[For a copy of this article in PDF format, which displays figures and equations, click here. Requires Adobe Acrobat Reader, free download.

The advent of software defined radio (SDR) has promised great hope for radio communications by offering ease of design and flexibility not previously possible.

The push is to place the analog to digital converter (ADC) as close as possible to the antenna, thereby performing a variety of receive functionalities in the digital domain. This ensures ease of design and flexibility.

Conventional Nyquist sampling requires that the sampling frequency FS must be more than twice the highest frequency of the signal to be digitized. However, for bandlimited signals at higher frequencies, this leads to using a very high sample rate.

For example, a 30 MHz band (including guard band) limited signal centered at 175 MHz would require a sample rate of 380 MHz. However, using the principle of undersampling, the same bandlimited signal may be sampled using a sample frequency of 80 MHz.

Images are formed as part of any sampling process and frequencies higher than the FS/2 are folded back in the 0 to FS/2 region. For proper undersampling (without aliasing), the signal must be bandlimited prior to sampling. The sample frequency FS must satisfy:

and

where TBW is the total bandwidth (including guard band), FL is the low end of the band, FH is the high end of the band and k is any non zero positive integer.

Figure 1 illustrates the undersampling concept. The signal of interest is of bandwidth 10 MHz centered at 175 MHz. The total bandwidth (including guard band) is 30 MHz. Thus the high end of the band is 190 MHz, while the low end of the band is at 160 MHz. A sample rate of 80 MHz satisfies the conditions discussed earlier.

After undersampling, frequencies higher than FS/2 fall in the 0 to FS/2 band:

Frequencies between FS/2 to FS are flipped and fall between FS/2 to 0; frequencies between FS to 3FS/2 retain their spectral shape. Table 1 explains the folding in of frequencies higher than Nyquist frequency.

The sensitivity of a radio is defined to the minimum input power needed to achieve a certain performance metric (such as SINAD for voice communications and BER for data communications). For FM radios, 12 dB SINAD is a typical metric. The signal to noise ratio (SNR for the input RF signal) needed to achieve this SINAD is usually budgeted at 8 dB to 10 dB. The typical sensitivity of a FM receiver is shown in table 2.

A 16 channel VHF receiver (135 MHz to 175 MHz) is chosen as a design example. The conventional design is compared to the software radio based design to demonstrate its effectiveness.

The block diagram for a conventional radio is shown in figure 2.

For VHF frequencies, a single stage downconversion is a difficult task due to the higher order products formed in the mixer.

For example, if a single stage downconversion was attempted to an intermediate frequency (IF) of 22 MHz, and the desired channel frequency is 150 MHz, the local oscillator (LO) would need to be set at 128 MHz. However the second harmonic of the LO mixes with the second harmonic of an interfering channel at 139 MHz (a 2-2 product) to produce an IF of 22 MHz, which will degrade the desired channel.

Usually, a 2-2 product is 40 dB to 50 dB below the 1-1 product. However, if the interfering channel (in this case 139 MHz) is a stronger interferer, the degradation of performance may be substantial.

A two-stage downconversion usually avoids this problem because the first stage upconverts the signal to a higher frequency. For example, if the first IF is chosen to be 900 MHz, the variable LO must be in the range of 765 MHz to 725 MHz (to cover the 135 MHz to 175 MHz band).

The interfering frequency band that causes the 2-2 product to fall at 900 MHz is in the 275 MHz to 315 MHz band, which will be filtered out by the front end bandpass filter. A second stage downconverter can now bring the signal back to a 22 MHz IF.

The diagram shown in figure 2 is for a single channel. The same architecture has to be repeated 16 times to implement a 16-channel receiver. This implies significant hardware, design costs and large size for the multichannel receiver.

A slow speed ADC is typically employed in a conventional radio, one that digitizes baseband signals. For direct digitization of VHF signals, a high speed ADC is essential.

An SDR-based implementation is shown in figure 3.

In the SDR-based implementation, a very simple RF front end is used. The RF front end consists of bandpass filters and gain stages, but no RF downconverters. The VHF band is broken up into three bands and digitized by high speed ADCs. The digital down converters (DDC) produce basedbanded complex output. Demodulation may now be performed on the I/Q data, producing a truly software defined radio.

The sample frequency of the ADC is chosen to match the undersampling sample frequency requirements discussed above. Table 3 shows the sampling frequency for each band.

Other channelization schemes are also possible. Table 4 provides an alternate channelization and sampling scheme.

A comparison of the SDR-based design method and the traditional implementation is summarized in table 5.

The performance of the SDR-based VHF receiver is tested with a 20 kHz tone-modulated FM signal with frequency deviation of ±75 kHz. The set up is shown in figure 4.

The demodulated output and the FFT of the downconverted basebanded I, Q data are shown in figures 5, 6 and 7. The 20 kHz tone is faithfully reproduced.

The FM sensitivity for the three bands is shown in table 6 (page 49). The items in bold and blue represent the 12 dB SINAD point. The FM sensitivities obtained for the SDR-based VHF receiver are better than the typical radio shown in table 2.

Figure 8 shows a plot of the RF sensitivity for all the frequencies in all three bands. The points in the graph represent the measured values and the line represents the regression line fitted through the data points. The sensitivity for 12 dB SINAD is obtained from the regression line as -105.4 dBm about 4 dB better than the typical VHF receiver.

Software-radio-based multichannel, multiband VHF receivers provide unprecedented flexibility and performance. This article demonstrated that a practical multichannel, multiband software-radio-based VHF receiver provides substantial performance enhancement.

The 14-bit high speed ADCs makes implementation of a truly software radio VHF receiver a reality. An actual VHF receiver is implemented using the RF conditioner, a four input, 105 MHz ADC board with on-board 16 digital tuners and FPGA.

The performance of the SDR-based VHF FM receiver is about 4 dB better than a typical VHF FM receiver. A multichannel, multiband VHF receiver may, thus, be implemented at a fraction of the cost of traditional receivers. The design is greatly simplified, as RF downconverters are no longer needed. The RF stage consists of bandpass filters, and automatic gain control blocks, channelizes and amplifies the signals.

As downconversion and tuning is performed in the digital domain, the same RF front end is used for the multiple channels, thereby reducing system cost. High performance ADC cards, with DDCs reduce the size of the receiver, making the receiver suitable for airborne or ship borne applications where space is a premium.

Further optimization of the RFC (in terms of noise figure and gain) will improve the overall performance of the receiver.

Angsuman Rudra, M.Eng., MBA, P.Eng, is the director of radio products for Interactive Circuits and Systems Ltd. (ICS at www.ics-ltd.com). He leads the development of software radio, and provides strategic direction for radio product evaluation. Interactive Circuits and Systems Ltd. designs and manufactures real-time data acquisition and processing products for the sonar, radar, communications and instrumentation markets. Rudra can be reached at arudra@ics-ltd.com.