What is direct digital SDR?
Software-Defined Radio (SDR) is a type of radio, where the analog signal is converted into the digital domain, and functionality is implemented in the digital domain employing signal processing algorithms. Conversion technology is limited in terms of bandwidth and frequency range, thus the right point for conversion has to be carefully chosen. Conversion can take place at the baseband, Intermediate Frequency (IF), or directly at the Radio Frequency (RF). In case conversion happens at the operating RF (likely after the pre-selector), we can talk about direct digital SDR.
Domain converter frequency parameters
Analog-to-Digital Converters (ADCs) and Digital-to-Analog Converters (DACs) are employed to bridge the analog and digital domains on the radio hardware platform. Converter parameters determine how we can use them in the radio implementations.
One of the most important parameters is the real-time bandwidth or instantaneous bandwidth. It is determined by the sampling frequency of the converter, and according to the Nyquist law, it is equal to the half of the sampling frequency.
The other very important parameter is bandwidth or frequency range of the converter itself. Usually, this is determined by the circuits involved: it starts with the analog components, and includes circuitry within the converter, like the sample-and-hold stage. The Nyquist criteria states that the bandwidth should be equal to the half of the sampling rate in order for a perfect reconstruction in both time and frequency domains. Hence, there is a possibility to generate and sample higher frequency signals too, if we keep the bandwidth inside half of the sampling rate. In other words, we can use upper half bands, called Nyquist bands. If we have a wider spectrum, we have to be sure not to alias or fold from higher Nyquist bands to the baseband. The anti-aliasing filter or SDR pre-selector is used for that propose. If we are talking about ADCs and receivers, the latter terminology is employed.
Frequency parameters of the DRU-244A SDR hardware
We’ve used 80 MHz as sampling frequency for our hardware platform, so, the instantaneous bandwidth is 40 MHz. We can tune to radio channels within this band using on-board hardware DDCs. The input bandwidth of the ADC itself is 650 MHz. This is the -3 dB point of the input stage, and it has no brick wall slope.
This means that we can use not only the 0-40 MHz first Nyquist band, but upper bands, like 160-180 MHz, too using an SDR per-selector filter. However, the bandwidth is degraded, because we have to use some other input analog circuits, like input low-noise preamplifiers and leveling attenuators. Still, it is possible to receive with good results up to 500 MHz. See this post about satellite signal reception at 435 MHz:
For more information, please see AN-835 application note from Analog Devices:
Designing SDR pre-selector filter
You can find a lot of different filter design tool kits on the net, which will approximate your requirements, and determine the right components for different realizations. I think, the best practice, – which I’ve used in the last decades – is to cascade a separate high-pass and a low-pass filter if the relative bandwidth is high. On the other hand, the band-pass approach will work for narrow band (<10%) filters. I always like to use standard components. E12 or E24 1% components will do good job for anti-aliasing and pre-selection filter implementations. Usually, the capacitors are the easier part, inductors may have to be manually wound and tuned.
Bandpass filter for VHF bands
Using the Dyonusos filter design software, I’ve designed a band-pass SDR pre-selector filters utilizing the capacitive coupled resonator structure, which is my favorite. The relative bandwidth is higher than 10%. During the approximation phase, I like to see ~40 dB attenuation at the Nyquist band corner. However, only 30 dB could be achieved by the high-pass filter at the lower band edge frequency if the insertion bandwidth was kept at 20 MHz. You can see the calculated filter response, the filter values, and the measured response after having very careful fine tuned the inductors in the circuits. Seems easy enough, but you need some practice to reach such results with a 5th order resonator filter. For beginners interested in designing and implementing filters, let me suggest to start with 3rd order structures and standard complements as close as possible to the calculated component values.