A by-product, strictly speaking

This image of NGC 6960 was never planned as a showcase project — it is the by-product of several test-runs between September 1 and 8, 2026. Over the past weeks I have been building an advanced sequence for N.I.N.A. to run my observatory through fully automated imaging nights, and the Veil Nebula happened to be the target I pointed at while putting that sequence through its paces. The development of the sequence itself — the logic, the pitfalls, and what finally worked — is a story of its own and will get its own post here. That test-run origin also explains a compositional flaw I readily admit: the framing was chosen quickly and is clearly not optimal, with the main arc sitting off to one side of the field rather than where a considered composition would have placed it. But the data that came down during those automated nights turned out to be far too good to leave sitting on the disk — so here we are.

The equipment used was the standard observatory setup: f/4 Newtonian at 1.019mm focal length, ASI 2600 MM Pro with the Optolong Filterset, EAF, EFW and CAA (the rotator not really touched…).

The Object, the introductory astrophysical background and hints where to find it

The Western Veil Nebula, NGC 6960, is the brightest western arc of the Cygnus Loop — the expanding remnant of a supernova that detonated roughly twenty thousand years ago, some 2,400 light-years away. What we see today is not the explosion itself but its aftermath: a vast shock wave, still sweeping through the surrounding interstellar gas, rendered visible as a network of glowing filaments. The two colors in this image are not decoration — they trace two distinct physical zones of that shock. The teal (=green/blue) [O-III] emission marks the razor-thin front where the gas has just been heated to extreme temperatures; the red Hα glow spreads out behind it, where the gas has already cooled and the hydrogen is recombining. In other words: the image is a snapshot of a cooling sequence, laid out in space. How exactly that works — and why the nebula looks like threads rather than a wall of light — is explained in detail below the final image at the end of this post.

One thing none of the images from that area shows is the corpse of the star itself. A core-collapse supernova — and the Cygnus Loop is thought to be one — should leave behind a neutron star. Yet despite decades of X-ray searches, no compact remnant has ever been firmly identified. A handful of candidates exist in the remnant’s southern reaches, including a point-like X-ray source that might equally well be a background galaxy, and none of them pulses. Perhaps the neutron star received a strong kick at birth and has long since drifted away from the geometric center; perhaps it is simply too faint, its lighthouse beam pointed elsewhere. Twenty thousand years after the explosion, the shroud is still magnificent — but the body is missing.

Finding the field is easy, by the way: NGC 6960 sits just under three degrees south of ε Cygni (Aljanah), the star marking the eastern wing of the Northern Cross — under dark skies, the nebula’s brightest section even shows up in binoculars, threaded right through the naked-eye star 52 Cygni.

Where my small field sits in the bigger picture

To put my framing into context: the image below shows the entire Cygnus Loop in a magnificent mosaic by Austrian astrophotographer Daniel Nimmervoll, from which I extracted the luminance — with my HOO field overlaid at its actual position. My Only now does the scale of the object become tangible: the complete remnant spans close to three degrees on the sky, about six full moons across, and my frame covers just its western rim. The great arc running through my image is one section of a shock front that continues far beyond the frame — north toward Pickering’s Triangle, south and east around the entire bubble to the Eastern Veil, NGC 6992/95. What looks like a self-contained nebula in a single field is, in truth, a small window onto the surface of one vast expanding structure.

The imaging train for this project works at 1,019 mm focal length with 3.76 µm pixels, which puts the sampling at 0.76″ per pixel. That is a comfortable match for the seeing at my site: with a typical FWHM of 2–3″, a star spreads across roughly 2.6 to 4 pixels — close to the Nyquist sweet spot, with hardly any oversampling. The full sensor of 6,140 × 3,917 pixels then covers a field of about 1.3° × 0.8°. Held against an object like the Cygnus Loop, which spans nearly three degrees … The overview mosaic of Daniel shows what the field would need to be — and why the full loop is a project for a very different focal length, or for a mosaic of my own one day.

