Radio Science

New Digisonde for research and monitoring applications

Authors


Abstract

[1] The new Digisonde-4D, while preserving the basic principles of the Digisonde family, introduces important hardware and software changes that implement the latest capabilities of new digital radio frequency (RF) circuitry and embedded computers. The “D” refers to digital transmitters and receivers in which no analog circuitry is used for conversion between the baseband and the RF. In conjunction with the new hardware design, new software solutions offer significantly enhanced measurement flexibility, enhanced signal selectivity, and new types of data, e.g., the complete set of time domain samples of all four antenna signals suitable for independent scientific analysis. With the new method of mitigating in-band RF interference, the ionogram running time can be made as short as a couple of seconds. The h′(f) precision ranging technique with an accuracy of better than 1 km can be used on a routine basis. The 4D model runs the new ARTIST-5 ionogram autoscaling software which reports in real time the required data for assimilation in ionospheric models. The paper highlights technical advances of the new Digisonde for research and monitoring applications.

1. Introduction

[2] High frequency (HF) radio ionospheric propagation has been attracting both academic and practical interests for the last 70 years, in spite of predictions for a gradual retirement of its operational aspects over the recent two decades. As years go by, the ionosonde observations and technology, incepted in 1931 by Sir Edward Appleton, held steady while increasing the maturity, reliability, and autonomy of the ionospheric diagnostics for a multitude of users. Such longevity reflects the fact that HF ionospheric sounding has reached a high level of significance for ground and space-based radio monitoring and communication systems. Ionogram-derived electron density profiles (EDPs) have become the ground truth reference for ionospheric specification, presenting the unrivaled accuracy of the data on continuous demand for validation of alternative ionospheric techniques, including radio occultation, ultraviolet, tomography, and research projects such as the ionospheric modification with HAARP [e.g., Djuth et al., 2006].

[3] Feasibility of multisite, continuous operations around the world is one of the key advantages that make ionosondes attractive for both research and monitoring applications. Unlike incoherent scatter radar systems [Evans, 1975], ionosondes must be built at low cost, easy to install, and be ready for unattended “lights-off,” remotely controlled operation with very little maintenance expense: a formidable challenge considering that ionosondes are active research instruments. The evolution of the Digisonde family of ionospheric sounders [Bibl and Reinisch, 1978; Reinisch, 1996a, 1996b] (Figure 1) exemplifies a systematic effort of refining the technology in order to operate at lower transmission power and operating voltages, lighter footprint, shorter down times, and greater flexibility of measurements, while keeping the evolving ionospheric science objectives in close attention. The new Digisonde-4D makes another step along the way.

Figure 1.

Evolution of the Digisonde over 3 decades.

[4] Ionosonde development has benefited from the ascent of digital techniques since the 1970s, when original analog ionosondes started to be replaced by digital instruments featuring analog-to-digital converters (ADC) and custom computer circuitry to perform digital integration and later digital spectrum analysis via lookup tables [Bibl and Reinisch, 1978]. With commercial availability of the digital signal processors, new DSP algorithms were developed to build the Digisonde Portable Sounder, DPS [Reinisch et al., 1997]. The Digisonde-4D is the latest digital ionosonde that the University of Massachusetts Lowell Center for Atmospheric Research (UMLCAR) developed during 2004–2008. While preserving the basic principles of the Digisonde family, the Digisonde-4D model introduces a number of important hardware and software changes that implement the latest capabilities of new digital RF circuitry and embedded computers.

[5] This paper describes hardware and software advances implemented in the Digisonde-4D, and highlights the new opportunities for sounding operations. First we summarize hardware design changes, including the digital transmitters and receivers, and new data/control routing solutions. Then we outline the new features of the data processing and commanding software, and conclude with a discussion of new science applications with the Digisonde-4D.

2. Digital Transceiver for the Digisonde-4D

[6] The “D” in the new Digisonde-4D model refers to the digital transmitters and receivers in the sounder that use down- and up-converter IC chips, the Graychip GC5016 and the Analog Devices AD9857, to implement the classic functions of radio transmitters and receivers (transceivers) by numeric techniques. Digital implementation of direct conversion between the baseband and the radio frequency not only greatly simplifies the architecture of the transceiver, it also brings higher precision of measurements, as well as reductions in size, weight, cost, and maintenance of the instrument. However, while digital transceivers have been in practice for years [Razavi, 1997], it is with a great deal of caution that direct-to-RF technology was considered for the HF receiver design in the Digisonde. Digital receivers that directly digitize the RF signals can easily be saturated by strong interferers, since ADCs have a limited dynamic range, rendering the receivers insensitive to smaller message signals. For ground-based ionospheric sounding strong interferers, especially AM radio stations in the 0.55–1.65 MHz band and the 6–24 MHz short wave communications bands, can pose serious problems. At some locations we have measured interference of >1 V at the receive antennas compared to <1 μV ionospheric echo signals.

