|Back to INAG Homepage||Back to UAG-104 Contents page|
The Canadian Advanced Digital Ionosonde: Design and Results
J. W. MacDougall, I. F. Grant, and X. Shen
Department of Electrical Engineering, University of Western Ontario, London, Ont., Canada N6A 5B9
A new digital ionosonde called CADI is in use at 5 Canadian sites for scientific studies. The design of this instrument is briefly described. The measurement of ionospheric convection using digital ionosonde measurements is discussed and it is shown that at high latitudes convection measurements are usually relatively easy and a variety of techniques can be used, whereas at mid latitudes the irregularity structure which is required for good convection measurements is not observed.
The Canadian Advanced Digital Ionosonde (CADI) was developed as a project of the Canadian Network for Space Research. There was a Network requirement to provide ionosonde support, at a number of sites, for collaborative experiments. These experiments involved observations by other Network instruments such as auroral radars or optical imagers and scanners. Since most of the Network sites were in the far north we needed a modern low-cost, digital ionosonde which could run essentially unattended, and had low power requirements.
In this paper the design of the CADI instrument will be briefly described. Following this description the remainder of the paper will be a discussion of our investigation of ionospheric structures to see if they provide a basis for convection measurements.
In this section a brief description of the CADI design is given. A more detailed design description will be published in a forthcoming issue of the Ionospheric Network Advisory Group bulletin.
The design philosophy of CADI was to make use of the capabilities of a modern PC as much as possible. Using the PC capabilities is very cost effective compared to providing similar capabilities using custom hardware. A PC can provide all the data storage, control, display and communications required by an ionosonde. The newer, faster, PCs can also do mathematical data processing, usually involving FFTs, with sufficient speed for 'real-time' data processing. This means there are also significant cost savings compared to satisfying these data processing requirements by means of dedicated processors.
Therefore we tried to utilise the PC capabilities as much as possible. The CADI design evolved over several prototype instruments. In the final design the receivers became PC plug-in boards, as did the frequency synthesiser. This had two advantages: the cost was decreased since no external cabinetry was required, and the system became very flexible and easy to maintain since one could plug in as many, or as few, receivers as required. In the CADI units we are using we have 5 plug-in boards: one control-and-frequency synthesiser board, and 4 receivers which are each one plug-in board.
The frequency synthesiser board uses a Qualcomm Direct Digital Synthesis 'chip' to generate the ionosonde transmitter frequencies. Frequencies can be specified to 1/100 Hz. There are a number of Motorola 68HC11 microprocessors which are used to control the system timing, and to sample the echoes. We usually use the system with 13 bit Barker pulse coding (an optional mode) and these 68HC11 microprocessors are used to generate the coded pulse, and to decode the echoes. Each receiver card has two of these microprocessors, one for the 'I' (in-phase) channel, and one for the 'Q' (quadrature phase) channel.
The radio frequency power amplifier is the only part of the basic system which is external to the PC. This produces 600W peak pulse power and is in a small chassis with a 'footprint' the size of a standard desktop PC and 95mm high.
Data storage uses standard PC backup tape units to which new data files are automatically transferred from hard disk once each day. The units we use store 120Mbytes. Storage capacity is increased by about 10% using the data compression provided in the backup software. Most of our CADI installations must operate for several weeks (typically 3) unattended so the system parameters are selected to give about 5 Mbytes of data per day.
The receiver antennas used at our field sites are based on an arrangement proposed by Wright and Pitteway . There are four receiver dipoles along the centres of the four sides of a 60m square. Each dipole is an untuned 'fat' dipole of overall length 19m and consists of a 'bundle' of 4 wires spaced approximately 30 cm apart. The centre of each dipole is fed to a balanced high input impedance wideband preamplifier. The transmitting antenna is a delta, using a relatively short mast (13m). The small size of the delta severely restricts performance at frequencies below about 2MHz, but is adequate for most of our studies.
Currently we have CADIs installed at Eureka (89oL magnetic latitude), Resolute Bay (84oL ), and Cambridge Bay (78oL ) . These form an approximate line from the centre of the (geomagnetic) polar cap to cusp latitude. The separation between these three stations is such that an F region feature moving along the line of these stations would become visible on the next station about the time it would become difficult to see from the adjacent station. Our experience is that dominant features can usually be tracked until they are of the order of 30-40o zenith angle.
Another CADI is at Rabbit Lake (68oL ) where it is used in support of auroral radar measurements. We also operate a CADI at a field site near our university (55oL ). This is used for studies of high midlatitude ionospheric phenomena.
In our routine measurements we usually record both ionograms and, interleaved, fixed frequency measurements on several frequencies. The fixed frequency measurements usually use a 64 point FFT data sample done every 30 seconds on each frequency. From these samples we have available the echo intensities, virtual height of reflection, Doppler shifts, phases, and calculated angles of arrival. Since a prime requirement of the CADI ionosondes is to measure ionospheric convection, in this section we will focus on some of the techniques which we have tried for convection measurements using the various measured parameters from the FFT samples.
