DSP eliminates the audio phase-shift and dc offset problems of analog Weaver-method modulators.
Carlos M. Puig, KJ6ST
Single sideband (SSB) remains the primary Amateur Radio HF communications mode. Of the three known SSB generation methods, the Weaver method has received the least attention. A recent design showed that a Weaver modulator can be readily built with current technology; however, its prototype suffered from performance limitations arising from the characteristics of high-order switched-capacitor filters. (See Note 5.) The Weaver SSB modulator design presented in this paper overcomes these limitations by using digital signal processing (DSP) for critical baseband signal-processing steps. The modulator's measured performance meets applicable ARRL and FCC guidelines and compares favorably with that of commercial units.
Single-sideband suppressed-carrier (SSB-SC, or just SSB) is the most commonly used modulation mode for Amateur Radio voice HF (3-30 MHz) communications. SSB is also the dominant mode for aeronautical, marine, and unencrypted military voice HF
764 Filbert Way Campbell, CA 95008
Baseband I Bandpass
Baseband I Bandpass
Fig 2—The phasing method in the frequency domain.
communications. SSB provides the best spectrum utilization and power efficiency of all analog communications modes and is less affected than AM by frequency-selective multipath fading.
Nearly all current SSB modulators use the filter method, in which an SSB signal is obtained from a double-sideband suppressed-earrier (DSB-SC) signal by use of a bandpass filter that selects only the desired sideband. Filter-method modulators offer good stability over temperature and time. However, they typically require the use of bandpass filters with eight or more accurately cut quartz crystals, and they are limited to operation at a single carrier frequency. Additionally, these crystal filters often suffer from poor phase characteristics. A second method of SSB generation, called the phasing method, can operate over a wide carrier frequency range. It is difficult to obtain good undesired sideband suppression (>40 dB) with the phasing method, and it is even more difficult to maintain that suppression level over time and temperature. Although the phasing method was popular briefly in the early days of SSB, it survives today primarily in textbooks (as an introduction to the Hilbert transform and analytic signals) and in digital signal processing (DSP) implementations, where aging and temperature stability aren't problems.
The third method of SSB generation was introduced by Weaver in 1956.1 The Weaver method is rarely mentioned in communications textbooks, although it has appeared in more specialized technical publications.2 3'4 Because it requires two closely matched, sharp-cutoff audio low-pass filters, it has historically not yielded good performance at reasonable cost. However, recent innovations in both analog and digital audio filtering have brought the Weaver method within the realm of practicality.
The project described in this report, which I did as my EE senior design project, was inspired by Anderson's
1991 revival of the Weaver method.5 Anderson's design relies on the Linear Technology LT1064-1 8th-order elliptic switched-capacitor filter, which gives him the required 1.28-to-l ratio between the stopband and passband frequencies. Although Anderson presented no sideband suppression measurements, his circuit appears to have achieved limited sideband suppression. Furthermore. Anderson reported only 40 dB of carrier and spurious-emission suppression, with higher carrier harmonic levels.
A review of Anderson's circuit suggests that Anderson's sideband suppression is limited by the phase mismatch between the two filters. The phase response of the two low-pass filters must be matched within 1 degree over the entire passband in order to achieve 40 dB of undesired sideband suppression, a standard signal-quality guideline for amateur SSB transmitters (Handbook, pp 18-4 and 18-8).6 Anderson was forced to use both coarse and fine manual dc-offset trims to cancel the filters' large dc offsets, which would otherwise have appeared in the modulator's output as carrier power.
The primary object of this project was to obtain improved opposite sideband and carrier suppression in a Weaver method SSB modulator by using DSP for the required baseband signal processing. A traditional analog design is used for the RF (bandpass) signal processing. A secondary objective was to develop a simple DSP development board that can be reused in future communications projects.
Fig 1 shows a comparison of the phasing and Weaver methods of SSB modulation. In each case the audio input is a speech signal that is bandlimited to approximately 3003000 Hz. The output is an SSB modulated signal with a suppressed carrier of angular frequency coc.
