The Heavy Lift Part 2A: The Speculative

We do not know what cameras fly on a Falcon or Starship booster. Not the model, not the sensor, not the housing, not the mount. That information is not public, and I will not pretend otherwise.

This is the first of three articles. This one is the speculative exercise, labelled as exactly that: two commercial cameras, held against the launch environment, by engineering reasoning from published specifications and standard physics. No camera was flown, tested or qualified here. The second article sets out what NASA actually flew, from the documented record. The third works through the remaining engineering: gimbals, stabilisation, radiation, and glass.

There is a second thing this series is, and it is really the first thing. It is about mathematics. Cameras are the occasion; the mathematics is the subject. Every question worth asking here (how wide, how fast, how cold, how bright, how steady, how strong) resolves into an equation, and most of those equations are older than the hardware they govern by a century or more. Where there is a formula, you will get the formula, set properly, with every symbol defined.

---

PART I. THE GUESS, MADE BEFORE THE RESEARCH

The sequence of events matters, so it is stated up front.

Before opening a single document, I was asked a simple question: if you had to bolt a camera to a rocket booster and bring back usable footage, what would you reach for?

The answer came from the film industry, not from aerospace. Two cameras:

The Panasonic VariCam Pure, the RAW-recording variant of the VariCam 35 cinema system. A known quantity on real sets, 4K, up to 120 frames per second, fourteen stops of latitude, and a dual native ISO that makes it genuinely useful in the dark.

The Nikon Z 9. A 45.7-megapixel full-frame stills camera that also happens to shoot 8K, has no mechanical shutter at all, and weighs 1,340 grams.

That guess was made cold; the research began only afterwards. The second article of this series holds the guess against the documented record.

A small correction on the way in, since precision is the currency here: the camera is a VariCam, not a VeriCam, and the specification sheet in question describes the VariCam Pure. The Pure shares its AU-V35C1G camera module with the VariCam 35 but replaces the recording section with the Codex AU-VCXRAW2, which writes uncompressed V-RAW and nothing else. No AVC-ULTRA, no ProRes, no proxy. That single design decision drives almost everything that follows.

---

PART II. THE TWO CANDIDATES, AND WHETHER THEY WOULD SURVIVE

2.1 The bodies, in numbers

Two figures in that table decide most of what follows.

The first is mass. The VariCam Pure is 3.8 times heavier than the Z 9 before you add a lens, a battery, a mount, or the Codex drive. On a vehicle where every kilogram is paid for in propellant, that is not a preference; it is a disqualification argument that has to be defeated by something.

The second is that Panasonic does not publish an operating temperature range on that specification page, while Nikon does, and Nikon's range is narrow: -10 C to +40 C. Not survival. Operating.

2.2 The lens, which is where the geometry starts

Before anything else, the single most fundamental relation in photography. Everything about framing is this:

Sensor size sits in the numerator, focal length in the denominator. It follows immediately that a larger sensor sees wider through the same lens, which is the entire practical argument for full frame over Super 35 in a fixed-mount application where you cannot go get a shorter lens.

Run both candidates at 14 mm:

Roughly eighty-seven degrees against one hundred and four. On a booster, where the camera is bolted down and the subject is a vehicle you are attached to, seventeen degrees of extra field is the difference between seeing the grid fin and seeing the grid fin plus the horizon it is steering against.

Candidate optics. For a wide, fast, high-frame-rate job on a moving vehicle, the requirements are: rectilinear (not fisheye, unless you intend to correct it; distortion correction is treated in the third article of this series), fast enough to keep exposure time short, mechanically simple, and ideally without a focus motor to fail.

For the Z mount: the NIKKOR Z 14-24mm f/2.8 S is the obvious first pick; constant f/2.8 across the range, and 14 mm at the wide end. The NIKKOR Z 20mm f/1.8 S is the alternative if you will trade eighteen degrees of field for a full stop and a half of light and a fixed, simpler barrel. For a locked-off mount I would take the prime, every time, because a zoom is a mechanism and a mechanism is a failure mode.

