7.2 Transmitter Architectures, CW Keying, and Output Filtering

7.2 Transmitter Architectures, CW Keying, and Output Filtering

A radio transmitter must perform three fundamental operations: generate high-frequency oscillations at a precisely determined carrier frequency, modulate that carrier with intelligence (audio, Morse code telegraphy, or high-speed digital data), and amplify the resulting modulated RF signal to the desired power level while eliminating all harmonic and spurious emissions.

Accomplishing this across multiple amateur frequency bands requires sophisticated architectural choices, meticulous control over keying transients, and robust filtering networks designed to comply with statutory spectral purity standards.

---

1. Transmitter Topologies and Frequency Generation

Modern amateur transmitters employ one of three dominant architectural configurations, each with distinct advantages and engineering constraints.

Heterodyne Up-Conversion Architecture

For decades, the heterodyne up-conversion transmitter has been the industry standard for high-performance multi-band transceivers.

Mic/audio -> Balanced modulator -> SSB crystal filter -> TX mixer
                    ^                                   ^
             Fixed carrier oscillator             Agile LO/VFO

TX mixer -> RF bandpass filter -> Linear driver/PA -> Output filter -> Antenna

1. Baseband Generation: The modulation signal (e.g. speech) is applied to a balanced modulator operating at a fixed intermediate frequency (such as $9.000\text{ MHz}$). A narrow crystal lattice filter strips the unwanted sideband and carrier, forming a pristine Single Sideband (SSB) signal at $9.000\text{ MHz}$.
2. Frequency Translation: The fixed IF signal is fed to a balanced transmit mixer, where it is mixed with an agile, synthesised local oscillator signal ($f_{\text{LO}}$).
3. Bandpass Filtering and Linear Amplification: A bandpass filter selects the desired sum ($f_{\text{LO}}+f_{\text{IF}}$) or difference product and attenuates LO leakage and the unwanted mixing product. Linear drivers and a power amplifier then raise its level, followed by output filtering. Filter rejection and amplifier linearity determine the residual unwanted emissions.

Advantages: Modulation generation and sideband filtering take place at one fixed IF, allowing the same optimized crystal filter to serve multiple bands. Its passband shapes the SSB signal on every band; later mixers, filters and amplifiers also affect the final response and suppression.

Direct Up-Conversion (Quadrature Modulation)

In a direct up-conversion (or quadrature modulator) transmitter, baseband intelligence is translated directly to the final operating frequency in a single mixing stage.

• Operation: Baseband audio or DSP-generated signals are split into in-phase ($I$) and quadrature-phase ($Q$, $90^\circ$ shifted) components. These drive two balanced mixers supplied by a synthesised local oscillator operating directly at the transmission frequency with a $90^\circ$ phase splitter. Summing the two mixer outputs cancels one sideband via phasor subtraction, producing SSB or complex digital modulations directly at RF.
• Challenges: Because mixing occurs directly at the transmit frequency, any phase or amplitude imbalance between the $I$ and $Q$ channels causes incomplete unwanted-sideband suppression. Furthermore, local oscillator leakage appears directly on the transmitted channel as unsuppressed carrier. Finally, strong RF energy from the power amplifier can radiate back into the local oscillator, causing oscillator pulling.

Direct Digital Synthesis (DDS) and Direct RF Generation

Modern Software-Defined Radio (SDR) transmitters synthesize the modulated RF carrier directly in the digital domain.

• Operation: A Digital Signal Processor (DSP) or Field-Programmable Gate Array (FPGA) synthesizes the complete mathematical waveform (carrier, amplitude envelope, and phase angle) using numerically controlled oscillators and digital filters. The digital samples are fed directly to an ultra-high-speed Digital-to-Analog Converter (DAC) operating at sample rates exceeding $120\text{ MSPS}$ to $1\text{ GSPS}$.
• Output Stage: An analog reconstruction filter suppresses DAC images before the driver and power amplifier. A low-pass filter suits first-Nyquist-zone generation; a bandpass filter can select a higher-zone output.
• Advantages: Direct RF generation removes the analog upconversion mixer and its LO-feedthrough path. Digital processing can give excellent sideband suppression, but finite numerical precision, DAC errors and clock noise still produce noise or spurs. Driver/PA nonlinearity also remains, so final spectral purity must be verified.

