visplot controls, in full¶
This page is the long form of the descriptions and warnings in the
visplot window’s status area: what a ⓘ there leads to. Pressing
? in the window (or clicking a warning’s ⓘ) opens this page at the
entry for what the status area is showing: see Help in the window.
Some entries are elsewhere in the documentation, where the subject is
already explained, for example Large selections.
The pages under Data formats and access and the other capability pages describe what each feature is for; this page describes each control.
Updated 2026-11-19 (ⓘ and ?).
Selection¶
Antenna and Baseline¶
Tick rows to select.
Antennas ticked: their baselines to every other antenna.
Baselines ticked: exactly those; they override the antennas.
Nothing ticked: all antennas, or all baselines.
The text box under each table sets the ticks from text: for antennas, a name or number such as
DA41,DA43~DA46,0,3,5or!DA42(all but); for baselines, two antennas joined by&(DA41&DV01), or the baseline’s number (12,10~20,!3). Enter replaces the ticks and clears the box; if any part matches nothing, nothing changes. Then press Plot.Prev / Next (◀ ▶) step through the rows one at a time: see Stepping through Field, SPW, Scan, Antenna and Baseline.
Stepping through Field, SPW, Scan, Antenna and Baseline¶
Field, SPW, Scan, Antenna and Baseline each have Prev, ✱ and Next beside their control (Scan: see Scan; it has a text box rather than a table, and its ▶ is the line under the box). Each step plots straight away, and steps wrap around at the ends.
The ▶ marks the row being shown. It is that axis’s current row, and the table scrolls to keep it in view. While a row is current, the status line gives the position, for example “Antenna 4/6: ANTENNA-3”.
Two or more rows ticked: Prev and Next step through only those rows. The ticks stay as they are.
One row ticked: Prev and Next step through every row that has data, and the tick moves with the ▶.
No row ticked: on Antenna and Baseline, none ticked means all, so the ▶ steps through every row with data and the ticks stay empty. On SPW the tick and the ▶ start at the first row and move together.
Stepping ends when the ticks are changed by hand (a click, the text box, ✱, a drill-down from the grid) or when Plot is pressed. Plot always plots what is ticked.
The plot titles name what is being shown when the selection holds one field, SPW, scan, antenna or baseline of several, for example “… — 1642+689, scan 7”: an exported PNG then says which it is.
The grid’s Play steps, in order of preference: the axis with a ▶ (the most specific one, if more than one has a ▶; a current scan counts, between Antenna and SPW); else an axis with exactly one row ticked; else the most specific axis with more than one row to step.
Scan¶
Scan numbers, applied by Plot. The status-area help lists the data’s scans.
1,3,7: those scans.1~7: scans 1 to 7. A range is written with~, not-.>30and<5: scans above or below a number.!5: all scans but 5.!can go before any of the forms above.Several parts are separated by commas.
Empty: all scans.
Stepping through scans. ◀ and ▶ beside the Scan heading plot one scan at a time: the scans the box selects (every scan when it is empty), and only those of the field shown when one field is chosen. A line under the box says which, for example “▶ Scan 7 · 3 of 5”; the box itself is left as it is. Typing in the box, Plot, or ✱ (all of the box’s scans again) ends it. A scan with no data for the chosen spectral windows, antennas or time range plots empty; the next step moves on. See also Stepping through Field, SPW, Scan, Antenna and Baseline.
The text is checked when Enter is pressed and when the box loses the focus. A problem shows as a ⚠ warning in the status area, which stays while other help is showing; text that selects no scan of the data is a problem too.
Time range¶
UTC, applied by Plot. The status-area help gives the data’s time range.
08:00:00~08:30:00: times of day, on the day of the data.2024-05-01/08:00:00~2024-05-01/08:30:00: with the date.>08:00and<08:30: after or before a time.08:00:00+00:10:00: a start and a length.1/00:10:00: day/time, as the time axis labels data that cross midnight.One time alone: the integration around it.
Empty: all times.
UV range¶
Projected baseline length, applied by Plot.
0~50klambda,100~300m,>200m,<1.5km; several with commas.Units: m, km, lambda, klambda, Mlambda (kl and Ml for short); metres when none is given.
Wavelengths are taken at each spectral window’s centre frequency.
Whole rows (one baseline at one time) are in or out.
Empty: all lengths.