For the EF or PL mount on the VariCam: the Canon CN-E 14mm T3.1 L F and the Zeiss Compact Prime CP.3 15mm T2.9 are the natural cinema equivalents, both built to be locked down and left alone.

Note the honest limit on all four: none of these is qualified for spaceflight, and none is claimed to be. They are excellent terrestrial glass.

2.3 Shutter, and the equation that decides everything

A cinema camera describes exposure as an angle, a habit inherited from the rotating disc shutter of mechanical film cameras. The disc is gone. The arithmetic remains:

Now the important one. This is the equation that decides whether a camera bolted to a rotating vehicle returns a picture or a smear:

Angular rate times exposure time gives you angle swept. Multiply by focal length and you have linear travel of the image across the focal plane. Divide by pixel pitch and you have the answer in pixels, which is the only unit that matters.

Work it for a booster rolling at a modest thirty degrees per second:

At a 45-degree shutter and 120 fps, about two pixels of smear: acceptable, arguably invisible at viewing distance. At a 360-degree shutter, sixteen pixels: destroyed. Same camera, same lens, same rocket. The only variable that changed was a number in a menu.

This is why shutter angle on a vehicle is not a cinematographic taste decision. It is a survivability parameter for the data.

2.4 The sensor: how it actually collects light and records data

This is worth doing properly; it is the least understood part of the instrument.

A sensor does not measure brightness. It counts events.

Light arrives as quanta. Each photon carries energy:

When a photon of sufficient energy strikes the silicon lattice, it can promote an electron across the band gap, leaving a mobile electron and a hole. That is the photoelectric effect, and it is the same physics Einstein was given the Nobel Prize for in 1921; every digital camera ever built is an application of that paper.

The liberated electron is held in a potential well beneath each photosite by an applied electric field. It stays there, accumulating with its fellows, for the duration of the exposure. A photosite is, quite literally, a bucket for electrons.

The fraction of arriving photons that succeed in producing a counted electron is quantum efficiency. A good modern back-illuminated sensor manages somewhere in the region of sixty to ninety percent across the visible band, an exceptionally high figure by the standards of any other detector class.

Reading the bucket. At the end of the exposure the accumulated charge must be converted to a voltage, amplified, and digitised. In the VariCam's MOS sensor and the Z 9's stacked CMOS sensor alike, each pixel has its own amplifier (this is what "active pixel sensor" means), and the conversion happens column by column in parallel, which is why modern sensors can be read out fast enough to sustain 120 frames per second.

The Z 9's sensor is stacked: the photodiode layer and the readout logic are fabricated on separate wafers bonded together, which shortens every signal path and makes the readout fast enough that Nikon deleted the mechanical shutter entirely. That fact matters for this application; Section 2.7 returns to it.

Colour. Neither sensor sees colour. Both have a colour filter array bonded over the photosites, in the Bayer pattern devised at Kodak in 1976: one red, one blue, two green per group of four, because human vision is most acute in the green. Every colour image either camera produces is an interpolation, a demosaic, reconstructing three colour values per pixel from one measured value per pixel. RAW recording preserves the un-demosaiced measurement; that is precisely what makes it RAW.

Noise, which is not a defect. Three sources, and only one of them is the camera's fault.

Photon arrival is a Poisson process. If a well collects ten thousand electrons, the standard deviation of that count is one hundred, and there is no engineering anywhere that removes it, because it is a property of light and not of the instrument. In good light this term dominates, which means a bright, well-exposed frame is limited by physics rather than by manufacturing.

Dark current is thermal: electrons promoted across the band gap by heat rather than by light. Read noise is electronic, introduced in the amplification and conversion chain.

Dynamic range is simply the ratio between the largest and smallest signals a photosite can distinguish:

The VariCam's advertised "14+ stops" is exactly this number. Fourteen stops is a ratio of about 16,400 to 1, or 84 decibels. It is a claim about the noise floor, not about the highlight.

