Seeing Monitor

Change log

3.0.11 2026-09-21

• Fixed: the upload frequency could not be changed. In Tools → Remote Upload Setup..., clicking the “Every ... minutes” option appeared to select it but left “Every cycle” selected as well, and the minutes box stayed greyed out so the number could not be edited. Underneath that the choice was never saved: uploads kept running after every cycle whatever you picked. If you set a slower upload frequency in 3.0.10 or earlier, it was not being applied — set it again after upgrading and it will stick this time. The setting has behaved this way since it was added in 3.0.2.

3.0.10 2026-09-19

• New: Image FWHM and Stacked FWHM are now saved in the seeing log, as two extra columns after the existing twelve. Image FWHM is the star's width in a single short frame; Stacked FWHM is its width once the last few seconds of frames are summed without re-aligning them, which is the value the Tonight graph plots. Both are in arcseconds, and both are left empty rather than written as 0.00 on a cycle where there was no measurement.
• The first twelve columns are unchanged — same names, same order, same number formats, same meanings. If you have your own scripts reading the log, have them accept at least twelve fields rather than exactly twelve. Rows written before you upgrade keep twelve fields and rows after have fourteen, and both can appear in one night's file: the header of a file already being written is deliberately left alone, because other software may be reading it at that moment. See the User Guide under Tools → Logging Setup... for the full column list and what the two new ones mean.
• The Seeing History window's database gains the two new fields in place, keeping every night you already have. Nights logged before 3.0.10 simply have no value for them, which is stored as “missing” rather than as zero — a zero would quietly distort any average taken across old and new nights together.
• Seeing History opens and switches views far faster. The window used to re-read every log file in your log folder each time you opened it, changed the dates, changed the grouping or switched between Table, Trend and Calendar — even when nothing on disk had changed. On a folder holding a few years of nights that was several seconds of frozen window every time. It now re-reads only the files that have actually changed since it last read them. Measured on a 515-file, 254,310-row log folder: a refresh that imports nothing went from about 8.8 seconds to well under a tenth of a second. The first import after upgrading is the usual full one.
• Cycles where no star was found are now kept. Some rows in older logs record a cycle that ran and found nothing to measure, and leave the star-count column empty. Those rows could not be read and were dropped, so a cloudy stretch looked exactly like a stretch where the monitor was not running. They are now imported and marked as having no valid seeing. On the log folder above this recovered 2,015 rows across 65 nights.
  Your seeing numbers do not change. These rows carry no seeing measurement, so they are never averaged, never plotted, and never counted as a sample. Checked across every night, week, month, season and year in that folder: every mean, median, standard deviation, P10, P90, best and worst value is identical to before. What does change is the Invalid count on affected nights, which now includes them, and the Time span for 28 periods, because the night is now known to have started earlier or ended later than the first or last successful measurement.
• Nights that were observed but measured nothing are now visible. Previously a night whose rows all failed the quality checks was omitted from Seeing History entirely, so it looked identical to a date you never observed. Such a night now appears in the Table with its stored-record count and an em dash in every seeing column, and in the Calendar as its own kind of cell — outlined, labelled with the record count, and deliberately given no colour, because there is no seeing quality to report. It is not given a seeing value of zero and no cause is claimed for it: the records show that the instrument ran and measured nothing usable, not why.
