17. A dysh glossary

Commonly used terms.

band

A coherent section of channels in frequency space, all with the same channel width. Sometimes called an IF band. Explain L-band, K-band, Q-band, W-band etc. Need a table. See also ifnum

beam

The footprint of one receiver horn on the sky. ARGUS has a 4x4 multi-beam receiver, numbered 0 through 15. Not to be confused with the FWHM. At 115 GHz the FWHM is about 16”, at 86 GHz about 21”. The beam separation is xx” for ARGUS.

Note that for some instruments beams are also interpreted while including other simulteanously taken data in another band/polarization See also fdnum

Beam Switching

This is a variation on position switching using a receiver with multiple beams. The “Main” and “Reference” positions on the sky are calculated so that the receiver is always pointing at the source. This is most useful for point sources.

blanking vs flagging vs masking

flagging defines the rules for blanking data, blanking actually applies them

caloff

Signal with no calibration in the signal path.

calon

Signal with a calibration in the signal path.

CALSEQ

Specific W-band calibration procedure involving a sky, cold1 and cold2 measurement. See GBT memo 305?

ECSV

(Enhanced Character Separated Values) a self-describing ascii table format popularized by astropy. See also https://github.com/astropy/astropy-APEs/blob/main/APE6.rst

fdnum

Feed Number. 0, 1, … Also used as the fdnum= keyword in getXX() See also beam

FITS

(Flexible Image Transport System): the export format for data-cube, although there is also a waterfall cube (time-freq-pixel) cube available. Unclear what we will use for pure spectra. SDFITS seems overly complex. CLASS needs to be supported.

flagging vs. blanking

FWHM

(Full Width Half Max): the effective resolution of the beam if normally given in FITS keywords BMAJ,BMIN,BPA. The term resolution

Frequency Switching

This is a variation on position switching using a receiver where the IF is changed. The “Main” and “Reference” positions on the sky are calculated so that the receiver is always pointing at the source. This is most useful for point sources.

horn

Another term used for beam or pixel.

ifnum

IF number (0,1,…) Also used as the ifnum= keyword in getXX()

intnum

Integration number. 0 being the first. Also used as the intnum= keyword in getXX()

masking vs. flagging blanking

noise diode

calibration gadgets at lower frequencies. See also VANE

OTF Mapping

On-The-Fly (OTF) mapping is where the telescope is continuesly scanned across the sky to sample the emission. The samples are then “gridded” into a map, after calibration.

plnum

Polarization number (0,1,…). Usually 0 and 1, but of course up to 4 values could be present for a full Stokes. Also used as the plnum= keyword in getXX()

Position Switching

This is a standard way to obtain spectra by switching between a “Main” and “Reference” position on the sky, usually using a single beam. For our multi-beam receivers see also Beam Switching

resolution

this term is used in the gridder, but it’s not FWHM, it’s lambda/D. Keyword –resolution= is used If selected this way, FWHM is then set as 1.15 * resolution. But if resolution is chosen larger, what is the effective FWHM? It would be better to have a dimensionless term for resolution/pixel and a different name for resolution alltogether.

RRL - Radio Recombination Line

Scan - GBT differentiates between different types of scans

(FSScan, PSScan, TPScan, SubBeamNod Scan)

ScanBlock - GBT. A container for a series of scan’s

SDFITS

Single Dish FITS format, normally used to store raw or even calibrated spectra in a FITS BINTABLE format. Each row in a BINTABLE has an attached RA,DEC (and other meta-data), plus the whole spectrum. This standard was drafted in 1995 (Liszt), and has been implemented by many telescopes (Arecibo, FAST, GBT, Parkes, ….)

SFL

Sanson-Flamsteed projection, used in LMT FITS files (the GLS - GLobal Sinusoidal is similar to SFL).

Spectral Window

In ALMA commonly abbreviated as spw, this is closest to what we call a bank, or band, a set of linearly spaced channels.

Spectrum

A coherent section in frequency space, with its own unique meta-data (such as polarization, ra, dec, time). Normally the smallest portion of data we can assign. A spectrum is defined by its own seting of (crval, crpix, cdelt) in a FITS WCS sense. See also Data Dimensions.