2.1. Gain Control and Saturation Sensors

[7] State-of-art ADC and down-converter chips with 160 MSPS clock rates, which are required to support operating frequencies up to 30 MHz, are only available at a maximum of 16 bit input resolution, corresponding to a dynamic range of ∼115 dB, still some 60 dB too low for operations in many noisy RF environments. The need to protect the ADC inputs from saturation warranted the design of an analog front end for the Digisonde-4D, a combination of “tracking band-pass filters,” suppressing powerful out-of-band interference, and gain-controlled wideband input amplifiers handling potentially saturating in-band signals. Figure 2 presents a simplified block diagram of the gain control and saturation sensor circuitry in the four Digisonde-4D receiver channels.

Figure 2.

RF Environment Sensor and gain control in the analog front end of the Digisonde-4D. The dynamic range of the 16-bit ADC is inadequate for direct-to-RF conversion in noisy RF environments. The analog tracking filters suppress out-of-band interference, and the variable gain amplifiers prevent saturation of the ADC by in-band signals. The analog RF Environment Sensor is tuned to detect high location-specific interference levels to recommend adjustment to the antenna preamplifier gains. The digital under/over voltage sensor provides status flags to the Autogain controller. Constant gain of the wideband antenna switch and tracking filters is programmable to follow day/night differences in the RF signal levels.

[8] A higher agility of the automatic gain control is required for the narrowband circuits that are sensitive to individual in-band RFI sources and their changes with time. Wide band circuitry is exposed to the cumulative environment where the dynamics of individual interference sources smear out the instantaneous interference level. To protect against input saturation, the Digisonde-4D implements three types of the gain control, (1) Fixed Gain in the antenna preamplifiers that is site-specific, adjusted at installation time, (2) Constant gain in the Antenna Switch and Tracking Filters that does not change during the individual measurement, but can be programmed to adapt to the larger timescale changes such as day/night difference, and (3) Autogain that changes from frequency to frequency. The analog RF Environment Sensor is tuned to detect high location-specific interference levels and recommend adjustment to the antenna preamplifier gains. The digital under/over voltage sensor monitors the traffic between ADC and Digital Receiver to provide status flags for the Autogain controller decisions. The optimum Autogain settings for each of the sounding frequencies are stored for future ionograms, and the global table of frequency-dependent optimal gains is refreshed every 5–15 min.

2.2. Phase Coherence of Converters

[9] Special design consideration was given to the task of sustaining phase coherence of the digital up-converters and down-converters whose built-in frequency synthesizers are independent of each other. The synthesized frequencies must have equal phase in order to be able to use the RF phases of the echoes received on the spaced antennas for calculation of the arrival angles. All converters are driven by the same 61.44 MHz master clock and in addition are synchronized by a start signal that clears the converters' phases to zero with high precision and minimum latency at the beginning of each new frequency transmission. This procedure also assures that consecutive frequencies are transmitted with the same (zero) phase, making it possible to use the phase difference of echoes at two closely spaced frequencies to accurately determine h′(f) [Reinisch et al., 2008a]; Figure 10 gives further details on precision ranging.

2.3. Cross-Channel Equalization

[10] Phase-aware multiple-channel RF instrumentation requires an additional design effort to have cross-channel differences in channel transfer functions calibrated out. In addition to the standard mode of transmitter operation when the output from the up-converters is routed to the RF power amplifiers, a cross-channel equalization (CCEQ) mode is also available for the loopback operations. In the CCEQ mode, a small calibration signal is routed to the antenna switch, instead of the RF amplifiers, for direct input to the receivers.

3. Digital Signal Processing in the Digisonde-4D

[11] Use of dedicated hardware for digital signal processing (DSP) continues to be part of our strategy to build low-power, lightweight, space-qualified radio sounding instrumentation. Thanks to the use of DSP techniques, the radio plasma imager (RPI) instrument [Reinisch et al., 2000] on NASA's IMAGE satellite [Burch et al., 2001] was capable of detecting echoes coming from 50,000+ km distances at transmitted power below 1 W. At the same time, for the ground-based applications that do not require radiation hardening, inexpensive small-format embedded computers present a feasible alternative to the custom-designed DSP hardware. Modern multiple core processors have adequate computing power to handle relevant calculations, and excellent software development environments are superior in terms of the development time and the learning curve. Thus, the current choice of the Digisonde-4D processing architecture balances both specialized and generic computing platforms, with the goal of first using embedded computers as the DSP test beds and then migrating the algorithms to their hardware implementations.

