BPSK Adventures

Bill Cantrell, “TEXAS” & WD5CVG

April `99

Converted to HTML (with permission) by K3PGP - rev 1.0d

What an incredibly fun experiment! During the Thanksgiving Holidays, I had the opportunity to take the van on a 300-mile road trip while receiving BPSK from the TEXAS beacon. The results were surprising. This is a report on my set-up and findings.

I have wanted to test the range of a LowFER beacon using differential BPSK ever since I read about Max Carter’s mobile experiments [1], [2] in 1991. Frequency accuracy is crucial. Rather than phase-lock to WWVB, I decided to use atomic clocks to discipline my TX and RX equipment. I have two Rubidium standards that operate at 10 MHz. The frequency error between the two is 0.3 parts per billion. One is used to derive the carrier frequency for the TEXAS beacon transmitter. It also provides the clock rate for the BPSK keyer [3], [4]. Therefore, frequency, modulation, and timing are all controlled precisely at the transmitter. The other clock is used in the van to provide the master oscillator signal for the Kenwood TS-850S receiver and the frequency reference for the 1.8432 MHz SD-ADC board [5].

The Equipment

My mobile test set-up is shown in Figure 1. Two large marine storage batteries provide all power for the receiving equipment in the van. One battery supplies 12Vdc for the antenna preamp, receiver, demodulator, and laptop computer. The second battery is connected in series to provide 24Vdc for the atomic clock. Note that no connections are made between the van’s electrical system and the receiving equipment. And only one ground connection is made between the two systems at the base of the antenna. Originally, I had planned to use the car’s 12V electrical system and build a 24Vdc switching power supply. That would have eliminated the need for the bulky marine batteries. But this set-up seems to pick up far less vehicle/engine/computer noise. Therefore, I decided to stick with the marine battery approach. They have to be charged periodically, but at least I don’t have to worry about running the van’s battery down. The atomic clock and the 1.8432 MHz synthesizer are powered continuously for optimum stability.

The antenna is a Burhans preamp mounted on the luggage rack. As previously mentioned, the preamp is grounded to the roof. This is the only ground point between the receiver and the vehicle chassis. A 75-ohm coax runs from the preamp output to the bias-tee located inside the van. The bias-tee sends +9.8 Vdc up the coax to power the preamp. A +9.8V (Motorola LM317T) regulator is mounted inside the bias-tee.

The receiver is a Kenwood TS-850S/AT. It operates from +12V and the receive frequency is set to the exact carrier frequency (189.900 kHz) in the CW mode. There are few, if any, computer “birdies” heard while tuning around the band. If the receiver were operated from the van’s 12V system, the birdies would be far stronger, and alternator-whine might be a problem. The TS-850 is an ideal choice for BPSK reception because all of its internal local oscillators are phase-locked to a common internal master oscillator. This includes the 800 Hz BFO. The TS-850 master oscillator operates at 20 MHz. This made it easy, for all I had to do was double the output of the 10 MHz standard to get the correct frequency, and then override the internal oscillator. No need for another synthesizer. I used a full-bridge rectifier consisting of four hot-carrier diodes followed by a 20 MHz LC tank-circuit to double the 10 MHz signal. The result is a full-wave rectified sinusoid at 20 MHz. Obviously it has some harmonic content, but the TS-850 doesn’t seem to mind at all.

The TS-850 can be tuned in 1 Hz increments by using the fine-tuning mode and the up / down keys located on the microphone. (The TS-850 readout only displays down to 10 Hz, so you have to count the number of button presses to keep track of the 1 Hz steps.) This receiver has the optional Y6_455_C-1 I.F. filter (BW = 500 Hz). The I.F. bandpass filters were configured as follows: 8.83 MHz 2nd I.F. filter selected for a 2.7 kHz BW, 455 kHz 3rd I.F. filter selected for 500 Hz BW. No variable cut-in filtering, No notch filtering, Noise Blanker #1 was “on”. No attenuation switched-in. No AIP.

