and it looks for the answer in pictures. Most of what is known about Jupiter’s nature came from somewhere else: from timing its moons, from spreading its light into a spectrum, from measuring heat and radio waves the eye cannot see, and from watching a star go behind it. None of those steps sharpens an image. Set aside everything a spacecraft has measured on the spot and this is the answer sheet as it stood from the ground and from orbit around Earth.

when	the measurement	what it says Jupiter is
1610 → 1687	Galileo times the four moons; Newton applies Kepler’s third law to Callisto’s orbit in the Principia	Its mass: 1/1,067 of the Sun by Newton’s figure, 1/1,047 today — 318 Earths. A dot and a clock, no picture.
1664–65	Hooke reports a spot; Cassini times it across the disc at about 9 h 56 min, and sees the disc visibly flattened — twenty years before Newton explains why	Its spin and shape: 6.5% oblate, the fastest-spinning planet. Mass and size together give the density, 1.33 g/cm³ — a quarter of Earth’s, and Newton derived it in the same proposition. Whatever it is, it is not rock.
1932	Rupert Wildt matches the dark bands in Jupiter’s spectrum to laboratory methane and ammonia	What is in its air: CH₄ and NH₃, among the first molecules identified on another planet — the same year Adams and Dunham found carbon dioxide on Venus. The bands themselves had been seen since the 1860s; Wildt’s step was naming them. A spectrum is a photograph of composition, and it is unsharpened by construction.
1955	Burke and Franklin pick up radio bursts from Jupiter at 22 MHz by accident; decimetric emission follows in 1959	It has a magnetic field and radiation belts, more than ten times Earth’s field; by the 1960s the bursts’ own period (System III) gave the rotation of the interior, not the clouds.
1960	Kiess, Corliss and Kiess detect the faint quadrupole lines of molecular hydrogen	It is mostly hydrogen, as its density had already implied; helium, invisible in the optical, is inferred from how the hydrogen lines are shaped and later measured at about one molecule in seven, a quarter by mass.
1966–69	Frank Low and colleagues measure Jupiter at 8–14 µm from the ground, then from a Learjet at 15 km with a broadband 1.5–350 µm bolometer	How hot it is, and that it makes its own heat: cloud tops near 125–135 K, and by the 1969 measurement nearly three times as much energy radiated as sunlight delivers — since revised to about 1.7 times, which is still its own heat. It is still cooling from its formation.
1971	Jupiter passes in front of the star β Scorpii; the star’s fading is timed from several sites	The temperature and scale height of the upper atmosphere, from how fast the starlight dims — a hydrogen–helium atmosphere at roughly 150–200 K, measured with a photometer on a point of light — this page’s whole thesis in one row.
1979	Peale, Cassen and Reynolds compute the tidal heating of Io from the Laplace resonance of the inner moons	A prediction about a moon: enough heat to melt Io’s interior and drive volcanoes — published days before a spacecraft photographed the first plume. Gravity said it first.
1980s–	Interferometers map Jupiter’s thermal radio emission at centimetre wavelengths; the VLA in particular	What lies under the clouds: ammonia’s distribution and the temperature tens of kilometres below the visible deck, from wavelengths that see through it.
1994	Comet Shoemaker–Levy 9 hits Jupiter; the plumes are watched spectroscopically from the ground and by Hubble	What comes up from below: sulphur compounds, carbon monoxide, water — the deep atmosphere sampled by an impact, on a schedule announced a year ahead.

Every entry is a dot, a clock, a spectrum, a thermometer or a radio receiver. The rotation, the mass, the density, the gases, the temperature, the internal heat, the magnetic field, the layer below the clouds: the answer to “what are these luminaries” was substantially written between 1687 and 1971, and the pipeline the book disputes contributed nothing to it. What sharpened imaging adds is weather.

The other side of this deserves saying. Every line in the table is an inference through physics — Kepler’s law for the mass, laboratory spectra for the gases, black-body radiation for the temperature — and a reader who rejects the physics rejects the inference. But that is a different objection from the chapter’s. The chapter says the pictures are processed and so the planets are unknown. The pictures are processed; the planets were known before the pictures were.

