This page contains detailed technical specifications for the SALSA Onsala telescopes. There are three telescopes: Torre, Vale, Brage (in order from left to right as displayed on the webcam). In principle they have identical hardware, although there are differences in cable lengths and connections. The diagrams to the right show the signal path from the sky to a spectrum (left column, top to bottom) and the control path from the webserver to the rotor (right column, bottom to top). The telescopes are located at Onsala Space Observatory, 57°24'N, 11°55'E, approximately 5 m above sea level.
SALSA can observe in two modes: a raw total-power mode, and a frequency-switched mode in which a reference spectrum at an offset frequency is subtracted. In frequency-switched mode the output is the normalized difference (sig − ref) / ref, which is scaled to antenna temperature via Ta = Tsys · (sig − ref) / ref. The scale factor Tsys was fixed by comparing a SALSA spectrum toward a bright HI region with the same region from the Bonn AllSky HI survey (GASS/LAB), convolved to SALSA's 6° beam. Since the Bonn survey is on a calibrated brightness-temperature scale, matching the two (allowing for main-beam efficiency) yields Tsys ≈ 300 K.
This is an approximate, single-point calibration and the absolute scale should be treated with caution. It assumes a fixed Tsys, whereas the true value varies with elevation, ambient temperature and time, and it depends on the poorly known main-beam efficiency of the small holey dish and on stray radiation entering the sidelobes from sky-filling Galactic HI. A further caveat is the convolution itself: the Bonn server specifies that the requested effective FWHM should be below 1°, so asking for SALSA's 6° beam goes outside the survey's intended range — the resulting profile may miss flux or rest on an unreliable spatial mapping, adding further uncertainty to the comparison.
Velocities measured from the 21 cm line are first relative to the telescope, which moves with the Earth and Sun. For galactic targets SALSA converts these to the Local Standard of Rest (LSR) frame so they are directly comparable with published HI data (see the experiments page for the concept). The correction is the sum of two terms: the Sun's 20 km/s motion toward the standard apex near Vega (RA 18ʰ, Dec +30°), projected onto the target direction, and the Earth's orbital motion about the Sun, approximated as a 30 km/s circular velocity. No correction is applied to non-galactic targets (Sun, satellites).
SALSA uses the classical "Standard Solar Motion" to define the LSR — the same convention as legacy HI surveys such as LAB — so velocities are directly comparable with those maps. The model is accurate to well within the standard ~1 km/s resolution.
The dish is 2.3 m in diameter. The surface has holes of about 5 mm in diameter, letting air through to reduce wind resistance. The holes do limit the frequency range (too short waves pass through rather than reflect) to less than about 6 GHz, but this is not a problem for the longer waves observed by SALSA. The horn is placed at a focal distance optimised for the 21 cm hydrogen line. Observing significantly off 1420 MHz degrades sensitivity more than the holes do. However, the frequency range is also restricted by filters, see below.
The incoming electromagnetic waves are reflected in the parabolic dish and summed coherently (in-phase for 21 cm wavelength) at the feed horn. Inside the horn is a small (5 cm) metal probe which is the actual antenna, sensitive to a single linear polarisation. The probe has an N-type connector on the outside of the horn, and a short (30 cm) CNT-400 cable connects it to a small box mounted on the horn. This box contains a Low Noise Amplifier (LNA) with a Bias-T that feeds the LNA 12 V DC through the coax. The LNA is custom built by Tommy Henderson at AGO in the US.
After the LNA, the signal travels down the telescope through approximately 10 m of RG-400 cable to a junction box. Here a second Bias-T injects the 12 V supply onto the cable feeding the LNA above. After the junction box, the signal continues via approximately 30 m of CNT-400 coax into the building where the sampler is located. In the rack, immediately before the sampler, the signal passes through a second amplification stage: a Radio Astronomy Supplies wide-band LNA (WBLNA) covering 1–2 GHz, specified by the manufacturer at 15 dB gain and 0.6 dB noise figure, with N-type connectors and powered from 12 V DC. This extra gain ahead of the sampler allows the USRP to be run at a lower gain setting, where the spurious signals it generates internally stay below the noise. The wide-band LNA has no filtering of its own, but it sits behind the bandpass filter in the horn LNA, so the out-of-band interference reaching it has already been suppressed. The sampler is a USRP N210 equipped with a DBSRX2 daughter board, which digitizes the waveform and sends samples via 1 Gbps Ethernet to the control computer. The DBSRX2 card covers 0.8–2.3 GHz, but due to strong interference above 1.8 GHz the LNAs include additional filters limiting the usable range to 1350–1620 MHz. The USRP supports up to 25 MHz instantaneous bandwidth; for standard HI observations the default 2.5 MHz is sufficient. Both bandwidth and channel resolution are selectable by the user. The DBSRX2 has a software-controlled gain range of 0–88 dB (73 dB RF gain plus 15 dB baseband gain); the ADC saturates at roughly −10 to 0 dBm at the SMA input depending on gain setting, so the total in-band power should be kept around −20 to −30 dBm for clean linear operation. The gain setting is nominal rather than exact: measurements on 2026-07-27 give about 1.14 dB of realised gain change per nominal dB — a nominal 10 dB step measured 11.25 dB on Vale and a nominal 20 dB step measured 23.0 dB on Brage. Levels taken at different gain settings should therefore be compared using the measured step, not the nominal one.
The telescope control and booking system is a web application written in Rust, running on a server at Onsala. The web framework is Axum on the Tokio async runtime, with HTMX for dynamic page updates and Tailwind CSS for styling. Observations and bookings are stored in a SQLite database, and authentication is via configurable OAuth2 providers or local accounts. The server communicates with each telescope's USRP sampler via the UHD driver to acquire spectra, and with the MD01 rotor controller to point and track targets. The webserver source code is openly available on GitHub.
The rotor and controller are manufactured by SPID Elektronik. The rotor model is BIG RAS/HR with 0.1 degrees resolution and 1.5 degrees / second rotation speed. The rotor is powered by a MD01 control unit which also reads the current rotor position. The MD01 accepts azimuth/elevation commands via the ROT2PROG protocol over a TCP connection. All coordinate conversions (galactic, equatorial, horizontal) are handled by the SALSA webserver before commanding the rotor.
The MD01 has its own software limits allowing roughly 10° of travel past each end of the azimuth range, to protect the feed cabling from being wound up. The SALSA webserver never commands an azimuth outside 0–360°, so this margin is not used in normal operation and the rotor never reports a position outside that range either. The practical consequence is for observers rather than for the hardware: a target whose azimuth crosses due north cannot be followed straight through 0°/360°, and the telescope has to unwind nearly a full turn the other way before tracking resumes.
Signal path ↓
Control path ↑