SubBeamNod

Another scan mode

VANE

That thing

VANECAL

Calibration procedure for ARGUS

Waterfall Plot

Tyically

Window

See Spectral Window

17.1. Overloaded Terms

Terms used in the code may not exactly match terms used by the develpers of the instruments. Here we clarify those overloaded terms in the form of a table

Table of some overloaded terms

code term	RSR term	SLR term	comments
beam	pixel?	pixel	multi-beam receiver
cell	n/a	cell	size of a sky pixel in gridding, usually 2-3 times smaller than the resolution
band	board	bank	spectrometer window
n/a	chassis	n/a	tuple of (pol,beam)
channel	channel	channel	with a simple FREQ WCS{crval,crpix,cdelt}

17.2. Data Dimensions

This section is not meant to describe either the data format, but the mental storage model we have in mind to be encapsulated in a Python class.

A unified data storage of LMT spectra would (naturally) break up the spectra, such that each spectrum has a different time, beam, band, polarization, etc. Each spectrum can be described as a set of sequential channels, described with a single (crval,crpix,cdelt)) WCS. In Python row-major array notation where the most slowly varying dimension comes first this could be written as an NDarray:

data[ntime, nbeam, npol, nband, nchan]

where we added the ntime and nchan as the slowest resp. fastest running dimension in this row-major (python/C) notation.

Note

For those used to GBTIDL plnum = npol, ifnum = nband, and fdnum = nband. Arguably different scans can act as as ntime, although each scan will often have several snapshots inside of them. ?? intnum

Overloaded words, including GBT lingo:

plnum   pol
fdnum   feed     beam    pixel
ifnum   window   band

Taking out those an observation can be seen as a set of spectra:

spectrum[nbeam, npol, nband]

This exactly matches the concepts used in an SDFITS file, although in the general definition of SDFITS there is no assumption of the data being able to be stored in an NDarray type array, where the more general

sdfits_data[naxis2, ndata]

where in general ndata=nchan, but dialect with ndata = npol * nchan are seen in the wild (FAST, Parkes). The FITS name naxis2 is the number of rows, which is the product of time,beam,band,pol in our case.

Taking an inventory of current and known future LMT Spectral Line instruments:

• RSR: two beams, two pols, 6 bands, though the term chassis is used to point at any tuple of (beam,pol). So here we have nbeam=2, npol=2,nband=6, nchan=256 and ntime it typically 10-20. Each beam happens to look at the same sky position here.

Note

If an instrument like RSR would multiplex the (beam,pol) pairs, this would be a challenge to the assumption of homogeneity, and the SDFITS model would be more appropriate.

• 1MM: one beam, two pols, two sidebands. So here we have nbeam=1, bpol=2, nband=2, nchan=2k

• SEQ: 16 beams (though 4 beams per roach board, and each roach board has its own time) in one band (they also call it bank) and one polarization. Thus nbeam=16, npol=1, nband=1. Once the 2nd IF will be installed, 32 beams will be recognized by the software, but organizationally it is easier to to think of 16 beams and 2 bands.

Note

The timestamps for the different roach boards make it impossible to store the data in a multi-dimensional array, unless (typicall one) integration is removed. Keeping all data would require data[ntime4, 1, 1, 1, nchan] for SEQ.

• OMA 8 beams, 2 bands (banks), 2 polarizations.

• B4R 4 XFFTS boards, 2.5 GHz/board: 1 beam, 2 bands (USB and LSB), 2 polarizations (XX and YY)

Note that FAST is the only known case that stores data as data[ntime, nchan, npol], where nchan is not the fastest running dimension, but npol. Technically this appears to be the case such that they can vary nchan per row.