The ladder does not stop at the amateur

The objection also contains a false assumption: that the processed amateur image is the thing our knowledge of these planets rests on. Instruments further up the ladder produce better pictures without the disputed step at all.

instrument	θ at 550 nm	elements across Jupiter	how it beats the atmosphere
60 mm	2.31″	18	it need not — the aperture is coarser than the seeing
100 mm (4 in)	1.38″	29	lucky imaging and stacking
200 mm (8 in)	0.69″	58	lucky imaging and stacking
356 mm (14 in)	0.389″	104	lucky imaging and stacking
508 mm (20 in)	0.272″	148	lucky imaging and stacking
Pic du Midi T1M, 1.05 m	0.132″	307	lucky imaging and stacking
Hubble, 2.4 m	0.058″	701	nothing — it is above the atmosphere
VLT, 8.2 m	0.017″	2,395	adaptive optics — hardware, not software
spacecraft in situ	kilometres per pixel	tens of thousands	no atmosphere and no distance

Across Neptune’s 2.2″ disc the same apertures give 1.0 elements at 60 mm, 1.6 at 100 mm, 3.2 at 200 mm, 5.7 at 356 mm (taking 0.39″) and 38.5 for Hubble. Seeing is 1–4″, so above about 100 mm every single frame is seeing-limited; stacking is the route to the resolution the instrument already has.

The disputed technique runs continuously from a garden telescope to a professional observatory. The 1.05 m telescope at the Pic du Midi images planets by capturing thousands of frames, keeping the sharpest and stacking them — the pipeline the book calls “how to manipulate a photo” — under formal professional–amateur collaboration. Then come two rungs that do not need it. Hubble matters most: above the atmosphere there is no seeing, a single exposure is already at the diffraction limit, and it resolves about seven times finer than a good fourteen-inch stack and twice as fine as the Pic du Midi metre. The VLT corrects the wavefront in hardware, a different mechanism with different failure modes.

Which makes the objection testable. A 356 mm stack resolves to about 0.39″; Hubble resolves 0.058″. So everything present in a good amateur stack must appear in the Hubble image of the same feature, and nothing in the amateur stack should be absent from it. Wavelet artefacts have no reason to coincide with what an instrument in orbit sees natively, still less with what a spacecraft photographed from four thousand kilometres. Show one that does not match, and this page is in trouble.

The professional images are processed too — Hubble frames are calibrated, drizzled (combined onto a finer grid from several offset exposures), composited and stretched — so there is no unprocessed image anywhere on the ladder. What matters is independent pipelines with unrelated failure modes converging on the same structure: the step the book disputes, selecting sharp frames through turbulence and sharpening the result, is absent from the Hubble pipeline and from a flyby, so an artefact of it has no route into either. Stacking is a booster, not a source.

We ran it

A simulation, and labelled as one. The target is a synthetic phantom — a banded field beside a ladder of line pairs at 1.6, 0.8, 0.4 and 0.2 arcseconds — so nothing real can leak in. Each frame is that phantom seen through the aperture’s diffraction and a randomly drawn moment of seeing, then stacked and deconvolved the way an amateur would (method notes). The test: if sharpening manufactured detail, the detail would not care how big the telescope was.

A grid of simulated images. Top: the synthetic ground truth, a banded field with four rungs of line pairs at 1.6, 0.8, 0.4 and 0.2 arcseconds. Below, three rows for 100 mm, 200 mm and 356 mm apertures, each showing a single frame, the best five per cent stacked, and the stack after deconvolution. The single frames are featureless mush at every aperture. The stacked and deconvolved 356 mm panel resolves the coarsest rung clearly, the 200 mm panel weakly, the 100 mm panel not at all, and no panel at any aperture recovers the finest rungs. — Contrast recovered at the 1.6″ rung: **0.005 at 100 mm, 0.105 at 200 mm, 0.214 at 356 mm**, against a true 0.224. The output tracks the telescope. Nothing finer than an aperture’s own diffraction limit is recovered at any frame count or iteration count.

A hundred-millimetre telescope and a fourteen-inch telescope, given the same target, sky, software and operator, differ by a factor of forty in recovered contrast — in the direction and by roughly the amount the aperture predicts. Detail that was being invented by the software would not know how large the mirror was.

The code is at stacking_demo.py and stacking_figure.py, seeded, and runs in about twenty seconds.

Then we ran it on somebody’s real data

A simulation shows what the mechanism can do, not what a working astrophotographer’s capture does, so we asked one. Christopher Go, who images Jupiter from Cebu with a 356 mm Celestron C14 and whose work appears in the professional literature, sent a complete raw capture with permission to use it, and a sentence worth more than the file:

“I really don’t have a single capture because we actually capture videos of planets that we process.”

