Direct Digital Synthess and DSP

Part five: Direct Digital Syntheses and Digital Signal Processing

Analog PLL's are being replaced by a pure digital means of generating any desired frequency. Computer circuits can generate millions of samples of sine waves which are converted to an analog wave form using a digital to analog converter. The process is almost the reverse of how a CD player works. The number of samples per second required is determined by the Nyquist theorem. Nyquest states that to re-construct a given wave form at least two samples for each 360 degrees of the wave form are required. Simply stated, the sample clock must be twice the highest desired frequency of the desired wave form. The device that performs this magic is called a Direct Digital Synthesizer or DDS. The DDS consists of a digital accumulator, a wave form function generator, and a digital to analog converter. The digital accumulator is an adder and latch circuit. On each sample clock the accumulator adds the previous sample to a constant. The value of the constant determines the frequency of the wave form generated. For example, if we used a 16 bit accumulator, and the constant was 256, it would take 256 clock samples to complete a full cycle. I.E.: 65536 / 256 = 256. The output frequency would therefore be sample clock freq / 256. The wave form function generator takes the output value of the digital accumulator and transforms it into a value of the wave form at a given point in time. In most cases the wave form function generator is a read only memory containing a sine function look up table. The sine value of the digital accumulator is then presented to the digital to analog converter. It is now only necessary to remove alias products from the wave form by passing it though a low pass filter. Several companies now make DDS components on a single chip. Analog Devices has several DDS devices complete with computer interfaces, clock generator, and digital to analog functions on a single chip that will function to as high as 300mhz.

Special microprocessors (digital signal processors or DSP) are now fast enough to analyze and modify wave forms in real time. If an analog signal is first passed though a low pass filter to bandwidth limit it, and then to an analog to digital converter such a microprocessor may operate on the resulting data and then output the modified data to a digital to analog converter. After again passing though a low pass filter the resulting signal is re-created, but in modified form. By the use of proper algorithms, it is possible to filter out QRM and QRN (interference and noise) from the signal. Several filters may be applied at once to the signal. If the DSP and analog converters are fast enough, and a low frequency IF is used, the IF signal may be processed and the signal can be demodulated in the digital domain.

In the early days of amateur single side band two methods of signal generation and demodulation were considered. The filter method used a high frequency filter (crystal or mechanical) to remove one of the unwanted side bands and to help suppress the carrier. The phasing method used a special mixer circuit (balanced modulator or image reject mixer) to remove the unwanted signal components. The filter method won out and is used today. However it is now possible to make use of the phasing method with digital means. The phasing circuit required that the audio signal and rf drive into the mixer (or doubly balanced modulator) be split into two 90 degrees apart. A DDS circuit with both sine and cosine function generators will generate precisely split quadature signals. The audio signal may be precisely split by use of a DSP. The doubly balanced modulator will work in reverse to demodulate a SSB signal and to select the desired side band. In this case the circuit generates a quadature audio signal pair, and the DSP must combine them into a single signal for only the desired side band. It is now possible to build a direct conversion SSB set which will generate and receive signals directly at the air frequency. Such a set has been described in QST, but it used analog components to generate and combine the quadature signals. Who will be the first to build such a digital radio?

As DSP circuits improve, and get faster a completely software radio is possible. Such a rig would have the antenna connected (via a tuned front end and perhaps some pre amplification) to an analog to digital converter. Tuning and demodulation would be done completely in software, and the output of the digital to analog converter presented (after a low pass filter) to the speaker. While such a radio is still in the future, I suspect it will find it's place in the ham shack of many of today's amateurs.