We propose to expand our development of digital backends based on the newest generation of
FPGA microchips and fast analog-to-digital converters. We aim to build a digital spectrometer
with a monolithic bandwidth of 2 to 4 GHz, and thereby provide the basis to assemble a
cost-efficient digital backend which allows for a total bandwidth coverage of more than 36
GHz at high resolution, filling a need for the present and future array receivers operated or
build by the partners in this SFB. The performance of the digital spectrometers will be
directly compared with the acousto-optical spectrometers designed within this SFB.
The rapid increase in the sampling rate of commercially available analog-to-digital converters
(ADCs) and the increasing power of field programmable gate array (FPGA) processors has
led to the technical possibility to directly digitize the down-converted intermediate-frequency
(IF) signals of coherent radio receivers, and to transform the digital signals into a power
spectrum in real-time. With simultaneous bandwidths of the order of 1-2 GHz and very high
spectral resolution, they are an important complement to the acousto-optical Spectrometer
technology (see SFB TP D6), which is limited to a resolution of about 1 MHz, but can cover
many GHz of instantaneous bandwidth within a compact, low power configuration. A first
generation of prototype digital fourier transform spectrometers (FFTS, more general: digital
spectrometer) has been assembled and is being commissioned at the Effelsberg 100m
telescope (50 MHz and 1 GHz bandwidths), at KOSMA (in collaboration with A. Benz, ETH
Zürich), at APEX (2× 500 MHz), and Arecibo (300 MHz). The efforts to build such FTS are
currently concentrated at U.C. Berkeley (D.Werthimer), Zürich (A.Benz/ETH & Acqiris), and
Bonn (RAIUB, MPIfR).
The currently most powerful card is commercially available from Acqiris; it employs two 1
GS/s ADCs which feed a XILINX Virtex 2 Pro70 FPGA chip. A software core that turns this
card into an FFT spectrometer is now either available commercially from RF Engines Limited
(500 MHz bandwidth, 16,384 channels), or from the ETH Zürich research group of A. Benz
(1 GHz bandwidth, 16,384 channels). A 50 MHz bandwidth, 1024 channel core running on a
PCIbus board (ICS-554C) from ICS was built through a RAIUB/MPIfR collaboration in 2004
as a prototype development for a 21cm multi-feed receiver for the Effelsberg 100m telescope.
The FFTSs now in use or being tested demonstrate that this novel technology is superior in
sensitivity as well as frequency resolution, simplicity, and cost compared with autocorrelators
and analogue filterbanks. Motivated by our immediate need for wide-band, high spectral
resolution spectrometers for the submillimeter array receivers at the new telescopes such as
APEX and NANTEN2, but also for future large HEMT/MMIC focal plane arrays with
hundreds of channels, and by the high cost of commercial digital spectrometer cards and
cores, we propose to aggressively expand our own development of digital backends. Using
the newest generation of XILINX Virtex 4 FPGAs – as they become available – and fast (at
least 2 GS/s) ADCs we propose to build digital spectrometers with a monolithic bandwidth of
2 to 4 GHz, well matched to the wide bandwidths with high velocity resolution and/or wide
velocity coverage, which are crucial for galactic spectral line surveys and extragalactic line
observations. Our central goal is to 1) thouroughly test these spectrometers in direct
performance comparison with the well established AOS and ACS, 2) build a cost-efficient
digital backend which allows coverage of a total bandwidth of ca. 36 GHz (composed of 2-4
GHz monolithic bandwidth spectrometer cards) at highest resolution for the present and future
array receivers operated and being constructed by the partners in the SFB 494.
See the home page of Prof. Bertoldi for more information.