Autocorrelations¶
Baselines from an antenna to itself.
Exclude (the default): cross-correlations only. Autocorrelations can be hundreds of times larger (VLBI data before amplitude calibration) and flatten every plot of all baselines.
Include: both. The Antenna × antenna matrix then has its diagonal.
Only: autocorrelations alone: the bandpass shape and the interference at each station.
A baseline ticked by name (for example
BR&BR) is plotted whatever is chosen here.Applied by Plot.
Raster panels¶
Raster axes¶
What runs up (Y) and across (X) the image. Everything else is averaged into each cell.
Baseline × Time: all baselines at once, to find when and where something went wrong.
Time × Channel: a waterfall. One baseline if one is ticked, otherwise the selected baselines combined (see Baselines combined).
Baseline × Channel: the band on every baseline.
Antenna 1 × Antenna 2: one cell per baseline over time and channels, mirrored, so a bad antenna is a whole row and column (the Antenna × antenna preset). The two Antenna axes only go together.
Y and X must differ.
Which correlation is shown is set by Correlation shown.
Correlation shown¶
A raster shows one correlation at a time.
First ticked: the first one ticked in the Data section, named in the title.
Parallel hands: the mean of the two parallel hands, (XX+YY)/2 or (RR+LL)/2: the Stokes I-like view AIPS editors start from. A sample counts only where both hands are unflagged, and a flag box here flags both.
A product: that one, whether it is ticked or not.
Two panels, on RR and on LL, show a problem that is in one hand only.
How it differs from the Data section’s Correlation ticks. The ticks decide what is read: a scatter draws one layer for each ticked correlation. Correlation shown decides what each raster draws, and a raster set to a product reads that product itself, so it can show a correlation the ticks leave out. That lets the two panels show RR and LL side by side while a scatter keeps a single layer, or a scatter show both hands while the raster shows one. With First ticked, the raster simply follows the ticks.
Averaging¶
How a raster cell combines the samples it covers (Amplitude and Phase only).
Vector: average the complex visibilities, then take amplitude and phase. The amplitude drops where the samples do not line up (noise, an uncalibrated phase, a delay across the averaged channels). This is the default in AIPS and plotms, which both offer scalar averaging too.
Scalar: average the amplitudes themselves. Shows the signal plus the noise level, with no loss from decorrelation.
Vector next to Scalar, on two panels, shows where coherence is lost.
See also Averaging.
Baselines combined¶
What a cell shows when it covers several baselines: any raster without Baseline on an axis, such as Time × Channel with no single baseline ticked. Averaging applies to the samples of each baseline; this is what happens to the baselines.
Mean: the average of the baselines’ amplitudes. The general picture of the selected baselines; a problem on a few of them is diluted.
Median: the middle baseline at each cell: what most baselines share, unmoved by a few bad ones. A feature that shows in Maximum but not in Median is on a few baselines (find and flag those); one that shows in both is on most of them.
Maximum: the largest baseline at each cell. Interference or a bad antenna shows even if only one baseline has it; then tick baselines to find which.
Vector (coherent): the baselines are added as complex numbers first. Only for a calibrated point source, where all baselines agree and the noise averages down; on anything else they cancel and the plot reads low. Offered only with Vector averaging.
Applies to Amplitude, Phase (Mean, Median and Maximum all give the mean direction), Amp V Diff and Phase Diff.
A flag box on such a plot flags every selected baseline.
Baseline order¶
How the Baseline axis of a raster is laid out. The baselines that have data are drawn side by side either way.
By number: in the order of the Baseline table; baselines to one antenna sit together.
By length: shortest first, and the ticks read the length. Trouble that grows along the axis is then about distance (the atmosphere, a source that is resolved); trouble in scattered columns is about particular antennas.
The cursor readout names the baseline and its length in both.
Time gaps¶
How a raster’s Time axis treats stretches with no data, such as the time spent on other fields between the scans of the one shown.
Shown (the default): time is drawn to scale. A field observed in short scans with long gaps between them shows as thin bands in mostly blank space. The Time axis then lines up with a scatter of X = Time beside it.
Closed: one row (or column) per time with data, side by side, with one thin blank row where there was a gap, much as the AIPS editors draw time. The ticks read the times, and the axis label says “(gaps closed)”. The cursor readout and flag boxes work in real time: a box drawn across a closed gap covers the scans on both sides, and the gap itself, which holds no data of the selection.
