Clifton Laboratories 7236 Clifton Road  Clifton VA 20124 tel: (703) 830 0368 fax: (703) 830 0711

E-mail: Jack.Smith@cliftonlaboratories.com
 

I've written about time variation of HF signals at my Signal Statistics page. An alternative view of the vagaries of HF radio propagation is to examine time shifts in signal arrival. If the path length changes, either due to changes in ionospheric height or changed reflection zones, the travel time from transmitter to receiver will correspondingly change.

In most cases, of course, it's not easy to determine how many milliseconds or microseconds it takes for the signal to travel from the transmitter to your receiver, or how this changes from moment to moment. However, it is possible to measure changes in transit time for signals from WWV. And, of course, there are a few special cases, such as LORAN-C, where the purpose of the station is to measure transit time. (If we can monitor the received signal phase, and if the transmitted signal is stable, we can determine timing changes by looking at phase shift, but this requires more work than the methodology I've used. WWV also transmits a 100 Hz binary encoded time signal that can be used to derive a 1 pulse-per-second signal with suitable electronics.

Reading the November-December 1959, HP Journal, I found an interesting article on comparing local crystal oscillators with WWV, using a method I had not previously read. Anyone who has listened to WWV is aware of the time "ticks." These ticks consist of 5 cycles of 1000 Hz tone (6 cycles of 1200 Hz for WWVH). Interesting, but how is that relevant to transit time? The key is that these ticks occur at a very stable interval, tied to the NIST's national time base, as is WWV's carrier frequency.

The figure below shows the tick modulation details.

Things have  changed in the precision timing world since 1959, and in 2008 the methodology laid out in the HP Journal can be used in reverse—if we have a sufficiently stable and accurate 1 PPS  time source, we can see how the propagation path changes second-to-second by watching the tick arrival time advance and retard. (It might be better to decode WWV's 100 Hz sub-carrier time data but that requires more than just general purpose test gear.)

For the precision 1 PPS timing reference, I use a Trimble Thunderbolt GPS disciplined crystal oscillator. The 1 PPS output is maintained with 20 ns or so of UTC time, and is automatically corrected for the latitude/longitude and elevation of my GPS receiver. Hence, changes in tick arrival time from second to second reflect changes in the ionospheric path, not changes in the local clock, with a high degree of precision.

The figure below shows my test setup. The TDS430A digital oscilloscope triggers on the GPS 1 PPS waveform. Not shown is a Prologix GPIB-USB adapter and a PC running KE5FX's 7470 plotter emulation program.
 

The image below shows quite stable propagation from WWV at 10 MHz. The worst change in tick arrival time is around 100 µs but most of the sweeps show close to zero arrival jitter.

When the path length changes this much over the space of 10 seconds, it's unlikely to represent a single point of reflection physically moving. Rather, it's almost certainly the result of a signal arriving in multiple rays with rays from different physical reflection points predominating during different sweeps.

This should also give us a crude view of how many paths simultaneously exist. Looking at the best 10 MHz data, there's likely only one predominant path during this 10 second interval. (It's possible that there are multiple stable paths and the signal is the stable vector sum of those multiple paths, but this seems unlikely.)  In other cases—particularly at 5 MHz during the day/night transition period—, it's apparent that there are multiple paths, perhaps as many as five or six, with each path's dominant period being a second or less. This implies that the signal statistics are more accurately described by Rayleigh statistics. At other times, two-ray Rician fading may predominate. And, at least for periods of 10 seconds, a single ray can predominate. I've discussed Rayleigh fading at my Signal Statistics page.

According to N. Hironaka and C. Trembath, The Use of National Bureau of Standards High Frequency Broadcasts for Time and Frequency Calibrations, Nat. Bur. Stand. (U.S.) Tech. Note 668, May 1975, 49 pp. (available at: http://tf.nist.gov/general/pdf/453.pdf), the leading edge of time ticks identify the start of each UTC second. Tech Note 668 recommends measuring to the second zero crossing, however, as the leading edge of the first cycle is not well defined. In this case, of course, the extra 1000 us (833 us in the case of WWVH) must be subtracted.

In addition, there is some delay within the WWV audio generation and processing chain  that must be accounted for. This information can be found in G. Nelson, M. Lombardi and D. Okayama, NIST Time and Frequency Radio Stations: WWV, WWVH, and WWVB, NIST Special Publication 250-67, January 2005 (available at http://tf.nist.gov/timefreq/general/pdf/1969.pdf). Although only WWVH measured data is provided, the report says WWV has similar delays. Since the mean audio processing delay is on the order of 17 µs, we'll ignore it.

There's also some timing uncertainty within the GPS chain, including the Trimble Thunderbolt. This error, however, is considerably less than the other errors in the system, being well below 100 ns, so we'll ignore that as well.

The last delay that must be considered is that within the 6790/GM receiver. I measured  the RA6790/GM's signal delay as 646 µs. (I used a signal generator AM modulated at about 500 Hz and measured the phase shift between the 500 Hz modulating signal and the  500 Hz recovered audio from the 6790. The phase shift is 119°and the actual modulating frequency is 511.5 Hz, as indicated in the image below. The modulating signal is on trace 1 and trace 2 has the recovered receiver audio.) Incidentally, the Publication 250-67 measured the delay of its test receiver, an Icom R10, as between 118 and 122 us, depending on frequency band selected. The 6790/GM has significantly more filtering than the R10 which accounts for the extra delay. A DSP-based receiver might be expected to have delay of several milliseconds or more, by the way.

(To verify that this shift is not actually N cycles plus 119° degrees, I started with a much lower modulating frequency.)

8164 µs corresponds to 2449 km or 1522 miles assuming the speed of light in the ionosphere is essentially that of free space.

Using the airline distance calculator at http://www.distance-calculator.co.uk/usa-distance-calculator.php#chosentown, the distance between Ft. Collins CO and Clifton VA (actually Centreville, VA, about 5 miles from Clifton) is 1466.42 miles or 2359.47 kilometers. A radio signal, of course, traverses a longer path as it is reflected from a vertical point perhaps 50 to 100 miles above the earth's surface. And, the reflection points do no necessarily occur on the great circle path, which also increases the total distance traveled by the signal.

My "time of arrival" measurement yields a propagation distance quite close to the real distance, considering the measurement method used and the uncertainty in the reflection point.