Unified FITS Editor (UFE)

Versión en español

The Unified FITS Editor is NightScribe's single place to view and work FITS images (ADR-044). In the interface it is called the «NightScribe Image Workbench» (it is the processing workspace by now, not just an editor) and opens from Tools → NightScribe Image Workbench…; "UFE" stays as the internal codename in code and docs.

The editor is the only door: the classic dialogs (blink, comparison chart, annotated FITS) retired once the UFE became the default, so there is nothing to switch back to. The only editor preference left is the look of its top bar, under Settings → Interface. Opened from a project, the UFE comes with the plate loaded, the right tab on stage and everything the project knows about the object already in place: name, coordinates and magnitude on the line under the top bar and in the title, and every tab pre-filled (Blink: name and coordinates; Photometry: target, magnitude and B−V when the record carries it; Annotate: label, the marker on the object's position and the extra visits). What it writes (annotated copies, blink GIF/PNG, sequence CSV/PNG) registers into the project just like the classics did. With no plate of your own, the Photometry tab (Sequence mode) downloads the survey field (DSS2/PS1) as a FITS with WCS and works on it directly.

The window

| [Open][Export][Solve] | [Fit][100 %][Zoom ▾] 100 % | [View ▾] | Image|Curve |
| [Blink][Calibrate][Annotate][Series]                                      |
|───────────────────┬────────────────────────────────────────┬──────────────|
|  visit             |                                        |  panel       |
|  · frame navigator |   IMAGE  (or the curve)                |  ────────    |
|  · FRAME PREVIEWS  |   · the object, over the plate          |  primary     |
|  · EXOTIC          |                                        |  [Settings▸] |
|───────────────────┴────────────────────────────────────────┴──────────────|
| Histogram ▸  (folded: 24 px; open: ~110 px, two rows of controls)         |
|──────────────────────────────────────────────────────────────────────────|
| ⓘ status: one line, fixed height, never grows                            |

The four errands (2026-10-06) are their own buttons in a row under the bar: Blink, Calibrate, Annotate and the photometric series. Each one opens a non-modal window over the plate, sized to what it holds, and while it is open it owns the plate (the blink frame, the annotation's clicks); closing it hands the plate back to the panel. One at a time. They are in their own row and not inside the bar for a measured reason: with the theme on, four icon buttons cost 232 px and inside the bar they left the project badge (the one thing that says WHICH project you are working on) without room from 1200 px down. They follow the bar's icon mode: a glyph with its tooltip by default, and icon plus label if you turn the labels on in Settings.

Image | Light curve is at the right end of the bar: it is a view of the centre, so it belongs to the bar and not to a row of its own above the plate (it used to cost a row of the plate's height).

The visit panel at the left holds the frame navigator and the frame previews. The photometric series and the EXOTIC transit reduction moved out of it into the series window (Series in the tools row): a transit fit is the series of a transit visit, so the reduction lives under the series block, and it still appears only in a transit visit with an armed comparison sequence.

Frame previews (2026-10-06): a vertical list that takes the whole column and the whole height of the panel, with one preview per frame of the visit, filled as they are read (a sampled read, ~2 MB per 2048² frame, in a worker). The preview is as wide as the column (140 px became ~250, and it follows the splitter) and the caption rides on the image: "Frame 12 · the_file.fits", small and in the object kind's colour (the same hue the chip, the hero button and the block spines wear), with a dark outline so it reads over black sky or over a bright nebula. A frame with something wrong wears a red border and says what it is, in red: unreadable (a cut capture leaves a 0-byte file) or not registered by the last astrometry run. Red means a problem and nothing else. The tooltip carries the frame's own numbers: size on disk, measured sky, filter, exposure and date. The header counts them ("207 frames · 3 with problems") and Only problems shows just those, which is the fast way through a night of hundreds of frames. A click opens the frame in the editor (the navigator follows both ways) and the context menu takes a frame out of the visit (the file is NOT touched) or moves it to the project's dscrd/ folder (nothing is deleted, it can be brought back).

The chrome takes what it NEEDS and the work area gets everything else: the plate is what the window is for. The panel column opens at 420 px and can be dragged to 560: its content needs 331 as a minimum (measured 2026-10-06, when a narrower column cut the messages and pushed a horizontal scrollbar into the panels), so the labels are read whole and the plate keeps the rest. The top bar is stable (it measured 25 px in a short window and 69 px in a tall one before: the layout's stretch was never applied by the loader), and the histogram strip is compact and folds with its state remembered. The object's name, position and magnitude are painted OVER the plate (and into the exported PNG), not in a row of their own.