For the demodulator, I use Bill de Carle’s sigma-delta (SD) board & DAC board [5]. The SD board has a 1.8432 MHz crystal to provide a clock for the demodulator. Obviously, this oscillator needs to be disciplined to the atomic clock, just like the TS-850. Therefore I modified the SD board clock circuit to create a 1.8432 MHz synthesizer. More on this later.

The computer is a Hewlett-Packard OmniBook 5700CTX laptop (166 MHz Pentium). I suspect that the van tends to form a Faraday Cage around the laptop computer, so that radiated emissions from it do not interfere with reception. I also use ferrite beads on all laptop cables to suppress common-mode radiation from the cables. The laptop is booted in DOS mode by pressing F8 at the “starting Windows 95…” announcement, and then selecting option “6”. I am running Bill de Carle’s AFRICA BPSK software, Version 1.8, available via [6]. Be aware that this program will not run properly if you try to enter DOS from Windows 95. Reboot in DOS only, not Windows 95. The AFRICA settings were as follows unless specified otherwise: MUTE: on, HANG: off, ET1: on, RCV: on, TC: 50, “SYNC 29”, DECO: COH, AUTOTRACK: on, GRAB 12:1 (no grab). The TEXAS Beacon was set to send a continuous twelve character repeating BPSK message at MS100, ET1, without CW interruption: “TEXAS(space) AGGIE(space)“

1.8432 MHz Synthesizer Design

The block diagram for the 1.8432 MHz synthesizer is shown in Figure 2. See Motorola Technical Application Note AN535 “Phase-Locked Loop Design Fundamentals”, available via [7]. It provides an excellent overview of PLL design techniques.

The various blocks shown above will be discussed in detail. A HC4040 divides the 10 MHz standard by 3,125. This results in a pulsed output at 3200 Hz, known as the reference frequency. Similarly, another HC4040 divides the buffered output at 1.8432 MHz by a factor of 576. This also results in a pulsed output at 3200 Hz. (Bench testing suggests that the HC4040 can be used for input frequencies up to about 25 MHz.) The two signals at 3200 Hz are compared using a Motorola HC4046 PLL/phase detector. Why choose 3200 Hz as the reference frequency? The choice is simple. This is the largest common-factor for both 10.000 MHz and 1.8432 MHz. The output of the phase detector drives the loop filter so that the crystal VCO becomes phase-locked to 1.8432 MHz exactly. Hence, we have a 1.8432 MHz synthesizer with the same accuracy as the 10 MHz standard.

The crystal VCO circuit is shown in Figure 3. The two varactor diodes in parallel allow the crystal to be pulled at least +/- 20 ppm. This forms a crystal VCO. The varactor diodes are M/A-COM part number MA4ST320-287T, but just about any varactor diode will do. Each diode’s capacitance is about 58 pF at 0.5 Vdc and 23 pF at 3.0 Vdc. This circuit has the frequency characteristics shown in Figure 4. Hence, Kv is 206 radians/second per volt (33 Hz/volt), as calculated from the slope. A crystal-pull range of +/- 20 ppm over temperature should be enough to maintain lock. The 1 mH coil and 6 pF cap are chosen to go parallel resonant at 1.8432 MHz.

The divider circuits were constructed as shown in Figures 5(a) and 5(b) using the HC4040 and common axial-leaded silicon diodes. Diode placement is determined by the binary representation of the divisor. The circuits shown can be socketed.

The loop filter is shown in Figure 6. The circuit was modified by adding an op-amp to increase the gain of Kv. Normally, Kv is ~3 MHz per volt when the HC4046’s VCO is used, but in this case, it is only 33 Hz per volt using the crystal VCO. This slows the lock-time considerably.

The gain of the op-amp (MC33072) was chosen to be Kx = +10. If larger values of gain are tried, the loop filter drives the op-amp into saturation, negating the benefit. A gain of 10 increases Kv to 330 Hz per volt. This reduces the lock-time by allowing for a larger design value of .

