The HF Setup

The success of Andy VK3YT's PS-x balloons down in the Southern Hemisphere which used WSPR, JT9 and JT65 transmissions to provide positional data from the middle of the ocean or across a half of the world led my interest towards the High Frequency (3-30MHz) part of the spectrum. Since I was in possession of only an RTL-SDR dongle which could operate just down to 24MHz, I turned my attention to upconverters.

Eventually I found DGK Electronics' nice and simple HF converter that I thought I could make on my own. The author provided a schematic upon which I designed a board in Eagle. I made only minor changes using USB micro connector to provide power and fitting the design inside a smaller PCB.

The device is built around Mini-circuits' ADE-1 mixer and a 100MHz local oscillator. The RF input first passes through an Elliptic low-pass filter with 65MHz cutoff frequency. The low frequencies that are allowed through then mix with the 100MHz signal generated by the oscillator. This essentially moves a signal, say, at 7MHz to 107MHz which is well within the RTL-SDR dongle's range. The output of the mixer is filtered by a 7th-order Butterworth filter with a passband of 100-171MHz. The local oscillator is powered by 3.3V (regulated USB supply) and equipped with a band-pass filter as well. The Eagle files can be downloaded from here Upconverter v1.0.sch and here Upconverter v1.0.brd.

I also came across nice little aluminium boxes on Ebay that could be used to shield the PCB. It took a little fitting since it hadn't originally been designed for the box and the hand drilled openings ended up a bit rough but in the end it served the purpose sufficiently.

This specific case was 25 x 25 x 40mm for maximum two PCBs of 40 x 22.5 x 1.6mm in size. Many more cases of different dimensions can be found on Ebay.

The next piece of the setup was an antenna. Generally, lower in frequency one goes larger the antenna tends to get. At HF the wavelength, which is a measure playing a core role in many antenna designs, ranges from 10m (30MHz) to 100m (3MHz). That often leads to 10s of meters of wire hung among trees and houses. Not the way I wanted to go. Some googling time later I found Magnetic Loop Antennas. With about 1/10 wavelength of circumference and a possibility to tune the antenna with a capacitor, this seemed like an ideal space-saving solution.

I wanted the antenna to cover the 20m, 30m and 40m bands so roughly from below 7MHz to about 15MHz. A great resource in the design for me was AA5TB's website with his experiences and all the necessary calculations included. A loop antenna consists of about 1/10 wavelength resonant loop, a tuning capacitor and a coupling loop which is responsible for the 50Ω output. I based the design of my antenna on knowing that the tuning capacitor I had managed to get hold of had a range of 540pF. That and the desired range served as inputs to the following computations.

I arrived at conductor length S for the resonant loop of 1.995m using copper tubing with diameter d of 15mm. That yielded area A of 0.317m². Unfortunately the equations don't use SI units. Conversions to MHz in case of frequency f, to feet in case of conductor length S, to inches in case of conductor diameter d and square feet in case of loop area A need to be done first. Taking that into account the antenna has efficiency of only 3.13% at 7MHz and 31.75% at 15MHz. Not great, but hey.

This set of equations gives the required capacitance of the tuning capacitor at specific frequency. First the loop inductance of 1.519uH is calculated. That yields inductive reactance of 66.81Ω at 7MHz and 143.17Ω at 15MHz. Using these values I arrive at 340pF at 7MHz and 74pF at 15MHz of required capacitance. Both values lie within the range of the tuning capacitor. The full range of the capacitor spans from 5.6MHz to somewhere close to 30MHz (depends on its minimum achievable capacitance).

The last calculation addresses the bandwidth of the antenna at specific frequency. First the quality factor Q is calculated to 1108.04 at 7MHz and 1142.85 at 15MHz. The bandwidth Δf then amounts to 6317Hz at 7MHz and 13125Hz at 15MHz. Quite narrow as well. Looking at the whole design now, I would probably focus more on efficiency and build a bigger resonant loop. However, what is done, is done.

The tuning capacitor I used was Tesla WN704-13 with two outputs of maximum 270pF and two more of 22.5pF. I wired the two 270pF outputs in parallel to get 540pF of total capacitance. The capacitor goes between the two ends of the tubing and connects them. As clearly visible in the image it wasn't easy to solder the leads to the copper. I tried to get hold of a knob that would fit the grip on the capacitor but unsuccessfully. Thus a bottle lid had to be given the job instead (it works fine).

