Building an HF Direct Conversion Receiver
As quarantine drug on into the summer of 2020, I began to reapply my interest in amateur radio. Having just gotten my technician license a few days after new years, I decided it was time to start getting into HF in preparation for obtaining my general license. Since I couldn’t transmit, save for a few select bands like 40m CW, I decided to first build a receiver that would allow me to listen in on the majority of the HF portion of the spectrum.
HF, standing somewhat ironically for “high frequency”, occupies the portion of the frequency spectrum from 3 to 30 MHz, although things like the AM broadcast band often also get lumped into the category despite being officially being classified as MF. Unlike radios operating at VHF and higher frequencies which are restricted to line of sight communication outside of a few rare circumstances, HF bands each have unique propagation characteristics that allow them to travel much further across the Earth, sometimes allowing them to reach the other side of the globe if conditions are favorable. This makes the band particularly interesting to hobbyists for DXing, or receiving signals from as far away as possible. However, world militaries still retain certain sections of the HF spectrum for cases where satellite communication may not be ideal or possible.
While FM typically reigns supreme in the VHF and UHF ham radio bands, single sideband (SSB) is king below 30 MHz. The concept of single sideband transmission came about not long after the dawn of the radio age, although it didn’t see serious use until after WWII, owing largely to its increased complexity over AM radio. In order to understand how SSB works, we must first understand the math behind AM modulation.
The concept of AM is fairly simple: by changing the amplitude of a carrier wave over time, it is possible to encode audio information inside a wave of any frequency, provided the carrier wavelength is substantially smaller than that of the modulating wave, which is no problem for audio information. Anybody well versed in basic trigonometry should then be able to deduce the equation of a simple AM radio wave as the following, where \(f_m\) is the frequency of the modulating wave and \(f_c\) is the frequency of the carrier:
\(\Large f(x)= \sin (2\pi f_{m} \cdot x) \sin (2\pi f_{c} \cdot x) \)
In the time domain (where time forms the x-axis of a graph), this will look like the following (use the sliders to play around with the frequency parameters):
This looks all nice and simple in the time domain, however things get more complicated when we enter the frequency domain. If we take x to be the frequency of a pure sine wave instead of time, we can determine what waves make up a given signal. This is particularly useful in radio and audio applications, where it is desirable to know the isolated frequencies that make up an incoming signal. This is where things get interesting for our AM signal. When we apply the following trig identity:
\(\Large \sin(b) \cdot \sin(a) = \frac{1}{2}\cos (a – b) – \frac{1}{2}\cos(a + b)\)
If we take \(\sin (a)\) to be our carrier wave and \(\sin (b)\) to be our modulating wave, it is clear to see what is happening. By modulating our carrier, we have produced two waves offset from the carrier frequency by the modulating frequency. For example, if we were to modulate a wave at 1 KHz with a 100 Hz signal, we would produce two pure waves with half of the amplitude of the carrier: one at 1.1 KHz and the other at 900 Hz. Recombining these sidebands, we obtain the original wave. This is illustrated visually below:
Now we arrive at the property that makes single sideband modulation so powerful. If you subtract the frequency of the carrier from the upper sideband, or subtract the frequency of the lower sideband from the carrier, you obtain the original modulating signal. Hence, we have just demonstrated that half of all the information carried in an AM signal is entirely redundant. If we were to eliminate one of the sidebands from the signal and therefore also eliminate the carrier wave, we would be able to cut the bandwidth of our signal in half while still retaining all of its information.
Now that we know how single sideband works, and why it is superior to AM in most cases, we must learn how to demodulate a SSB signal, or obtain the information contained within the modulated signal. Most simple AM radios use a method called envelope detection. By using a single diode and a simple low pass filter, the envelope of an AM wave can be restored, similar to how power supplies use rectifiers to turn AC power into DC. However, this method will not work with SSB because there is no carrier wave to rectify, as we eliminated it at the source. Instead, we will have to use something called a product detector.