The Processing at a glance

The workflow for this image runs in two strictly separated branches that only meet at the very end. The first branch builds the nebula: Hα and [O-III] are combined into an HOO composite, calibrated, freed of gradients, sharpened, stretched — and then stripped of its stars entirely, leaving nothing but the filaments. The second branch exists for one purpose only: honest star colors. Narrowband stars can never show a star’s true color — two 3 nm slices of a continuous spectrum simply don’t carry that information — so the stars come from a separate set of broadband RGB exposures, calibrated spectrophotometrically along the same principles, and are finally extracted as a stars-only image. In the last step, the starless HOO nebula and the RGB stars are screened together into the final image. One deliberate rule governs both branches: every measurement-dependent operation — flux calibration, gradient correction, color calibration, deconvolution — happens while the data is still linear; the stretch comes as late as possible. The flowchart below shows the complete sequence of both branches; each step is then covered in detail in the sections that follow.

Branch 1 – HOO – Nebula Processing

CC – Channel Combination

The two narrowband masters are combined with ChannelCombination, assigning Hα to the red channel and [O-III] to both green and blue.

There is a physical reason behind this mapping: H-α emits at 656 nm, which is deep red, while [O-III] emits at 501 nm — a blue-green that sits right between the green and blue bands. Assigning [O-III] to both G and B therefore reproduces the „natural colors“ of the emission lines themselves. This is therefore not a decorative false-color scheme: every color in the image has a defined physical meaning — red traces hydrogen emission, blue-green traces doubly ionized oxygen [O-III] . A palette is valid precisely when it states what each color represents, and this one does.

SPFC – Spectrophotometric Flux Calibration

Before any reference-based gradient correction (MGC) can be applied, the image needs a spectrophotometric flux calibration (SPFC). Flux calibration establishes the relationship between the pixel values recorded by the camera and a standard photometric reference system. Using the plate-solved star field, SPFC measures the brightness of stars in the image and compares them to catalog data, taking the spectral response of the sensor and filters into account. The result is an image whose brightness values are expressed on a physically meaningful scale — the prerequisite for comparing it to external reference data such as the MARS survey, which is the basis for Multiscale Gradient Correction (MGC).

Since we are working with emission-line data, the tool is switched to Narrowband Filters Mode: the red channel is defined by the H-α line at 656.3 nm, green and blue by the [O-III] line at 500.7 nm. The bandwidths should match the actual filters in use — either taken from the Filter Manager database or from the manufacturer’s data sheet. The quantum efficiency curve can be left at „Ideal QE“. This step ties the measured pixel values to a photometric reference system, which is exactly what MGC in the next step will rely on.

After running SPFC, the tool reports its Flux Calibration Functions. Each graph plots the individual scale factor derived from every measured star, sorted in ascending order, for one channel: almost 1,800 stars from the Gaia DR3/SP catalog were matched in each. The median scale factor (about 4.07 × 10⁻² for Hα in red, 6.57 × 10⁻² for [O-III] in green and blue) is calculated, while the shaded band in the diagrams shows the dispersion around it. What we want to see is exactly this shape: a long, nearly flat central section, meaning the vast majority of stars agree on the same calibration factor, with only the tails at both ends curving away — those are outliers such as variable stars, saturated stars, or blended sources, which are rejected robustly by using the median. The very small dispersions (around 2–4 % of the scale factors) confirm a consistent, well-constrained calibration. Note that green and blue are identical here by construction, since both channels carry the same [O-III] data.

MGC – Multiscale Gradient Correction

A personal note before diving in: the entire gradient correction approach described here is something I had the privilege of learning directly from Vicent Peris at the PixInsight intensive workshop in Valencia in July — here is probably no better source, since Vicent is the very person who created the MARS survey that this method relies on.

For this particular target, the choice of gradient correction method is not a matter of physics. NGC 6960 is embedded in extended, real hydrogen emission that pervades this part of the constellation Cygnus. Any gradient model built solely from the image itself cannot tell the difference — it will happily classify large-scale genuine nebulosity for example as an instrumental gradient and remove it. MultiscaleGradientCorrection (MGC) avoids this trap by comparing the image against an external reference: the MARS survey (LINK), a calibrated wide-field sky atlas. With that reference, the tool can distinguish what actually belongs to the sky from what the instrument and light pollution adds to the image.