We thus arrive at the following summary for the multi-dimensional data[] array:

data[ntime, nbeam, npol, nband, nchan]

in the table we leave out the ntime dimension

Table of data dimensions of LMT SLR instruments

data	nbeam	npol	nband	nchan	comment
RSR	2	2	6	256	(pol,beam) tuples are the 4 chassis. 6 overlapping bands make one final spectrum
SEQ	16	1	1 (2)	2k, 4k, 8k	beams have time issue, perhaps ntime ~ ntime * nbeam, and nbeam=1. Future will have 2 bands
OMA	8	2	2	2k, 4k, 8k	Future instrument, with 4 more roach boards (USB+LSB)
1MMRx	1	2	2	2k, 4k, 8k	band: 2 IF’s in USB/LSB
B4R	1	2	2	32k	Japanese 2mm receiver

17.3. Single Dish Math

The meat of Single Dish math is getting the system temperature

\[T_{sys} = T_{amb} { { SKY } \over { HOT - SKY } }\]

or

\[T_{sys} = T_{cal} { { <SKY> } \over { <HOT - SKY> } } + T_{cal}/2\]

where the \(< >\) operator averages over the center 80% of the spectrum. This way \(T_{sys}\) is a scalar. The routine meantsys computes this.

and using this system temperature, calculating the signal by comparing an ON and OFF position, assuming there is only sky in the OFF:

\[T_A = T_{sys} { { ON - OFF } \over {OFF} }\]

All of these have values for each channel. How exactly the \(T_{sys}\) is computed (scalar, vector, mean/median) is something we generally leave open.

17.4. Observing: ObsNum / SubObsNum / ScanNum

An observation with a single dish such as LMT is done via proposals, which gets assigned a proposal ID, associated with the P.I. name. An example of such is 2018-S1-MU-46

An observation is that divided in a set a ObsNum ‘s, which can be hierchically divided up in SubObsNum’s and ScanNum’s. When an observing script executes, each source will gets its own ObsNum, though calibration data often gets another ObsNum.

17.5. Band Designations

Nomenclature comes from the IEEE radar band names, but there is also a NATO nomenclature standard. See e.g. https://en.wikipedia.org/wiki/Radio_spectrum

There is also a list of instruments at GBT clarifying this list on https://greenbankobservatory.org/portal/gbt/instruments/

• L: 1-2 GHz

• S: 2-4 GHz

• C: 4-8 GHz

• X: 8-12 GHz

• Ku: 12-18 GHz

• K: 18-27 GHz

• Ka: 28-40 GHz

• V: 40-75

• Q: 33-50 GHz or 6-9.1mm

• W: 84-116 GHz (also: 75-110)

17.6. Common Parameters

• gbtfits~fits.sdfitsload.SDFITSLoad

  input SDFITSLoad object

• fdnumint

  the feed index

• ifnum: int

  the IF index

• IF bla bla

• intnum

• plnum: int

  the polarization index

• feedsint list of two beams

  the two nodding beams

• bintableint

  the index for BINTABLE containing the scans

• selection~pandas.DataFrame

  selection object

• observer_location

• fitsindexint

  the index of the FITS file contained in this GBTFITSLoad. Default:0

• scansint or 2-tuple

  The scan(s) to use. A 2-tuple represents (beginning, ending) scans. Default: show all scans

• scansarray-like

  list of one or more scans

• channelnumber, or array-like

  The channels to operate on

• weights: str

  ‘tsys’ or None.

• overwritebool, optional

  If True, overwrite the output file if it exists. Raises an OSError if False and the output file exists. Default is False.

• # TPScan(

• scan: int

  scan number

• smoothref: int

  the number of channels in the reference to boxcar smooth prior to calibration

apply_flags : boolean, optional. If True, apply flags before calibration.

• # PSScan

• scansdict

  dictionary with keys ‘ON’ and ‘OFF’ containing unique list of ON (signal) and OFF (reference) scan numbers NOTE: there should be one ON and one OFF, a pair

• # NodScan

• scandict

  dictionary with keys ‘ON’ and ‘OFF’ containing unique list of ON (signal) and OFF (reference) scan numbers NOTE: there should be one ON and one OFF, a pair. There should be at least two beams (the nodding beams) which will be resp. on source in each scan.

• # FSScan(ScanBase)

• scanint

  Scan number that contains integrations with a series of sig/ref and calon/caloff states.

17.7. Finally

we shuld ahev some closing words.