Where do my FWHM and HFR numbers come from?
Every Light frame gets a star catalog: a list of detected stars, each with position, FWHM, HFR, eccentricity, and SNR. From that catalog EigenFrame reports a star count and the median FWHM, HFR, eccentricity, and SNR for the frame, plus two trail figures, directionality and elongated fraction. These are the values you rank by in Cull by metric and see summarized on a frame’s Stars tab.
Directionality measures how strongly the detected stars line up along one axis, from 0 for an ordinary field to 1 for a field where every star’s nearest neighbours sit on one line. It catches trailed frames: a trailed star breaks into a chain of small detections along the trail, each one round enough to count as a star, so the frame reads more stars at a smaller FWHM than its sharp neighbours and its eccentricity looks normal. Directionality is high on frames with straight trails and low on sharp fields, diffraction spikes, and Newton’s rings. Elongated fraction is the share of bright detections that are stretched rather than round, measured before the acceptance settings discard them. It catches trails that wander as well as straight ones, and smeared frames generally. Both appear in the tile’s i details, as tabs in Cull by metric, and as quality thresholds.
There is also a transparency figure: a relative measure of atmospheric transmission, built from differential photometry across matched stars in the set. It catches haze and thin cloud that leave stars looking sharp while the sky dims. Transparency is relative within a set, never an absolute figure, and it lives on the Transparency tab of Cull by metric. It depends on alignment, not just detection: it only appears once a frame has an alignment fit with enough matched stars against the reference, so a frame that’s been detected but not yet evaluated for alignment shows it blank.
When detection runs
Section titled “When detection runs”Detection runs the first time a frame needs its full-resolution detail: opening it in the viewer, preparing it for a stack, or a Targets-page metrics pass across a whole session. That first pass reads the raw, uncalibrated pixel data.
It runs again, this time against calibrated pixel data, once you assign or resolve calibration masters for the frame.
It also runs on demand: open a frame whose catalog isn’t ready yet and its Stars tab shows Detecting stars… while detection completes, typically a few seconds.
Screenshot: Viewer Stars tab with the star overlay and the acceptance sliders.
The star overlay
Section titled “The star overlay”Every detected star draws as a circle sized to its FWHM. Stars that pass the current acceptance settings draw in accent colour; stars that fail them draw muted, so you can see at a glance what the sliders are excluding without leaving the frame. Candidates that fail on eccentricity draw in amber instead, so on a trailed frame the fragments along each streak stand out as the reason the frame was flagged. The Stars tab lists the frame’s directionality and elongated fraction beside the median values, in amber when over the rig’s limit.
Reading the raw-data note
Section titled “Reading the raw-data note”If a frame’s catalog was built before calibration masters were assigned to it, its Stars tab carries an amber note: Detected on raw (pre-calibration) data. The FWHM, HFR, and eccentricity you’re looking at were measured on the frame as captured, not as calibrated, so treat them as provisional. Assigning calibration masters to the frame triggers a fresh detection pass against the calibrated pixels, and the note disappears once that lands.
The acceptance sliders
Section titled “The acceptance sliders”Four sliders on the Stars tab set which detected stars count toward the summary: min fwhm, max fwhm, max ecc, and snr ≥. Moving any of them re-filters the catalog immediately, no re-detection involved, and the accepted count and the median FWHM/HFR/eccentricity below update live as you drag.
Save for rig persists the current slider values to the frame’s equipment group, so every frame on that rig opens with the same acceptance from then on. The app confirms with Saved. Existing frames refresh on re-detection.
Quality thresholds and flagged frames
Section titled “Quality thresholds and flagged frames”Flagging is advisory. It marks frames for your attention and leaves every decision to you: a flagged frame stays in the catalog, stays in your stacks, and stays in alignment exactly as it was. What integration and alignment act on is rejection, which is always something you do yourself.
Quality thresholds live on the Equipment page, per rig: Max FWHM, Max HFR, Min stars, Max ecc, Max directionality, and Max elongated. The first four are blank by default, since a good FWHM or star count depends on the rig, so they flag nothing until you fill one in. Max directionality and Max elongated start at 0.10 and 0.40 on every rig, because trailing looks the same on any optics; clear either field to turn it off. Any frame whose stored metric falls the wrong side of a threshold is flagged, and you see it three ways: an amber ring and a badge on the frame’s tile, an N flagged count on the group header, and a Reject flagged (N) action on the group. Hover the badge to read which thresholds the frame failed, with its value and the rig’s limit; a trail glyph marks frames the trail thresholds caught. The same values turn amber in the tile’s i details. That action is a one-click way to reject exactly those frames, described in culling frames.
A frame is flagged only once stars have been detected on it, since the thresholds are compared against its measured metrics. Rejecting a flagged frame takes it out of the count, because its tile carries the REJECTED overlay in place of the flag marker. Restore it and the flag comes straight back.
Screenshot: Equipment page, the quality thresholds for a rig with Max FWHM and Min stars filled in.
Two settings blocks, two behaviours
Section titled “Two settings blocks, two behaviours”The Equipment page carries both blocks, and they answer different questions.
Acceptance settings (SNR ≥, Min FWHM, Max FWHM, Max ecc) define what counts as a star. They shape the measurement itself, so a frame picks up a change the next time it is detected.
Quality thresholds (Max FWHM, Max HFR, Min stars, Max ecc, Max directionality, Max elongated) define what counts as a good frame. They are compared against metrics already stored, every time the catalog is read, so a change takes effect the moment you save it.
That is why loosening a threshold un-flags frames straight away, while loosening an acceptance slider waits for a re-detection.
Frames you rejected while a threshold was tighter stay rejected when you loosen it. Rejecting is your decision and it stands until you take it back with un-reject.