The coupling loop is made of a 50Ω coaxial cable and is 1/5 of the resonant loop in diameter. It is terminated in an SMA connector.

I wanted the antenna to be self-standing and easily portable so I equipped it with a simple stand. It puts the lower edge of the loop at 0.7m above ground.

Tuning can be done looking at the frequency spectrum in SDR# or a similar software that visualizes the data.

Recently I bought the more stable TCXO equipped RTL-SDR Blog v3 to replace the cheapest dongle available I had been using. Here it is in my completed HF setup.

A few examples of reception with the antenna in a room inside a house. These are AM radios at 7MHz.

Several WSPR receptions on 40 meters. The furthest I've managed was from The Canary Islands at about 3700km.

JT9 and JT65 receptions work fine as well. Although I've managed receptions in all three bands: 40m, 30m and 20m, 7MHz seem to produce the most successful decodes for me.

The Antenna Tuning

Some time has passed since I built the 2 meter Dipole, 70cm Yagi and 70cm Helix antennas and wrote about them in The New Antennas. At that time I tried rather unsuccessfully measure their SWR and somewhat tune them. They have been working quite alright but I still was curious about getting some specific data out of them. Recently I've come across a couple of interesting approaches. Namely Measuring filter characteristics and antenna vswr with an rtl-sdr and noise source article and Adam 9A4QV's Youtube videos. They both use affordable equipment so I thought I would give it a try.

The first device I needed was BG7TBL noise source. Ordered from Ebay for $13.87. It requires a 12V power supply and generates additional excess noise on top of the omnipresent temperature noise. According to the seller it can generate 60dB at 100MHz, 55dB at 500MHz, 52dB at 1GHz, 48dB at 1.5GHz, 38dB at 2GHz, 30dB at 2.5GHz, 27dB at 3GHz and 20dB at 3.5GHz. This would be used in combination with an RTL-SDR dongle and a directional coupler as a low cost spectrum analyzer to visualize the frequency response of the antennas.

The second device was a board equipped with AD8318 demodulating logarithmic amplifier, again from Ebay, for £11.46. Since the RTL-SDR dongle doesn't provide absolute power measurements, I looked for something that should have been capable of doing so. This detector takes in RF signal and outputs voltage inversely corresponding to the input signal's power level. The output voltage ranges from ~1.1V (5dBm) to ~3.9V (-55dBm) - frequency dependent (1MHz to 8GHz).

The spectrum analyzer setup uses a directional coupler with the noise source at the OUT port providing wideband noise to the antenna connected at the IN port.

Due to impedance mismatch at different frequencies a portion of the power gets reflected and is coupled at the CPL port to which an RTL-SDR dongle is connected. The dongle then samples the input signal and with the right software shows the relative power at individual frequencies.

To visualize the data I eventually settled on RTL-SDR Scanner. It allows to scan over wide ranges of frequencies, however, using RTL-SDR dongle with its maximum usable sample rate of ~2.4MSPS it takes a while (+10min) to sample, say, 1GHz worth of  bandwidth. The software also allows to export the data into a .csv file.

Now, the spectrum analyzer's output is useful visually and allows one to see the frequencies at which the antenna reflects the least power, but it doesn't say anything about the absolute powers at play. For that I used the directional coupler again with TT7F at the OUT port this time while the antenna stayed at the IN port.

The CPL port was connected to the AD8318 whose output was sampled by Arduino MEGA's ADC. I had to power the amplifier board externally with 5V from an L7805CV regulator because the Arduino provided only about 4.52V when powered via USB which impacted the amplifier's output.

The Arduino ran a simple script that averaged 100 ADC measurements every 250ms and output the result via the serial interface. Concerning the conversion from the measured value in mV to a value in dBm, I solely relied on a lookup table provided by the seller. Unfortunately, I don't posses any other power meter nor any signal generator of known strength to compare and verify the board's and lookup table's accuracy. A few test measurements of TT7F's output showed more or less expected values, but a proper calibration would be handy.

The Si4060 transmitter fitted on TT7F is limited by its specific matching circuitry and the lower limit of its capabilities to several MHz of bandwidth around the frequency of interest. Luckily I  had versions for both 2m band and 70cm band as well. The graph above shows the measured power as TT7F goes through cycles of 4s transmissions with 4s breaks each time increasing the values in PA_PWR_LVL register that sets the transmitter's output level.