• New: Image FWHM, Stacked FWHM and star brightness on the night chart. Drill into a single night in Seeing History and a Show: row under the chart lets you add any of the three alongside seeing. Seeing is always drawn and is still the primary series. The two FWHM values get their own scale on the right of the chart, so they cannot be mistaken for the seeing measurement, which is a different quantity measured a different way. Star brightness draws as a shaded band showing variation within that night only — it is not a calibrated transparency figure and is labelled as relative.
  Nights logged before this release have no FWHM values, so on those nights the two FWHM options are switched off and say why. They will start filling in from now on.
• New: click a point on a night's chart to see the whole stored measurement. A panel beside the chart shows that cycle's time, seeing, both FWHM values, star counts, background, X and Y centre, X and Y RMS, the 5×5 and 9×9 box values, whether the reading was valid, and which log file and line it came from. Anything the row did not record says “not recorded in this row” rather than showing a zero.
  If two stored records sit at the same point — which happens when a spare copy of a log file is left in the log folder alongside the live one, since both get read — the panel lists them by file and line so you can look at each. Nothing is merged, hidden, or picked as the correct one.
• The Seeing History table's “Coverage” column is now called “Time span”. The number is unchanged: it is the time from the first stored observation in the period to the last, including every gap in between. It was never a measure of how much of the period was observed, which is what “Coverage” suggested. Hovering the column header now says so.
  This also renames the column in the CSV export. If you have a script or spreadsheet that looks for a column headed “Coverage”, it needs updating. Every other column keeps its name, position and format.
• Fixed: hovering a point on a multi-night chart showed the period's median value under the label “Average”. The table has always listed Mean and Median as separate columns, so the tooltip disagreed with the table about the same period. It now says “Median”.
• Worth knowing: the ◀ and ▶ arrows beside the date range jump to the nearest night that has a seeing measurement, keeping your date range the same length. They step over a night that recorded no valid seeing, so use the Calendar or Table view to find those.
• New: Help → ASCOM Driver… The optional ASCOM driver — which lets N.I.N.A., PHD2 and other ASCOM clients read your seeing value as an ObservingConditions device — now has its own entry in the Help menu. It reports whether the driver is installed, which version you have, and whether a newer one is available, and offers to open the download page. The driver remains a separate download with its own installer; Seeing Monitor never installs or updates it for you.
• Check for Updates now also mentions the ASCOM driver, but only if you already have it installed and a newer version exists. If you don't use ASCOM, nothing changes.
• Fixed: Tools → Night Schedule could refuse to open, with the error “Value of '60' is not valid for 'Value'”. Saving a latitude or longitude whose seconds were zero — 32° 55′ 00″, say — left the dialog unable to reopen: the saved position was being split back into degrees, minutes and seconds as 54′ 60″, and the seconds field will not accept 60. It now rounds and carries correctly, so the dialog opens on any position. The seconds fields also accept 59.1″ to 59.9″ now, which they previously refused — the last nine tenths of every arcminute were untypeable. Nightly auto start/stop itself was never affected — it reads the saved position directly and kept running normally; only the settings dialog would not open. This affected earlier versions too, not just 3.0.10.

