S30, S30 Pro, S50: optics and models for science

A thin line beside a passage says how far its claim is proven: measured, inferred. Colored numbers carry the same mark. A passage with no line is stated. How claims are marked.

The models on paper

These are ZWO's own numbers for the main (telephoto) camera of each model, from its product pages, checked on 20 October 2026. I've added the S50 Pro because one of the units in my bright-star measurements is an S50 Pro.

ZWO gives the S30, S30 Pro and S50 Pro fields as one number, which matches the frame's diagonal; the S50's as width × height (inferred).

Two numbers matter more for science than the ones in the table, and you can work them out from it. The pixel scale is how much sky one pixel covers, in arcseconds (1/3,600 of a degree): 206,265 × pixel size ÷ focal length. ZWO lists a 2.9 µm pixel for the S50 Pro; for the other three it isn't on the product page, and 2.9 µm is Sony's published pixel size for those sensors (its product flyers, linked under the sources below). That gives:

Worked out from ZWO's focal lengths and frame sizes and a 2.9 µm pixel. A 50 mm lens has (50 ÷ 30)² ≈ 2.8 times the collecting area of a 30 mm one, so it gathers that much more light from each star in the same time.

The paper numbers are close but not exact. On my S30 Pro, plate solving (working out exactly where a frame points from the stars in it; a Note explains it) measures 3.672 to 3.674″ per pixel (measured). That comes from the S30 Pro optics Experiment, which says it agrees with the pixel size and focal length in the frames' own headers. It's about 2 % finer than the 3.74″ that 2.9 µm at ZWO's 160 mm works out to (3.7″ in the table, rounded), so I use the solved scale, not the paper one (inferred).

Where green clips, model by model

Every pixel records light as a code from 0 to 65,535 in the 16-bit raw files a Seestar saves for each sub (one single exposure); these are FITS files, the standard astronomy image format. In my Experiment on bright-star colors (9 October 2026, Exploratory), I looked at raw frames from 8 Seestars of four models: 3 S30s, 1 S30 Pro (mine), 3 S50s and 1 S50 Pro, the seven others shared by community contributors. On six of them, red and blue filled at 65,535, the largest number the file can hold, while green stopped lower. Where green stopped, its ceiling, depended on the model (measured):

The ceilings are ranges over the six units with what the Experiment calls normal color handling. The Units column counts every unit of that model I looked at, including the two odd ones below (measured).

The conditions that go with the table (measured):

• Two units didn't fit the pattern. One S50 applies no color scaling, so every color stops in a narrow band ending at 65,520. One S30, on one firmware version, scaled red barely more than green, so its red stopped in a band too, and its green ceiling was 56,400, the highest I saw.
• On one S30, green's ceiling rose about 850 to 1,050 codes after a firmware update; on two S50s it moved by under 200 between nights. A ceiling belongs to a unit on a firmware, not to a model for life.
• The frames were 17 sets on 16 nights between August 2025 and October 2026, mostly through the IR-cut filter (the Seestar's standard filter, which blocks infrared), and every night was also used to work out the method, so nothing has been checked on fresh data yet. With one or three units per model, these are ranges for the units I had, not model constants.

What it means for photometry (measuring star brightness): keep the star peaks you measure under about 40,000, which is 0.80 of the lowest green ceiling, the S50 Pro's (measured). On three of the four models the response was straight up to there; on the S50 Pro, until its low reading (below) is understood, check your result with fainter comparison stars too, peaking under about 20,000 (measured). Below the ceiling, on the S30, S30 Pro and S50, green stayed within 5 % of the brightness Gaia's catalog predicts up to 0.9 of its ceiling; red and blue scattered more, 0.81 to 1.12 in single bands, mostly a little low, with fewer stars per band. The last band before the ceiling, 0.9 to 1, is less clear (measured). The one S50 Pro read low from 0.4 of its ceiling up, green and red by about 5 to 8 % and blue by 10 to 23 %, on one night, and I don't know why (measured). Why green clips first, and what it does to star colors, is in Why bright stars turn green or purple.

Two S50s don't collect exactly the same light

In the Outpost Experiment (8 October 2026, Exploratory), my own S50 at a remote dark-sky observatory (the telescope that page calls the Outpost) collected 0.85 to 0.93 as much starlight as a community S50 on three sets (measured). The qualifiers travel with it:

• The comparison is green pixels only, through the light-pollution filter on both units, measured against catalog stars, and it includes the air above each site (measured).
• The first 13 minutes of the first night read 0.64, while the measurement was wandering for a reason not yet pinned down; an earlier, quicker check on those minutes alone had put it at about 0.54. The rest of both nights read 0.85 to 0.93 (measured).
• It's two units of one model. It says the Outpost isn't a weak unit, not how S50s compare in general (measured).

The lesson I take from it: don't carry a brightness scale from one Seestar to another, or from one night to the next. Tie each night's frames to catalog stars in the same field, which is what comparison stars do in photometry anyway.

What a stack's corners tell you

The corners are where a lens shows its flaws. My S30 Pro optics Experiment (5 October 2026, Exploratory) measured my own S30 Pro over seven sets and frames from three more S30 Pros whose owners shared them, matching thousands of stars against the Gaia star catalog. Most of what follows is a fraction of a pixel, far too small to see in a picture, and it matters only if you measure positions or brightnesses (measured).