For our stations in the polar cap region, convection measurements are usually relatively easy. As has been shown by Cannon et al. , if the angles-of-arrival for significant peaks in the Doppler spectrum are plotted as a 'sky map', then the drift velocity can be obtained from this sky map by fitting of the equation:- ds = (1/p ) v.ks where s is an index number for the individual points on the sky map, d is their Doppler shift, v is the drift vector, and k is the wave number vector directed from the ionospheric reflection point towards the ionosonde.
This method works well for the polar cap measurements and for each of our polar cap stations we routinely produce plots of convection direction as a function of time that are similar to those shown in Cannon et al. .
The reason the 'sky-map method' works well for polar cap stations seems to be because the polar cap ionosphere is usually highly structured into various scale sizes rather than being just a relatively uniform medium with small density perturbations. The sensitivity of an ionosonde is usually set for specular (mirror) reflection. Therefore to get strong echoes from features at various sky positions requires large blobs of enhanced ionisation at these positions. These blobs must have surfaces that are comparable with the Fresnel zone size (a few kilometres) and these surfaces must be normal to the raypath. Figure 1 shows a plot of 4 MHz echoes recorded for 6 hours at Resolute Bay. This figure illustrates the spread nature of the polar cap echoes. In the spread can be seen occasional discrete echoes from the larger scale structures. The many 'U' shaped features are signatures of ionospheric perturbations that are convecting overhead. An apparent speed estimate is easily obtained just using the range versus time curves of these features, and for Figure 1 the speeds are about 180 m/sec. Thus when the ionosphere is highly structured, as it usually is in the polar cap, many techniques can be used for velocity estimates.
At midlatitude sites the ionosphere is usually not highly structured and therefore measurement of velocities is not as easy. Also the convection velocities are an order of magnitude smaller than they are in the polar cap, and this also makes measurement more difficult.
If, however, one could observe convecting discrete ionisation structures, even though weaker than those in the polar cap, then it should be possible to make reliable convection measurements. Therefore we used the capabilities of the digital ionosonde to look for evidence of convecting structures.
The most obvious evidence of some sort of structuring in mid latitudes is the spread-F which is commonly observed, particularly at night. However midlatitude spread-F may be due to 'corrugations' of a relatively flat surface rather than due to discrete structures [Bowman, 1991]. Thus even spread-F is not conclusive evidence that there are discrete structures.
Much of our search for discrete structures centred on searching for their signatures on plots of Doppler shift versus time. If a discrete reflector is convecting with a horizontal velocity, V, then its Doppler shift, fD, as a function of time is will be, fD » -2V2t/l h ( where l is the radio wavelength, and h is the height: this result assumes the reflector passes overhead at time t = 0). Thus on a plot of Doppler shift versus time a discrete reflector would show as a negative sloping trace. We searched many hours of Doppler versus time plots for evidence of such sloping traces. Most of the plots looked like Figure 2 and revealed no evidence of any convecting reflectors. We tried varying the Doppler and time resolution but produced no better results.
Many of our plots showed features like those shown in Figure 3. Here there are 3 cycles of what is probably a gravity wave effect. The midlatitude ionosphere is continuously perturbed by gravity waves propagating from various sources. These gravity waves can cause large perturbations of the Doppler shift and angle-of-arrival of the echo. If there were weak convecting structures these gravity wave effects would make it difficult to observe them. For Figure 3 the peak-to-peak variation in Doppler shift is about 0.55Hz and using 20 minutes as the period of the wave, the variation in height is about ±1.8 km. This is not a large effect, and the maximum velocity perturbation corresponding to the Doppler shift is only about 10m/sec. However, because the perturbation is in vertical height, it is therefore in a direction that is easily detected by the ionosonde radar. A 10 meter per second horizontal speed would usually produce only a quite small Doppler shift if the reflector is near to overhead.
Before leaving Figure 3, note the slanting feature just to the left of the line which has been drawn on the figure. The apparent speed of approach of this feature can be determined from the slope of the Doppler shift versus time. For this feature the speed is 201 m/sec. This is a typical speed for gravity waves so presumably this slanting feature is due to reflection from a part of the wave where the geometry is conducive to returning a reflection rather than from a convecting discrete structure.
The only strong convecting feature that we observed from many hours of recordings of Doppler shift versus time is shown in Figure 4. This is an E region feature at 110km height. From the slope of the trace the speed can easily be determined and is 102 m/sec. This result assumes that at closest approach the feature passed overhead. For this same feature a plot of position of the feature, using angle-of-arrival measurements as a function of time, gave another measure of the velocity, and this measurement gave the speed as 99 m/sec and the direction of motion was almost exactly towards magnetic south. These two measures of the velocity are in good agreement and confirm that if we could observe the signatures of convecting structures we could easily measure their velocity. We are suspicious that even the feature seen in Figure 4 is not a convecting structure but is associated with a wave. This is because the velocity is inconsistent with the known properties of convection at this location [MacDougall, 1981].
