1 Introduction

  Pulsars are relatively weak radio sources with steep spectra, with signals which are highly polarized and variable on time-scales from sub-microseconds to years (for a recent compendium, see Johnson et al. 1996). High time-resolution observations have resulted in a deep insight into the degree of coherence of the pulsar radiation and the physical emission mechanism itself. Most importantly, such observations can be made for sensitive searches for companions of pulsars like planets (Wolszczan 1994) and perhaps black holes, and for the investigation of several issues in fundamental physics. In particular, timing measurements have revealed that millisecond pulsars are extremely stable clocks rivaling in their long-term stability cesium clocks on Earth, on which terrestrial time standards are based (Taylor 1991). Accurate timing of binary pulsars allows studies to be made of the nature of gravity and tests of general relativity (Taylor 1994). When pulsar timing is combined with VLBI measurements, it can be used to tie the solar system reference frame directly to the quasi-inertial extragalactic reference frame for improved determinations of earth orbital parameters, tests of general relativity, and spacecraft navigation (e.g. Bartel et al. 1996). And finally, timing permits searches of the cosmological background of gravity waves predicted as a remnant from the early universe (Stinebring et al. 1990).

There are three phenomena that limit the time resolution obtainable for pulsar signals: interstellar scattering, interstellar dispersion, and the rise-time of the receiving system. Interstellar scattering broadens pulses through multipath propagation. The broadening time, $t_{{\rm ISS}}$, is proportional to $\nu^{-4.4}$ for observing frequency $\nu$, and is in general negligible for $\nu \gt$ 1.4 GHz. For instance, for pulsars with a dispersion measure DM $\le$ 150 ${\rm pc ~cm^{-3}}$, $t_{{\rm ISS}}$ is still smaller than the shortest S2 sampling time of 31.25 ns at $\nu \ge$ 1.4 GHz (Hankins 1996). In contrast, interstellar dispersion has, in most cases, a more severe effect on the obtainable time resolution. It changes the relative phase of the frequency components of the pulsar signal which is equivalent to smearing of the signal over a time $t_{{\rm DM}}$ across the filter bandwidth $\Delta\nu$. This dispersion time is given as $t_{{\rm DM}}\, {\rm =\, 8.3 \ 10^{15} ~ DM ~ \Delta\nu ~ / \nu^3}$. The highest time resolution can be obtained by choosing $\Delta\nu$such that the filter rise-time, $\sim 1/\Delta\nu$, equals $t_{{\rm DM}}$. In the case of our pulsar example above, the highest time resolution would be $\sim$ 20 $\mu {\rm s}$, obtainable with a filter of 50 kHz bandwidth, chosen appropriately for the combination of pulsar DM and observing frequency. This time resolution is still almost three orders of magnitude coarser than $t_{{\rm ISS}}$ and the shortest sampling time of the S2. Observations at other frequencies and/or of other pulsars generally require filters with different bandwidths to match $t_{{\rm DM}}$ with $1/\Delta\nu$. The signal from the filters is typically applied to square-law detectors where the relative phase of the frequency components is inevitably lost. The most effective and flexible way of correcting for the signal phase distortions induced by the interstellar medium (ISM) is by recording the baseband signal and convolving it with the inverse impulse response of the ISM before square-law detection. This way, the obtainable time resolution of the pulsar signal is limited only by the broadening time of the interstellar scattering and the Nyquist sample time of the baseband recording system. This technique was pioneered by Hankins (1971) and used subsequently for particular studies of pulsar radiation (see Hankins 1996 for a review). At the time, only relatively small data sets could be analyzed because of the need for a huge sampling rate and the limitations in storage capacities and computing power. Today these capacities and powers are orders of magnitude larger than those of two decades ago, and a new approach towards phase-coherent pulsar observations and subsequent processing with a general-purpose computer is now feasible.

In this paper, we describe a phase-coherent baseband processing system for radio pulsar observations. Short reports on this system were given earlier by Wietfeldt et al. (1995, 1996a). Subsequently, a report on another new baseband processing system was given by Jenet et al. (1997). Our system is based on the S2 VLBI recorder (Wietfeldt et al. 1996b) developed at the Institute for Space and Terrestrial Science (ISTS) at York University in Toronto, Canada. At present, approximately 20 radio astronomical observatories worldwide are permanently equipped with an S2 recorder. Each of these stations could in principle be utilized for phase-coherent baseband recording of pulsar signals on commercially available SVHS tapes. After the observations, the recorded tapes may be processed with a computer/workstation at particular processing centers. In the remainder of this paper, we give an overview of the S2 baseband system in Sect. 2, discuss considerations for the design of the Tape-to-Computer Interface (S2-TCI) in Sect. 3, describe the software control system of the S2-TCI in Sect. 4, present results of the first observations in Sect. 5, and derive our conclusions and present future prospects for baseband processing systems in Sect. 6.