The unit of planetary imaging is a video, and the picture is what you compute from it; asking for the raw photograph is asking for a thing that was never made. What arrived: 3,215 frames at 107 per second — thirty seconds of Jupiter on 22 July 2022 at 20:41 UT, raw sensor data straight off the camera. Our own pipeline: register every frame against a fixed reference, average — no commercial software, no wavelet sliders, no deconvolution. The file checks out as the object, date and instrument stated (method notes).

Three views of Jupiter from the same thirty-second capture. Left: a single frame, grainy and low in contrast, the belts barely separable from the noise. Centre: the sharpest five per cent of frames stacked, showing crisp equatorial belts, festoons along their edges and a clear oval storm. Right: all 3,215 frames stacked, essentially identical to the centre panel. — **Nothing in the stacked panels is absent from the single frame** — it is buried in noise there, and averaging uncovers it. The sharpest 5% and the whole run are almost indistinguishable, because the seeing that night was steady.

Eyes are poor judges of this, so we measured the power at each spatial scale, single frame against stack.

Log-log plot of power against detail size for a single frame and for the stack of 3,215 frames. The two curves coincide at coarse scales above about two arcseconds, then diverge sharply: the stack falls away steeply while the single frame flattens into a noise floor. A dashed line at 0.389 arcseconds marks the telescope's diffraction limit, beyond which the stack has essentially no power and the single frame still has a great deal. — At coarse scales the curves lie on top of each other — the structure survived. At fine scales they part by **2,140× in power**: 46× in amplitude, against the 56.7 that averaging 3,215 independent frames predicts (frames are not perfectly independent, registration leaves a residual, and the sensor has fixed-pattern noise).

Now look at where the dashed line falls. A 356 mm telescope cannot deliver detail finer than 0.389″. Beyond that line the stack has almost nothing, and the single frame has a great deal — all of it necessarily noise, because the instrument could not have resolved it. Stacking removed precisely the part that could not have been real, and kept the part that could.

That is the answer to “Image stacking + Deconvolution = Reality?”, measured rather than argued, on a working amateur’s own data. One limit: the seeing was unusually steady — sharpest and worst frames differed by only 7% on our metric — so frame selection contributed almost nothing and averaging did the work.

One capture can also be walked down the ladder: impose on each of Go’s frames the diffraction limit and reduced light grasp of a smaller mirror, then stack as before. Only the telescope is hypothetical, and only in the first three columns.

A four by two grid of Jupiter. Columns are 102, 150, 250 and 356 mm apertures. The top row shows one frame at each: the 102 mm frame is noisy and soft with only the main belt visible, improving steadily to the 356 mm frame. The bottom row shows 1,200 frames stacked at each aperture: all are far cleaner, and detail increases from left to right, with the 102 mm stack still visibly softer than the 356 mm stack. — **Down each column** is what stacking does: the noise goes and the structure stays. **Across each row** is what aperture does: no amount of stacking lifts the 102 mm column to the 356 mm column, because 1.36″ is the finest that mirror can deliver. Derived from a single capture rather than photographed by four telescopes. The photographed ladder is below.

Eight photographs of Jupiter in a grid, each disc scaled to the same size. Top row: a 150 mm Maksutov (belts and zones, soft), a 152 mm Meade with the Great Red Spot visible, a 250 mm single unstacked infrared frame that is the softest of the eight, and a 356 mm stack with belts and a faint oval. Bottom row: a 408 mm infrared stack with fine belt structure, Hubble in colour with the Red Spot and white ovals, Gemini North in thermal infrared with the belts glowing and the zones dark, and a JunoCam close-up of the Red Spot and the turbulent region beside it. — **Eight instruments, eight pipelines, one planet.** Each image is shown as its author released it — the amateur frames carry whatever stacking and sharpening their photographers applied, the professional frames their own pipelines, and ours (356 mm) is a plain stack. The only thing done here is to find each disc and resample it to a common diameter: no sharpening, no wavelets, no contrast, nothing added to any of them. Two things to read off it. First, the ladder is not monotonic in aperture, and that is the point: the 250 mm frame is a single exposure through poor seeing and is the softest of the eight, below two stacked 150 mm images — aperture sets the ceiling, and stacking through the atmosphere is how a telescope reaches it. Second, the rungs that do not use the disputed step — Hubble above the atmosphere, Juno at the planet — show the same belt structure, and the Red Spot in the place the ephemeris gives it for each date, as the rungs that do; and Gemini, an 8.1 m professional instrument, reaches its own resolution by the very method the book calls manipulation, keeping the sharpest 10% of frames and stacking them. Images used with the photographers’ permission and credited on the figure; Hubble and Juno frames CC BY. Assembled by build_jupiter_ladder.py, which finds each disc, resamples it and adds the labels; the source files are the photographers’ and are not redistributed.