The target value for zeta () is 0.707. This should result in a slightly underdamped (ringing) response as the circuit attempts to lock. An overdamped response is not desirable. If the circuit is somehow perturbed from a locked state, an underdamped response will average “around” the desired frequency, as lock is obtained again. Based on the design equations in [7], the loop’s natural frequency () should be chosen to be less than one-tenth of the reference frequency. Normally, would be 2 320 rad/sec (fn = 320 Hz) in this case. However, this results in negative values for the loop filter resistors (because Kv is so small). A Mathcad spreadsheet was created to determine an appropriate value for the natural loop frequency. was finally chosen to be 2 0.32 rad/sec (fn = 0.32 Hz). With this value, its gonna take a while to lock!

The design parameters are as follows:

Kv = 206 rad/sec, Vdd = 6.0V, = 0.707, = 2 0.32 = 2.01 rad/sec, N = 576,

Kx = 10, = (Vdd / 4 ) = 0.477 volts per radian, Let C2 = 2.0 uF = 2.0 E -6 farad

Using this circuit, the equations for the loop-filter resistors are:


Therefore R2 and R1 are computed to be 59k and 152k, respectively. No doubt, this circuit could be optimized further. The actual loop frequency appears to be ~ 0.166 Hz, and the circuit appears to be very much underdamped. After the synthesizer is powered-up, it actually takes about 3 to 4 minutes for the circuit to ring-out and “lock”. However, the circuit remains locked, and performance is excellent.

The purpose of Rx and Cx is to form a low-pass filter with a corner frequency (38 Hz) that is much higher than the natural loop frequency (0.32 Hz). This filter eliminates higher-frequency noise generated by the loop filter resistors and the op-amp, without changing loop filter characteristics. This should result in less phase-noise on the 1.8432 MHz waveform.

On the Road

Several hours prior to leaving, I fired-up the equipment and allowed the 1.8432 synthesizer to lock. A quick check with a scope verified that it was operating okay. I allowed the recovered message on the laptop to synchronize, turned-off the sync, and made sure that the message was correct:


The AFRICA software has a relative signal strength indicator called the “measuring amp”. It was reading ~100 in the driveway. I disconnected the antenna from the receiver. This resulted in loss of copy with the measuring amp dropping to ~0.8. Next, I used the GRAB feature. Sure enough, after a few minutes, the proper message emerged from a garble of characters on the screen:


I then reconnected the antenna, set GRAB back to no-grab, and we started out that morning. We headed south from Fort Worth. Destination: the Texas Gulf Coast, about a five-hour drive of 300 miles. The weather was perfect for the trip, 70 and sunny, without a cloud in the sky.

The first 100 miles was uneventful -- solid copy the entire way to Waco, Texas. The measuring amp was reading ~25 by now. After stopping for fuel just a few miles from the infamous site of the Branch-Davidian compound, I activated sync to give the signal a chance to re-sync. But the Startbit remained on the same number that it had been on for the duration of the trip - Startbit “11”.

What about Doppler effects? Max seemed to have run into problems with Doppler a few years ago. Well, I had been travelling at 80 mph, which would have imparted a 23 millihertz frequency shift to the received signal at 189.900 kHz. Nothing to worry about unless using extremely large values of GRAB at slow baud rates. Reception was solid the whole time - just the occasional garble of a character when passing under an overpass or high-tension power line:


By the way, 80 mph on the interstate is slow by Texas standards. I was in the slow-lane to keep from impeding traffic. This is a big state and we don’t wanna take all day to get there. All of the mileage listed is straight-line (airline) distance from the transmitter.

At 145 miles, near Calvert, Texas, the signal suddenly went from almost perfect copy to a garbled mess. I noticed that the power lines along one side of the road were closer now. We were no longer on Interstate 35, but on the farm roads. Perhaps line noise or a PLC was causing problems, but more likely, it was the Faraday Shield effect of the overhead wires. I tried using some GRAB to recover the signal. At 152 miles, GRAB 12:2 was tried, but it wasn’t good enough for solid copy. GRAB 12:3 was used at 155 miles and good copy was obtained. But at this point, the power lines were gone.