The photometric series has its own window (Tools ▾ → Series) and keeps what an observer touches while measuring (the grouping of frames, Measure the sequence, live mode and the progress) with the rest behind two doors: Chart and quality… opens the chart's own non-modal window (scale, error bars, binning, mean curve, outliers and exclusions) and Series ▾ holds the occasional actions (undo, ExoClock, the night's figures, save the chart, period and phase, the guide). Before this the same panel showed some thirty controls stacked in a column.

• Image: takes up most of the window. The wheel zooms anchored at the cursor; dragging pans; double-click returns to the fit. Hovering shows a tooltip with the pixel, its DN value and the RA/Dec when the plate carries a WCS. In the marking tabs (Photometry, Annotate) the cursor becomes a crosshair with a centre-gap reticle that snaps to the source's centroid under the mouse: the click is born centred. The detection is local and robust (it sees faint sources even on a galaxy's glow) and the snap's reach is capped at 9 plate px: the reticle never jumps to a bright star far away.

• Panels: two, Photometry and Astrometry, plus the four tool windows (Blink, Calibrate, Annotate, Series) behind Tools ▾. Only the panel (or the tool window) that is on stage answers clicks on the image.

Blink

The Blink tab blinks your plate against the PanSTARRS DR1 g reference (supernovae and transients; the plate is the one you loaded with "Load FITS…"):

• Type the SN name (or tick "Manual coordinates" and give RA/Dec) and press Prepare pair: it resolves the target, downloads the reference in your plate's geometry and starts the live blink.

• The stretch is the common one (the histogram strip drives the blink too); Balance multiplies the reference to match the sky background (Auto button); Fine alignment is the cross of arrows that walks the reference in half-pixel steps when the registration is not perfect, just like the legacy blink dialog.

• Blink (alternates at the chosen interval) or Fade (static blend with the slider); the amber marker sits at the SN position mapped through your plate's WCS.

• GIF… / MP4… / PNG… export the pair (side by side for PNG), with the crop zoom on the SN of your choice.

Calibration in the editor

The Calibrate button in the tools row opens the Calibration window (its own tab): the visit's recipe resolved against the master library, and the Calibrate the visit run. It reads the recipe for the visit's first frame, says which master each piece uses (one this project does not own is marked "(shared)") and what is missing, with the criterion a rejected master failed and the numbers. The library and the policy (the vignetting model, the dark's temperature tolerance) live in Settings → Measurement, and Calibration settings… is the door there.

Calibrate the visit applies the recipe to every frame as a one-off run (offset first, flat second); it works in memory, and it writes FITS copies only if Export the calibrated frames (FITS copies) is ticked. Those copies land in the project's calibrados/ subfolder and register to the visit like any other product; each carries the NSCALIB card, so a run that reads it back leaves it alone instead of subtracting the dark twice, and the report says so. When the library has no flat for a filter and the model policy is on, the run also writes the vignetting model next to the visit, and See the model opens it: it has to be a smooth surface, with no star and no mote.

The per-engine switch is separate: the Astrometry tab's Apply the calibration is what makes the track & stack calibrate each frame as it is read. Chapter 05. Capture has the physics, the matching rules and the warnings in full.

Photometry

The Photometry tab does both jobs in the same place, in two stacked halves (there is a draggable divider between them): the Sequence radio on top picks which half receives the clicks, and the other stays in sight with its rings and labels, so you can switch between building and measuring without losing track of what is on the plate (ADR-044, layout revision, 2026-09-24).

Sequence (the comparisons block)

Builds the photometric sequence on your plate (a WCS is needed; if it is missing, the configured solver gets you one, ASTAP or nova):

With the visit open, the left panel also carries the frame navigator (previous/next, frame i/N, "first frame": the open frame is the reference) and, in transit projects, the Transit reduction (EXOTIC) block: the two buttons (reduce, and export the inits.json), and under them the visit's last reduction in one line (T_mid ... +/- ... · Rp/Rs ... +/- ... and when it ran) with its two doors: "See the result..." and "Open the folder". The result opens in its own non-modal window (the fitted numbers, the light curve EXOTIC drew and every file the run wrote: field figures with the apertures and the comparisons, diagnostics, CSV, JSON and the AAVSO report), each one a double click from the system; the window opens by itself when the reduction lands. With no reduction yet the line is empty and the buttons are off: never zeros. The project's saved sequence loads by itself when the visit opens.