With the gaps closed the axis no longer runs at a constant rate, so it does not line up with a scatter of X = Time beside it. The linked cursor still follows: over the raster it marks the same time on the scatter, and over a scatter’s time in a closed gap it shows nothing on the raster.
Zooming in magnifies the cells already read rather than reading finer ones.
Each panel has its own setting, in its gear tab. Applied by Plot.
Time window¶
The stretch of time each Phase RMS or Coherence value is measured within.
Auto: one integration where Time is a plot axis, one scan where it is not.
Off: no windows.
Scan, or a length: each scan, or pieces of that length; never across a gap.
For Amp V Diff and Phase Diff it is the stretch the reference mean is taken over (Auto and Off mean one scan).
Longer windows are steadier but blur changes in time. On the plot they show as blocks.
Custom… opens a box under the list for any length: a number of seconds, or with a unit (
45,45s,2min,1.5h). Enter adds it to the list and selects it. A value that is not a positive length is reported as a warning and changes nothing.A scatter panel has its own Time window, for its Phase RMS and Coherence points: Auto is one integration when X is Time, otherwise one scan.
Channel window¶
How many channels each Phase RMS or Coherence value is measured within.
Off: on a raster, the whole band where frequency is not a plot axis and single channels where it is; on a scatter, single channels when X is Frequency or Channel, otherwise the whole band.
A number: blocks of that many channels.
Custom… opens a box for any whole number of channels, 2 or more (
12,12ch); Enter adds it to the list and selects it.On a Time × Channel waterfall, Phase RMS and Coherence need a channel window, a time window, or both: without one, each cell holds one sample.
Color scaling¶
How values are spread over the colors. It changes the picture, never the data.
eq_hist uses every color equally and shows faint structure, but distances between colors mean nothing.
linear is honest about size, and lets a few strong values hide the rest.
log and sqrt sit between.
Min and max (or a drag on the histogram) clip the range; threshold highlights everything above a cutoff.
Phase is fixed at −180..180°, linear, with a cyclic colormap.
See also Colormaps and color scaling.
Scatter panels¶
Averaging (scatter)¶
One point per baseline and correlation for each stretch of time and block of channels, instead of one per sample: what a spectrum or a time series needs.
Average over time: Off (every integration), Scan (a spectrum per scan, with X Frequency or Channel), or a length (from the start of each scan, never across a scan boundary).
Average over channels: Off, a number of channels, or All (one point per spectral window: a time series, with X Time).
Vector: average the complex visibilities, then take amplitude and phase. The amplitude drops where the samples do not add up (noise, an uncalibrated phase, a delay across the averaged channels).
Scalar: average the amplitudes.
Amplitude, Phase, Real and Imaginary only. Flagged samples are left out, and a flag box takes every sample behind the points it encloses.
Presets¶
All-baseline waterfall¶
Time × Channel with every selected baseline in one image, like AIPS FTFLG.
Each cell shows the largest amplitude among the baselines (Baselines combined is set to Maximum).
The scatter panel shows Amplitude vs Channel.
A quick survey for interference: one picture instead of one per baseline. A flag box here flags every selected baseline, and the message says how many.
What looks bad here may be on only a few baselines: tick antennas or baselines to find which before flagging them all.
Antenna × antenna¶
One cell per baseline, averaged over time and channels, like the matrix of AIPS LISTR.
The matrix is mirrored, so an antenna that is bad on every baseline shows as a whole row and column.
Hover a cell for its baseline. The diagonal holds the autocorrelations when Autocorrelations is Include.
The scatter panel shows Amplitude vs UV Distance, where that antenna’s baselines stand apart.
Averaging applies: vector averaging over a long stretch reads low where the phase is not calibrated yet. Use Scalar, or narrow the time range.
Flag on Baseline × Time once the antenna is found.
Flagging¶
See also Flagging.
Flag reaches¶
How far each flag (or unflag) goes beyond the box that is drawn. The settings stay until they are changed; the line underneath turns amber while any is wider than the box, and the box is outlined in amber as it is drawn.
Baselines: As drawn; or all baselines to the antenna they share (drag across two or more baselines of the bad antenna; refused if they do not share exactly one); or all baselines to every antenna drawn; or all baselines.