---

2. Continuous Wave (CW) Keying Dynamics and Spectral Purity

Continuous Wave (CW) Morse code telegraphy is conceptually the simplest transmission mode: an unmodulated sinusoidal RF carrier is switched on and off using a telegraph key (On-Off Keying, OOK). However, the physics of switching an alternating current waveform introduces severe spectral challenges.

The Fourier Physics of Key Clicks

In the time domain, turning an RF carrier on and off is equivalent to multiplying a continuous sinusoid $\cos(\omega_c t)$ by a rectangular gating pulse train $p(t)$:

$$s_{\text{CW}}(t) = p(t) \cdot \cos(\omega_c t)$$

According to the Fourier convolution theorem, multiplication in the time domain corresponds to convolution in the frequency domain. The spectrum of the transmitted signal is the frequency spectrum of the gating pulse $P(f)$ shifted to the carrier frequency $f_c$.

For a centred rectangular pulse of duration $T$, the Fourier transform has a sinc-shaped spectrum. Normalized to its value at zero frequency:

$$\frac{P(f)}{P(0)}=\frac{\sin(\pi f T)}{\pi f T}$$

The sidelobes of a $\text{sinc}$ function decay very slowly—at a rate of only $20\text{ dB}$ per decade ($6\text{ dB}$ per octave). Consequently, instantaneous on-off switching creates violent high-frequency Fourier transients called key clicks. Key clicks splatter broadband noise across tens of kilohertz on either side of the operating frequency, destroying adjacent-channel communications.

Instantaneous (Sharp) Keying Envelope:             Properly Shaped Keying Envelope:

+-----------------------+                        /---------------------\
|                       |                       /                       \
|                       |                      /                         \
|                       |                     /                           \
+                       +                    +                             +
Fast, abrupt edges: strong sidebands       Smooth 4-5 ms edges: reduced sidebands

Waveform Shaping and Bandwidth

Rounded envelope edges, such as raised-cosine or Gaussian shaping, reduce the high-frequency sidebands responsible for key clicks. Rise and fall times around $4\text{ ms}$ to $5\text{ ms}$ are a common practical choice. Excessively slow edges can impair fast Morse timing; abrupt corners can leave strong sidebands even when a rise-time number looks reasonable.

Necessary bandwidth and occupied bandwidth describe different things. ITU-R SM.1138 gives the Morse necessary-bandwidth estimate:

$$B_n=BK$$

Here $B$ is the modulation rate in baud; the table uses $K=3$ for non-fading circuits and $K=5$ for fading circuits. Its example uses $25\text{ WPM}$, $B=20\text{ baud}$ and $K=5$, giving $B_n=100\text{ Hz}$.

The 99% occupied bandwidth instead comes from integrating the actual transmitted power spectrum, leaving $0.5\%$ of total mean power below its lower limit and $0.5\%$ above its upper limit. It depends on keying pattern, speed and complete envelope shape. A $5\text{ ms}$ rise time does not establish a universal $50\text{ Hz}$ width or confine every sideband inside a fixed channel.

Transmitter Chirp: Causes and Mitigation

Chirp is an audible, instantaneous shift in transmitter carrier frequency that occurs during the key-down transient, causing the Morse code tone to sound like a bird chirping.

• Mechanism 1 (Supply Voltage Pushing): When the key is closed, the power amplifier switches from drawing zero current to drawing several amperes. If the DC power supply has poor voltage regulation, the DC rail voltage dips. If the master oscillator shares this unregulated supply rail, the voltage drop alters the internal junction capacitances of the oscillator transistor, shifting its resonant frequency.
• Mechanism 2 (Oscillator Pulling / RF Feedback): When the final power amplifier conducts, high-level RF current flows through chassis ground and radiates near the master oscillator tuned circuit. The shifting impedance reflected back into the oscillator pulls its resonant frequency during the dot/dash envelope.
• Engineering Remedies:
  1. Rigidly regulate the DC voltage feeding the oscillator using dedicated linear voltage regulator ICs or Zener diodes.
  2. Insert one or more high-isolation Class A buffer amplifier stages between the master oscillator and subsequent driver/power stages.
  3. Never key the master oscillator directly; allow the oscillator to run continuously and execute keying in a downstream low-level buffer stage or digital multiplier.