Even so: is the stacking legitimate?

Mostly yes, for checkable reasons.

Stacking is averaging, and averaging cannot add information. Signal repeats from frame to frame; noise does not, so signal-to-noise improves as the square root of the frame count — ten thousand frames buys a factor of a hundred.
Lucky imaging is selection, not synthesis. Frames are ranked by a sharpness score — gradient energy or local contrast — not by resemblance to any template, so keeping the sharp ones cannot manufacture agreement; the retained frames agree with one another before anything is combined, and the noise-phantom test above is the direct check.
Deconvolution is auditable, and rarely audited. The telescope’s point-spread function — the blur it makes of a single star — is measurable. Deconvolving with the measured PSF undoes a known distortion; dragging a wavelet slider until the picture pleases does not, and this is where the book’s suspicion is well placed.
Invention cannot hit a target set in advance. The Great Red Spot appears at a longitude computable years ahead; moon shadows arrive at a computed minute; the Cassini division falls where it must; and a planetary ephemeris fitted in 2008 places the seven planets of February 2025 to a tenth of an arcsecond of one fitted in 2021 (The Eight-Planet Photograph).
Independent operators converge. Different observers, telescopes and software on the same night produce the same features in the same places.

Or skip the software and look

The chapter opens, at p. 154, by recommending a video to anyone who has never seen the planets first hand. The better recommendation is an eyepiece. A 60 mm telescope, the first row of the ladder table, resolves 18 elements across Jupiter’s 40″ disc with no camera and no software; the two dark equatorial belts are each several of those elements wide and are visible live, with the four Galilean moons and the rings of Saturn. “Jupiter 4.0” is a sharper version of a thing anyone with a small refractor can see from a back garden.

Where detail does matter

It mattered enormously once. The phases of Venus are a crude shape observation, and they killed the Ptolemaic system. The claim here is only that the dynamical results do not depend on detail.
Some of our own arguments use resolved measurement. The Jupiter page measures shadow sizes on the disc; Is the Moon a Ball? is an appearance argument throughout; and Mirrored Reflections had to correct its source photograph for post-production stretching. A processed image is not worthless; it is an image whose processing has to be accounted for.
“Does it look like a ball of gas?” is not answerable by astrometry. That question at p. 161 is about appearance, and this page does not address it.

What would change our mind

A stack of noise that produces a planet, or a known-target stack that disagrees with the target — a resolution chart at distance, through the same pipeline, showing detail the chart does not have.
Raw frames that separate by aperture. Every single frame above roughly 100 mm should be seeing-limited, so a raw frame from a 356 mm and one from the Pic du Midi metre should look much the same. If they differ the way their apertures suggest, seeing is not the limiting term we say it is.
An amateur stack showing a feature the orbital image does not — same target, same rotation phase, against a Hubble frame. The cheapest decisive test on the page.
Astrometric positions that fail the field-star check. If planetary positions measured against catalogue stars in the same frame do not reproduce, this page’s reframing goes with them.

Where this page could be wrong

“Averaging cannot add information” is true of averaging and not of every step. Deconvolution genuinely does add assumptions, and a badly chosen PSF genuinely does create features. The argument is that the astrometry does not route through that step, not that the step is harmless.
One capture, one night, one instrument. The Go data is a single thirty-second run in good seeing, not a survey.
The convergence claim is shown as a ladder across dates, not on one date: a same-date amateur–Hubble overlay, feature by feature, is still outstanding. The historical table is a summary, and the tutorial quoted at pp. 155–157 is cited from the book’s pages rather than given a stable reference of its own.

Method notes

Simulation (stacking_demo.py): per frame, aperture diffraction at 550 nm, a seeing draw around a 2.5″ median, tip-tilt shift, photon noise; four hundred frames per aperture; the sharpest five per cent registered, averaged, then deconvolved. Recovered contrast at the 1.6″ rung is read against the phantom’s true 0.224.

Go capture (go_capture_stack.py): 3,215 frames, 720×620, eight-bit, raw RGGB sensor data; demosaic, register by phase correlation against a fixed reference, average. Provenance check: the planet measures 501 pixels across and the ephemeris gives Jupiter an apparent diameter of 43.82″ on 22 July 2022, so the image scale is 0.0875″ per pixel, consistent with the object, date and instrument stated. Power spectra are from an identical patch of the disc in a single frame and in the full stack; the diffraction line is at 0.389″ for 356 mm at 550 nm.