All channels: the drawn time range over the whole band (a bad integration).
All selected spectral windows: the same channels and times in every selected window.
Whole scan: every integration of the scans the box touches.
All fields: matters only on a plot without a Time axis.
All correlations: not just the one plotted.
Baselines, windows, scans and fields are extended for a raster box with the All selected (immediate) filter; other boxes say so.
For one box only, without changing these settings: Shift while dragging gives the full height of the plot as shown, Alt (Option on a Mac) the full width. A key held as the drag starts, or pressed during it, counts for the whole drag, even if it is let go first. Esc before letting go of the button drops the box.
Flag filters¶
What a flag box takes from the samples it encloses.
A box from which the filter takes no sample adds nothing to the pending flags, and the status area says Nothing to flag (or unflag), with the reason: usually “no samples matched filter …”. The filter’s cutoff is then beyond every value in the box: lower it (for Value range, check the numbers against the color bar or axis, whose units it uses), or draw the box with All selected.
All selected (immediate)¶
Every sample in the box: AIPS-style immediate flagging. The other filters take only some of the samples in the box, so a generous box can be drawn round a bad stretch.
Value range¶
Samples whose value, of the quantity the panel shows (Amplitude, Phase, Real or Imaginary), lies in the range. For dropouts (High only) and strong interference (Low only); CASA’s clip.
Low: flag samples at or above this value; empty for no lower bound. In the units of the color bar or axis.
High: flag samples at or below this value; empty for no upper bound. Low and High together take the samples between them.
Outlier from neighbors¶
Amplitudes far from the running median of the 9 samples around them, per baseline and correlation, within each scan. Finds spikes that a smooth change in gain or a bandpass does not explain; CASA’s tfcrop or AIPS’s FLAGR in spirit.
Compare along: time, each sample against the integrations before and after it (spikes in time, brief interference); channel, against the neighboring channels (narrow-band interference, a bad channel).
Cutoff (sigma): how far from its neighbors a sample must be, in robust standard deviations of the baseline’s scatter about them. Lower takes more.
Z-Score above cutoff¶
The robust per-baseline joint (real, imaginary) Z-Score, as the Z-Score coloring computes it: what the Z-Score view shows bright.
Cutoff: the per-sample Z-Score cutoff (3.5, after Iglewicz and Hoaglin).
Match per: sample, each sample against the cutoff; cell, a whole raster cell (the maximum over its reduced dimensions, with a Šidák-corrected cutoff), as the raster coloring shows it; auto, cell on a raster showing Z-Score, else sample.
See also Z-Score.
Amplitude outlier (MAD)¶
|amp − median| / (1.4826 MAD) above the cutoff, per baseline over the whole selection: a baseline’s samples that are off its usual level.
Sigma: how far from the baseline’s median amplitude, in robust standard deviations.
Phase deviation¶
Phase further than the limit from the per-baseline median direction. For a calibrator, where the phases should agree.
Max deviation (deg): phases further than this from the baseline’s median direction are taken.
Grow around flags¶
Unflagged samples next to flagged ones (on disk or pending). Interference has weaker edges that a cutoff misses: flag the core with another filter, then draw a box with this one; CASA’s extend.
Grow along: time, the integrations before and after a flagged sample; channel, the channels either side; both, either.
Samples: how many samples either side of a flagged one are taken.
Export / commit¶
What to do with the pending flags. Nothing is written until Go.
Save as JSON: the flags as a file that can be loaded again here, with their reasons.
Write flags to the MS or Processing Set: exact, checked afterwards, and with a backup.
Write flags with CASA flagdata (when CASA is installed): the CASA route. CASA may apply large or scattered selections incompletely, so prefer Write flags for many points. The result is checked either way.
After writing, visplot reads the flags back and compares them with what it showed. A difference is reported as a warning with the number of samples not flagged, not unflagged, or changed that should not have been; the flags as they were before are kept in the backup file it names (see the next point).
Restore flags from a commit backup (.npz) puts back the flags a write replaced; give the backup file in File.
Writing flags is unavailable when visplot cannot write to the data: no write permission for it, or a reader that cannot write. The warning gives the reason. Save the flags as JSON and apply them where the data can be written.
Load flags from JSON: the flag files saved in the working directory or next to the data are listed, this data’s first.
File: where to write, or what to load or restore. Empty takes a name made from the data name and the time.