The ultimate SDR guide

The first receivers worked very simple: The incoming radio waves from the antenna were fed to a rectifying element and a headphone was connected to it. This simple detector technology can only detect amplitude modulation (AM) and is very insensitive due to the lack of amplifiers. In addition, "all transmitters at once" are received – apart from the antenna, any frequency-selective component is missing. This was only possible in the early years of radio technology, when there were hardly any transmitters at all.

To increase the input sensitivity and raise the output signal to loudspeaker level, amplifiers are used at various stages in the receiver. In order to select individual stations on the receiver, a filter is needed that must be set exactly to the transmitter frequency. In order to achieve a reasonable filter "selectivity", there are multiple filters necessary that have to be operated synchronously. Moreover, with higher amplification and feedback, the whole arrangement becomes more selective, but also unstable, easily oscillating and thus unintentionally becoming a transmitter that interferes with other receivers. Thus, the capabilities of such a so-called straight receiver or tuned radio frequency receiver are limited.

More useful and therefore standard nowadays is the receiver principle published by Edwin Armstrong, called superheterodyne or superhet for short. At the same time, Lucien Lévy in France and Walter Schottky in Germany also discovered this principle. Here, after a slight pre-amplification, the signal from the antenna is passed to a superheterodyne, a mixer, which mixes it with an internally generated superheterodyne frequency from an oscillator. The two signals are not simply added but multiplied, for which either a component with a non-linear characteristic or one with two inputs (multi-grid tube, dual-gate semiconductor, mixer module) is used. This produces a signal that still contains the modulation with a frequency that corresponds to the sum of the received and oscillator frequencies and one that corresponds to their difference.

In this way, the received signal is converted to a lower so-called intermediate frequency. This can be amplified and filtered much more easily: the filters are now set to a fixed frequency and no longer have to be tuned synchronously. To tune to a different reception frequency, only the superposition frequency is changed. The risk of feedback is also reduced, as the signal is no longer amplified on the same frequency from the antenna to the detector.

If very high frequencies are to be received, this mixing and down-converting can also be implemented multiple times. For example, the receiving part of a satellite antenna, the LNB directly at the antenna, converts the original signal of around 10 GHz to 1 to 2 GHz. This signal is then sent to the satellite receiver via normal (i.e. affordable) coaxial cables.

The only disadvantage of this reception technique: Since not only a difference but also a sum of reception and superposition frequency is formed, the superhet in principle receives on two frequencies simultaneously, which leads to interference if both are occupied. To prevent this, pre-filtering is required in the circuit section that is located before the mixing stage. This filter lowers signals outside the intended reception band. In more powerful receivers, multiple conversion is also common; a first higher intermediate frequency ensures better image frequency rejection by the pre-filters, a second lower intermediate frequency then ensures better selection of the station to be received.

A special case is a superheterodyne receiver where the receive and oscillator frequencies are the same, so that the receiver mixes down directly to zero – to the audio frequency level. It is also called homodyne or synchrodyne

Here, the problem with the far-off mirror frequency does not apply. Instead, complex signals are created in the LF range, consisting of an I and a Q component. For conventional analogue receivers, this technology could not establish itself on a larger scale due to problems with strong signals and strong noise. In addition, the image frequency occurs here directly in the low-frequency range; it is not possible to distinguish between, for example, 1 kHz below and 1 kHz above the reception frequency. This is different with digital SDR receivers, because here the I and Q signals can be processed directly, so it is possible to distinguish clearly between the desired and the image frequency.

The simplest version on the digital side uses a direct conversion receiver to mix down to the low-frequency level, just a few kHz. The signal may then be processed by a standard sound card and a normal PC without any further specialized hardware. This is how, for example, the first demo receivers for DRM – Digital Radio Mondiale – worked, which was to replace AM broadcasting on long, medium and short wave. Since a normal PC is not very well suited to such a task, the CPU is under heavy load it and when other demanding programmes are started, the decoding stutters. Sometimes the load also causes the PC's fans to turn up, which will make it necessary to use headphones, especially with notebooks. In addition, although special modulation procedures can be processed, the limitation of conventional reception of only one frequency, one station at a time, remains – the bandwidth that can be processed by the sound card and PC hardly goes beyond the normal low-frequency range.

At first glance, this method is inexpensive, requires the least amount of special hardware, but de facto blocks a PC for signal processing and should rather be regarded as an entry into SDR technology. However, it is also popular independently of a PC, whether in mobile phones or cheap current shortwave receivers, as it is the cheapest solution. If a radio amateur wants to receive in the long-wave range, it has become the standard, as here the sound card is absolutely sufficient for signal processing – models that work up to 192 kHz clock frequency can even decode the time signal transmitter DCF 77 on 77.5 kHz. For this purpose, the German radio amateur DL4YHF has written the software SpecLab Operating modes such as WSJT also use the decoding of a signal in the AF range via a PC sound card.