At 164 miles, I was back to partial copy again. Something was going wrong fast. I tried GRAB 12:5 but no improvement. That amount of grab should not be needed yet. It looked like the display was jumping out of sync for just a few characters, as some were correct but others were not. I tried switching off the AUTOTRACK feature. This solved the problem and solid copy returned. I reverted back to GRAB 12:1 (no grab). Remarkably, the copy was still solid without error! This was getting spooky. The TEXAS beacon is a powerhouse [8] by LowFER standards with its whopping 4.6 milliwatts ERP, but gee, solid copy at 175 miles, real-time, no processing gain? Wow!

At Bryan, TX, (180 miles) the screen was showing an occasional missed character - not quite solid copy. The measuring amp was reading ~14. I went to GRAB 12:3 and the copy was solid again. And it remained solid as we drove through the middle of College Station, home of world-renowned Texas A&M University. Gig `Em Aggies!

At 200 airline miles, the measuring amp was reading ~9 to 12. Partial copy without grab, solid copy with GRAB 12:3.

At 222 miles, Kennedy, the measuring amp was ~7 to 11. Partial copy without grab, solid copy with GRAB 12:3.

At 231 miles, Bellville, it was necessary to go to GRAB 12:7 while in town due to all of the power lines. The measuring amp reading was as low as 6.

At 245 miles, Sealy, and at 271 miles, Hungerford, copy was solid using GRAB 12:7.

Finally at 304 miles, my destination in Bay City, TX, copy was solid using GRAB 12:13, provided that I was parked away from the house because of power lines and 40 ft trees nearby. The Startbit remained at “11” for the whole trip. No need to re-sync at all. Copy degraded considerably after sunset (skywave/QSB effects?)

The next afternoon we decided to drive to the Gulf of Mexico. This would give me a chance to test the range a little farther. We stopped at the Intracoastal Canal (325 miles). The drawbridge was up to allow a barge to pass on the waterway. Copy was solid with GRAB 12:13.

Our final stop was Matagorda Beach. This is where the famous French explorer LaSalle landed in 1685 to claim the region for France. His tale is a tragic one, however. Storms destroyed his ships and most of his supplies during the landing. Hostile Indians and disease finished the job. Only 6 members out of his crew of 100 survived and he was killed by his own men.

This year, four hurricanes have swept the area, resulting in the destruction of the docks. A lot of debris has washed-up on the beach. The airline distance from the TEXAS beacon to this location was 330 miles. For me, there was no way to go any farther south without a paddle. It was now 5 PM local time the beginning of a pretty sunset. The beacon was partial copy with GRAB 12:7, and solid copy with GRAB 12:9. The measuring amp was indicating ~ 5.5 to 7. I noted that GRAB 12:1 (no grab) would sometimes produce solid copy for a few words, and then become garbled for a few words - sorta like slow QSB or maybe skywave multipath. I decided to try the AFR decoder, now that we were stopped. (By the way, you can press “Shift Tab” to toggle between the two decoders.) Amazingly, the AFR decoder was solid copy with GRAB 12:1 (no grab). That means that I was receiving the beacon without any processing gain. I was receiving real-time, character for character. Simply amazing! Of course, the beach is a perfect DX location no trees, no buildings, and no power lines. I also tried the AFR decoder with GRAB, but this made things worse, and garbled the message regardless of autotrack setting. At this particular location, AFR no-grab was equivalent to COH GRAB 12:9.

A few days later, we started on the return trip. There were local thunderstorms, so I had no copy for the first 40 miles or so. It wasn’t until I reached a farm road devoid of all power lines and trees that I was able to copy the signal again (260 miles). And it was solid copy at this point!

Later in the return trip, near College Station (185 miles), I was having trouble with the COH decoder. I switched over to the AFR decoder and solid copy returned. This was an example of where the COH decoder would not give good copy, but the AFR decoder worked well using GRAB 12:3. I was in motion at about 70 mph at the time, with a large power line nearby. The AFR decoder did a fine job in this instance.