Distortion: smooth, but it changes every night

Distortion is a lens putting stars slightly off where a perfect lens would. My S30 Pro's lens stretches the field a little toward the corners: over seven nights, a corner star lands 1.8 to 2.4 pixels farther out than a perfect lens at the best-fitting scale would put it, about 0.1 % of its distance from the center (measured). Most of that looks like a slightly different pixel scale. The curved part left over was 0.59 to 0.82 pixel at the corner, and it changed from night to night more than the measurement errors allow (measured). This is my unit only: from the other units' frames I couldn't compare their lenses properly.

Coma: each camera leans its own way

Coma makes star images lopsided, with a faint tail on one side, so a star's light-weighted center sits a little off its peak. On each of the four S30 Pros, the direction of the lean stayed put across targets, filters and nights; on mine, within 23 degrees over seven sessions. On the other three units it pointed 60 to 140 degrees away from mine, and on one of them the lean was almost zero: as far as this measure goes, its optics were the best centered of the four (measured). Two of the other units differ from each other only barely, so the claim that each camera has its own direction rests on three units, and it hasn't been checked on a fifth (measured). The size of the lean depends on focus: on my unit, its growth from center to corner went from 0.06 pixel on a sharp night to 0.32 pixel on my softest night (measured). The Experiment puts it in one line:

“So the direction belongs to the camera, and the size belongs to the night.” (measured)

The S50 has a one-sided flaw of its own. In the Outpost Experiment, the optics stretched stars slightly on the sensor's right side, about 1.4 to 1.55 times longer than wide there, and with the field fixed on the sensor that stays in the stack (measured). That's one unit, and I haven't mapped an S50's lean the way I did the S30 Pros', so I can't say whether it's the same kind of effect.

Errors that don't average away

Each star also carried a small position error that stayed the same in every frame, so averaging the night didn't remove it. For moderately bright stars it ranged from 30 to 260 thousandths of an arcsecond (0.01 to 0.07 pixel) over 12 nights on the four units, largest on soft-focus nights; on the alt-azimuth nights where it was large, it grew toward the edges (measured). What causes it is still open: in my data the mount, the unit and the focus can't be told apart.

Vignetting: why corners matter for photometry

Vignetting is the optics delivering less light to the corners than to the center. For a picture it's a gradient you can remove. For photometry it's a bias: the same star reads fainter near a corner unless the frame is divided by a flat (a frame of an evenly lit surface that records the falloff), and a comparison star far from your target sits under a different falloff. On the Seestars I've profiled, corners read 0.93 to 1.05 of the center (measured), a figure from How calibration works, which also covers flats. So I don't assume a sub arrives already flat, and I keep the target and its comparison stars close together (inferred).

What I do about the corners

• Keep what you measure in the inner field. Distortion, coma and the fixed per-star errors all grew toward the edges on my S30 Pro nights, so put the target and its comparison stars near the middle (measured). On my unit the best-behaved spot isn't the exact center of the chip, and on yours it will be somewhere else (measured).
• Focus carefully, and refocus as it cools. Soft focus made the lean and the fixed errors worse on my nights (measured).
• Dither (shift the pointing a little between subs) in both directions, so an error tied to one spot on the sensor averages out.
• For positions, solve every frame and fit the lens per night, with a solver that fits distortion, rather than reusing one night's fit on another.

What the differences mean for science

In my own words, how each difference plays out, without a verdict on any model:

• Aperture sets how faint you can go in a given time. The 50 mm models collect about 2.8 times the light per star, which helps faint variable stars and shallow transits.
• Field sets how many comparison stars land in the frame with your target and how big a target fits. By the numbers above, the S30 Pro sees the most sky, then the S50 Pro and the S30, and the S50 the least.
• Pixel scale sets how finely each star is sampled. Finer pixels (the 50 mm models) separate close stars better in a crowded field and help positions; coarser pixels pack more of a star's light into fewer pixels.
• The clip level sets the brightest star you can measure in a given sub length, and it differs by unit and firmware as well as model. Measure your own, or stay under about 40,000 (measured).
• The unit and the night matter as much as the model: throughput, coma direction and the lens's leftover curve all moved between units or nights in my data (measured). That's why careful measurements calibrate every night against catalog stars, whatever the telescope.

Where these numbers come from

• ZWO's Seestar product pages, checked 20 October 2026: the spec table, and the S50 Pro's 2.9 µm pixel. The pages are the Seestar S30, S30 Pro, S50 and S50 Pro. ZWO's S30 Pro FAQ describes the S30 Pro's cameras but gives none of the table's numbers. The 2.9 µm pixel for the S30, S30 Pro and S50 is the unit cell size in Sony's product flyers for the IMX662, IMX585 and IMX462, checked 20 October 2026. These are third-party numbers, not mine.
• Why a Seestar's bright stars turn green or purple (Experiment, 9 October 2026, Exploratory): the green ceilings, the odd units, the firmware shift, the straight response below the ceiling, the 40,000 limit and the S50 Pro's low reading.
• First nights from a remote Seestar S50 at a dark-sky observatory (Experiment, 8 October 2026, Exploratory): the 0.85 to 0.93 throughput and the S50's one-sided stretch.
• What my Seestar S30 Pro's optics do to star positions (Experiment, 5 October 2026, Exploratory): the measured pixel scale, distortion, coma, the fixed per-star errors and the advice on corners.
• How calibration works: the 0.93 to 1.05 corners.

The lines beside passages mark how far each claim is proven; see how claims are marked. More Notes and guides are on the Learn page.