Introduction

The new radar receiver system is based on newly developed data standards: Digital RF (DRF) and Digital Metadata (DMD). These are HDF5 based storage formats aimed for efficient storage and fast random access of raw voltage radar data and arbitrary data products with an inherent temporal dimension. This data is time stamped accurately with a concept of samples since epoch, which allows data with various different sample rates to be naturally aligned. These data formats are already used at various observatories and universities for geophysical instrumentation. The specifications for DRF and DMD are available.

An ion line overview plot is generated using the wide band 25 MHz data in the same way as the plasma line data, by multiplying a range delayed and conjugated transmit pulse with the radar echo. For \(N\) transmit pulses at a time interval of \(\tau\), we form an average with: $$ S_{r,\omega, m} = \frac{1}{N} \sum_{n=0}^{N-1} |\mathcal{F}_D\{ \epsilon^*_{n \tau + m\tau N + t-r} m_t + \xi_{n\tau + m\tau N + t}\}|^2, $$ where \(\epsilon_t\) is the transmit waveform multiplied with a window function, \(m_t\) is the measured radar echo, and \(\xi_t\) is the receiver noise. In addition to the average range-Doppler profile, we also calculate the sample variance estimate \(\hat{\sigma}^2_{r,\omega,m} = \mathrm{Var}( S_{r,\omega, m})\) using the Welford 5 online estimation method. The quantities \(\hat{\sigma}^2_{r,\omega,m}\) and \(S_{r,\omega,m}\) are averaged over one second to provide sufficient statistics, but to not smear time transient interference too much.

Finally, these 1 second intervals are averaged using sample variance weighting and aligned to either one minute or the length of the measurement cycle. $$ S_{r,\omega} = \frac{1}{\sum_{m=0}^{M-1} \hat{\sigma}^{-2}_{r,\omega,m}} \sum_{m=0}^{M-1} \frac{1}{\hat{\sigma}^2_{r,\omega,m}} S_{r,\omega,m}, $$ where \(M\) is a sufficient number of individual spectra to cover the measurement interval. Currently, \(M\) is set to cover either a minute or a full integration period.

The data is processed using a matched filter and a filterbank of transmit pulses matched in range and Doppler, which are incoherently averaged. No other attempt to deconvolve the true incoherent scatter spectrum is made.

Due to the large amount of radio interference in UHF bands at Haystack, we estimate receiver noise using two independent methods: mean power over a 1 MHz band, and median power over a 0.1 MHz band. These data products are calculated for a 5 second interval and recorded as digital metadata in the rx_noise_temperature channel. The mean estimator has smaller variance due to a larger number of samples going into the estimate, but is susceptible to power line interference appearing e.g. as impulsive power spikes in the receiver. In the plots above, the mean estimator is shown using a light color and the median estimator is shown with a darker color. The plot is also a good indicator of the severe effects induced by strong power line interference. The more divergence in the mean and median estimates, the more interference the receiver is experiencing.

The receiver noise can also be used to determine the gain of the antennas. A recent survey using strong radio sources Cygnus A and Cassiopeia A with known flux densities at 440.2 MHz 4,5 showed that the gain of MISA is \(43.5 \pm 0.5\) dB and Zenith is \(44.5 \pm 0.6\) dB. Zenith gain has more errors associated with it because the antenna's pointing direction was set for optimal Cygnus-A transits back in 1962, and therefore it is not completely certain that Cygnus-A still passes through the center of the beam.

This plot is useful for diagnosing the state of the radar and to identify possible problems, such as radio interference 1. The most common source of interference is power line interference, which is though to be generated by a faulty insulator in the transformer yard next to the radar. This faulty insulator causes arcing on dry weather, which is seen as strong sub microsecond spikes.

RF System Interconnection Diagram

A high level plot of the data system configuration is shown in the figure below. The main source of RF is the MISA and Zenith receiver taps, which originate from the grey wall panel at the GSC. Other important signals are transmit waveform from the combined MISA and Zenith transmitter waveguide taps, the TTL level receiver protection signal (TR), and the identification code for each pulse (ID), which is an amplitude modulated 10 MHz carrier.

Software System Diagram

The system is implemented in Python, C, and C++. Most of the code is in fact Python, with only the performance intensive portions written in very C-like C++ or pure C.

Digital Receiver System Usage

The system is divided into minimal services, which run indefinitely in the background with nohup with the midasop user account permissions. All of the production services are in directory ~/mho_production. All data is also owned by midasop. Service specific log files are located in directory ~/mho_production/log. These logs are rotated, so the previous three log files survive restart, and can be used to determine causes for potential failures.

The services can be restarted with the start_all.sh script. The services are shutdown with the stop_all.sh script. The scripts are designed to be relatively tolerant of various failure modes and to recover from them. This is mainly achieved by having each service implement a very minimal set of functionality.

Acknowledgements

Thanks to members of the Atmospheric Sciences Group for technical and scientific support of this work. Radar operations and analysis at the Millstone Hill Geospace Facility are supported by cooperative agreement AGS-1242204 between the US National Science Foundation and the Massachusetts Institute of Technology.

References