FLOAT
EN FR

Block

Report

FLOAT 0.2.0 · beta of 15 November 2026

What it is
The record of a processing, set out as a short scientific report: what was shot, what was measured, what was done.
What it shows
The shooting details, the star size and sky of every frame through the night, the figures each block measured, and the settings that were changed.
What can be exported
The whole report as a PDF or a PNG, the shooting details as text, and any section or night card as a picture.

The Report tab sets the measures of a workflow out as a short scientific report: the shooting details, the star size, signal and sky level of each frame, the figures each block measured, the settings that were changed. It computes nothing of its own. Every number is a fact a block already wrote, and the report names the block it comes from. It is read in the tab as a PDF, exported as PDF or PNG, and its parts can be copied as text or as pictures for a forum.

  1. Stars/background
  2. Post
  3. Report

1What the report is for

A processed picture does not say what was shot, what the sky was like that night, or what was done to get it. The report is that record: what was shot, what was measured, what was done. It is made to be kept with the project or posted with the picture.

It is a document, not a dashboard: a white page, numbered sections, tables with thin rules, black figures with their captions. The tables are the blocks’ own facts, the same ones Inspect shows. The figures are drawn for the report; Inspect keeps its interactive charts.

The tab has two parts. On the left, a form for what the files do not say, and the buttons that copy. On the right, the report itself in a PDF reader: what is seen is what is exported. It zooms with − and +, with Ctrl and the mouse wheel, and Fit width returns to the page width. When the project holds several workflows, a list above the form chooses which one the report describes.

Until the workflow has been run, the report says so and shows nothing else. If a block is outdated (a night was added since the last run), the report is made from the results it had and says so in red italics, instead of claiming nothing is computed.

2The first page

The first page opens like a journal article. It holds the target’s name (the project’s target, else its name), the name of the workflow and of the Post version when there are several, and one line of position: the centre of the field in right ascension and declination (ICRS), the size of the field and the scale in arcseconds per pixel, all three from the Astrometry block. Then comes the final image, as Post last showed it (figure 1 of the report), and one small line with the program’s name, “Processing report” and the date.

Two tables follow.

Table 1, Acquisition lists the shooting details in the order of a forum post. Each row appears only when known.

  • Camera, with its Gain or ISO speed. Both come from the frames’ headers; Gain is written when the header names a gain (an astronomy camera) and ISO otherwise.
  • Instrument, Mount, Filter, Location and Date. The files do not say them: they are typed in the form on the left of the tab, kept in the project, and the Project block’s form edits the same values. Undo takes an entry back.
  • Frames: the frames kept, grouped by exposure (for example 90 × 180 s), with their total exposure to the second. Darks, Flats and Bias: their numbers.
  • Sampling, the scale of the final pixels, when it is not already the one of the line under the title.

The focal length, the star size and the field are left out of this table when they are given elsewhere on the page.

Table 2, Figures measured on the frames of this session gives, for each figure, its value and how it was measured. A figure is listed only when the block that measures it has run. Table 1 of this page lists them.

Table 1.The figures of the head of the report, and the block each comes from.
FigureValueHow it is measuredBlock
Focal length, measuredin mmFrom the solved sky: the number of stars used and the residual of the fit, in arcseconds.Astrometry
Star FWHM, measuredmedian, in arcsecondsMedian star width, Moffat fit, over the frames kept, with the best frame. Given in pixels when the scale on the sky is not known.Registration
Field driftin arcsecondsHow far the field moved on the sensor during the session (tracking and deliberate shifts), with the rate per minute when the first and last times are known.Registration
Colours offset, correctedin pixelsThe larger of the red and blue offsets from green, with both, and the number of stars they were measured on. Listed only when the correction was applied.Stacking
Sky noise÷ nThe noise of the green channel before Denoise over the noise after it, each 1.4826 times the median absolute deviation of the linear image minus its Gaussian blur (σ 2 px).Denoise
Sky gradient, removedin %The spread of the sky across the field against its level, modelled on a number of samples.Gradient
Gaia DR3 stars in the fielda countThe stars of the catalogue in the field, and how many of them have their measured spectrum.Catalogs (the stage the Color block reads)
Sensor response, rebuilta countThe stars whose spectra rebuilt the sensor’s colour response (section 4).Color
Transparency changesin %The signal of the stars from frame to frame: the gap between the tenth and the ninetieth percentile, over the ninetieth.Registration

3The night, frame by frame

Each session of the workflow has one figure, with up to three panels on one horizontal axis. Every point is a frame.

  • FWHM: the star size, from the Moffat fit, in arcseconds when the scale on the sky is known, else in pixels.
  • Relative signal (%): the signal of the stars in the frame relative to the best ones (100 is the best tenth, and a frame can go a little above).
  • Sky level (ADU): the sky level of the frame after calibration. The panel is drawn only when the calibration’s statistics are there.

The frames kept are black dots. The frames left out are open grey circles. The size, the signal and the times come from the Registration block, the sky level from Calibration. The frames left out are the ones of the Selection block, or those excluded by hand.

When the project has several sessions, each gets its own figure, titled with its name. A small table of the session’s figures goes with it: Frames kept, Total exposure, Median, Best and Worst FWHM (the median also in pixels when it is in arcseconds), and Median roundness. With a single session it is shown only if the roundness was measured.