3.0.9 2026-09-06

• A new star-centre measurement. An ST-i user running this app alongside their previous seeing-monitor software saw this app report 11.06″ where the other read 2.39″ on the same star, same night, identical optics settings. The cause was this app's hot-pixel filter, which on a tightly-focused (undersampled) star discarded the star's own brightest pixel - differently from frame to frame - and that flicker was being read as star motion. The measurement now uses a star-sized measuring box (sized in arcseconds, so it fits any camera), a background sampled right beside the star, and a plain brightness cut-off instead of a neighbour test. Validated over five nights and more than a thousand cycles against the previous method running on the very same frames.
• Expect a small step down in your numbers. On rigs like the project's own QHY5L-II the new method reads about 3-4% lower than the old one on clear nights (it no longer adds a little of its own noise to the real atmosphere), and a good deal lower on rigs where the star covers only a pixel or two, such as the ST-i with a short lens. Historical entries are unchanged, so there will be a visible discontinuity at this version.
• Cloud can no longer produce wild readings. The "is there really a star here?" test now works in the camera's native units, detected from the data (12-bit cameras such as the QHY5L-II deliver values 16x larger than they look). On frames where cloud has blinked the star out the measurement now correctly says "no star" - the same condition SBIG-era log files mark with a 2000.00 entry - instead of inventing a position. Within a cycle, frames where the star was under 30% of its usual brightness are dropped before the result is computed, and a cycle where the star never tracked steadily (more than 10% of positions rejected as jumps) is skipped rather than logged. All brightness thresholds also scale to the camera's measured pixel noise, so an empty patch of sky can never add up to a "star".
• Hot pixels can no longer capture the star search. Isolated defective pixels are masked before the star is located, the search looks near where the star was last seen rather than sweeping the whole frame every time, and after any skipped or "star too dim" cycle the star is re-found from scratch. A wrong lock can no longer survive more than one cycle.
• The "star too dim" check adapts if your equipment changes. Its reference for Polaris's unobscured brightness was previously the brightest cycle ever recorded and could only climb; a refocus, dew, or a gain change would leave it permanently too high. It is now learned from a rolling window of recent well-tracked cycles (about two to three clear nights' worth), and it carries over between nights. On first launch after upgrading it re-learns from scratch, and the check pauses until it has seen twenty good cycles - about twenty minutes of clear sky.
• Fixed for 16-bit cameras (ST-i): logged StarCts and Bgrnd. Since 3.0.8 these were divided by 16 for every camera, which is right for the QHY but put ST-i users' values 16x below their own SBIG-era logs. They now divide by the detected scale. A status line at the start of each run names which case your camera is.
• New: Help > Diagnostic Logging. Off by default. When ticked, a second file is written beside each night's log (SeeingLog_MMDDYY_centroid_compare.csv) recording per-cycle diagnostics: the result the previous star-centre method would have given on the same frames, frames with no star, frames rejected, frames dropped as too dim, the measured pixel noise, whether the dim check fired, and why any cycle was skipped. Your main log is unchanged in format. Leave it off unless you have a question about your readings; then turn it on, run a night, and send the file with the log - it is how every problem in this release was found. About 50-100 KB per night.
• Hardening: the once-per-cycle settings save can no longer discard a cycle's measurement if the file is momentarily locked, and settings dialogs can no longer collide with it. User Guide: the "Reject hot-pixel spikes" option now correctly says it applies to the comparison method only.

3.0.8 2026-08-27

• Logged StarCts and Bgrnd are now in the same units, measured the same way, as the values in SBIG-era log files, so they're directly comparable again (typical healthy values: Bgrnd ~175-220, StarCts ~15,000-22,000). They are now measured on the sensor's native 12-bit scale, with the background sampled locally right next to the star - this app had been logging raw 16-bit values (16x larger) from a whole-frame estimate, which is why its numbers looked wildly inflated by comparison even though the camera and sky were fine. Seeing measurements were never affected - this was purely about the two logged brightness columns.
• Camera gain for monitoring is back to 8 - v3.0.7-beta1's revert to 1 was based on comparing this app's raw-unit star counts against 12-bit-unit historical values, which made gain 8 look 25x too bright when it was actually right. Verified against real hardware on the stable QHY driver: gain 8 with the camera's own default black level reproduces the trusted historical numbers almost exactly.
• The camera black-level (Offset) is no longer written at connect. The camera's factory default IS the trusted baseline. (The brief 3.0.7-beta Offset management was a countermeasure for a register drift that only exists on QHY's beta driver - if you're on a beta driver from 2025 or later, downgrade to the recommended stable release, v24.12.27.10 as of this writing.)
• The dim-star reference brightness re-learns from scratch on first launch (the previously learned value was in the old units). The Focus/Alignment window's gain control now starts at the monitoring gain instead of 1, so the star you see while focusing is the star monitoring will actually measure.

3.0.6 2026-08-27

• Fixed the QHY camera gain used for monitoring defaulting to 1 instead of 8 - an end user comparing star counts against their historical logs from the same QHY camera found this app's readings noticeably off; gain 8 reproduces those values. Existing installs migrate automatically on next launch (no user action needed).