Dual native ISO deserves a moment because it is widely misunderstood as a marketing term:

It is not one sensor amplified twice. It is one sensor with two physically distinct readout paths of differing conversion gain. The ISO 800 path keeps the full well and maximum highlight headroom. The ISO 5000 path applies higher conversion gain before the noise-adding stages of the chain, so read noise is suppressed relative to signal. Both are optically honest base sensitivities. For a night launch, this is a real and substantial advantage, and it is the strongest single argument in the VariCam's favour.

Sampling. Pixel pitch sets the finest detail the sensor can represent:

The Z 9 samples at 3.76 micrometres, giving a Nyquist limit near 133 cycles per millimetre. The VariCam Pure, with 8.9 megapixels spread across a Super 35 frame, samples at roughly 6.4 micrometres and reaches about 78 cycles per millimetre. Coarser, yes; but each of those larger wells holds far more electrons, and by the dynamic range equation above, that is exactly where the VariCam's fourteen stops come from. Resolution and latitude are traded against one another on the same piece of silicon. There is no configuration that wins both.

And there is a ceiling neither can pass, set by the aperture rather than the sensor:

Stop down for depth of field and the Airy disc grows until it is larger than the pixel. At f/11 the disc is roughly 14.8 micrometres, which is about four Z 9 pixels across. Physics takes back what the aperture gave.

2.5 Recording the data, which is a bandwidth problem

Four thousand ninety-six by two thousand one hundred sixty, twelve bits per pixel, one hundred twenty times a second. About 1.59 gigabytes per second; roughly ninety-five gigabytes per minute of running time.

That number is the entire reason the VariCam Pure records to a Codex Capture Drive rather than a memory card, and it is the reason the recording module is a separate box. It also means that a two-minute ascent sequence is roughly 190 gigabytes of data that has to survive whatever happens next, which makes the storage medium a structural component and not an accessory.

The Z 9 sidesteps this by offering compressed options, including N-RAW and ProRes 422 HQ, and by writing to solid-state CFexpress cards with no moving parts and a far smaller physical footprint.

2.6 Temperature, which is where the guess starts to fail

Dark current roughly doubles for every five to eight degrees Celsius. Take a sensor forty degrees above its reference temperature and the dark signal is up by a factor of about fifty. That is not a subtle degradation; that is a black frame turning grey and a shadow turning into noise.

Nikon's published operating range for the Z 9 is -10 C to +40 C, with humidity under 85 percent and no condensation. That is a consumer and professional terrestrial envelope. It is not a claim about vacuum, it is not a claim about vibration, and it is certainly not a claim about a booster.

Now, what does a camera on the outside of a launch vehicle actually face?

In vacuum, there is no convection. None. A camera in orbit cannot be cooled by moving air, because there is no air to move, and the internal fans of a cinema camera become decorative. Heat leaves only by conduction into whatever the camera is bolted to, and by radiation from its outer surfaces. The equilibrium temperature is set by a balance of absorbed sunlight against emitted infrared:

Rearranged for a flat plate facing the sun, the whole problem collapses to a single ratio:

White paint, with a solar absorptance to infrared emittance ratio around 0.25, sits near minus twenty Celsius. Bare aluminium, at a ratio near four, runs to well over two hundred. This is not a small effect and it is not adjustable after launch. It is why flight hardware is white, and why NASA wrapped its EVA cameras in white thermal covers rather than fitting them with cooling.

On the way up, the problem inverts. Two mechanisms:

Dynamic pressure peaks at max Q; it is the structural load case for anything mounted externally, and it is a pressure acting on every exposed surface of a camera housing. Meanwhile, air brought to rest against a fast-moving surface heats by compression alone, whether or not the surrounding air is hot. At Mach 6 in cold upper atmosphere the stagnation temperature is on the order of 1,800 kelvin. Real recovery temperatures on a swept surface are lower, and a booster spends little time at that condition, but the order of magnitude tells you why external cameras live behind windows or inside fairings and never in the open air.

Then there is the vibration.