---

3. Transmitter Output Stages, Harmonics, and ACMA Regulatory Standards

RF power amplifiers operate at high electrical efficiency by driving active devices (BJTs, LDMOS, or valves) in Class AB or Class C modes. However, because the active devices conduct over only a portion of the AC cycle, their output current waveforms are non-sinusoidal, containing rich harmonic content at integer multiples ($2f_0, 3f_0, 4f_0, \dots$) of the operating frequency.

Australian (ACMA) Spectral Purity Requirements

For operation under the Radiocommunications (Amateur Stations) Class Licence 2023, section 15 compares spurious power with the power within the emission's necessary bandwidth, supplied to the antenna transmission line. Required attenuation is:

For the HF formula, $pX$ is peak envelope power in watts. In the second formula, $P$ is power in watts supplied to the antenna transmission line. Lesser means the smaller required attenuation; there is no $25\text{ W}$ breakpoint or separate $50\ \mu\text{W}$ ceiling in this provision. The duty to avoid harmful interference also applies.

At $400\text{ W}$ PEP below $30\text{ MHz}$, $43+10\log_{10}(400)=69.02\text{ dB}$, so the lesser value is $50\text{ dB}$. The worked example below converts this into allowed spurious power and an illustrative filter requirement.

Switched Low-Pass Filter (LPF) Banks

Ideal balanced push-pull operation cancels even harmonics at the combiner, while odd harmonics are not cancelled by that symmetry. Practical residual levels depend on balance, operating conditions and the output network. Measure them to determine the additional filter rejection needed.

Many multiband transmitters use switched multi-pole low-pass filters (LPFs) to reduce PA harmonics. Choose the required rejection from the measured unfiltered harmonic levels and the applicable final-output limit:

• Filter Topologies: Typically 5-pole or 7-pole Chebyshev or Cauer (elliptic) low-pass filter networks.
• Elliptic Filters: Finite-frequency transmission zeros can provide steep rejection near the passband. An RF design may place zeros near troublesome harmonics, but their frequencies, attenuation and insertion loss depend on component values and losses. Verify rejection at each required harmonic across the filter bank's operating ranges.
• Sub-Octave Grouping: Filters are grouped to cover discrete frequency segments (e.g. 160m, 80m, 40m, 30/20m, 17/15m, 12/10m) switched via low-loss power relays or high-voltage PIN diodes.

---

4. Speech Processing and Automatic Level Control (ALC)

SSB carries audio through the amplitude and phase of its RF signal. A single audio sine tone produces one constant-envelope RF tone; ordinary speech produces a varying RF envelope. RF power therefore does not simply follow the squared instantaneous value of the raw audio waveform.

Voice Dynamics and Peak-to-Average Power Ratio (PAPR)

The RF peak-envelope-to-average-power ratio depends on the voice, processing and averaging interval. For this example, assume a PAPR of $12\text{ dB}$. This is an illustrative input to the calculation, rather than a fixed property of all unprocessed speech.

In a transmitter adjusted to deliver $100\text{ W}$ Peak Envelope Power (PEP) on voice peaks, the average transmitted power ($P_{\text{avg}}$) is remarkably low:

$$P_{\text{avg}} = P_{\text{PEP}} \times 10^{-\frac{\text{PAPR}}{10}} = 100\text{ W} \times 10^{-1.2} \approx 6.3\text{ W}!$$

Speech processing aims to improve copy in noise by raising useful average speech power at the same PEP while preserving intelligibility. Excessive processing can instead increase distortion and background noise.

RF Speech Clipping and Compression

To boost average talk power without exceeding the transmitter's peak power rating, speech processors are employed:

1. Audio Compression: Fast-acting audio AGC reduces dynamic range. While simple, excessive audio compression introduces room noise and unpleasant breathing effects.
2. RF Speech Clipping: An SSB signal at IF, for example $9\text{ MHz}$, can have its RF peaks limited by a symmetrical diode clipper. After filtering and level adjustment, this can raise average power relative to the same PEP. An achieved $6\text{ dB}$ average-power increase is about fourfold; $10\text{ dB}$ is tenfold. The actual gain depends on speech and settings.
3. Post-Clipping Filtering: Clipping creates harmonics and intermodulation products. A following SSB filter attenuates products outside the wanted channel before upconversion and amplification. In-band distortion and filter-induced peak regrowth can remain, so check the final envelope and spectrum while setting drive.