The configuration follows the HOO channel assignment: the MARS H-α data serve as reference for the red channel, the MARS [O-III] data for green and blue. (MARS covers both emission lines; only S II is missing from the survey — irrelevant for the actual HOO-project.)

I tested two different approaches to gradient removal on this data: GraXpert and PixInsights MGC. The difference is striking, particularly in the lower right of the field — with MGC, the extended hydrogen emission survives largely intact, whereas GraXpert „eats“ noticeably into these real large-scale structures, treating them as gradient. This is exactly the fundamental limitation of any image-internal gradient model: without an external reference, it cannot know which large-scale signal belongs to the sky.

Settling on the final MGC parameters took some experimentation — about an hour and a half of it, to be honest. I eventually chose a gradient scale of 2,048 pixels, which removes only a very large-scale, smooth gradient component and leaves everything below that scale untouched.

SPCC – Spectrophotometric Color Calibration in narrowband mode

With the gradients removed, one last calibration step remains before leaving the linear stage: SPCC, switched to Narrowband Filters Mode. The filter definitions match those used for flux calibration — 656.3 nm for the red channel, 500.7 nm for green and blue, each with a 3 nm bandwidth, and the ideal QE curve.

One might reasonably ask whether color calibration makes any sense for a „false-color“ palette like HOO — but for this project it is essential. In narrowband mode, SPCC does not attempt to produce „correct“ star colors; instead, it removes the instrumental signature from the ratio between the two channels, correcting for filter throughput, sensor QE at 656 vs. 501 nm, and the unequal exposure times of both stacks. Only after this step does the red-to-teal (=blue-green) balance in the image encode the true flux ratio of the two emission lines rather than the properties of my equipment — the difference between a palette chosen by taste and one that can actually be interpreted physically.

What about the White Reference in Narrowband?

When I first ran SPCC on the narrowband data, I left the white reference at its default, Average Spiral Galaxy. That was a mistake, and unpacking why is worth a few paragraphs, because it goes to the heart of what color calibration actually means. The following is also what I learned at the PixInsight Intensive Workshop in Valencia this year. 

At the workshop Vicent Peris devoted an entire lecture to a question most of us skip past: what does „white“ even mean for an astronomical image? His starting point was that the reference is never a cosmetic choice. A white reference sets the relative scaling of the color channels — and if that reference is contaminated by a gradient, by strong emission, or by an inappropriate object population, the resulting balance is systematically wrong, no matter how pleasing it looks. His central formulation stuck with me: the choice of reference determines what the calibration factors mean. The numbers SPCC hands out to you are only as meaningful as the reference you measured them against.

He also explained why his preferred broadband reference is the average spiral galaxy. A spiral galaxy contains both old and young stellar populations and, depending on the model, varying amounts of dust; by averaging over a statistical sample, the reference stops depending on any one unusual star or any single galaxy in the frame. He was careful to stress what this is not: it is a defined, statistical model — not a claim that spiral galaxies are somehow intrinsically white, and not a demand that I find one perfect galaxy in my own image. It is a sensible, physically grounded definition of neutral for broadband light. Which is exactly why I use it, correctly, for the RGB star branch of this image.

But what does that mean for my narrowband channels? Strictly speaking, an HOO image does contain continuum — every star in the field is a continuum source, and it is precisely this stellar continuum, sliced through my 3 nm Optolong filters, that SPCC measures against Gaia spectra to calibrate the data in the first place. The object signal, however, is another matter: the nebula itself emits in two lines, Hα and [O-III], with no continuum worth speaking of — the optical emission of a supernova remnant is almost pure line radiation. For a two-line spectrum, the question „what should be white?“ has no meaningful answer. Handing SPCC the Average Spiral Galaxy reference here tells it to scale the channels so that a galaxy spectrum would come out neutral — and a galaxy spectrum has nothing physically to do with these two lines. In Peris’s own terms, this is precisely the „inappropriate object population“ case: the factors still come out, they just no longer mean anything. The result is an arbitrary reweighting of Hα against [O-III].