The popular RTL-SDR, a USB stick actually produced for DVB-T reception, also uses this technology with the E4000 as DSP, while the R820T is the next variant with an intermediate frequency of 3.57 or 4.57 MHz.

It is more effective to work at intermediate frequency level. For this, however, special samplers and DSPs are needed in the receiver; a sound card and a normal PC do not manage to work at least at some 100 kHz. On the other hand, it is possible to sample at a wider bandwidth, which can be useful for special types of modulation such as spread spectrum modulation. However, these are more likely to be found in WLAN modules, for example; in amateur radio they are not represented much due to the limited frequency allocations on the classic bands. On the other hand, several transmitters can be received and processed simultaneously with processing on a wider bandwidth level.

Of course, the intermediate frequency stages must also be more wide banded for this. Higher intermediate frequencies are thus used even for the shortwave bands and accordingly faster scanners and more powerful DSPs are necessary. The good news: This technology has been available for several years. Likewise, the technology of mixing to an intermediate frequency has proven itself; powerful modules and years of experience are available. Band-pass filters prevent the module stages from being exposed to extreme levels outside the desired reception range. The results in terms of large-signal immunity and intermodulation are therefore just as good as with analogue receivers. In the GHz range, no other technology is applicable at all – no DSP can process such high frequencies directly. Some receivers therefore process some reception ranges hybrid, others direct, such as the Icom IC-9700 processes 2 m and 70 cm direct and 23 cm downmixed to an intermediate frequency of around 341 MHz.

Bandwidths up to the two-digit MHz range can still be processed with a normal PC, which then works together with a special interface, for example a receiver on a plug-in card, and no longer with the PC's own sound card. The PC as a control interface, not necessarily as a signal processor, is very popular with SDR as a whole, and even some stand-alone SDR solutions use an integrated PC for this purpose. With the combination of proven analogue technology and new digital technology, very good characteristic data can be achieved and the system can also be implemented well for the broadcasting branches.

The most consistent SDR technology is the direct sampler. Here, the input signal is directly digitally processed instead of being downmixed. Of course, this is the most flexible way and the entire reception range may be processed in one block. However, the demands on the sampler and DSP are also the highest. This working principle can be difficult if there are no input band filters and the receiver therefore has to process all signals at once, whether shortwave, FM broadcasting or weak amateur radio signals. Often, therefore, the AM range, although technically the easiest to process, is filtered away with a high-pass filter in view of the high useful and interference levels produced by this range.

With a high resolution of at least 12 bits, a direct sampler is superior to analogue concepts because it can circumvent their problems such as intermodulation distortions in analogue amplifier stages. The only disadvantage is the higher price and the higher power consumption caused by the high computing power, which is why this technology is less suitable for portable devices.

The input signal for digitizing should be sampled with at least twice the maximum input frequency if possible, in order to be able to reconstruct the signal reliably. With an A/D converter with sampling frequencies of up to 3.6 GSPS (Giga-Samples Per Second) at 12 bit resolution, reception ranges of up to 1500 MHz were already possible in 2013. However, low-cost SDR modules intended for experimentation, such as HackRF [https://www.wimo.com/de/hackrf-one] and SDR sticks, which were actually developed for DVB-T reception, only use 8-bit resolution for the analogue-to-digital converter, which is not quite sufficient for clean radio signal processing according to today's standards. When transmitting, this causes excessive spurious emissions. Such modules can therefore without further technical measures only be used in amateur radio for reception, where they naturally also lead to increased false reception points where there is actually no signal at all. However, this remains the personal problem of the radio amateur and does not disturb others; for simple portable receivers that do not have to process high signal levels, this is acceptable. But there are also – of course a bit more expensive – USB SDR sticks with higher resolution such as the Colibri-NANO with 14 bits or the Funcube SDR with 16 bits.

Systems for lower sampling rates such as sound cards do not have these problems; in the 90s, they also started with 8 bits, but 16 and 24 bits have been available for many years because this is indispensable for qualitatively acceptable music production and reproduction.

If the signal frequency to be processed is higher than twice the sampling rate, reception is still possible, but it becomes ambiguous: As with the superheterodyne receiver, reception now occurs simultaneously on a sum frequency and a difference frequency, and unexpected mixed products arise. This is called "aliasing". One is familiar with this from scans of pictures from newspapers or with newscasters wearing petty jackets on television, where unexpected interference patterns ("moiré") become visible due to the superposition of the sampling rate and the picture content.