Ladder-down figure. Each frame is given the extra blur of a 102, 150 or 250 mm aperture (the quadrature difference between that aperture’s diffraction limit and the C14’s) and its noise is raised so that per-frame signal-to-noise falls with collecting area, the native noise being measured from the difference of two registered frames; 1,200 frames per column are then registered and averaged. The 102 mm limit in the caption is 1.36″.

Sources & further reading

Levi Miller, Globe Deconstruction? (prerelease) — pp. 155–157 (the imaging-process argument), p. 161 (“deconvoluted and sharpened images”), and pp. 158–163 for the two arguments that depend on it.
Neptune’s discovery within about one degree of Le Verrier’s predicted position, 23 September 1846: Galle, J. G., “Account of the discovery of Le Verrier’s planet Neptune, at Berlin, Sept. 23, 1846,” Monthly Notices of the RAS 7, 153 (1846); Le Verrier, U., Comptes Rendus 23, 428 (31 August 1846). First ground-based cloud features: Smith, B. A., Reitsema, H. J. and Larson, S. M., Bull. AAS 11, 570 (1979).
The 1.05 m telescope at the Pic du Midi and its planetary programme, including the professional–amateur observing collaboration: t1m.omp.eu.
Raw capture 2022-07-22-2041_2-CG-L-18.avi — 3,215 frames, Celestron C14, courtesy of Christopher Go, used with his permission. Our processing code accompanies this page in the repository.
Diffraction limits computed as 1.22λ/D at 550 nm. ESA/Hubble imagery is CC BY 4.0 and NASA imagery public domain, so the matched comparison can be published.
Mercury’s perihelion advance, planetary versus relativistic contributions: Fitzpatrick, Celestial Mechanics.
The 405,000-year Jupiter–Venus eccentricity cycle, its g₂ − g₅ origin, the Newark, Hartford, Chinle, Japanese chert and Bristol Channel records, and the geological-orrery programme: Olsen et al., Mapping Solar System chaos with the Geological Orrery, PNAS; and Kent et al., Empirical evidence for stability of the 405-kiloyear Jupiter–Venus eccentricity cycle, PNAS 2018.
Planetary masses from pulsar timing: Champion et al., Measuring the Mass of Solar System Planets Using Pulsar Timing, ApJL 720 L201 (2010).
Apparent disc diameters computed from JPL DE440 via Skyfield for 2025-02-22, using IAU equatorial radii.
Cross-references on this site: Jupiter’s Shadows and The Eight-Planet Photograph (the two arguments this one sits underneath), Mirrored Reflections (where processing in our own source image had to be accounted for).
Photographed ladder: Silvia Kowollik (150 mm, 25 Oct 2013), Javier Peralta (152 mm, 24 Jan 2015), Gregory T. Shanos (250 mm, 24 Jun 2026) and Trevor Barry (408 mm, 17 Jul 2016), each via the PVOL planetary observation database and used with permission; NASA/ESA, A. Simon (GSFC), Hubble WFC3 image heic1410a, 21 Apr 2014, CC BY 4.0; International Gemini Observatory/NOIRLab/NSF/AURA, M. H. Wong et al., Gemini North NIRI lucky-imaging mosaic noirlab2011a, 2020; NASA/JPL-Caltech/SwRI/MSSS, JunoCam perijove 21 (JNCE_2019202_21C00036), processing by Thomopoulos via the JunoCam gallery, CC BY.
Newton, I., Principia (1687), Book III, Proposition 8, weighing Jupiter from Callisto’s orbit; Wildt, R., “Absorptionsspektren und Atmosphären der grossen Planeten,” Veröff. Univ.-Sternwarte Göttingen 22 (1932); Burke, B. F. and Franklin, K. L., “Observations of a variable radio source associated with the planet Jupiter,” J. Geophys. Res. 60, 213 (1955); Kiess, C. C., Corliss, C. H. and Kiess, H. K., “High-dispersion spectra of Jupiter,” Astrophys. J. 132, 221 (1960); Aumann, H. H., Gillespie, C. M. and Low, F. J., “The internal powers and effective temperatures of Jupiter and Saturn,” Astrophys. J. 157, L69 (1969); Hubbard, W. B. et al., “The occultation of Beta Scorpii by Jupiter and Io,” Astron. J. 77, 41 (1972); Peale, S. J., Cassen, P. and Reynolds, R. T., “Melting of Io by tidal dissipation,” Science 203, 892 (1979); de Pater, I. et al., “Peering through Jupiter’s clouds with radio spectral imaging,” Science 352, 1198 (2016).