The phasing method requires a Hilbert transformer to obtain a -90° phase shift at unity gain over the entire speech band. The resulting I Cm-phase") and Q ("quadrature") baseband signals are applied to a pair of doubly balanced modulators that act as four-quadrant analog multipliers. The local (carrier) oscillator signals applied to these modulators are also generated in quadrature. The desired SSB output is the sum of the two DSB-SC signals produced by these modulators.
The Weaver method uses exactly the same bandpass processing as the phasing method. However, the method for generating the baseband I and Q signals differs significantly. Instead of a Hilbert transformer, the Weaver method uses a pair of doubly balanced modulators followed by low pass filters. The baseband modulators operate with quadrature audio subcarriers at a frequency (ft>s) of 1.5 kHz, in the middle of the speech band. The low-pass filters must have an essentially flat frequency response from 0-1.2 kHz, with high stopband attenuation at 1.5 kHz.
Fig 2 illustrates the operation of the phasing method in the frequency domain. For ease of discussion, the input baseband signal at the top is assumed to have a real spectrum. Note that the label F(m) on each of the vertical axes does not imply that each graph represents the same signal; instead, these labels merely indicate a frequency spectrum representation of the signal at that stage. The output of the Hilbert transformer has a purely imaginary spectrum, so that the corresponding graph shows F(co)lj rather than F(o>). The suppressed RF carrier is denoted by a dotted line. In this illustration, a lower sideband (LSB) signal is obtained at the output because the upper sideband (USB) components at the summer's inputs have equal magnitudes and are 180° out of phase.
Fig 3 shows the operation of the Weaver method in the frequency domain, with the same notational conventions as Fig 2. Note that the output of each subcarrier modulator shows a 50% overlap between the translated copies of the input spectrum. This overlap is a form of controlled aliasing that is an essential feature of the Weaver method. Signal components above the subcarrier frequency of 1.5 kHz are removed by the low-pass filters. The output modulators and summer operate as they did in Fig 2, but yield the USB.
In both the phasing and Weaver methods, any amplitude or phase errors at the summer's inputs will result in an undesired sideband of finite power. A comparison of Figs 2 and 3 reveals two important differences between the phasing and the Weaver modulators:
(1) In the phasing method, the undesired sideband appears outside of the communications channel, on the other side of the suppressed carrier. Thus, the inevitable loss of modulator I/Q amplitude or phase balance that occurs with aging and temperature variations leads to spurious signals that may cause adjacent channel interference. This weakness is the main reason why the phasing method quickly lost ground to the filter method. In the Weaver method, the undesired sideband appears within the communications channel, as a frequency-inverted copy of the desired sideband, so that the undesired sideband is heard as low-level distortion of the desired signal.
(2) In the phasing method, a partially suppressed carrier appears just outside of the SSB signal. The partially suppressed carrier is rejected at the receiver and, therefore, is not heard. In the Weaver method, the partially suppressed carrier appears at the center of the SSB signal, so that this carrier is heard at the receiving station as a low-level 1500-Hz tone.
Thus, the phasing method's potential for harmful out-of-channel emissions is absent in the Weaver method, which exhibits only a potential for benign in-channel audio distortion products.
Modulator Design and Operation
Fig 4 shows a block diagram of the RF board, which implements the bandpass portion of the Weaver method, as designated in Fig 1(b). The circuit was built on a 4" x 5" prototyping board with a single ground plane, using a mixture of wire-wrap and point-to-point soldered wiring. The schematic is shown in Fig 5.
The two balanced modulators use the classic MC1496 in a circuit adapted from a Motorola data sheet.7 The fixed-frequency quadrature local oscillators are implemented with a 28.322-MHz TTL oscillator module and a dual D flip-flop (74AS74), using the simple circuit described in the Handbook (p 1810) and note 4. With this circuit, the carrier frequency (7.08 MHz) is one quarter of the oscillator frequency.
Was this article helpful?