Reprinted from Amateur Radio (Australia) February 1998 - without the editor's permission (I must get round to getting it)
This article describes a phasing exciter (figure 7) which is easy to get going, offers excellent performance and has only three adjustments - all for carrier nulling. The exciter uses a digital counter to generate the 90 degree RF phase shifts, a "sequence network" (Figure 4) to produce the 90 degree audio phase shifts and a quad analogue switch to perform the modulation. The exciter produces clean SSB at frequencies up to 3.7 MHz, using easily obtainable parts.
The exciter uses a variation on the phasing method of generating SSB. Four equal amplitude audio sources of relative phase 0, 90, 180 and 270 degrees are sequentially selected by an analogue switch. Each source is connected through to the output for one quarter of an RF cycle. The sequence repeats at the carrier frequency, producing SSB. There is, in theory, no carrier and no opposite sideband in the output and the nearest unwanted output is at three times the carrier frequency.
The 4 phase audio source can be treated as two push-pull sources, differing in phase by 90 degrees.
When one source is sampled, taking alternate samples of the "push" and "pull" signals, a series of double sideband (DSB) signals results. The first DSB signal is centred on the sampling frequency and others are at odd multiples.
Similarly, sampling the other (quadrature) audio source produces another series of DSB signals.
Providing the two sampling signals are a quarter of an RF cycle out of step, adding the two DSB signals produces SSB exactly as in the phasing method.
In theory, a family of SSB signals will be all that is produced. The first will be at the sampling frequency, the next at three times the sampling frequency, then five times etc. There will be no carrier, no baseband, no unwanted sidebands and nothing at even multiples of the sampling frequency.
In practice, clean SSB is produced at the sampling frequency. The third order distortion products, carrier and unwanted sideband are all more than 50 dB below PEP. There appears to be no other "rubbish" near the wanted SSB signal. The nearest unwanted signal is at twice the sampling frequency.
Audio from a dynamic microphone is first amplified by one quarter of a TL074 connected as a compressor. The audio is then band pass filtered to restrict the range of frequencies to those handled by the "sequence network", so that reasonable sideband suppression is achieved. Another section of the TL074 acts as a phase inverter to produce the second half of the push-pull drive required by the "sequence network". The output from the "sequence network" is buffered by another TL074 with offset voltage adjustments provided on three of the op-amps so that all four audio signals can be "aligned" to minimise the resultant carrier. The outputs from the op-amps are RC filtered to present a low RF impedance to the 4066 analog switch and to avoid upsetting the TL074.
The VFO signal, at 4 times the final carrier frequency, is amplified by two sections of a 74LS00, biased into the linear region. The signal is gated by another section of the 74LS00 so that the clock signal to the counter may be disabled while allowing the VFO to run continuously for minimum drift. The Johnson (ring) counter is clocked by this gated signal. The counter output is decoded by 4 AND gates to produce the 4 sampling signals. The use of the Johnson counter and symmetrical decoding gates is aimed at matching the switching times as closely as possible. The counter circuit could all be programmed into a PAL or similar to simplify this part of the hardware.
The 4066 analog switch then sequentially selects, at the carrier rate, pieces of the four audio signals and presents them to the output buffer. A "roofing" filter (not included) is required to extract the desired SSB signal. A single tuned circuit or a low-pass filter would be sufficient.
Note: The overall board dimensions are (as best I can measure with a crooked ruler ;-) 142mm by 74.5mm.
Here is the easy part.
First, with no audio input and using a multimeter, adjust the DC outputs of three of the buffer amplifiers to match the fourth one. The SSB exciter must be switched to SSB (not AM). It doesn't matter which sideband is selected.
Then tune a receiver to the output frequency (one quarter on the VFO frequency), switch to SSB and adjust all offset trimpots for minimum carrier (again with no audio input). Repeat a couple of times to get minimum carrier.
Figures 1 and 2 show spectra of the SSB generator output with a two-tone input signal.
The narrow-band spectra, figure 1, shows worst distortion products more than 50 dB below the PEP level of the desired output (PEP is 6 dB above the level of one of the tones).
