“SDR (software radio) is highly adaptable, allowing the mode or waveform to be changed at will. This design scheme focuses on the “excitation source” part of the moderate bandwidth SDR (Figure 1). The RF carrier or transmitter IF is added to the integral regulator, and according to the design details, the modulated output is withdrawn for further frequency conversion or expansion. Half of the DSP part is used to analyze the baseband signal-in this case, the signal is divided into real and imaginary parts.
Author: John Wendler; Ray Tremblay
SDR (software radio) is highly adaptable, allowing the mode or waveform to be changed at will. This design scheme focuses on the “excitation source” part of the moderate bandwidth SDR (Figure 1). The RF carrier or transmitter IF is added to the integral regulator, and according to the design details, the modulated output is withdrawn for further frequency conversion or expansion. Half of the DSP part is used to analyze the baseband signal-in this case, the signal is divided into real and imaginary parts. These signals are derived from the output of voice through a microphone with ADC, or data from a computer. Regardless of the signal source, the DSP completes a series of digital calculations, realizes filtering, may increase the signal tone or packs the data, and converts the data string to the final I and Q modulated signals. For moderate bandwidth, stereo Σ-ΔDAC or encoder provides conversion to analog signal, and implements some additional filtering on the signal. The integral modulator consists of a pair of mixers, so these filters are usually necessary. These mixers convert any noise at the baseband frequency directly to the modulator output.
Output noise is a problem. The FCC (Federal Communications Commission) regulates the power ratio requirements of some equipment, such as terrestrial mobile radios, spectrum masks or adjacent channels. These requirements change according to the channel bandwidth and transmission frequency, and control the spectrum allowed to be transmitted. Their functions are usually the same, but they are restricted from conflicting with other users on adjacent channels of the transmitter. Satisfying the spectrum mask is a modulation requirement; there is no guarantee that the radio does not meet the requirements, and without this certification, it cannot be sold legally. Figure 2 shows an example of a spectrum mask, 47 CFR 90.210 G, the X axis shows the offset from the channel center, and the Y axis shows the unmodulated carrier output. This mask is applied to 800MHz SMRS (dedicated mobile radio service), the channels are 25kHz, and the signal only occupies 20kHz.
The unmodulated carrier is first transmitted to the center and top of the mask, which is adapted to the output power of the corresponding transmitter. Then, start to modulate and spread the spectrum. The resulting spectrum must fall below the mask line position at all positions.
The closed inspection in Figure 2 shows some interesting features. On the carrier trace, a glitch of the sampling frequency appears at ±19.2 kHz from the center. The modulated spectrum is also very interesting. The filter in the Σ-ΔDAC causes an almost vertical drop at approximately ±10kHz. There is a bulge at about ±12kHz, and it gradually decreases with the increase of the frequency spectrum, which is caused by the nonlinearity of the high-power amplifier.
Many moderate bandwidth SDRs need to be converted between the single-ended output of the Σ-ΔDAC and the typical balanced input integral modulator. It is often necessary to follow the output of a DAC with a hardware filter, which eliminates the high frequency noise of the DAC and ensures that the spectrum mask needs are met. To make things more complicated, the optimal common-mode and differential-mode output voltages of the DAC are likely to be different from the modulator needs. The simple scaling factor has nothing to do with common mode and differential mode voltages.
Considering all conventional methods, each I or Q channel requires four operational amplifiers with multiple filters. The filter needs to be matched with precision components to ensure that the carrier and single-sideband suppression-the key to an ideal integral modulator-cannot be degraded as the baseband frequency. On the other hand, Linear Technology’s LTC1992 uses a single device to solve the problem. In its data sheet, Linear shows a completely smooth method (Reference 1).
However, its closed, completely smooth method is not necessary. The circuit in Figure 3 has an excellent phase and amplitude balance between the output channel and the elimination of some dangerous device matching requirements. Pin 2 sets the required common-mode output voltage, and the reference voltage of the DAC is connected through the input resistance of pin 8. Note that any mismatch between the input and the output of the reference voltage will cause an asymmetric swing. This application bypasses pin 7. The filter is an active single-pole circuit, cascaded with reverse Sallen-Key filters, but other topologies are also feasible.
Figure 4 shows the measured frequency response of the ground positive channel. Obviously the 6dB gap is the result of focusing on only half of the differential output voltage; when checking the fully stable output, the network gain is 0dB. Figure 5 shows the measured variance of the 180° phase shift between ideal equal amplitude and positive and negative output. In the critical range of 300Hz to 3kHz, both are less than 0.1dB and 0.1°. Even at 50kHz, the error is less than 0.5dB and 1°.