The correct setting for narrowband in SPCC is Photon Flux. It calibrates the channels so that each one’s brightness corresponds to the actual photon flux of its emission line — stripped of the distortions introduced by the filters, by the sensor’s quantum efficiency at 656 versus 501 nm, and by the unequal integration times (3 h 50 m of Hα against 2 h of [O III]). Where the Veil genuinely emits twice as many Hα photons as [O III] photons, that ratio is now what the image records. This is the same principle Peris teaches for broadband, carried honestly into narrowband: pick the reference whose factors actually mean something for the light you captured. For a purely aesthetic image the choice is a matter of taste; for an image meant to represent real line ratios, Photon Flux is the only reference under which the relative channel brightnesses carry physical meaning. So SPCC goes back in — same filters, same bandwidths, same Ideal QE — with the white reference set to Photon Flux this time:

True Physics instead of pretty pictures!

The comparison of the two SPCC reports is instructive. The linear fit — the actual measurement of my instrument against 2,920 Gaia stars — is identical in both runs, down to the last digit: the measurement never changes, only what it is measured against. But the white point moved. With the galaxy reference, catalog flux ratios of R/G = 1.29 were defined as neutral — an average spiral galaxy simply delivers about 29 % more flux through a 3 nm Hα band than through the [O-III] band, and the calibration obligingly balanced my image toward that spectrum. With Photon Flux, the neutral point sits at exactly 1.0: equal photon flux in both lines means equal channel values. The white balance factors tell the story in one number — 0.7823 before, 0.5932 after. The galaxy reference had been weighting my [O-III] channels roughly 32 % too high. The teal in my earlier rendition was almost a third stronger, relative to Hα, than the physics of the nebula warrants. Not a subtle difference — and exactly the kind of error that looks perfectly fine on screen until you check what the numbers mean.

Same data, same stretch — only the white reference differs. The 32 % over-weighting of [O-III] in the galaxy-referenced version (right) pushes the entire field toward a cool teal-gray and visually swallows the extended Hα emission between and behind the filaments. The Photon Flux version (left) restores the actual line ratio — and with it the physical structure of the remnant: a thin [O-III] front tracing the current shock, with recombining hydrogen glowing red across the field behind it. The right-hand image is arguably the more conventional HOO look. The left one is what the nebula actually emits.

Background Neutralization?

After the recalibration, the left part of the field shows a distinctly red background — which raises the question of background neutralization. The answer is yes, but only after checking what that red actually is. SPCC rescales the channels multiplicatively, and that rescaling also acts on the background offsets: a previously neutral residual level can acquire a color cast that the later stretch would amplify dramatically. So a neutralization step is part of the plan. The caveat is the location: this field sits inside the Cygnus emission complex, where genuine large-scale Hα is nearly everywhere, and a neutralization fed with real emission as its reference would subtract exactly the physics the Photon Flux calibration just restored. Two quick tests separate the cases: real emission shows structure under a harsh screen stretch, while a calibration offset is featureless and flat; and channel medians measured in several small previews reveal whether the red excess is constant everywhere (an offset) or varies with visible texture (emission). The neutralization then runs against a Region of Interest on the darkest, verifiably emission-free corner.

The HOO-Image after SPCC and BlurXTerminator (BXT)

The stellar adjustment settings are of no consequence here anyway — the stars will be removed from the HOO image in the next processing steps, so any effort spent sharpening them would be wasted. BXT was applied with 0.85 at sharpen non-stellar (result tested on several preview-runs before).

Star Removal – … but not using StarXTerminator!?

My first attempt at removing the stars used the AI-tool StarXTerminator — and the result seemed not that bad – but only at a first glance:

Should the Veil Nebula be the kind of field where AI-based star removal reaches its limits? The filaments are full of compact knots and condensations that, to a neural network, they probably look like stars. Anyhow, StarXTerminator removed them along with the actual stars, eating visibly into real nebular structure.