The wide-band spectra, figure 2, shows no undesired signals (to 80 dB below PEP) near the wanted output. The nearest rubbish is at twice the frequency of the desired output.
Figure 3 shows the performance of the switching modulator with a clean audio source. The carrier has not been nulled (to make it easier to see); it would normally be 20 dB lower.
If you used the PC layout, Figures 5 and 6, then fault finding is easier.
1. Check DC voltage levels at the op-amp outputs match those shown on the circuit diagram (figures in brackets). Minor deviations, say plus or minus half a volt are OK. The four buffer amplifier outputs (the ones that drive the analog switches) should be within a milli-volt of one another.
2. Check that the 4 digital inputs to the analog switch are active. These should be selected sequentially, one per cycle of the VFO. You could try replacing the VFO by a very low frequency (audio) oscillator and use a logic probe to check for activity.
3. Check the 4 audio inputs to the analog switch. They should all be the same amplitude: about 330 mV pp. If not, check the push-pull driver op-amps. The outputs here should be equal amplitudes too: about 1.25V pp. If you have access to a dual trace CRO, check for 90 degree phase shift between adjacent audio signals at the inputs to the 4066 analog switch.
4. If both the preceding checks are OK, then the 4066 is probably faulty. With no audio input, check the DC level at the output from the analog switch. Then connect the output to earth via a 1 K resistor. If the 4066 is OK you shouldn't see much change in DC voltage. The voltage should be within 20 mV of the voltage at the wipers of the offset voltage trim-pots (somewhere near 5.9 volt).
If you have made your own pc board, or otherwise lashed the SSB generator together, then check carefully that you have followed the circuit diagram (I hope I have drawn it correctly). The circuit is critically dependant on the 4 digital signals and the 4 audio signals arriving at the 4066 in the correct sequence! If the wiring is correct, you get "perfect" SSB. If the wiring is wrong, you get perfect rubbish.
This switching modulator should be capable of acting as a demodulator. This would require reversing of the direction of signal flow through the circuit. The problems expected in doing this are firstly that attenuation through the sequence network would prevent the reception of microvolt signals. Secondly, to obtain 40 dB of opposite sideband suppression, the signal level through the analogue switch would have to be held below 0.05 volt peak to peak. This would result in a poor dynamic range. This may be acceptable if the demodulator is used at the output of an AGC controlled IF amplifier.
Semiconductor List:
Device Quantity TL074 2 7805 1 74LS00 1 74AC74 1 74ACT08 1 74HC4066 1The op-amps are all TL074s. Please don't use LM324s as substitutes for the TL074s. LM324s usually have bad cross-over distortion, lower gain and more noise than the TL074, all of which directly degrade the performance.
For most of the digital ICs, use the fastest CMOS types you can get. At a pinch, LS series devices will work fine, but will limit the upper RF frequency a bit (and degrade the carrier suppression too). The one exception is the 74LS00. Fast CMOS '00s sometimes consume a heap of supply current when biased into their linear region or oscillate uncontrollably - avoid using them in the VFO amplifier.
The capacitors in the sequence network should ideally be matched in groups. This matching influences the opposite sideband rejection. Matching between groups is not so important. The 1nF capacitors at the output of the sequence network should also be matched as these will influence the opposite sideband rejection at lower frequencies.
The SSB generator presented is easy to get going and produces clean SSB up to 3.7 MHz. Only common parts are used.
On the negative side, the circuit is rather complicated and its upper SSB frequency is limited to about 7 MHz.
The same switching modulator should be useable as a direct conversion receiver by reversing the direction of signal flow through the RF and audio sections.
1. J.R. Hey G3TDZ "Simple SSB generator" Electronics Today in August 1979 pp 48-51.
2. J. D. K. West - COMSIG 1991 Proceedings. South African Symposium on Communications and Signal Processing Published by IEEE, New York USA
3. M. J. Gingell, "Single Sideband modulation using sequence asymmetric polyphase networks" Electrical Communications, Vol 48, No 1-2, 1973, pp 21-25
