Chapter 22: The Transmitter

To generate a modulated RF signal for an antenna

A transmitter must produce a radio-frequency signal that carries information, through modulation, ready for the antenna. A carrier with no modulation carries no information, and recovering audio is the receiver’s job.

VFO, buffer, PA, lowpass filter

The VFO sets the frequency, the buffer shields it, the keyed PA builds the power and the lowpass filter comes last. A buffer after the PA could no longer shield the VFO from the PA’s changing load.

To give the VFO a light, steady load

With a light, steady load the VFO is unaffected as the PA is keyed, so its frequency holds and there is no chirp. Changes in supply voltage need a regulated supply; a buffer cannot cure them.

The power amplifier (PA)

The keying block drives the PA, while the VFO and buffer run all the time, so the frequency stays steady. Starting and stopping the VFO for every character would make its frequency wander as it settles.

A keyed carrier needs no linear amplification

Keying only switches the carrier on and off, so the PA need not copy the RF waveform faithfully, and efficient class C suits it. Class C makes plenty of harmonics, which the output lowpass filter removes.

It weakens the harmonics that the PA produces

The PA’s distortion creates harmonics at whole multiples of the signal frequency, all above it, so a lowpass filter passes the signal and stops them. Blocking the fundamental and passing harmonics is a highpass filter.

3,66 MHz

Harmonics lie at whole multiples of the signal frequency, so the lowest is the second: 2 × 1,83 = 3,66 MHz. Class C makes even harmonics too, so the third, 5,49 MHz, is not the lowest.

About 2,2 MHz

The cutoff must lie above 1,85 MHz, so the signal passes, yet well below the second harmonic at 3,62 to 3,70 MHz. A 3,5 MHz cutoff sits just under that harmonic, so it would barely weaken it.

Key clicks

Switching the carrier on or off too sharply spreads keying sidebands, heard as clicks on nearby frequencies, so the shaper ramps it. Chirp, a shift of frequency as the key closes, is prevented by the buffer.

About 5 ms

Chapter 21 gives about 5 ms as the best rise and fall time: slow enough to keep the keying sidebands narrow, fast enough for Morse. A 50 µs edge is almost a square switch and would click.

10 MHz

The manual advises keeping VFOs below about 10 MHz, since the higher an LC oscillator runs, the harder it is to hold steady. 30 MHz, the top of HF, would be far too high.

80 m or 40 m, keeping the VFO low enough to be stable

Here the VFO runs on the output frequency, and VFOs are best kept below 10 MHz, so 3,5 MHz or 7 MHz suits. On 20 m the VFO itself would have to run at 14 MHz.

75 W

Efficiency = RF output ÷ DC input, so the input is 50 ÷ 0,65 = 76,9 W, and 75 W is nearest. Multiplying instead, 50 × 0,65 = 32,5 W, would give less input than output, which is impossible.

They must all be linear amplifiers

An AM signal holds the carrier and its sidebands at once, so non-linear stages would cause intermodulation distortion; every amplifier after the modulator must be linear. Class C suits only signals of constant amplitude: a plain carrier, CW or FM.

Low-level AM uses the same circuits that make SSB

Making SSB is simplest at low level, and the same circuits give AM too, given a filter wide enough for both sidebands. AM cannot pass the narrow SSB filter, which keeps only one sideband.

Its RF stages can be efficient class C amplifiers

Until it is modulated the carrier is a single frequency, so class C stages can amplify it efficiently, and the audio is added at full power. Either way the AM signal’s bandwidth is the same, set by the audio.

The RF is a single frequency until it is modulated

Intermodulation needs two or more frequencies to mix; the unmodulated carrier is just one, so class C only adds harmonics, which the lowpass filter removes. Intermodulation products near the signal could never be filtered.

Preamplifier, filter, audio PA, modulator

The weak microphone signal is pre-amplified, filtered to the speech band, built up by the audio power amplifier and then drives the modulator. The manual filters at low level, ahead of the audio PA, not after it.

To remove audio above about 3 kHz and keep the signal narrow

Communications speech needs only about 300 Hz to 3 kHz, and each audio frequency reappears that far either side of the carrier, so higher audio would widen the signal. Speech without its lows would be unintelligible.