3.0.5 2026-08-25

• Added a transparency band and a Transparency % stat to the Refined Line and Instrument Panel chart styles (independently toggleable in Choose Chart Style...) - the same relative star-brightness band the Seeing & Transparency style already had, plus a percentage reading of current star counts against the reference value from Configure Camera / Optics....

3.0.4 2026-08-23

• The seeing plot's vertical scale is now fixed (default 0–7", configurable in Tools → Options...) across the Tonight graph, Seeing History Trend chart, the Seeing Detail drill-down popup, and the plot snapshot, instead of auto-fitting to each night's data - a reading above the top now clips flat rather than stretching the range.
• The Tonight graph's stacked FWHM line can now be hidden (Tools → Options...) if it isn't useful for a given setup - the underlying value is still measured and logged either way.
• Added a fourth plot snapshot chart style, Seeing & Transparency, mirroring the main window's own Tonight graph: a Seeing trace over a relative star-brightness/transparency band.
• The Instrument Panel snapshot style now plots discrete points instead of a connected line.
• The Median/Average/Lowest/Highest reference-line checkboxes moved from Tools → Options... into Choose Chart Style..., next to the Refined Line and Instrument Panel styles they actually apply to.
• Updated the User Guide's explanation of Stacked FWHM, and added ASTAP star-database sizing guidance for a typical field of view.

3.0.3 2026-08-22

• Remote Upload Setup's plot snapshot and seeing log uploads can each now optionally be uploaded under a fixed name (e.g. always "SeeingGraph.png" instead of a dated filename) so a server always has one stable file to point at - the local file's own (dated) name is never affected. The extension is fixed to .png/.txt so it can't be pointed at the wrong file type.

3.0.2 2026-08-21

• Added a Plot Snapshot feature (Tools → Configure Plot Snapshot...): automatically saves tonight's chart as a PNG to a folder of your choosing after every cycle, for publishing current conditions to a website or comparing across a small network of sites. Choose from 3 chart designs (Refined Line, Instrument Panel, Stacked Tile), 2 export sizes, and a snapshot theme independent of the app's own Dark Mode setting - all previewed live before you commit to a combination.
• Added Remote Upload Setup (Tools → Remote Upload Setup...): pushes the plot snapshot and/or the current seeing log straight to FTP, FTPS, SFTP, or S3 after every cycle (or throttled to every N minutes), with independent toggles for what gets uploaded and encrypted credential storage.
• Added a nightly NightlySummary_MMddyy.txt file (cycle count, mean, median, min, max) written automatically whenever monitoring stops, matching the nightly summary file found alongside SBIG-era logs.
• Added "Launch Seeing Monitor automatically when Windows starts" (Tools → Options...) - registers the app to start at login, no admin rights required. Pairs with the existing Night Schedule auto-start.
• Added a stacked image FWHM diagnostic to the Tonight graph - averages a configurable window of recent frames for a truer sense of what a real multi-second exposure would show, versus any single ~0.01s frame.
• Dark Mode now covers every settings dialog in the app (previously only the Seeing History window), including native window title bars.
• The Seeing History window gained a five-band color-coded quality scale with a clickable filter legend, Trend/Calendar hover tooltips, and Save as PNG / Export to CSV on every view.
• Refined the chart color palette for better contrast and colorblind accessibility across Seeing History, Trend, Calendar, and the plot snapshot.
• Fixed a cycle where the star was too dim to trust ("dim lock") plotting as a fabricated 0.00" reading instead of an honest gap in the live graph and history.
• Fixed the Seeing History table sorting oldest-first by default; now sorts newest-first.

3.0.0 2026-08-07

• Added local historical seeing storage backed by SQLite.
• Added the Seeing History window with date filtering and historical summaries.
• Existing text logs remain unchanged and are imported safely for long-term analysis.