The top bar carries the two astrometries together, side by side: Solve astrometry... (this plate) and Solve the visit... (every frame of the visit in one go). The pair explains itself: the first solves the plate in front of you, the second the whole night's field.

Solve the visit... is preparation, not measurement: it solves every frame of the visit in one go, and what needs it are the visit's products (the astrometry report and the EXOTIC reduction of a transit), not the series, which measures on the reference plate and registers the rest. With no frames in the visit the button is not shown at all (there is nothing to solve). A visit is one field, and the project knows where it is, so each frame takes a moment instead of a minute of blind search: frames that already carry a WCS are skipped, and each solution is written into its own FITS (it needs "Save the solved WCS in the FITS" in Settings; with it off the button explains instead of leaving solutions that would die with the session). If the project has no coordinates, the first frame is solved blind and the rest follow its field. One frame that fails does not stop the batch: it is counted and named in the status line. A plate opened from a project is also solved with its field, so the Solve button of the top bar answers in a moment.

Build the sequence (comparisons)... is the big button of the Photometry tab (ADR-038 rev): on entering you see it, the line saying what it will do, the tab's own guide ("Click a star...", which lives right under it) and the column's cards, all closed except the result's: "The comparison sequence", "The photometry recipe" and "Measurement" (the last one appears with the first measurement). It does the whole pipeline in one click: if the plate has no WCS it solves it first (the project's field points the solver, so it is a moment), then it queries the catalogue (the field) and proposes. The manual window's Propose sequence fills in the missing step the same way, so the order of the buttons is never something to remember. A rebuild that cannot deliver (the query fails, nothing lands on this plate, no usable comparison star) keeps the sequence you already had and says why.

• Target and Target mag pre-fill what the project knows; the approximate magnitude guides the proposal, and where it comes from is said: the project records whether that figure is a measurement (from an earlier stack), a prediction from the planner (an ephemeris, a catalog, an alert) or yours (you typed it), and the field declares it in its help and in the proposal's own line. It is not a detail: the proposal picks the comparison stars around that figure, and a proposal anchored on a measurement is not the same thing as one anchored on a guess.

• Generate field queries the catalog (Gaia EDR3 or APASS DR9) and the VSX variables around the plate centre: the brightest stars come out labelled ("show catalog magnitudes" checkbox) and the known variables with a red ring (they can never be comparisons). DSS2... on the same row downloads the survey field (PS1-g, fallback DSS2-red) as a FITS with WCS when you have no plate: you work on it directly.

• Click a star to add or remove it from the sequence, as Comparison (cyan) or Check (pink, "on click, add as" radio).

• Propose sequence automatically picks isolated, non-variable stars matched to the target's brightness.

• Sequence (N)... opens the table in a small non-modal window (N is the current number of stars, and it updates itself): you can rename, retype, edit the band and the magnitude by hand (a doubtful catalog value is fixed there: the measurement uses the manual value) and remove rows, and leave it open while you keep picking stars on the plate. Hovering tells you each star's catalog, magnitude and colour, with the window open too.

• Remove all empties the sequence and Export CSV... writes it (fixed columns plus every band); the chart PNG goes through the shared "Export PNG..." button in the top bar: plate, rings, labels and the north arrow / scale bar, exactly what you see.

Measure (the recipe block)

Turns one click into a catalog-calibrated magnitude (single-plate differential aperture photometry):

• It needs the plate with a WCS (if missing, it solves it by itself) and a sequence with the big button (if there is none, a "Go to the sequence" button takes you there).

• Click on the star or the SN: sub-pixel centroid, aperture and sky annulus visible on the image, and the panel tells the full story: instrumental magnitude, the zero point with its error and how many comps were used (and why any was refused), and the calibrated magnitude +/- error. The sequence stays in sight (rings and labels): you measure WITH it; and changing any option (band, radii, sky, sigma-clip, colour term, B-V, subtraction) re-measures at once.

• The daily flow is Band (default V) and the three aperture radii (aperture, inner and outer annulus): touching a radius re-measures the point at once, and your hand edit wins over the seeing auto-scale until you load another plate (or re-arm the checkbox).

• Manual centre: the checkbox opens a small non-modal window with the arrows ↑ ← → ↓ (0.1 px steps) and the 0 button (back to the click). With it checked the measurement uses exactly the centre you place, with no centroid search, for very faint objects or SNe the algorithm would drag to a neighbour; unchecking closes the window and the automatic centroid returns. A new click starts at (0,0).

• Advanced... opens the full recipe in another small non-modal window (you can leave it open while measuring): the sky model (flat median or a tilted plane for galactic cores), sigma-clip of the sky (two 2.5-sigma passes), the seeing-following aperture (FWHM of the comps on your plate, aperture at 1.35 times), the colour term fitted with the comps' and the target's B-V, host-galaxy subtraction with the PS1 reference registered on the frame's stars (scale included), its point spread matched to yours and its masked pixels excluded (for SNe on cores, one download per field), and the Suggest button that proposes the radii with the target's growth curve and its neighbourhood, with the reasons in plain words.

• Report brightness (in the same recipe block): what the band over the image shows, the measurement of this plate or the ephemeris' prediction. It is part of the recipe because it decides what a reader sees; the track & stack writes it into the stack's own band, so a measurement not worth reporting (too few comps, a trailed object) is replaced by the labelled (eph) figure without hiding the measurement, which stays in the run and in the astrometry point.

• Quality controls (phase H, in the background): the real saturation ceiling (SATURATE or ccd_saturate), internal vs. total error (photons + scatter + scintillation + colour + flats) and the check star as the measurement's traffic light. When the header lacks the gain, the panel warns that the error is the comps' scatter only.

• CSV... exports the measurement as one row, AAVSO EFF... in the AAVSO's format (sequence in CNAME/CMAG/KNAME/KMAG) and, when the editor opened from a project, Save… records it in the light curve (source "measure", visit attached). When the sequence lacks the target's magnitude, it is looked up in the project (planner, VSX, or the last saved sequence); and exporting the sequence writes it into the project for next time. The plate on disk is never modified.

Both halves share: the annotations on load (if the plate already carries ANNOTATE cards, written by NightScribe or AstroImageJ, they are drawn on load with their plate-pixel sizes and labels readable at any zoom), the north arrow and scale bar (the "N" and "Scale" buttons in the top bar, with a WCS) and Solve astrometry... (blind-solves with the configured solver, ASTAP or Astrometry.net; a progress dialog with a Cancel that stops the solver; the solution lands in memory and is stored into the FITS itself, atomically, so the plate stays solved for any other program).

To understand how photometry is then measured with these sequences: docs/PHOTOMETRY.md.

A visit's curve: one night, one pass

The series block measures one visit (one night) by default. When the project has more than one visit with frames, the "this visit" / "all visits" selector appears next to the frame counter: with "all visits" the engine measures the frames of every visit in one pass, each night is filed in its visit (one run per night) and the chart shows the project's curve, the union of the nights. Live mode and "discard the curve" are per visit: with "all visits" they step aside and say why.

When the editor is opened from a visit, the chart in the centre draws the curve the visit already has, read from the project: nothing is measured again.

A visit can hold several passes (you measured the series again with another band, with another sequence, or to check something). Every one keeps its points, but the chart draws only one, the one the visit has marked. Measuring again makes the new pass the curve; the earlier ones are not drawn, and the panel says so: how many points the one you see has and how many more passes the visit holds. The door Series ▾ → Passes of this visit… lists them all (time, band, points, stretch of night, state) and lets any of them be the curve without deleting anything, or undo a pass (its points go, its row stays marked, and the chart falls back to the pass before it).

The band in the legend and in the AAVSO file is the band it was calibrated with (the one the engine used on the comparisons), not an invented "V" when the frames' header carries no FILTER. And the detrended curve is refitted on load (it is deterministic: the same points with the same airmass give the same coefficients), so the detrended switch has something to show after reopening the visit too.

The plate's band and the marker style (ADR-046)

The plate says what it knows in the band at the top of the image, on screen and burned into the exported PNG, in two lines:

• Line 1, who it is: the object, the target's sexagesimal RA/Dec and its brightness.

• Line 2, the context: UT date, exposure, filter, the kit that took the frame (from its own header), the MPC station, the plate scale in ″/px and the FOV of what is shown (the last two only when the plate is solved).

