Instrumentos de recolección de datos 56 

The meteor radar used for our study is the Canadian Meteor Orbit Radar (CMOR) described by Jones et al. (2005) and Webster et al. (2004), and is based on the commercially avail- able SKiYMET system (Hocking et al., 2001). CMOR is a three frequency system located at 43.264◦N and 80.772◦W, operating at 17.450, 29.850, and 38.150 MHz, although only 29 MHz is used in our study. Basic technical specifications of the radar system pertinent to this study are given in Table 2.1. While recent upgrades have changed the system specifications (specif- ically an increase of the transmitter power on 29 MHz), the data here were recorded when all frequencies were operating at 6 kW peak power, and no further measurements have been made since the 29 MHz system changed to 12 kW to avoid introducing configuration specific biases. The system features a five channel interferometric array on 29 MHz with reception blanked in phase with the transmitted pulse. The antennas are set up as two linear and orthogonal

arrays, following Jones et al. (1998). The main station is named Zehr (located near Tavistock, Ontario), while two remote stations, named Thames (8 km from Zehr) and Gerber (6 km from Zehr), are used to compute time-of-flight velocities (see Jones et al., 2005).

To avoid issues of mutual coupling, the antennas are placed 2.0λand 2.5λapart, although this means that any specific pair cannot uniquely determine the meteor echo angle-of-arrival, but combinations can, as discussed in Section 2.2.4. LMR400 cable is used to connect each antenna to its receiver. According to the cable specifications (Times Microwave, online), at 29 MHz this cable has 2.2 dB loss per 100 m with a phase stability of less than 10 ppm per degree Celsius.

Table 2.1: CMOR technical specifications for 29 MHz system

parameter value frequency 29.850 MHz receive cable LMR400 range interval 15−255 km range resolution 3 km coherent ints. 1 pulse length 75µs pulse frequency 532 Hz peak power 6 kW skynoise floor −107 dBm dynamic range 33 dB daily meteors 8000 daily orbits 2500

beam gainGT +GR 9.8 dBi

beam size 55◦to−3 dB point

2.2.2

Radar Detection

For our experiments, the radar system was configured to stream all data directly to disk. This was done to bypass the SKiYMET detection software which is well optimised to reject meteors that are not suitable for mesospheric wind measurements, but which may otherwise be suitable for astronomical measurement purposes. Examples include overdense echoes, where the elec-

Figure 2.2: An example radar meteor echo (event 20070421 034559) at rangeR0= 108.11 km

withθ= 27.7◦(echo direction angle measured from the zenith) andϕ=175.3◦(echo direction azimuth measured North of East). The time-of-flight offsets between the main and remote sites are shown as vertical lines, as well as Fresnel amplitude oscillations (see Ceplecha et al., 1998). Refer to the text for a more complete description.

tron line density is sufficiently high to prevent any penetration into the trail. These events should correspond to the brighter video events. It is also possible (although rare) to have two echoes overlap in the same range gate and produce an unusual echo profile. Recording the raw data allows for these cases to be easily identified.

As will be described later in Section 2.4.2, we required the ability to detect radar meteors independently of having video data. To automatically find echoes in the raw data stream saved during each observing session, our detection program (independent of the SKiYMET software) incoherently stacks 14 consecutive samples at each range gate at the main site (named Zehr). This stacked window size was arbitrarily chosen, but happens to be an integer divisor of our pulse repetition frequency (PRF) and was originally needed due to a software limitation. Each sample is one in-phase and quadrature pair per channel. We then search for an increase of 8.0σ above the background (determined from average background noise before a meteor echo). This is the same detection threshold that the SKiYMET software is configured to use during normal

system operation, and past experience with our system indicates this is a reasonable threshold to avoid false triggers due to noise. The background is computed (after the receiver DC biases were removed) from the average in-phase and quadrature values in the range gates correspond- ing to ranges of 195−225 km and is re-computed for each one second of data. It is possible for overdense echoes at these ranges to affect the background value, however, it is recomputed each second to minimise such effects. Non range-aliased underdense echoes found in this range interval will have an echo zenith angle greater than 60◦. We do not filter the detection list based on interferometry phase residuals, echo rise time speeds, or underdense echo decay times, in order to maximise the number of possible detected simultaneous meteors.

Fig. 2.2 shows an example of a radar meteor echo as an amplitude versus time plot for a single range gate for the main station and two remote stations. Note that both pre-t0and post-t0

Fresnel oscillations (see Ceplecha et al., 1998) are visible. The frequency of these oscillations depend on the radar wavelength, echo range, and meteoroid speed, so knowledge of the first two allows for an independent determination of the third. The plots are offset vertically for ease of viewing, and plotted as relative amplitudes since the scales for each station do not have the same amplitude-power calibration. The bottom most profile shows the amplitude for Zehr, while Thames (located 8kmfrom Zehr) is plotted above Zehr, and Gerber (located 6 km from Zehr) is plotted above Thames. The vertical lines represent thet0 point for each station which

correspond to the specular point along the trail (see Weryk and Brown (2004) for details on how these are determined).

2.2.3

Receiver Calibrations

While the pulse repetition frequency (PRF) (and hence, the relative time error) is very stable, we have noted from experiment that the absolute time error of the system can be offby up to one second. To account for this, the radar observations are synchronised against a GPS receiver by transmitting small pulses at ten minute intervals from a simple dipole located near the RX antennas, and these pulses are then detected in the streamed data. This calibrates the time

results (discussed in Section 2.2.4). The phase delays are measured by disconnecting each an- tenna from its receiver cable, and then connecting the cable to the frequency synthesizer unit (FSU) using a common test cable that goes out to the antenna. It is not practical to calibrate the receiver phases after each observing session, but they are generally stable over long periods of time as verified by radiant measurements of the major showers. In the absence of cable/con- nector hardware changes, the phase calibrations are generally consistent to a few degrees over yearly timescales.

Another source of uncertainty is the slightly non-coplanar antenna arrangement. Seasonal variations have caused each antenna to shift in location which causes discrepancies in the inter- ferometry solutions for common echoes between the frequencies using our chosen interferom- etry algorithm (described in Section 2.2.4). Table 2.2 lists the relative positions of our antenna array for 29 MHz, measured with a theodolite accurate to seconds of arc, and these positions are used by our interferometry algorithms.

The initial range calibration is performed through the vendor software and was done at the time of the radar installation in 2001, and verified in 2009 to be unchanged. The range calibration works by transmitting a known pulse into the receiver and computing the time offset from the expected time.