Ascent is not a shake. It is broadband random vibration across roughly 20 to 2,000 hertz, quantified as the square root of the integral of the acceleration power spectral density. Eight or nine g RMS is an unremarkable figure for a launch environment, and the important word is random: every resonance in the structure will be excited, because the excitation contains every frequency.

Every camera contains resonances. Lens elements in their barrels. Circuit boards on their standoffs. A rotating mirror. A memory card in a friction-fit slot. And on the Z 9, the sensor itself, deliberately mounted on a compliant five-axis stabilisation stage, which is to say deliberately mounted on springs.

2.7 Verdict: would either of them survive?

Straight answers, in order.

The VariCam Pure: no, not without a housing that would weigh more than the camera. Five point one five kilograms of body, a separate recording module, a separate viewfinder, an EF mount that is a friction-and-spring interface never intended for 9 g of random vibration, and a data rate that demands a physically large storage device. Its optical merits are real and its low-light performance is genuinely superior. None of that survives the mass and mounting argument.

The Nikon Z 9: closer than you would expect, and interesting for one specific reason. It is light, it is sealed reasonably well, it stores to solid-state cards with no moving parts, and critically, it has no mechanical shutter. There are no shutter blades to shear off their pivots under launch load, because there are none. For a device that must endure high-g broadband vibration, deleting the single fastest-moving mechanical assembly is an enormous structural advantage that Nikon achieved for entirely unrelated reasons.

Against it: the published -10 to +40 C envelope, a lithium-ion battery chemistry with its own thermal and vacuum problems, and that five-axis sensor-shift stage, which in a launch environment is not a stabiliser but a mass on a spring inside your instrument. It would have to be locked or removed.

Neither would fly as bought. Both would need what every flown camera has ever needed: an enclosure, thermal control, conducted heat paths, a mount designed to the vibration spectrum, and requalification of every lubricant and elastomer inside. That is not a modification. That is a different product.

---

Where this goes next. The verdict above is reasoning, not test data, and the record is the check on it. The second article in this series sets out the cameras NASA actually flew, from the Hasselblad magazines of the early Shuttle programme to the digital bodies of the 2000s. The third works the engineering that any flown camera answers to: gimbals and their mathematics, stabilisation, radiation damage, and the best documented precedent in the subject, the Space Shuttle windshield.

---

DISCLAIMERS

1. Nothing here describes any current launch vehicle's camera systems. No claim is made, implied or intended about the cameras aboard any SpaceX vehicle. That information is not public.
2. Manufacturer specifications for the Panasonic VariCam Pure and the Nikon Z 9 are quoted from the manufacturers' own published specification pages, linked below. Neither product is offered by its manufacturer as spaceflight hardware, and no such claim is made here.
3. The survivability assessment in Part II is engineering reasoning, not a test result. It is my analysis, from published specifications and standard physics. No camera was flown, tested or qualified in the preparation of this article.
4. This article is not engineering advice and must not be used as a basis for design.

---

---

FAIR USE NOTICE: Under Section 107 of the Copyright Act of 1976, allowance is made for fair use for purposes such as criticism, comment, news reporting, teaching, scholarship and research.

Source Citations

Ordered by authority: government agencies first, then manufacturers, then academic and reference literature. This series cites per article; a source used in more than one article appears in each.

I. Government agencies and public research

• National Air and Space Museum, Smithsonian Institution. Collection records for NASA flight camera hardware, including EVA camera thermal covers. airandspace.si.edu

II. Manufacturers

• Panasonic. VariCam Pure: Specifications. pro-av.panasonic.net/en/cinema_camera_varicam_eva/products/v
• Nikon USA. Nikon Z 9: Technical Specifications. www.nikonusa.com/p/z-9/1669/overview

III. Standards, academic and reference literature

• Bayer, B. E. Color Imaging Array. US Patent 3,971,065, filed 5 March 1975, issued 20 July 1976. (The colour filter array, Section 2.4.)

Figures set by the author. Formula plates are original work.