LAST bit mask description

The images are accompanied by 32-bit mask images. Each pixel contains information about possible problems. In addition, the bitmask are propagated to the source catalogs and stored in the FLAGS column. For each source, the propagation is done using the OR operator for all the pixels found within a 3 pixel radius from the source position.

Bit mask description

Bit name	Bit index	Used	Description
Saturated	0	T	Pixel is saturated.
LowRN	1	T	Pixel noise is low, Var < median(Var) * 0.03; probably a dead pixel.
HighRN	2	T	Pixel noise is high, Var > 20 * median(Var); noisy pixel.
DarkHighVal	3	T	Pixel bias/dark value is high, > 2 * Mean.
DarkLowVal	4	T	Pixel bias/dark value is low, < 0.3 * Mean.
BiasFlaring	5	T	Pixel is possibly flaring, > 20 sigma.
NaN	6	T	Pixel is NaN due to illegal arithmetic operations.
FlatHighStd	7	T	Flat standard error is high, Std / sqrt(N) > 0.01.
FlatLowVal	8	T	Flat normalized value is low, < 0.3.
LowQE	9	T	Pixel has low quantum efficiency, < 0.5 of the nominal value.
Negative	10	T	Negative pixel after processing.
Interpolated	11	F	Pixel value was interpolated.
Hole	12	T	Hole, or anti-star.
Spike	13	F	Diffraction spike.
CR_DeltaHT	14	T	Cosmic ray identified using hypothesis testing against a delta function (see the note below).
CR_Laplacian	15	F	Cosmic ray identified using a Laplacian filter.
CR_Streak	16	F	Cosmic-ray streak.
Ghost	17	F	Ghost.
Persistent	18	F	Persistent charge.
Xtalk	19	F	Cross talk.
Streak	20	T	Streak.
ColumnLow	21	F	Bad column with low values.
ColumnHigh	22	F	Bad column with high values.
NearEdge	23	T	Pixel is near the image edge.
NonLinear	24	T	Non-linear pixel response.
Bleeding	25	F	Possible bleeding/blooming.
Overlap	26	T	Pixel is in an overlap region (of any crop).
SrcNoiseDominated	27	T (coadd)	Source-noise-dominated pixel, (Val - Back) > Back.
GainHigh	28	F	In dual-gain detectors: high-gain state.
CoaddLessImages	29	T	Number of images contributing to the coadd is less than X percent.
SrcDetected	30	F	Source detected; all pixels within a 1-FWHM radius from the detected source.
	31	F	Unused

Note on the Overlap bit (issue #1180)

Since Aug 2026 the Overlap bit is symmetric: every pixel covered by more than one crop carries it in every covering crop (the flagged region is the complement of the EXCLSEC header section). It is therefore purely geometric — do not use it to reject duplicate sources, since it flags all copies; de-duplicate with the catalog primary column instead. Products from before this change carry the opposite convention under the same bit (flagged only in the non-owning crop); the presence of the EXCLSEC keyword / primary column tells which semantics applies.

Note on the CR_DeltaHT bit (issue #1333)

The bit records the outcome of the DeltaHT test — the source is flagged when the S/N measured with a narrow, delta-like template exceeds the S/N measured with the PSF template (SN_1 > SN_2, with no margin). Within the pipeline that test is the definition of a cosmic ray, but it is a sharpness criterion, not a physical identification of a cosmic-ray event: anything sharper than the PSF satisfies it, including hot or noisy pixels and marginal detections close to the detection threshold.

Measured on one LAST visit (LAST.01.07.01, 2026-04-22, 20 epochs, 6 crops): 634–1197 sources are flagged per epoch and crop, against ~600 sources kept in the epoch catalog; 40% of the flags sit on pixels flagged in at least 3 of the 20 epochs; and 826 pixels are flagged in every epoch, of which ~88% already carry HighRN and DarkHighVal. Cosmic rays do not repeat at a fixed detector pixel, so a large part of the flagged population is fixed-pattern detector defects and noise peaks.

Consequences for users:

• To estimate a realistic cosmic-ray rate, this bit alone is not enough. Discard positions that also carry HighRN, DarkHighVal, FlatHighStd or LowQE, and positions that recur across the epochs of a visit; a genuine cosmic ray appears in a single exposure only.
• Sources satisfying the test are removed from the catalog (imProc.mask.maskCR is called with RemoveFromCat=true), so the pixel keeps the bit while the detection itself is gone.
• Bits propagate from the individual images into the coadd with an OR, so a coadd pixel carries the bit if any contributing epoch did. In the visit above this puts the bit on ~20% of the coadd catalog rows, and imProc.calib.fitPhotCalibTrans excludes those sources from the photometric calibration (BadBitNames).

Image quality bit mask (PSTATUS)

A second, unrelated bit mask describes the processing status of a whole image rather than of single pixels. It is written as an integer into the PSTATUS header keyword of every epoch image and of every coadd, and its bits are defined in config/BitMask.ImageQuality.Default.yml (issue #1318). PSTATUS = 0 means that every step listed below produced a result.

Bit name	Bit index	Dec	Set per	Description
NO_BKG	0	1	image	No background and variance: they are empty, or the estimation failed and left them all NaN (issue #1226).
NO_SRC	1	2	image	No sources in the image (an empty catalog).
NO_PSF	2	4	image	No PSF was built for the image.
NO_ASTR	3	8	image	No astrometric solution.
NO_PHOTCAL	4	16	crop	The photometric calibration of the crop produced no zero point.
NO_MERGE	5	32	crop	No merged (matched sources) catalog for the crop.
NO_COADD	6	64	crop	No visit coadd for the crop: fewer good epochs than MinNumCoadd.
NO_RELZP	7	128	crop	No relative photometric zero point: the fit either never ran or returned nothing, and the matched-source magnitudes of the crop are uncorrected (issue #1339).
NOT_GOOD	8	256	epoch	The image did not take part in the coadd and in the matched sources.
FEW_SRC	9	512	epoch	Fewer sources than the coaddition threshold MinNstars (50).
HIGH_BKGRAD	10	1024	epoch	The background gradient over the epoch exceeds MaxFracGrad (1.0).

Reading the values:

• NOT_GOOD is exactly the negation of the pipeline's IsGood criterion, and it always decomposes into at least one of NO_ASTR, NO_BKG, FEW_SRC, HIGH_BKGRAD — the four terms of that criterion.
• NO_SRC means zero sources, FEW_SRC means below the threshold, so an image with no sources at all carries both.
• HIGH_BKGRAD is a property of the epoch, not of the crop: it is the spread of the background over the sub images of that epoch, so it marks all 24 crops of a bad epoch at once.
• The three per-epoch bits are never set on a coadd, so a coadd PSTATUS of 0 does not mean that all of its epochs were good.
• A crop-level bit is copied onto every epoch of the crop as well as onto its coadd.