How to Plot Meteor Power Profiles from BRAMS Data

I’ve already written about meteor detection for BRAMS, so please refer to that article for background information. This post is specific for BRAMS, but it is also applicable to comparable forward scatter systems. I reiterate the filtered signal power from the mentioned article in Figure 1.

Figure 1. Filtered signal power.Figure 1. Filtered signal power.

When you have this filtered signal power graph, the next step is obviously to look at the power profile of the individual meteors (the sharp peaks). The reflected power is directly related to the physical properties of the plasma trail that was created by the meteor, so it can provide all kinds of information about it. Figure 2 shows the strong reflection that is almost exactly in the middle of the above image. It is a selection of 0.8 s in the time dimension, with the meteor itself being about 0.1 s long.

Figure 2. Original meteor power profile.Figure 2. Original meteor power profile.

This image is not what we expect, since the power profile should be a single line that indicates the received power at each point in time. The fact that the region under the graph is completely blue indicates that it is full of unexpected oscillations. If we stretch the meteor itself, we get the result in Figure 3 (the illustration shows a selection of 0.11 s around the meteor).

Figure 3. Stretched meteor power profile.Figure 3. Stretched meteor power profile.

These oscillations have a technical reason. The received carrier signal is mixed with a local oscillator to shift its frequency to about 1 kHz. This causes meteor reflections to contain several frequencies in a small range around that frequency. The implication is obviously that these frequencies are not part of the “meteor signal”, since they are simply a consequence of the experimental setup. This effect is actually quite close to what happens in amplitude modulation, although in that case the signal only contains a single carrier frequency, while the meteor reflections contain several different ones.

A straightforward way to get rid of the unwanted frequencies is to low-pass filter the signal. I’ve used a simple running average of length 41 in this case. Seen as a filter, a running average is simply a rectangular filter, and the coefficients of this filter can be computed as follows (in Python).

import numpy as np
 
h = np.ones(41) / 41

This filter can then be applied to the signal power as

s = np.convolve(s, h)

The filtered signal is shown in red in Figure 4.

Figure 4. Meteor power profile with average.Figure 4. Meteor power profile with average.

Hence, the true power profile, or at least an approximation limited by the low-pass filtering, is the following.

Figure 5. Final meteor power profile.Figure 5. Final meteor power profile.

This looks a lot more like what us old hands in radio meteors would expect…

Submitted by Tom Roelandts on 4 February 2015

Add new comment