300 Hz to 3 kHz

Speech stays intelligible with only 300 Hz to 3 kHz, so the filter removes the rest and keeps the signal narrow. The voice itself spans about 50 Hz to 20 kHz (chapter 21), far more than communication needs.

5 kHz

Each audio frequency appears both above and below the carrier, so the signal spans 2,5 kHz either side: 2 × 2,5 = 5 kHz. Only a single-sideband signal would fit into 2,5 kHz.

On the output frequency itself

The VFO feeds the buffer and the class C PA directly, with no mixer or multiplier, so it runs on the output frequency. A doubler would need the VFO at half the output, which this design does not use.

Two-thirds

At 100% modulation the two sidebands together carry half as much power as the carrier, so the carrier holds 1 part in 1,5: two-thirds. One-third is the sidebands’ combined share.

30 W

The audio supplies the sideband power, half the carrier at 100% modulation: 0,5 × 60 = 30 W, more in practice to cover the PA’s losses. Taking a third of the carrier gives the wrong 20 W.

45 W

Full modulation needs audio of half the carrier power, so the carrier can be 2 × 23 = 46 W (less after real PA losses): 45 W is nearest. Halving instead gives 11,5 W, nearest 12 W.

105 W

At 100% modulation the sidebands add half the carrier power: 70 + 35 = 105 W, two-thirds of it in the carrier. Adding a third of the carrier instead gives 70 + 23,3 = 93 W.

The audio power amplifier

In high-level modulation the sideband power comes from the audio power amplifier, through the modulator, so it must be large. The RF power amplifier only turns it into RF; its own DC supply provides the carrier.

Distortion and splatter onto nearby frequencies

Beyond 100% the carrier is cut off on the troughs and the modulation is clipped, which distorts it and spreads splatter onto nearby channels (chapter 21). The signal grows wider, not usefully louder.

It varies the final stage’s supply with the audio

Its secondary carries the supply to the final stage, so the audio adds to and subtracts from the supply voltage and the RF output’s amplitude follows. It carries audio, not RF; varying the frequency would give FM.

At the transistor’s base

The RF drive goes to the base, while the audio drives the transformer’s primary, whose secondary feeds the collector supply through the RF choke. The emitter is earthed, and +VCC is the supply end.

It passes DC and audio but blocks the RF

A choke has little reactance at DC and audio but a high reactance at RF, so the modulated supply reaches the collector while the RF stays out of the supply. Passing RF and blocking DC is the output capacitor’s job.

It passes the modulated RF and blocks the DC supply

A capacitor’s reactance is low at RF and infinite at DC, so the modulated RF reaches the output while the collector’s DC supply is kept off it. Passing DC and blocking RF is the choke’s job.

Use an unbalanced modulator and a wider crystal filter

An unbalanced modulator lets the carrier through, and a crystal filter wide enough for both sidebands passes the whole AM signal. The narrow filter would keep one sideband and reject the carrier just outside its passband.

The crystal filter

The crystal filter follows the balanced modulator, then come the mixer, bandpass filter, PA and lowpass filter. The crystal filter works only at the fixed 9 MHz, so it cannot follow the mixer, whose output moves with the VFO.

Both sidebands, with the carrier suppressed

Balancing cancels the carrier and leaves the upper and lower sidebands: a double-sideband suppressed-carrier signal. An unbalanced modulator would keep the carrier as well, giving AM.

The balanced modulator

The balanced modulator cancels the carrier and leaves only the two sidebands. The crystal filter’s job is to remove one sideband, though in practice it also trims any carrier left over.

A mixer, which multiplies two signals

The circle with a cross is the mixer symbol; the manual uses it for the balanced modulator and for the mixer, since both multiply two signals. Simply adding two signals would create no new frequencies.

It passes the upper sideband and blocks the lower

The carrier has already gone at the balanced modulator; the very narrow crystal filter passes only the upper sideband, 300 Hz to 3 kHz above 9 MHz, and rejects the lower. This is how the filter method makes SSB.