One reason why we were unable to observe evidence for convecting discrete structures in plots such as Figures 2 and 3 might be that the quite strong specular reflection, and its spread over neighbouring Doppler shifts, due to the finite sample length, might be obscuring the echoes if they are quite weak. We therefore looked to see if there might be weak partial reflection echoes in the ionosphere. To do this we chose a frequency a bit above the E region critical frequency, but usually below the F region critical frequency, and looked to see if, at a height between E and F region, there were any weak echoes. Although we used many samples and processed them using various techniques we found no evidence for partial reflections unless they were weaker than -40 dB with respect to specular reflection.
From our measurements we concluded that there was no evidence that in mid latitudes there are readily observable discrete convecting reflectors in the ionosphere. Of course if the reflectors were extremely weak, or were short lived then we would not have been able to detect them. The lifetime of F region irregularities should be of the order of hours so it is unlikely that short lifetimes would be making them unobservable by the methods we used.
Mid latitudes scintillations are evidence that sometimes there are discrete irregularity structures present in the ionosphere. Unfortunately we were not making scintillation measurements during the time of this study, and therefore could not investigate whether there are observable discrete structures during intervals when scintillations are observed.
We conclude that discrete reflecting structures are, at best, rarely observable. Therefore methods of measuring convection based on the presence of such structures should not usually work. Both the standard methods of measuring convection are based on the assumption that convecting structures are present. The two methods are: correlation analysis [Briggs et al., 1950; Phillips and Spencer, 1955] and 'Doppler imaging' [Adams et al., 1986]. The second method is the same as the 'skymap' method used for the polar cap measurements mentioned earlier.
We have used both these methods to analyse many hours of data and both methods give reasonable velocity measurements a large fraction of the time. The Doppler imaging method works better than does the correlation method, but we suspect that this is because our antenna spacings favour the Doppler imaging method. Reinisch et al.  also show reasonable ionospheric velocity measurements from mid latitudes using the Doppler imaging method. Thus although there is no evidence of discrete convecting structures, velocity analysis based on the assumption that such structures are present appears to often produce satisfactory results.
This immediately raises the question: Since the velocity measurements would not work unless there were convecting structures, why were we unable to detect such structures? Possible explanations are that reflections from the structures were either too weak to be observed by our methods, or that at any time there are such a large number of echo components that one cannot resolve individual reflectors. In hopes of solving this enigma we plan on repeating the search for convecting structures with increased resolution.
The CADI ionosonde is in use at a network of 5 Canadian stations. We now have more than a year's data from some of these instruments. Convection measurements at polar cap locations are usually possible using Doppler shift and angle-of-arrival data. This is because of the highly structured nature of the ionosphere at these locations. At mid latitudes our measurements fail to identify any signatures of discrete convecting structures. However standard convection measurements using methods based on the assumption that such structures are present work well a large fraction of the time.
The Canadian Network for Space Research supported this research.
Adams, G. W., D. C. Brosnahan, D. C. Walden, and S. F. Nerney, Mesospheric observations using a 2.66 MHz radar as an imaging doppler interferometer: Description and first results, J. Geophys. Res., 91A2, 1671-1683, 1986.
Bowman, G. G., Ionospheric frequency spread and its relationship with range spread in mid-latitude regions, J. Geophys. Res., 96 (A6), 9745-9753, 1991.
Briggs, B. H., G. J. Phillips, and D. H. Shinn, The analysis of observations on spaced receivers of the fading of radio signals, Proc. Phys. Soc. London Sect. B, 63, 106, 1950.
Cannon, P. S., B. W. Reinisch, J. Buchau, and T. W. Bullett, Response of the polar cap F region convection direction to changes in the interplanetary magnetic field: Digisonde measurements in northern Greenland, J. Geophys. Res., 96 (A2), 1239-1250, 1991.
MacDougall, J. W., Measurement of ionospheric electric field convection by the Long-Line technique, J. Geophys. Res., 86(6), 4781-4789, 1981.
Phillips, G. J., and M. Spencer, The effects of anisotropic amplitude patterns in the measurements of ionospheric drift, Proc. Phys. Soc. London Sec. B, 68, 481, 1955.
Reinisch, B. W., K. Bibl, D. F. Kitrosser, G. S. Sales, J. S. Tang, Z. -M. Zhang, T. W. Bullet , and J. A. Ralls, The Digisonde 256 ionospheric sounder, WITS Handbook Volume 2, Ed. C. H. Liu, SCOSTEP, 1989.
Wright, J. W., and M. L. V. Pitteway, Real-time acquisition and interpretation capabilities of the Dynasonde 2. Determination of magnetoionic mode and echolocation using a small spaced receiving array, Radio Sci., 14, 827-835, 1979.