[12] Figure 3 shows the Digisonde-4D block diagram identifying the data processing elements. With new DSP concepts in mind, the sounder provides two preprocessor cards mated to the output of the digital receiver. Targeting space applications, we selected field-programmable gate arrays (FPGA) as the CPU of the preprocessors, whose design admits implementation in various FPGA families, including generic Altera chips suitable for the Digisonde-4D, and radiation-hardened Actel versions that can operate in the harsh radiation environments. At this time the Digisonde-4D sounders are equipped with one preprocessor card that performs all data formatting, routing, double buffering, and output function selections, leaving plenty of space for future DSP algorithms like the RF interference mitigation (RFIM), which is currently performed on the Data Platform, and the twin-frequency operation (dashed lines). The Control Platform is an embedded computer that manages hard real-time tasks under control of the RTEMS operating system [Acuff et al., 1994]. The digital receivers are designed to appear to the Control Platform as an IDE hard disk drive whose contents can be read using a standard HDD driver. The Data Platform is another embedded PC that performs processing of the acquired sample data to create data products. The Data Platform runs a suite of processing algorithms ranging from pulse compression to ARTIST ionogram autoscaling and EDP calculation, and the derivation of ionospheric tilt angles from Doppler skymaps.

Figure 3.

Digital data processing architecture of the Digisonde-4D. The system combines dedicated preprocessors based on the configurable FPGA logic with two generic embedded PCs for real-time tasks and data processing/analysis applications. The dashed lines are used to indicate future extensions of the architecture, including hardware implementation of the RF Interference Mitigation (RFIM) algorithm and twin-frequency operation.

[13] Ability to operate in noisy RF environments has been a very important factor influencing the HF sounder technology and research applications. As the line between “monitoring” and “scientific” ionosonde instruments [Davis, 1990] is fading, the attention is drawn to new systems that combine reliable low-power operations suitable for ionospheric monitoring with high measurement cadence, resolution, contents, and flexibility needed for research. It is, however, the signal-to-noise ratio (SNR) considerations that have been standing in the way of rapid ionogram measurements at low transmitted power. With the new patented RF interference mitigation algorithm [Bibl, 2005], 2-s high-resolution ionograms with a low transmitted power of 300 W became a reality.

[14] The 33 μs wide chips of the Digisonde-4D transmitter pulses imply a transmission bandwidth of 30 kHz that requires a receiver bandwidth of at least 30 kHz. This large bandwidth makes the receiver susceptible to a variety of “contaminating” radio emissions that are received together with the signal. These emissions consist of (1) impulsive RF sources appearing as a wideband background noise in the 30 kHz receiver bandwidth, and (2) narrowband interferers, typically strong signals from other transmitters, in many cases broadcasting stations. Mitigating the background noise can be accomplished via signal integration/accumulation, interpulse phase switching, pulse modulation, or increased transmission power. Though effective, such measures oppose the goal of rapid measurements at low power. Removing narrowband interferers poses a great challenge, as such interferers occur unpredictably at a priori unknown frequencies.

[15] It was the immunity to narrowband interferers that earned the chirp-sounders [Barry and Fenwick, 1969] its success in the 1970s, accomplished by using frequency-modulated continuous wave (FMCW) signals that allowed the receiver bandwidth to be as small as 200–500 Hz during the frequency sweep. For the pulse sounders, various techniques for suppression of the in-band interferers by tunable notch filters were explored. Our new digital RFIM technique [Bibl, 2005] introduces several innovations for adaptive suppression of the interferers.

[16] Figure 4 illustrates a common problem with using notch filters in the frequency domain to eliminate a monochromatic interfering signal. Once the receiver output is digitized, the spectral amplitudes are obtained via conventional DFT algorithms for the integer-indexed frequencies that are multiples of 1/T where T is the coherent integration time. In general, the interferer frequency fI will not be a harmonic of 1/T, i.e., fIm/T where m is an integer. When the digitization rate is not exactly equal to the interferer frequency, the monochromatic interfering signal actually contributes to several neighboring spectral components. The width of the main spectral peak is inversely proportional to the length of the time period over which the spectrum is calculated, which in the 4D is typically 5 ms or 10 ms for a pulse repetition rates of 200 or 100 pps.

Figure 4.

Spectrum of a truncated monochromatic CW signal. The continuous line shows the analytical continuous spectrum, the black circles show the DFT spectrum amplitudes, and the dashed line gives the actual interferer frequency and amplitude.