Lets look deeper at the lower part of the nebula (right in the middle of the image) zoomed at 1:1:

A different (but too complex) Idea to remove the Stars …

Instead of letting an AI guess which structures are stars, I build a star mask directly from the Gaia DR3 catalog: since the image is plate-solved, the exact position of every real star in the field is known. A synthetic star image rendered from these catalog positions should serve as a mask, and a MorphologicalTransformation then maybe shrinks the stars away — but only at the masked positions. By construction, this method should not touch a single nebular knot or filament, because it only operates where the catalog says a star actually is.

Measuring the Median FWHM

PixInsights FWHM-Eccentricity script only accepts grayscale images, while the HOO image is RGB. The quickest workaround is using just the Luminance Data (extracting the CIE*L component). The measured median FWHM is the only number we need: it will define the PSF width of the synthetic stars rendered from the Gaia catalog in the next step. In my case the median FWHM is 1.532 pixel.

Rendering a Synthetic Star Image

With the FWHM measured, the synthetic star image can be rendered. The CatalogStarGenerator script (Script → Render) takes the plate-solved HOO image as reference, which means the output inherits its exact geometry and astrometry — every rendered star lands precisely on its real counterpart. As catalog I use Gaia DR3 from the same local XPSD files that SPCC already relies on. The magnitude limit is set generously deep (G = 18) so that even the faintest stars visible through the 3 nm filters are covered; extra catalog stars that are invisible in the image do no harm, as the later erosion will merely touch background at their positions. The PSF is set to Gaussian with the measured FWHM rounded up slightly, ensuring the synthetic disks fully cover the real stars. The result is a black frame with white points at every Gaia star position — blinking it against the HOO image under an aggressive screen stretch confirms that the patterns align perfectly.

The rendered catalog image is not yet a usable mask —a finishing touch: Convolution (StdDev 2–4 px) widens the hard rendered points into soft profiles that cover not just the star cores but their flanks too. Before the mask does any work, it gets inspected as an image in its own right: every visible star must be covered, the coverage must extend beyond the star flanks — and, the whole point of this method, not a single nebular knot may appear in it. Using MAS (Multiscale Adaptive Stretch) the synthetic stars are delinearized afterwards.

Applying the Mask

Now the mask goes to work. Dragged onto the HOO image — not inverted, so the white star disks define where the process may act while the nebula sits safely under black — it restricts the following MorphologicalTransformation entirely to known star positions.

I used several parameters at the Morphological Selection …after tryin nearly every combination: …I finally gave up! The stars grew smaller, but never (neither the big ones nor the small ones) vanished completely. It seems that the Morph-Tool is actually too complex for me …

Star removal, second attempt — and a lesson learned

After delinearizing the image with Multiscale Adaptive Stretch, I gave StarXTerminator another chance — and the difference could hardly be more dramatic. On the stretched image, the star removal works almost perfectly: the stars vanish cleanly, and the filaments — every knot, every thin shock front that the first attempt had shredded — remain intact to the vast majority. The three-way verification (original, starless, and the difference between them under a harsh stretch) now shows exactly what it should: stars in the removed layer, and nothing but stars.

The explanation seems straightforward. The AI behind StarXTerminator is trained predominantly on stretched images. On linear data, the thin, sharp shock fronts of the Veil present as narrow, spiky structures — statistically almost indistinguishable from stars. After the stretch, the same filaments reveal their extended, connected nature, and the network suddenly has no trouble telling gas from star. The practical takeaway for supernova remnants and similarly filament-rich fields: if StarXTerminator eats your nebula, don’t fight it with sensitivity settings — delinearize first, then remove the stars!

HOO stars + nebula & starless

Branch 2 – Processing RGB-stars

With the HOO processing branch complete, the star processing begins — the whole point of capturing broadband alongside narrowband: the HOO image will carry the nebula, but its stars, recorded through two narrow emission lines, can never show true stellar colors. Those come from 50 minutes each of R, G, and B. The first step mirrors the narrowband workflow: ChannelCombination of the three drizzled masters — importantly, from the same drizzle_1x_autocrop set as the HOO branch, since only that set carries an astrometric solution for all five filters, and pixel-perfect registration between the two branches is what will later allow the starless nebula and the RGB stars to merge without any offset. No creative channel mapping again: red to red, green to green, blue to blue. The result looks unimpressive under an automatic stretch — flat, tilted by gradients, nothing like the narrowband data — but that is expected from ten frames per filter. Everything this branch needs to deliver is physically correct star colors; the calibration steps that follow will take care of exactly that.

