At present, digital auto-correlation and acoustic-optic spectrometers are in common use as spectrum analyzing systems for radio astronomy. The digital auto-correlation spectrometer provides a flexible selection of bandwidths and resolutions and has convenient digital components. However, its practical application for wide bandwidth and high resolution is limited by the micro-electric technique and the processing speed. The alternative acoustic-optical spectrometer also has its own limitations. As an optical system, its stability is susceptible to influence by the environment. Attempts have been made to develop new type of spectral analyzing system with better stability and reliability in structure and with simplicity of design. The SAW CZT-based spectrometer described below may have the desirable characteristics.

The CZT algorithm for Fourier transform was developed in 1960's. Since then research on the algorithm from theoretical principle to practical application has flourished. Comprehensive discussions can be found in various review articles (Jack et al. 1978; Jack et al. 1980). To understand the structure of the SAW spectrometer in our spectral receiving system, we summarily depict the CZT algorithm blow.

For a spectral analyzing system, a signal in frequency domain may
be expressed in terms of a Fourier integral:

(1) |

(2) |

This transform can be realized by a pre-multiplication (*M*) of input signal *f*(*t*)
with a down chirp waveform e, followed by a convolution (*C*)
with a up chirp signal, and finally multiplied by a down
chirp signal (as shown in Fig. 2) (Zhang et al. 1996).
The signal processing system is named by the *M*-*C*-*M* configuration of CZT or
the SAW chirp filter. The post-multiplication, used to correct the phase of
the output, can be omitted when only power spectrum is required and reduced
to the *M*-*C* structure. The CZT algorithm appears to be a frequency axis replaced
by a time axis. The spectral components in an input signal are dispersed into
a set of pules with different delay and amplitude corresponding to the frequency
and the input power, respectively, when they pass the SAW chirp filter.

A diagram of the SAW CZT-based spectrometer consisting of three parts
(I, II and III) is shown by Fig. 3 (Chen et al. 1990).
Pre-amplifiers and mixers constitutes the first stage with two highly
stable L.O. (*f*1 = 710 MHz, *f*2=690 MHz).
The operating frequency and bandwidth of SAW spectrometer are limited
by the technology of SAW devices. So input signals are first split into
two bands, *A* and *B*, 20 MHz each and then they are down-converted to
IF 171 MHz. Part II is the SAW CZT processor containing four parallel
CZT subsystem (*A*1, *A*2, *B*1 and *B*2). Each subsystem operates in a
bandwidth 20 MHz with 50% working duty cycle and a frequency resolution
39 KHz. The SAW CZT processor, therefore, has a bandwidth of 40 MHz
and 100% duty cycle. The equivalent transformation rate is greater 1024 points/66.6 s.
The dynamical range of SAW CZT processor is
determined by the sidelobes of impulse response. Hamming weighting
is employed to reduce the sidelobe to 35 dB.
Part III is a digital processor to carry out the A/D conversion
and digital integration. The longest integration is about 300 ms.
The integration time and the beginning of the integration can be
controlled by an I/O interface. Table 2 gives the properties of
the SAW CZT processor. Performances of the spectrometer,
such as bandwidth, frequency resolution, accuracy of frequency,
dynamical range, and temperature behavior were carefully examined
and these results have been published in relevant articles.

Copyright The European Southern Observatory (ESO)