Product detectors get their name from the fact that they detect the modulation products (aka the sidebands) of a given signal. Since we know that the frequency of a given modulation product is offset from its original frequency by the frequency of the carrier, we may simply apply the following formula to obtain the original signal:
\(\Large f_m = |f_p – f_c| \)
In order to achieve this, we must use something called the heterodyne principle, more commonly referred to as the superheterodyne principle when applied to radio. The basic concept again calls back to fundamental trigonometry and constructive and destructive interference. In the time domain, when two signals are combined in a mixer, the result is the product of the two signals. However, things get interesting in the frequency domain. As we learned earlier from AM modulation, the resulting outputs in the frequency domain are sine waves with frequencies equal to the sum and difference of the two inputs, as shown below:
Here we see that when the demodulating input is equal to the carrier frequency, the output signal from our mixer (\(O\) in the above graph) is equal to the combination of the sum and difference of the input frequencies. Applying a lowpass filter to isolate \(O_1\) outputs a waveform of the exact same frequency and phase as the original modulating signal. Hence we have just demodulated a single sideband transmission using product detection. By playing around with the graph, you may see that we are able to perfectly reconstruct the information contained in our SSB signal for any value of \(f_m\) as long as \(f_d = f_c\). However, if \(f_d\) begins to diverge from \(f_c\), our reconstructed signal will become pitch shifted relative to its source, resulting in distorted sounding speech on the output. Because all frequencies will be distorted by the same amount by an out of tune detector, this warps the harmonics of human speech and produces an alien sort of sound. Funnily enough, this is actually how the sound effects for the probe droids in Star Wars were created:
If we shift this audio down uniformly by 1.2 KHz, we are able to reconstitute the original human speech:
Yes that’s right, the imperial probe droid was really talking about mummies this whole time.
Now that we know how SSB demodulation works, we can finally go about building a suitable receiver. Product detectors are also able to demodulate AM and CW (morse code) transmissions by nature of their principle of operation, so in theory I would be able to hear the majority of the HF spectrum save for some digital modes. The chip that lies at the center of this build is the NE602. The NE602 is a gilbert cell mixer chip that was originally made for use in cordless telephone receivers, but it has found a new life in the modern era among ham radio enthusiasts. To build a direct conversion receiver, all we need to do is connect the mixer’s main input to an antenna as connect its secondary input to a signal generator of some sort for tuning. Follow this up with a low pass filter and an audio amplifier and you should have a fully working radio. The block diagram of this system is shown below:
Strictly speaking, the RF preamp isn’t a necessary component for a barebones DC receiver, although I found it useful to include as it would offer me some gain control before the mixing stage. I chose to use an LT1252 wideband video amplifier IC as my preamp as it had good linearity over the entire HF spectrum. After this, the output of the preamp was impedance matched to the mixer input using a small toroidal transformer that I wound myself. For the local oscillator, I first used my signal generator to tune the circuit and ensure it was working. For the audio amp, I used the venerable LM386 to drive a small speaker. While this worked well, I found that the output at maximum gain still wasn’t quite loud enough for my liking, so I inserted a class A preamp made from a single 2n3904 transistor immediately after the mixer stage. This improved the audio output to my liking. The schematic for the final receiver is shown here:
When assembling the receiver, I chose to use a combination of through hole protoboard and single-sided FR-4 copper sheets. It’s very important to have a good ground plane for RF applications in order to eliminate noise. In order to facilitate this, I wired all non-ground connections on the protoboards and then connected all grounds to the copper. The assembled receiver (with some ground wires on the preamp still unsoldered) is shown below:
After attracting a long length of wire to the antenna terminal (seen here in the bottom left of the image), I was able to tune the receiver using my signal generator, although the noise was extremely prevalent and almost washed out any discernable audio. In order to fix this, I soldered a large nail to a grounding wire and set it on the metal frame of my workbench. This provided enough of a solid ground reference for the radio that it more or less eliminated my noise problems. My first test station was a local AM broadcaster:
After this, I tried my hand at picking up something on the ham bands, and I was eventually able to hear some ragchews on 80 meters:
Finally, a glamor shot of the whole setup:
