Backend Spectrometers at SORAL
Detecting an astronomical signal at the heterodyne receiver stage is only one of the challenges that submillimeter receiver arrays must face. The IF signal must be processed to extract the spectrum and be sent on to the acquisition and data reduction computer(s) for storage and on-screen analysis. Single spectrometer systems are by no means trivial; their complications get multiplied in array systems. New technologies must be developed to tackle these challenges. Following is a short description of generalized heterodyne spectrometer types.
Filterbanks
Filterbanks are perhaps the most straight-forward means of obtaining a power spectrum from the IF signal. The IF signal is passed through a power splitter, followed by a series of many narrow (1 MHz or less) bandpass filters, each of which sends the resulting power to a square-law detector, A/D converter, and then to a computer for reassembly. In practice, filterbanks can be quite cumbersome. Even a moderately-sized filterbank takes up an entire equipment rack. Since each filter (out of several hundred) has a different central frequency, they must be designed individually. The parts count is extremely high, so there's a lot that can go wrong!
Digital Autocorrelator
The IF signal can be digitized electrically ("correlated"; typically with 1 or 2-bit quantization). This digital signal can be clocked and sampled, resulting in the so called autocorrelation function. The Fourier Transformation of the autocorrelation function is the power spectrum. Autocorrelators are used extensively in centimeter-wavelength interferometric arrays (like the VLA) but are well-known to be huge and power-hungry. Their adaptation for use in submillimeter astronomy is still in its infancy.
Acousto-Optical Spectrometers (AOS's)
The third spectrometer design, Acousto-Optical Spectrometers, is perhaps the most contorted conceptually, but is clearly the most popular. The IF signal is brought to a piezoelectric transducer, which mechanically modulates a Bragg cell, through which a collimated laser beam is passed. The laser light is diffracted through an angle that depends on the laser wavelength and the acoustic wavelength of the piezo-induced modulation. That is, different RF frequencies present simultaneously in the Bragg cell will result in multiple diffraction beams. These diffraction patterns are detected by a light detector (typically a CCD), and the spectrum is digitized and stored on a computer.
SORAL will have access to a 4x1000 channel AOS originally manufactured by Photonics Systems Incorporated, and refurbished by the University of Cologne. Some of its specifications are highlighted here:
| Characteristic | Specification |
| IF center frequency | 1.55 GHz |
| Instantaneous bandwidth | 4 GHz |
| Number of channels | 4000 |
| Channel bandwidth | 1 MHz |
| Dynamic Range | 30 dB |
| Sensitivity | -50 dBm |
| Power | 200 W |
Large-format spectrometer systems will be the target of substantial work in the near future. The following description of our proposed 2 THz, 16-pixel heterodyne array (STAR) for the airborne SOFIA telescope highlights this need.
In its final form, such an array system will produce 16 independent channels of receiver output with an instantaneous bandwidth of > 1 GHz. To create spectra from these channels, each must be processed by a backend spectrometer. A high resolution THz receiver requires a spectral resolution of about 1 MHz. This translates to at least 1000 spectral points per array element. With a heterodyne array of 16 mixer elements, we gain efficiency by a factor of 16, but now require 16 x 1000 channels to be taken simultaneously. Such a large number of channels is difficult to achieve with even the latest generation of autocorrelators and would require far more power and space than is available on SOFIA. Where wide bandwidths and high spectral resolution are required, Acousto-Optic Spectrometers (AOSes) are an attractive alternative technology.
AOSes have proven to be very capable backends for single-mixer receivers. The Submillimeter Wave Astronomy Satellite (SWAS) uses an AOS that provides 1400, 1-MHz channels in a very efficient, space-qualified package. AOSes from the University of Cologne are also successfully used at KOSMA on Gornergrat, at the IRAM 30-meter telescope on Pico Veleta, and at the Antarctic Submillimeter Telescope (AST/RO). Our experience is that a properly designed AOS is a very powerful backend for heterodyne receivers. In the past year, the University of Cologne AOS group has extended the technology of the SWAS AOS and constructed a 4-channel, array-AOS (aAOS). The optical layout and performance of the prototype aAOS is shown below.
| Characteristic | Specification |
| Center Frequency | 4 x 2.1 GHz |
| Bandwidth | 4 x1 GHz |
| R.F. Drive Power | < 4 x 10 mW |
| Gain Flatness | < 2 dB equalized |
| Total No. of Pixels | 5600(1), 4000(2) |
| Noise Dynamic Range | 13 dB |
| Crosstalk | < -30 dB |
| Amplitude Linearity | Better than 1% |
| Frequency Linearity | Better than 0.1% |
| Allan Variance Minimum | > 600 sec |
| Reception Bandwidth | 1.5 MHz(1), 2.3 MHz(2) |
| Resolution Bandwidth | 1.0 MHz(1), 1.5 MHz(2) |
| Weight | 10 kg(3) |
| Power Consumption | 12 W |
(1): at 0.7 MHz channel spacing
(2): at 1.0 MHz channel spacing
(3): electrical and optical unit, without rf.-PAs
Tests of the prototype aAOS are now underway at the IRAM 30~m. Below is a spectrum taken with one channel of the aAOS compared to spectra taken simutaneously with the IRAM filterbank and autocorrelator spectrometers. The quality of the aAOS spectrum is excellent and compares well to the spectra taken with the other two systems.