An anti-aliasing filter tries to eliminate the "wrong" reception frequency; normally it is a low-pass filter that filters out "too high" frequencies above half the sampling frequency. In special cases, the "lower" receive frequency is also filtered out – for example, in the Icom IC-9700 , if it is to receive in the 2-m band with 144 to 146 MHz or in the 70-cm band with 430 to 440 MHz: It’s A/D converter has a sampling rate of only 122 MHz. This is called "undersampling": The sampling rate is actually too low for clear decoding. Here, a high-pass filter is used for anti-aliasing, which filters out signals below about 130 MHz; in the 2-m band, even a band-pass filter is used to also filter out the equally possible reception on even higher frequencies. The situation is similar with the Colibri NANO USB stick, which also scans at 122 MHz and was initially developed for direct-sampling shortwave reception from 9 kHz to 55 MHz. Here, the anti-aliasing filter can also be switched off, then reception up to 500 MHz is possible with undersampling and the associated restrictions, i.e. multiple reception.

Conversely, if the sampling rate is much higher than the signal frequency, i.e. if the signal is sampled much more often than is actually necessary, this is called "oversampling". This acts like an increase in the bit rate and increases the quality, but also the amount of data to be processed. For this reason, audio technology now works with sampling rates of up to 196 kHz, although the human ear can only hear sound waves up to about 20 kHz. Oversampling also increases the signal quality of radio equipment, but is also associated with increased effort (costs, power consumption). On the other hand, it is easier to construct a clean anti-aliasing filter and the noise decreases, similar to an increased number of bits of the A/D converter. In this way, signals that are actually below the noise limit become readable, as with WSJT.

Anti-aliasing filters as low-pass or high-pass filters have already been mentioned. In order to avoid overdrive, it makes perfect sense to restrict the continuous reception range of an SDR with switchable band filters to a specific frequency band, for example an amateur radio band, even without aliasing problems. However, this prevents the entire shortwave range from being scanned and stored as a file or several users from using an SDR receiver on different frequencies and frequency bands via the internet ("WebSDR"). The Kiwi SDR, which was developed for exactly this use, therefore does not use band filters for preselection. The Icom IC-7300, the shortwave counterpart to the IC-9700, on the other hand, does, because this device, as a transceiver, should of course not be used by several people at the same time via the Internet, but only by its owner, and offer him the best possible result. It achieves this with band filters switched according to the selected frequency band. In contrast, the expensive quartz filters previously used in amateur radios for sufficient selectivity in narrowband modes have become obsolete due to SDR.

Harmonic reduction through predistortion: Predistortion can be used to compensate for the harmonics-generating unlinearities of the transmitting output stage. In some cases, significantly better values can be achieved than with analogue technology, but the technology has its limits with higher harmonics, which can sometimes even increase if the compensation is incorrect. Predistortion therefore does not replace output filters in the transmitter!

"Higher, faster, further!" – this principle has been valid in computer technology for decades, and PCs can do more every year. SDR solutions also benefit from this, but so far it is not possible to go arbitrarily fast. The sampling frequencies are only just above 100 MHz and then with 12-bit resolution – if you want more, you have to make do with fewer bits, and vice versa. Shortwave can thus be processed completely digitally (direct sampling); solutions for transmitting and receiving frequencies of up to 6 GHz are available on the market via intermediate frequencies. However, analogue receivers normally do not go higher without conversion either, as in the satellite LNB, which mixes down from 10 ... 12 GHz to 800 MHz ... 2 GHz. It is therefore also possible to operate via AO-100 by SDR.

In terms of bandwidth, with direct sampling it is theoretically possible to process the entire frequency range in one piece. In fact, however, it usually makes sense to limit this to a few 10 MHz, because otherwise the data rates become so high that recording in real time is no longer possible. The current top devices of the amateur radio manufacturers work in this way up to the 23 cm band, experimental solutions even higher, but then with reduced quality/resolution. In this range, however, there are no longer any ready-made analogue devices available either; instead, do-it-yourself construction is still in demand, which can of course also use digital signal processing. For example, the Langstone SDR by the English radio amateur G4EML can transmit and receive from 70 MHz to 5.7 GHz.

Overall, SDR has rather unnoticedly replaced much of the analogue technology in radio-technological devices in the entire range from cheap to high-end in the last 20 years. In most cases, the costs have fallen and the performance and flexibility of the devices have increased.