The same figure exists as a night card: the picture with a heading (“Night card”, the target, the session’s name when there are several) and a foot of the same figures, made to be posted with the final image. Post’s Export writes it next to the exported picture, as a PNG, one per session. When the picture’s facts are known, the card also states what the whole picture is made of: its number of frames and total exposure, then the same for each session or workflow when there are several.

4One section per block

After the first page, the report has one numbered section for each block of the workflow, in the order of the chain: Calibration, Registration, Selection, Stacking, Astrometry, Gradient, Denoise, Color, Stars/background. A block the workflow does not have has no section. Each section opens with one sentence on what the block does. If the block has not run, or is outdated, its state is written under that sentence.

Then come the block’s figures, one row each, with the meaning of the figure beside it. These are the figures Inspect shows for the block, so a reader of the report does not need the manual. Two sections add a figure.

Registration: drift and residual

A figure of two panels against the frame number. The upper one gives the offset of each frame from the reference frame, in x (solid line) and in y (dashed), in arcseconds when the scale is known, else in pixels. The lower one gives the alignment residual of each frame, in pixels. The figure is left out when no offset was measured.

Color: approximate spectral response of the sensor

A figure of the red, green and blue curves of the sensor, relative to their maximum, against wavelength in nm. The positions of Hβ (486.1 nm), [OIII] (500.7 nm) and Hα (656.3 nm) are marked with dotted lines. Its caption gives the number of stars used.

The curves are rebuilt from the Color block’s measure of its stars in the camera’s colours, and from the Gaia XP spectra of the same stars. A star’s flux in a channel is the integral of its spectrum times the channel’s curve, and each curve is a non-negative sum of smooth Gaussians (a centre every 15 nm from 380 to 1000 nm), with a penalty on its roughness, as in the recovery of camera sensitivities from known spectra.[1] The stars the fit cannot explain are rejected over four passes. The curves change no colour: they are a diagnostic. The computation takes about a second and nothing is stored. It needs at least 50 colour stars, and the catalogue’s spectra; without them the figure is left out.

5Recipe and appendix

Recipe is the section after the last block. It writes the chain as a line, for example Calibration → Registration → Stacking, and below it a table of the settings that differ from the defaults, as “block · setting” and its value. When nothing was changed, it says that all the settings are the defaults. Values that are not plain (a number, a text, yes or no) and values that name a place on a disk are left out.

Appendix: settings follows, with one small table of settings per block. It lists the changed settings, and the unchanged ones except those of the advanced level, which change nothing a reader needs. Of the two stacking methods, it lists only the settings of the one that ran. The appendix pieces have no title of their own, so they are not offered as pictures to copy.

6Copy and export

Everything that leaves the tab is made from what the reader shows.

  • Copy text. The shooting details as plain text, shown in the box of the form: the target, then one line per known detail (“Camera: …”, “Mount: …”, “Frames: …”, “Processing: FLOAT”), nothing for what is unknown. A list chooses the language of this text, French or English, whatever the language of the screen.
  • Copy image. A list chooses the part, and the button puts it on the clipboard as a picture: the session card (the first page), the night cards, then each numbered section. Pictures are made at twice the page’s size.
  • Export PDF. The whole report, on A4 pages, with a foot on each page (the title, the program’s name, the page number). A piece that does not fit the rest of a page goes whole to the next one; none is cut.
  • Export PNG. The whole report as one tall picture.

Both exports propose a file named after the target, ending in _report, in the project’s folder. The status line says when the file is written. The figures use a decimal comma in French.

7Contents, section by section

Table 2.What the report holds, in the order it is read.
PartWhat it holdsWhere it comes from
HeadTarget, position, field and scale; the final image.Astrometry; Post
Table 1, AcquisitionCamera, instrument, mount, filter, location, date, frames and exposure, darks, flats, bias, sampling.The frames’ headers and inventory; the form of the tab
Table 2, Figures measuredUp to nine figures, each with how it was measured.Table 1 of this page
The night, frame by frameStar size, relative signal and sky level of each frame, per session.Registration; Calibration
Calibration to Stars/backgroundA sentence, the state, and the figures of each block of the workflow, each with its meaning.Each block’s own facts
Drift and residualOffset and residual of each frame, in Registration’s section.Registration
Spectral responseRed, green and blue curves, in Color’s section.Color; Catalogs
RecipeThe chain, and the settings changed from the defaults.The stages’ saved settings
Appendix: settingsThe settings of each block, except the advanced ones left at their default.The stages’ saved settings

8Limits, and what was tried

  • It measures nothing. The report reads the files the blocks wrote. A figure that was not measured is not listed, and a section of a block that has not run says so.
  • Details the files do not give stay empty until typed. Instrument, mount, filter, location and date are never guessed. The pasted text holds only what is known.
  • The spectral response is approximate. Star spectra are broad and alike, so different curves explain the same stars about equally well. On synthetic stars the shapes come back close to the true ones, but the response at the Hα line came out 0.76 for a true 0.59, about 30 % off. No error band is drawn: neither resampling the stars nor varying the smoothing measured that error.
  • Sizes in pixels without a scale. If the scale on the sky is not known, the star size is given in pixels, and the figures say so.
  • Outdated results. After a change to the workflow, the report shows the results the blocks had at their last run until the workflow is run again.

References

  1. Finlayson, G., Hordley, S. & Hubel, P. 1998, Recovering device sensitivities with quadratic programming, IS&T/SID Color Imaging Conference 6, 90. doi:10.2352/cic.1998.6.1.art00020