[17] The first RFIM innovation is to use the DFT spectral amplitudes A and B in Figure 4 to calculate the precise value of the interferer frequency:

equation image

where fA is the frequency of the stronger of the two strongest neighboring components in the spectrum, and A and B are their amplitudes. The second RFIM innovation is to suppress the interferer in the time domain instead of the frequency domain where its spectral components are numerous. To synthesize the monochromatic interfering signal in the time domain, its amplitude and phase need to be known. They are obtained using a single-line discrete Fourier transformation for the precisely determined frequency fI:

equation image

where equation imageI is the complex spectral amplitude, equation imagen are the complex signal time samples, and N is the total number of samples. The inverse transform of (2) gives the precise time domain presentation of the interferer

equation image

The function equation imagen can now simply be subtracted from the source data. Finally, the RFIM procedure is applied iteratively: it removes interfering signals one at a time, starting with the strongest signal, so that the strongest distortions to the spectral amplitudes are removed at each step of the iteration.

[18] The RFIM performance has been tested with the Digisonde-4D by infusing a coherent interference signal at 20 dB above the loopback test signal (Figure 5). Figure 5 (top) displays the Fourier spectrum of the Digisonde signal (16-chip phase-coded pulse). Figure 5 (bottom) shows the spectrum of the same signal with the added interferer producing a spike near 6 kHz and affecting the entire spectrum to various degrees.

Figure 5.

Spectrum of the received signal (top) without interferer and (bottom) with one in-band interferer.

[19] Figures 6 and 7 demonstrate the RFI mitigation effect in a field experiment with the Digisonde-4D operating at Millstone Hill. The “waterfall” presentation of the signal spectra in Figure 6 is used to plot the linear spectral amplitudes in 18 abutting ranges staggered on top of each other, the unprocessed data are displayed in Figure 6 (top), and the RFIM-processed data are displayed in Figure 6 (bottom). Comparison of Figure 6 (top and bottom) demonstrates the extraordinary efficacy of the RFIM technique. Actually, as described earlier, the RFIM “cleaning” is done on the time domain data as illustrated in Figure 7. Figure 7 (top) presents the unprocessed linear amplitudes (vertical axis) versus time (horizontal axis, shown as virtual height in km), and it is not possible to see the ionospheric echoes. After RFIM has subtracted the interferers, the echo location is easily identified at ∼240 km.

Figure 6.

Waterfall presentation of the measured spectra demonstrating the effectiveness of RFIM. The vertical axis is linear amplitude, and range increments are in 2.5 km steps; the horizontal axis is Doppler frequency in Hz. (top) Spectra prior to RFIM. (bottom) Spectra after RFIM is applied.

Figure 7.

Received signal in the time domain (top) contaminated by interferers and (bottom) after RFIM processing.

[20] Routine ground-based ionogram observations demonstrate the advantages of RFIM processing as illustrated for a Millstone Hill ionogram in Figure 8; the data in Figure 8 (left) are recorded before, and in Figure 8 (right) after applying RFIM. The RFIM advantage is most visible at frequencies or frequency bands where the SNR is small. At the low-frequency end of the ionogram, the inefficiency of the transmit antenna and the presence of powerful commercial radio broadcast transmissions (AM radio) make echoes below 1.7 MHz undetectable without RFIM. At the high-frequency end, near the foF2 cusp (around 6 MHz), pulse dispersion, deviative absorption, and interference in the 6 MHz broadcast band lead to an ill defined foF2 cusp (red echo trace) without RFIM processing (Figure 8, left), and an improved signature with RFIM processing (Figure 8, right).

Figure 8.

Digisonde-4D ionograms (left) before and (right) after RFIM processing. The RFIM-processed trace starts at 1.15 MHz (compared to 1.6 MHz in Figure 8 (left)) and shows a better cusp signature in the O trace.

[21] Up to 40 dB improvement in signal-to-noise ratio due to RFIM processing allows shorter integration times and therefore faster measurements. Figure 9 presents a sequence of 2-s ionograms taken at Millstone Hill with a 10-s cadence. Rapid ionogram measurements offer substantial interest to experimental campaigns that study ionospheric plasma dynamics under influence of natural or artificial disturbances like magnetic storms or powerful HF heaters.

Figure 9.

The 2-s ionograms taken every 10 s at Millstone Hill on 20 September 2006 with RFIM processing.

4. Enhanced Science Applications With the Digisonde-4D

[22] In contrast to the earlier Digisondes, which output only Fourier-transformed data, the 4D has an additional time domain data output. The two 16-bit quadrature samples for each sampled range and all four receiver channels are stored. For a typical ionogram with 500 frequency steps and 512 h′-samples per frequency, this results in 262 MB per ionogram of raw time domain data, considering complementary interpulse phase coding, ordinary and extraordinary polarization, dual frequencies for precision range measurements, eight repetitions for Doppler analysis, and 4 receive antennas. While this data volume is generally too large for routine operation, it can be afforded for special campaigns. This feature will make it possible for the 4D users to process the data anyway they like with their own algorithms suitable for specific scientific research. In addition to these “raw” data the standard packaged data products described in Figure 10 are also generated and stored.

Figure 10.

Summary of key Digisonde-4D measurement modes and data visualizations. Ionogram image example courtesy of L. A. McKinnell. Daily directogram image example courtesy of A. M. Abdu.

Figure 10.

(continued)

Figure 10.

(continued)

Figure 10.

(continued)

5. Campaign Planning With the Digisonde-4D

[23] Modern monitoring ionosondes have progressed beyond the hourly ionogram observations into the realm of operational campaigns in support of complex, multimode experiments involving research instrumentation on the ground and fly over spacecraft. Some of the campaigns require targeted monitoring of plasma regions synchronously with other instrumentation, such as the recent HAARP campaign in October 2008; other campaigns involve multiple Digisonde stations providing ground truth data for many thousands of spacecraft overpasses, such as the “CalVal project” for the calibration/validation of the ultraviolet sensors on DMSP F-16 [Paxton et al., 2002] and radio occultation instruments on COSMIC/FORMOSAT-3 [Anthes et al., 2008]. Without appropriate tools, the planning of Digisonde measurements with increased cadence and resolution at multiple locations and times would quickly become overwhelming and error-prone.

[24] Figure 11 shows a screenshot of the DCART planning software [Reinisch et al., 2008b] that manages design and commanding of the Digisonde-4D measurement schedules. Figure 11 displays a real-life example of programming 15 min in the life (15MITL) of the Digisonde, where 15MITL is plotted with time on the horizontal axis, and operating frequency on the vertical axis. During the 15MITL two ionogram measurements are made (shown as black rectangles, program 8) and six Doppler skymap measurements at various groups of fixed frequencies (shown as black stripes, programs 4, 7, and 58). In this example the science measurements are programmed at the cadence of 8 times an hour, with each of the eight runs including one ionogram running from 0.3 to 10 MHz, one Doppler skymap covering frequencies around 4 MHz to target the F region, a high-resolution skymap at 2 MHz for E region study, and a wide coverage 1 to 10 MHz skymap to capture reflections from E and sporadic E layers. In addition to the science programs, the schedule includes three housekeeping measurements: one autogain evaluation (program 27), system Built-In Test (program 11), and cross-channel equalization (program 25). Figure 12 shows an example of Day–In-The-Life (DITL) of the Digisonde-4D at Millstone Hill observatory. During the DITL two main schedules are used, one for daytime (schedule 041) and one for nighttime (schedule 004), plus two additional campaign requests for high cadence measurements in support of a spacecraft overpass (schedule 006).

Figure 11.

Measurement schedule: 15 min in the life (15MITL) of the Digisonde-4D.

Figure 12.

Measurement timeline of one day in the life (DITL) of the Digisonde-4D, with two regular daytime and nighttime schedules 041 and 004 interspersed with two campaign mode requests of schedule 006.

6. Discussion and Summary

[25] The new Digisonde-4D offers unprecedented accuracy and resolution as well as increased measuring flexibilities for routine ionospheric observations and scientific research. Users can design their own scientific applications by processing the raw time domain data samples collected at 512 ranges for all four antenna signals, and for the O and X polarizations. The ARTIST 5 autoscaling program now calculates the EDPs together with uncertainty limits for each height, making the data products suitable for ingestion in assimilative ionospheric models. The SAO data exchange format was expanded into the SAO-XML format [Reinisch and Galkin, 2008] to accommodate the expanded data content. In August 2008 during the URSI General Assembly in Chicago, Commission G of URSI accepted SAO-XML as the standard format for ionogram data exchange. All Digisonde stations are currently updated to SAO-XML. By using this standard format the Digisonde network has truly become a real time Global Ionospheric Radio Observatory (GIRO). With the recently started U.S. NEXION program, which will install thirty or more Digisonde-4D systems, GIRO is expected to grow rapidly. Since SAO-XML can easily accommodate data from any digital ionosonde, other ionosonde models can become part of GIRO by archiving their data in DIDBase.

Acknowledgments

[26] Parts of this research were supported by AF grant FA8718-06-C-0072 to University of Massachusetts Lowell. The authors also thank the many Digisonde users who have made valuable suggestions for the improvement of the DPS-4.

Ancillary