I was setting up for a jt9 session on 15 meters, when I heard the familiar quick chirp of an oblique ionosonde sweep by. I'd played with these signals many years ago, programming my hp65 calculator as a timer to sort out the various stations in their mostly regular repeating schedules. It struck me that these signals would make an interesting subject for my continuing phase analysis project.

Many hams interested in propagation are somewhat familiar with the results of vertical incidence ionosondes, which shoot a pulsed signal vertically and time the echo returns to measure the height and density of the various ionized layers in the ionosphere. From the resulting ionograms you can measure the height of the F layer, for example, and use this to help predict propagation - if your signal happens to bounce off the ionosphere not too far from the ionosonde station.

There is also a world network of oblique ionosondes which transmit frequency-swept signals at oblique angles to receiving sites other than themselves. Analysis of these signals yield information on the state of the ionosphere between the transmitting and receiving stations. The advantage is that with n stations, you would get n vertical profiles, but n^2 paired midpoint locations. Additionally these midpoints can be in the middle of an ocean where it is difficult to put a permanent station. These days ALE is more likely to be used for propagation optimization of communications links, and the ionosphere is more easily measured using the GPS satellite signals, but the ionosondes continue to be used.

My Flex receiver was set at 24.932 MHz and with a little listening I determined that there were a number of chirps that repeated almost exactly 12 minutes apart. I selected the strongest one and recorded about 20 seconds around the expected signal times for three successive 12 minute times, getting three nice sweeps through the 48 KHz bandwidth of the receiver. Waterfall plots of these three sweeps:

Waterfall of three chirps, 12 minutes apart.

Interesting. Back in the old days when the period of the chirps was usually 15 minutes, they used analog setups in which the frequencies were continuously swept from somewhere around 1 MHz to maybe 25 MHz at a rate of 100 KHz per second. This looks like they must now be using more modern digital devices since the sweep stops a couple of times and resumes again a fraction of a second later. It took me quite a while to convince myself that these interruptions were real and not an artifact of my receiver or analysis chain. I eventually found a chirp signal that was continuous and so reassured myself that all was well.

Now we cover a frequency range of almost 48 KHz, which is much wider than the Khz or so that we were able to get on the WWV signal phase analysis in an earlier post. The downside is that this gives only a single value at a frequency in the instant that the chirp produces it before moving on. But at least we can get an almost instantaneous slice of the phase dependence on frequency. Indeed the point of the ionosondes is that the receiving station at the end of the frequency sweep has a complete map of the ionosphere's response from 2 MHz to 25 MHz (or whatever the range of the sweep is). With some fancy programming control of the Flex we could also sweep it to match the signal and get a complete sweep ourselves. But that's a project for another day.

Onward with the phase analysis of what we do have. Now I needed to set up a sweeping local oscillator that was synchronized with the signal and matched the sweep rate. I decided to analyze each of the two complete segments on the three signals. It was a tedious couple of hours, but finally I had the six segment results. Here are plots of the those results:

Some comments are in order. There are a lot of free parameters chosen to produce these plots. The starting phase for each segment is arbitrary, as is the starting time of the segment. The sweep rate was chosen to be the same for all segments ( 104.791 KHz/sec ) but this particular value was chosen to give nice plots. "Nice" in this context means that the phase on average was flat.

On all plots (including those in the previous posts) if my software local oscillator frequency is off, then the phase exhibits a slope linearly dependent on time as the phase error steadily accumulates. If the sweep frequency of the local oscillator is off on these signals, then the phase exhibits a curvature - quadratic dependence on time - as the phase error accumulates at an increasing rate as the frequency error increases with time. Judicious choice of the sweep frequency and its starting value will yield relatively flat plots, and that's how I chose the values to give the above plots.

For example, I chose the starting phase of the second segment to make it continuous across the gap between the segments. This is probably reasonably accurate - notice that the sinusoidal oscillations seem to be nicely continuous if the gap is simply ignored. But I chose the starting time (and hence frequency) to make the plots flat on average. Actually I stopped once I got reasonably close as you can see from some of the segments that aren't very level.

These arbitrary parameter choices of course means that we can't really learn much about how the actual values of these parameters change with propagation. In the above plots it could be that there is a different doppler shift induced frequency shift for the different chirps, but I've taken that out with the choice of starting frequency and so I'd never see it.

It might appear that I could at least get the frequency correct by measuring it from the signal. The problem is that the Flex5000 tunes in an irregular fashion, so exactly what frequencies the 48 KHz passband covers is not accurately determined. This irregularity arises from the discrete tuning characteristics of the DDS chip they use for the local oscillator. I've never seen a detailed explanation of the algorithm used. The DDS chip is known and its tuning characteristics are nicely specified, but just what Flex does in the software control of the signal lines is not known. Some day I may have to take the time and figure this out, but I'm not looking forward to it.

I tried tuning the receiver to a frequency so that one of the WWV signals was included in the 48 KHz passband, and since I knew that frequency accurately I could calibrate the digitized signal, as well as read off accurate times. Unfortunately it appears that the chirp signals are good citizens and they apparently suppress their signals around the WWV frequencies. I could have tuned the second receiver to a WWV station and hence at least got the time (but not the frequency of the other receiver) accurately.

My hope is that the new Flex6700's improved capabilities will allow precise determination of some of these parameters. The GPS will presumably at least give very accurate time and frequency references. Its a long shot, but it might even provide GPS carrier phase information that would allow measuring TEC (Total Electron Content) of the ionosphere between the satellites and my station directly. That would be cool !

Tangled Spaces