CC – Channel Combination

The faint nebula filaments shimmering through the broadband channels are of no concern; this branch only has to deliver stars, and everything else will be removed at the end.

The freshly combined RGB image makes a useful case study, because it contains both kinds of large-scale signal side by side. The greenish parts along the top edge and the general brightness tilt toward the left are to my opinion true gradients — structureless, sweeping across nebula and empty sky alike, the classic signature of airglow and light pollution. The reddish-brown glow on the right, however, is Hα emission which is visible in narrowband data too – the broadband red filter naturally picks it up as well. This mixed situation is why the gradient correction again runs with MGC against the MARS model: a purely image-based method would probably read that genuine emission as a gradient and subtract it, while MARS knows it is real and leaves it alone. The green parts on the upper area has to go completely; the red glow should survive. Let`s judge upon in the next chapter!

SPFC in Broadband Mode

Flux calibration for the star branch runs in broadband mode, and this changes what the process needs to know: instead of idealized narrowband parameters, SPFC now folds three real transmission curves — the Optolong R, G, and B filters — with the actual quantum efficiency curve of the IMX571 sensor and the Gaia spectrum of every measured star. 

MGC – Multiscale Gradient Correction

The before/after pair below shows the result. The green airglow veil that draped the entire top edge is gone, and the warm brightness tilt in the lower left has leveled out into a uniform neutral background. The decisive detail, however, sits on the right-hand side: the extended reddish glow — genuine Hα emission that the broadband red filter legitimately records — has survived the correction untouched.

SPCC – Spectrometric Color Calibration

Color calibration for the star branch, executed: white reference Average Spiral Galaxy — correct here, for the reasons the narrowband exercise showed to me — with the real Optolong filter curves and the IMX571 quantum efficiency curve doing the instrument modeling and the background neutralization referenced to a small preview on the darkest, verifiably emission-free corner of the field. The report to check afterwards now contains two genuine linear fits: a star field in which cool giants glow orange, hot main-sequence stars blue-white, and 52 Cygni „convincingly“ warm is a field whose colors can be believed.

BlurXTerminator – Correct only

In correct only – mode BXT repairs what the optics and guiding left behind — slight coma residues in the corners, minute elongation, chromatic fringes — but performs no deconvolution, no star reduction, no sharpening. The reasoning is simple: these stars are meant to look natural. The visual check is subtle by design — rounder stars in the corners, hardly any change in the center.

MAS & Star Removal

The star branch now passes through the same gate as the nebula did: the Multiscale Adaptive Stretch (MAS) — executed with the identical parameters, straight from the saved process icon of the HOO run. This is not convenience but a requirement: both branches must carry the same tonal transformation, so that when stars and nebula are later screened together, they live on one common brightness scale. The check afterwards concerns the brightest stars — 52 Cygni may saturate in its core, but the warm color must survive in the flanks.

Then StarXTerminator runs one more time, with its purpose inverted: „generate star image“ is switched on, and the product this time is not the starless view but the extracted star layer itself — every star, in its calibrated color, on black.