2.2.0.0 2026-07-31

• Added support for Player One USB cameras via a new Connect → Player One... menu item, built on Player One's official SDK. Like SBIG (and unlike the bundled QHY/ZWO SDKs), Player One's own USB driver needs to be installed separately first - download it from Player One Astronomy's website. Verified connecting and streaming live frames against a real Player One camera in Focus/Alignment mode at the observatory.

2.1.0.0 2026-07-31

• Added a minimum star-brightness gate. A geometrically plausible lock can still be riding on a thin cloud or twilight glow dimming the star - in that case the RMS is mostly sky noise, not real atmospheric jitter. If a cycle's average star brightness falls below a configurable floor (Tools → Configure Camera / Optics..., "Min star counts", off by default), that cycle's seeing/RMS values are zeroed rather than trusted, and flagged "dim lock" in the recent-cycles list. The row is still logged (same convention as the existing noise-floor/seeing-floor values) - no log column changes, since StarCts is already a logged column that makes a zeroed-but-dim row distinguishable from a genuinely calm-sky one.
• The hot-pixel guard (Tools → Options...) now also masks isolated stuck/hot pixels anywhere within the centroid window, not just at the initial seed search - a spike elsewhere in the box could previously still skew the weighted centroid even after a clean star was correctly found.

2.0.0.0 2026-07-31

• Replaced the seeing calculation with the real Kolmogorov single-aperture image-motion formula instead of an empirical approximation. Earlier this same night this project tried an empirically-fit constant (K), then a bare plate-scale conversion, then rolled that back - the formula now used is the standard Fried/Sarazin-Roddier single-aperture derivation: r₀ = λ(θ²/0.179 × (D/λ)^(1/3))^-0.6, seeing = 0.98λ/r₀, then zenith-corrected.
• Added Aperture Diameter as a new required optics parameter (Tools → Configure Camera / Optics...) - a genuine input the formula needs that this project never captured before. Defaults to 30mm, matching the ZWO 30mm Mini Guide Scope already in use.
• Added a "Use combined X+Y motion" option (off by default). The formula's 0.179 coefficient is calibrated for single-axis motion, so X-axis-only is the direct default; combining both axes (theoretically sound, not just an average of the two RMS values) is available as an option. Both modes are always computed regardless of the setting (see "Alt. axis mode" in the recent-cycles list) so the two can be compared against real conditions before committing to either one exclusively.
• Zenith correction is back to being applied directly as part of the primary seeing value rather than a separate diagnostic - the earlier "double-application" concern from this same night doesn't apply to this formula, which has no zenith correction implicitly baked in anywhere else.
• Retired the photon-noise/gain-based dynamic noise floor diagnostic added and removed earlier this same night.

1.8.0.0 2026-07-30

• Added Help → Check for Updates..., which checks a small version manifest hosted online and offers to open the download page if a newer release is available. Requires an internet connection - if the check itself fails (no connection, server unreachable), you'll see an explicit error message rather than a false "you're up to date."

1.7.0.0 2026-07-29

• Added support for SBIG USB cameras (developed against the ST-i) via a new Connect → SBIG... menu item. Requires SBIG/Diffraction Limited's own driver installed separately - it must be the 64-bit driver, since Seeing Monitor runs as a 64-bit process and can't load a 32-bit one. If unsure which is installed, download SBIGDriverChecker64 and run its update checker. Unlike the bundled QHY/ZWO SDKs this driver isn't shipped with Seeing Monitor, and its installer doesn't register it on the system search path either - Seeing Monitor automatically finds and loads it from wherever the Driver Checker installed it, so no manual file copying is needed. Verified that the driver loads and responds correctly on a real installed Driver Checker; not yet tested against a real ST-i.

1.6.0.0 2026-07-24

• Fixed Focus/Alignment's Solve & Overlay: re-solving after adjusting alignment could silently reuse the previous solve's leftover result and leave the overlay frozen in its old position when the new solve failed, instead of reporting the failure.
• Fixed the night-schedule auto-start being able to kick in while Focus/Alignment mode was open, colliding with its live view on the same camera - monitoring now defers auto-start until Focus/Alignment is closed.
• Focus/Alignment's image view no longer runs flush against the window edges.