The colour of each datum says how much to trust it, and that is the point of the band:

• the position in ink when this plate's own solution places it, and dimmed with a cat when it is only the catalogue's (an unsolved plate);

• the magnitude in a scale of its own: green when the measurement is clean (error up to 0.05, comparisons and check star in order), orange when it is usable but not clean (up to 0.15, or a light caveat: only three comparisons, a magnitude derived from a colour, no check star in the sequence, a flag on the point), red when it is not worth reporting without looking (error above 0.15, too few comparisons, a check star that fails, a clipped core), and white with a cat when it is only the project's or the catalogue's value, which is not a measurement of this plate. The magnitude shown is the one measured on THAT frame (the visit's curve when there is one), then a single-plate measurement, and only then the catalogue. On a track & stack, what the band shows is the run's own Report brightness choice (see the Photometry recipe): the measured figure or the ephemeris', labelled either way;

• everything else (date, exposure, filter, kit, station, scale, FOV) in the quiet colour: it is context, not a judgement.

On a track & stack the project's object mark is switched off when the plate loads: the stack already carries the run's measured position (its cross and the ANNOTATE circle), and two marks, at the plan's coordinates and at the measured ones, read as an error. The top bar's toggle brings it back.

The same code is on the curve's points (the chart's Quality colours button, on by default) and on the measurement's panel: green clean, orange usable but not clean, red doubtful, white for a catalogue value. See the series guide for the thresholds and what each figure needs.

The band never cuts a word: when the window is narrow it drops whole fields (the FOV first, the date last) and, in the extreme, the context line goes and only the plate's name is left. The top bar's "Data" button turns it off (Settings → Interface → "Plate band" sets the default), and the compass and the scale bar keep their classic corners. The blink GIF/MP4 and the sequence chart keep their own metadata boxes, with their own switch ("Other charts" in the same Settings group).

In the same Settings group, Object marker picks the look of the object's mark: ring with ticks (classic) or full-frame cross with a box (applied in Photometry, Annotate, Blink and the sequence chart).

Annotate

The Annotate tab saves AstroImageJ-compatible annotated FITS copies (the original file is never modified):

• Click on the image drops the marker; the arrows ↑ ← → ↓ move it in 0.5 px steps with instant feedback (the readout counts the offset since the last click); size and colour are yours.

• Label and notes travel in the ANNOTATE and NS_NOTES cards; RA/Dec, plate scale and north PA are written from the plate's WCS (NS_RA, NS_DEC, NS_SCALE, NS_NORTH).

• Also annotate (visits): a list of extra plates receives the same annotation; the marker lands on each through its own WCS.

• Save annotated copy… writes the chosen copy plus, next to every visit, its <name>_annotated.fits.

• Histogram: 256 log-scaled bins computed on the display frame. The shaded zones are what the stretch throws away.

Fine stretching

Built for subtle targets (supernovae against galactic cores):

• Handles: blue (black) and orange (white) draggable lines on the histogram, AstroImageJ style; a plain click moves the nearest one.

• Black / White in absolute DN with a fine step adapted to the plate's range; they never cross (white always stays above black).

• Gamma: below 1 lifts the mid-tones; above 1 sinks them.

• Auto: back to the 1 / 99.5 percentiles.

• Invert: swaps black for white; faint objects pop against the sky.

• Keep stretch on load: the next plate keeps your black, white, gamma and invert values instead of the auto percentiles. This is what you want when reviewing a same-camera series of frames.

Zoom

The Fit / 50 / 100 / 200 / 400 presets set the absolute scale: 100 is one plate pixel per screen pixel, and at 200/400 you inspect real, uninterpolated pixels. The zoom survives window resizes, and the view pans 25 % past the plate edge.

Keyboard

Exporting

Export PNG… saves exactly what is on screen (with the NightScribe watermark), ready to attach. Data exports (annotated FITS, charts) always use the original file, never the screen pixmap.

When something does not work

The window has one status line at the bottom, with the whole text in its tooltip (the result of a measurement stays in its own box, next to the action that produced it). A build that cannot deliver keeps your sequence and says why; a solve that fails says what it was doing.

And the application writes what it does to a file while it runs: Help > Open the log. A GUI launched from a menu has no console at all, so a report like "a dialog appears and disappears and I do not know what happens" has its answer in there.