Screening the Stars Back In

One small refinement before the merge: a gentle saturation boost on the star layer. This needs some justification, because the calibration was the point of the whole branch. The key distinction is between hue and saturation — SPCC fixed the hues and their ratios, and a saturation curve does not touch them: 52 Cygni, the K0 giant blazing in the middle of the front, becomes more richly golden-orange, never differently orange. The adjustment belongs to the same category as the stretch itself — a change of rendition, not of the underlying measurement. There is even a physical case for it: compressing the luminance range, as any strong stretch must, perceptually desaturates colors, especially in bright star cores — and 52 Cygni is the prime victim in this field, its warm hue washed toward white by the stretch more than any other star’s. A moderate boost restores the color impression the linear data actually contained. The crucial craft decision is where to apply it: on the star layer alone, before the merge. There it affects exactly what it should and provably leaves the calibrated Hα-to-[O III] balance of the nebula untouched. The curve itself follows the my Valencia-workshop learning“ — smooth, concentrated on the low-saturation range, returning to the diagonal at the top. And 52 Cygni doubles as the control star for both failure modes: if its core turns artificially amber instead of staying a white-saturated center inside a warm rim, the curve has been pushed too far into the highlights; if colored rings begin to bleed into its flanks, the overall boost is too strong. As long as the brightest star in the field passes both checks — and the faint stars have not started glowing like confetti — the field still looks like a photographed stellar population, not an artificial rendered star catalog.

The two branches finally meet in a single line of PixelMath: the screen blend — invert both images, multiply, invert back.

A quick word on the tilde: ~ in PixelMath simply means „invert“ — turn bright into dark and vice versa (1 − x). The formula does three things: invert both images, multiply them, invert the result back. Why that complicated? If you just added the two images, a bright star on a bright filament would exceed the maximum and burn out. The invert-multiply-invert trick makes that impossible — the result can never overshoot. Where there is only nebula, the nebula stays as it was; where there is only a star, the star stays; where both meet, the image brightens gently instead of clipping hard.

Why the detour through inversion matters: The same star layer combined with the same starless nebula — once by naive addition (left), once with the screen blend (right). Around 52 Cygni, addition pushes star plus bright nebula beyond the maximum: the core clips into a burned-out white block, the diffraction spikes blaze, and the filaments behind the star’s halo drown in the overflow. The screen blend cannot overshoot by construction — the core saturates gently, the warm rim survives, and the shock front stays legible right through the star’s glow.

The final inspection walks the workshop checklist: no dark rims around stars, no burned-out white block where 52 Cygni should glow warm, faint stars alive and colored inside the filaments, a neutral background with a free histogram foot — and, under a merciless screen stretch, no ghost rings where bright stars once stood in the narrowband data. If all of that holds, the image is done: hydrogen in red, oxygen in teal, and every star wearing its own calibrated color.

Before you enjoy the final image: a thought on beauty

I owe this thought — and the book it comes from — to Vicent Peris. Somewhere in one of his image descriptions he quoted the philosopher Byung-Chul Han, and the passage stayed with me long after I had forgotten where exactly I read it. It led me to Han’s essay The Scent of Time (Der Duft der Zeit, on the art of lingering), and to an idea that has shaped how I approach these images ever since: Han argues that immediate enjoyment never truly reaches beauty — that the beauty of a thing reveals itself only later, in the light of another, in the careful act of remembrance. Beauty, in other words, is not consumed in a glance; it requires lingering, context, and return.

I believe this applies to astrophotography with unusual force. An image like this one can be scrolled past in two seconds as a pleasing arrangement of red and green and blue — and at that speed, that is all it is. Its actual beauty only opens up when you know what you are looking at: that the thin teal line is a shock front at this very moment, that the red behind it is the same gas remembering its own heating, that the entire structure is a twenty-thousand-year-old event still unfolding. The knowledge is not an accessory to the image; it is the condition under which the image becomes beautiful at all.

This is also, ultimately, the reason this entire processing report insists on physical honesty rather than a pretty picture. What follows, then, is my attempt to hand you that light: the astrophysics of this image, in the plainest language I can manage.

Take your time with it. The nebula has.

Reading the final Image: the beauty of the physics in NGC 6960

Now that the processing is done, it is worth stepping back and reading the picture — because with a calibration that preserves the real line ratios, the physics is genuinely legible.

A shock front seen edge-on

The great curved front running through the frame is the supernova blast wave itself, caught nearly edge-on — not a wall, but the surface of a huge bubble. Wherever our line of sight grazes that curved sheet tangentially, the emission of many path lengths stacks up and we see a bright, sharp line; this „rippled sheet seen edge-on“ interpretation goes back to Hester (1987) and has been confirmed by two decades of Hubble proper-motion imaging, which even resolves individual sight-line tangencies within single filaments [1, 2]. Measured shock speeds in the optical filaments lie in the range of roughly 100–180 km/s [2].