1.5.0.0 2026-07-22

• Fixed the "Recent cycles" history box not clearing on (re)start - a scheduled restart no longer leaves the previous session's rows sitting above tonight's new ones.
• Fixed monitoring being able to lock onto the wrong thing on a fresh start or reacquire (a stray hot pixel, satellite, or twilight glow) and stay stuck there indefinitely - since that kind of lock never technically "fails," it never tripped the existing lost-lock recovery. A freshly acquired lock now has to be confirmed by the next frame at roughly the same spot before it's trusted.

1.4.0.0 2026-07-21

• Main window's night graph now also plots Seeing and star brightness alongside Image FWHM, spanning the whole night (dusk to dawn) on a horizontal time axis that's reserved up front rather than growing/rescaling as the night goes on.
• Fixed the night graph silently drawing a straight line across any gap in the data (e.g. a cloud-out) - gaps now render as genuinely empty stretches instead of looking like smoothly interpolated data.
• Fixed monitoring getting permanently stuck rejecting the real star as an "implausible jump" after a long stretch of lost lock (e.g. clouds clearing after an hour or more) - the star-position reference now resets after a couple of consecutive lost-lock cycles so it reacquires automatically instead of needing a manual restart.
• Monitoring is now resilient to an unexpected error in a single cycle - it's logged and skipped rather than silently ending the whole night's run. If monitoring does ever stop unexpectedly, the status bar now says so explicitly instead of looking identical to ordinary cloud cover.

1.3.0.0 2026-07-20

• Camera capture switched to continuous/streaming mode for both QHY and ZWO cameras instead of a fresh single-shot exposure every frame - fixes both a ~2fps Focus/Alignment live view and monitoring cycles that had drifted to ~144s instead of the original ~66-67s.
• Monitoring cycles are now explicitly paced to still span ~66.5s, matching the cadence the seeing calibration constant was originally fit against, rather than racing through as fast as the faster capture now allows.
• Added a persisted camera gain setting for monitoring - previously monitoring used whatever gain the camera was last left at.
• New "Show Polaris/NCP reticle (predicted)" option in Focus/Alignment mode: a polar-scope-style circle showing Polaris's orbit around the true pole and where it currently sits on it, calibrated from a plate solve (or typed in manually) and predicted forward without needing to keep re-solving.
• Fixed several plate-solving accuracy issues found via real-sky testing: a missing field-of-view hint that let ASTAP occasionally lock onto the wrong star pattern, a parsing bug that mishandled ASTAP's scientific notation, a position search hint that was actively steering solves to an incorrect result, a sky-to-screen projection error, and a coordinate-epoch mismatch between Polaris's catalog position and the plate solver's reference frame. Solving now takes a few seconds longer but the result is verified accurate against independent tools.

1.2.0.0 2026-07-18

• Focus/Alignment gain control now has a numeric up/down box paired with the slider, same as the exposure control, for typing an exact gain value.
• Default gain changed from 20 to 1.

1.1.0.0 2026-07-17

• Focus/Alignment exposure control reworked: range extended from 1ms-1000ms to 0.01ms-10,000ms (10 seconds), with a logarithmic slider so low exposures (where most time is spent) aren't squeezed into a sliver of travel, plus a paired numeric box for typing an exact value directly.
• Gain controls moved to their own row below Exposure, and the raw pixel min/max and 5x5/9x9 focus-quality readouts moved to a row of their own below that, fixing a layout clipping issue.
• Reworded the "plate solver not configured" status message to point at the actual menu (Tools > Configure ASTAP...) instead of an internal setting name, and wrapped it so it isn't clipped.

1.0.0.0 2026-07-17

• Initial deployment.