First I took a look at the 2-meter band dipole antenna initially mounted on an aluminium tripod  with two meters of coaxial cable between the antenna and the coupler.

The dimensions were 496mm in length and 4mm in diameter for each of the antenna elements. The first scan put the resonant frequency at 141.5-142MHz.

This shows a wider bandwidth of the same setup.

However, upon scanning the frequency response of the noise source with the antenna disconnected it was clear the output wasn't equal at all frequencies.

So the final graph of the actual reflected power had to take this into account (antenna reflected power - noise source power).

Now, to get absolute power measurements I setup TT7F to do a series of CW transmissions on nine different frequencies in 1MHz steps (142MHz unfortunately being the lowest Si4060 could do). The blue lines depict measurements with the antenna disconnected while the red lines the reflected power with the antenna plugged in. The original values are represented by the dotted lines while the solid lines show the final values after adding 14.46dB for the coupling factor of the directional coupler and twice 0.7dB for the coupler's mainline loss.

Using these equations to first calculate the reflection coefficient Γ from the ratio of the reflected power P_r to the incident power P_i, I arrived at these SWR values:

Since my main intention for this antenna was APRS reception at 144.8MHz, I decided to cut the antenna elements a little to push the SWR curve higher in frequency.

However after shortening each element by 5mm to 491mm in length, the resonant frequency allegedly decreased.

I then tried sweeping the antenna in different setups (mounted further away from the tripod, held in a hand as far from the body as possible) and found out that the aluminium tripod was responsible for detuning the antenna. Generally I came across recommendations to measure SWR as close to the antenna as possible with the antenna placed where it was intended to be operated to account for the specific environment variables. Since I don't plan on any permanent installation of the antenna, I decided to leave it with the current dimensions (491mm length per element) keeping in mind its behaviour as shown by the data. When held in a hand at arm's length from the body, the dipole showed resonant frequency at around 144.5MHz.

Moving on to the 70cm-band yagi. As said in the original blog post, I had shortened this antenna too much shifting its resonant frequency supposedly higher then intended (434MHz).

The original RTL-SDR Scanner data had to be adjusted for the noise source output variability again.

The 'spectrum analyzer' sweep confirmed the suspicion.

The more detailed frequency response showed the antenna least impeding the noise at around 439MHz. In this case with the antenna on the aluminium tripod.

Once again I tried scanning in different setups. First with the antenna in my hand at arm's length and the coaxial cable in parallel with my arm. After that similarly except for the coax now hanging freely underneath the antenna. Unlike in the case of the dipole the resonant frequency didn't seem to be affected as much.

The absolute power measurements, this time done with a 434MHz version of TT7F, generated data responsible for the SWR calculations above. It shows SWR of 2.82 at my frequency of interest (434MHz) which corresponds to 22.8% of power being reflected. That isn't good, however, the following wider bandwidth measurement quite shook my confidence in the method/execution because the shape of the data doesn't seem to correlate with the frequency sweeps. I am not certain whether it is a problem of TT7F's output at different frequencies (the matching and output filter should be able to cope with this range) or the approach itself.

Last but not least, the 70cm-band helix antenna. It has already proven itself worthy by tracking TT7F's first flight all the way until the signal disappeared suddenly in the later stages of descent.

The wideband frequency response suggested that the most power got transferred at around 395MHz with 434MHz located close to the third lowest trough.

These three detailed frequency responses show the antenna on the aluminium tripod (blue) and two attempts to hold it in hand at arm's length (red and green).

I then tried shortening the length of the helix's tubing by 2cm from 298cm to 296cm, however it moved the resonant frequency up by just 5MHz. I didn't want to get to crazy with the cutting especially since there were other parameters to helix design that weren't easily modifiable (diameter of helix, spacing between individual helixes).

The main reason I didn't bother too much with trying to move the resonant frequency closer to 434MHz was the computed SWR values. The first graph shows quite low values across the whole tested bandwidth (blue - original, green - shortened helix). Being somewhat suspicious about such results I then used the 2m-band TT7F to get some numbers further away from the resonant frequency (second graph).

In conclusion, I consider mainly the frequency response data to bear some value. On the other hand I am quite uncertain about the absolute power measurement setup. Comparison measurements with proper instruments would be desirable to evaluate the setup's credibility.