Why teal at the edge and red behind it

When the shock slams into the interstellar gas, it heats it abruptly to several hundred thousand kelvin — initially far too hot for optical emission; the freshly shocked gas radiates in the ultraviolet and X-rays (I link the NASA source in IV: „Ultraviolet image of the Cygnus Loop Nebula“). Then the gas cools, and as it does, it descends through the ionization states one after another: first through the zone where oxygen is doubly ionized and [O-III] at 500.7 nm lights up — the teal — and further down, at around 10⁴ K, into hydrogen recombination, which produces the Hα glow — the red in 656.3 nm — behind the front. This stratification of the cooling, recombining post-shock flow is classic radiative-shock physics [3, 4] and is precisely what the edge-on viewing geometry lays open [4]. The image is a cooling sequence translated into spatial coordinates: the teal edge is the shock’s present, the red expanse behind it is its past.

Why filaments and not a smooth sheet

Two reasons. First, the interstellar medium is clumpy — the front runs faster through thin gas and slower, but brighter, through dense pockets. Second, the thin emitting sheet itself wrinkles; where our sight line pierces a wrinkle tangentially, we see a luminous thread, and where it passes flatly through, only diffuse glow [1, 4]. Many of the „filaments“ are thus folds of one and the same sheet, and apparent crossings of threads are projection effects of different fold depths along the line of sight [2].

The bright star

The dominant star embedded in the front, 52 Cygni, has nothing to do with the nebula: its parallax places it at roughly 60 pc — more than ten times closer than the remnant — a foreground star standing coincidentally in the line of sight.

The distances and the age

The remnant’s distance is remarkably well pinned down by now: 725 ± 15 pc (2.360 ± 50 light years), anchored by Gaia parallaxes of stars demonstrably in front of and behind the shock [5]. Its age is estimated at about twenty thousand years [5, 6].

The red expanses on the right

Emission from deeper layers of the remnant — partly the projected far side of the bubble, partly regions the shock traversed long ago, where hydrogen now recombines across large areas. The faint teal veils at the far right belong to neighboring sections of the front, toward Pickering’s Triangle.

The missing corpse

A core-collapse supernova should leave a neutron star, yet no compact object has ever been firmly associated with the Cygnus Loop: the candidates in its southern reaches — an ASCA point source that might equally be a background AGN (active Galactic Nucleus), and a second object with a candidate pulsar-wind nebula — show no pulsations [7]. Radio mapping has even raised the possibility that the „Loop“ is two overlapping remnants [8].

References

[1] Hester, J. J. 1987, ApJ 314, 187 — the rippled-sheet interpretation of SNR filaments.
[2] Sankrit, R., et al. 2023, ApJ 948, 97 — third-epoch HST imaging of a Cygnus Loop shock: https://iopscience.iop.org/article/10.3847/1538-4357/acc860
[3] Raymond, J. C. 1979, ApJS 39, 1 — radiative shock models.
[4] Levenson, N. A. & Graham, J. R. 2001, „Shocking Clouds in the Cygnus Loop“: https://arxiv.org/abs/astro-ph/0106378
[5] Fesen, R. A., et al. 2021, MNRAS 507, 244 — Gaia EDR3 distance 725 ± 15 pc: https://academic.oup.com/mnras/article/507/1/244/6324593
[6] Loru, S., et al. 2021, MNRAS — radio observations, age 1–2 × 10⁴ yr: https://arxiv.org/abs/2009.09948
[7] Katagiri, H., et al. 2015, „Fermi/LAT Study of the Cygnus Loop“: https://arxiv.org/abs/1502.03053 (and Miyata et al. 2001: https://arxiv.org/abs/astro-ph/0011535)
[8] Uyanıker, B., et al. 2002, A&A 389, L61 — „Is the Cygnus Loop two supernova remnants?“: https://www.aanda.org/articles/aa/abs/2002/26/aaec262/aaec262.html

The Paper in English and German