For differential BPSK, each bit-time is 100 milliseconds long using the MS100 baud rate. There are 16 bits per printed character using ET1 error correction. It should be noted that Bill’s software looks at the BPSK phase change from one bit to the next, on a bit-by-bit basis, using the COHerent decoder. Since each bit is 100 msec, this is about how long the signal’s phase must stay stable for COH to work.

The AFRica decoder is more sophisticated. It looks at the entire 16-bit frame (for ET1) as if it were a single configuration of bits or a “waveform signature”, so to speak. The software selects the best match from an internal table of 128 possible bit-configurations or “signatures”. Each signature represents a different character to display. The phase must stay stable during the entire length of the frame, a factor of 16 increase.

AFR should be superior to COH, unless there is a frequency offset between transmitter and receiver, or a variation in propagation path delay during a frame. (Frequency errors were eliminated by using the atomic clocks.) The reader may have surmised that there should be far more than 128 signatures available for 16 bits. The few signatures that were actually selected for usage were chosen for maximum orthogonality with respect to each other. This is a form of bit-redundancy, and it provides error correction [9] during the decoding process. It’s the classic trade-off -- extra bits take more time to transmit, but result in fewer errors during receive.


A few conclusions can be made regarding mobile BPSK reception at LowFER frequencies:

1) Use separate battery power in the mobile to reduce noise problems. Do not use the vehicle’s electrical system. Tie to chassis ground at the base of the antenna only.

2) Try to stay on the Interstate highways. This forces utility right-of-ways and trees farther away from the road surface. Power lines are your enemy. Trees are your enemy. I found places where overhanging trees would destroy copy, although no power lines were in the vicinity. These objects act like a Faraday Shield at LF.

3) Hilltops and height are not necessary for good reception.

4) Try to phase-lock all local oscillators in your receive-chain to a common master oscillator. Get the best master reference that you can obtain.

5) Try both the COH and AFR decoder, with and without GRAB, whether in motion or parked. (You can press “Shift Tab” to toggle between the two decoders.) One works better than the other at times, but it is not yet clear as to which is best under certain conditions. Some bench-test simulations are needed to determine Eb/No performance, which I plan to do this summer. Signal path conditions (multipath, fading) and frequency error will no doubt be determining factors. Bill de Carle’s BPSK decoder software is an incredibly powerful tool.

Of course, only a “radio-guy” would be in the van playing with the radio, instead of enjoying the beach, but there I was, parked on the beach, with the laptop printing on the screen in real-time:


I knew that my pipsqueak transmitter was sending out this message from 330 miles away. There was no discernable audio in the speaker, for the signal was several dB below the noise floor, yet there it was... This is the incredible mystery and fascination that is “Radio” -- what a wonderful hobby we have!


[1] The LOWDOWN, Mar 91, “BPSK BITS”, pp. 12-13, by Max Carter.

[2] The LOWDOWN, May 94, “BPSK BITS”, pp. 30-31, by Max Carter.

[3] The LOWDOWN, Apr 96, “Simple Beacon Keyer”, pp. 30-33, by Bill de Carle.

[4] The LOWDOWN, May 97, “A Dual-Mode Beacon Identifier”, pp. 21-22, by Lyle Koehler

[5] QST, Jan 92, “A Receiver Spectral Display Using DSP”, pp. 23-29, by Bill de Carle.

[6] Bill de Carle’s AFRICA Ver 1.8 available at: http://www.ietc.ca/home/bill/bbs.htm

[7] Motorola SPS Technical Application Note AN535 “Phase-Locked Loop Design Fundamentals”, available via: http://mot2.mot-sps.com/cgi-bin/dlsrch

[8] The LOWDOWN, Oct 98, “LowFER Antenna Equivalent Circuit Measurements”, p. 21, by Bill Cantrell.

[9] The LOWDOWN, Dec 95, “Latest Coherent Uses Lattice Coding”, pp. 28-29, by Bill de Carle.

Contents of this website are (c)1997-99 of K3PGP and of the originating authors.

NOTE: You may link to this page but you may NOT copy without permission from the author and K3PGP

To contact the author of this article: Email Bill Cantrell