Automatic Level Control (ALC)

Even with speech processing, excessive audio drive will over-drive the final power amplifier into saturation, causing flat-topping (waveform clipping of the RF envelope peaks). Flat-topping generates severe intermodulation splatter that spills into adjacent channels.

To prevent flat-topping, transmitters incorporate a closed-loop Automatic Level Control (ALC) system:

• Sensing: Depending on the design, ALC derives a control signal from detected RF output or an amplifier operating current. Some valve amplifiers use grid-current sensing, but the permitted current is equipment-specific: Class AB2 operation can normally draw grid current.
• Feedback Loop: The sensed peak voltage is rectified into a negative DC control voltage. This negative ALC voltage is fed back to the variable-gain intermediate-frequency (IF) amplifier stages in the exciter.
• Gain Reduction: The feedback reduces exciter gain as the sensed level approaches its setpoint. The loop has finite response time, so it can limit sustained overdrive without guaranteeing that every transient remains below the linear threshold. It cannot undo clipping already produced upstream.
• Proper Adjustment: Operators must adjust the microphone gain so that the ALC meter deflects only within the manufacturer's recommended scale. Excessive ALC drive forces the loop into heavy gain-pumping, compressing background noise and increasing audio distortion.

---

5. Architectural and Filtering Comparison Tables

---

6. Worked Engineering Examples

Worked Example 7.2.1: Keying Timing and Necessary Bandwidth

Compare a keyer with abrupt $0.5\text{ ms}$ envelope edges and one with smoothly shaped $5.0\text{ ms}$ edges. Both send a continuous dot-space sequence at $B=24\text{ baud}$.

1. Dot timing: Each mark or space lasts:

$$T_{\text{dot}}=\frac1{24}=41.7\text{ ms}$$

The complete dot-space period is $83.3\text{ ms}$, so its fundamental repetition frequency is $12\text{ Hz}$. Keying sidebands are spaced around the RF carrier according to this envelope spectrum; their amplitudes depend on the edge shape.

1. Necessary-bandwidth estimates: Applying the ITU Morse expression $B_n=BK$:

$$B_n=24\times3=72\text{ Hz}\quad\text{(non-fading circuit)}$$ $$B_n=24\times5=120\text{ Hz}\quad\text{(fading circuit)}$$

These estimates allow different numbers of useful keying sidebands for the stated circuit conditions. They are not a measurement of the power inside the actual transmitter's spectrum.

1. Compare the envelopes: The abrupt-edge setting generally produces stronger distant sidebands and more key clicks. Smooth $5\text{ ms}$ shaping reduces them. Neither the two rise-time values nor a guessed corner frequency supplies a 99% occupied-bandwidth calculation. To obtain that width, measure or calculate the complete power spectrum and find the limits leaving $0.5\%$ of total mean power in each outer tail. Sidebands outside those limits can still exist and cause interference.

Worked Example 7.2.2: Spurious-Power and Filter Budget for 400 W PEP

An Advanced amateur operates an SSB amplifier at $400\text{ W}$ PEP on $7.100\text{ MHz}$ under the class licence. Treat this as the necessary-bandwidth reference power at the antenna-line connection.

1. Reference power:

$$P_{\text{reference}}=10\log_{10}(400000)=56.02\text{ dBm}$$

1. Required attenuation below 30 MHz:

$$A=\min[43+10\log_{10}(400),50]=\min[69.02,50]=50\text{ dB}$$

1. Maximum spurious power under this requirement:

$$P_{\text{spur,max}}=56.02-50=6.02\text{ dBm}$$ $$P_{\text{spur,max}}=\frac{400\text{ W}}{10^{50/10}}=0.004\text{ W}=4\text{ mW}$$

1. Illustrative filter budget: Suppose the second harmonic at $14.200\text{ MHz}$ is measured $30\text{ dB}$ below the wanted output before the LPF. It needs at least $50-30=20\text{ dB}$ additional attenuation relative to the fundamental. This is a measured starting condition, not a guaranteed push-pull cancellation figure. Allow for passband loss and practical tolerances, and verify the complete transmitter at the antenna-line connection.