The filter method

A balanced modulator removes the carrier and a narrow crystal filter then removes one sideband: the filter method. The phasing method, which cancels the unwanted sideband instead, is a different technique, not used here.

8,998 to 8,999 MHz

The lower sideband lies below the carrier by each audio frequency: 9,000 − 0,002 = 8,998 MHz up to 9,000 − 0,001 = 8,999 MHz. Adding instead gives the upper sideband, 9,001 to 9,002 MHz.

2,7 kHz

One sideband spans the speech band itself, shifted in frequency: 2,9 kHz − 250 Hz = 2,65 kHz, so 2,7 kHz is nearest. Doubling the top audio frequency gives 5,8 kHz, nearest 6 kHz, the width of AM.

Its crystal filter cannot be retuned

Filters selective enough to pass one sideband and reject the other are fixed-tuned, so the signal is formed at their frequency and then shifted by a mixer. Balanced modulators work over a wide range of frequencies.

Digital signal processing (DSP)

A footnote says growing use of DSP will soon make the fixed-tuned rule untrue: a digital filter can be retuned in software. Better quartz crystals would still give fixed-tuned filters.

14,215 MHz

The sum product is 9,000 + 5,215 = 14,215 MHz, in the 20 m phone segment. Subtracting gives 9,000 − 5,215 = 3,785 MHz, the difference product, which the bandpass filter rejects.

5,000 to 5,350 MHz

The VFO supplies the difference between output and 9 MHz: 14,000 − 9,000 = 5,000 MHz up to 14,350 − 9,000 = 5,350 MHz. Adding 9 MHz to the output gives 23,000 to 23,350 MHz, the wrong way round.

3,830 MHz

The difference product is 9,000 − 5,170 = 3,830 MHz. 5,170 MHz is the VFO’s own signal, which the bandpass filter also rejects; the manual (p199) calls that VFO range the unwanted product.

It moves down, from 3,800 to 3,700 MHz

The difference falls from 9,000 − 5,200 = 3,800 MHz to 9,000 − 5,300 = 3,700 MHz as the VFO rises: the inversion the manual mentions. Only the tuning runs backwards; the signal is still USB.

It passes the transmitter’s 20 m output, blocking the rest

Passing only 14,100 to 14,350 MHz, this LC filter keeps the sum and rejects both the 3,65 to 3,9 MHz difference and the VFO’s own 5,1 to 5,35 MHz signal, which p199 calls the unwanted product.

The unwanted products lie MHz away from the wanted one

The nearest unwanted mixer output lies several MHz from 14,1 to 14,35 MHz, so a gentle LC filter can reject it. Sidebands only kHz apart need a crystal filter; the signal is still weak here, so power is no concern.

A linear amplifier, probably in class AB

An SSB signal’s amplitude follows the speech, so the PA must be linear; the manual suggests class AB, linear enough yet reasonably efficient. A class C stage would distort it into splatter and intermodulation.

A multiplier’s distortion would ruin the SSB signal

A multiplier works by distortion, which would intermodulate the many frequencies in the SSB signal into products too close to filter out; a properly designed mixer shifts it cleanly. Multiplying swaps no sideband.

The PA creates new harmonics of its own

Even a linear class AB PA adds harmonics, from 28,2 MHz up, after the bandpass filter, so a lowpass filter must follow it. A VFO leak at 5 MHz lies below the signal, which a lowpass filter passes.

28 MHz

The lowest harmonic is twice the lowest frequency: 2 × 14,170 = 28,34 MHz, so 28 MHz is nearest. The third harmonic, 3 × 14,170 = 42,51 MHz, is not the lowest; even harmonics occur too.

A VFO is hard to keep stable enough at VHF

At VHF an LC oscillator drifts too much for channelised FM, while a synthesiser holds its output almost as steady as its crystal reference. A VFO can be frequency-modulated quite easily, with a varicap.

Its channels lie far apart, so coarse steps will do

With channels 12,5 or 25 kHz apart the reference can be that large, and one loop then steps quickly. Multiple loops are needed only where small steps are wanted, since a low reference slows a single loop (chapter 20).

To keep the audio below about 3 kHz, so the FM stays narrow

Speech needs audio up to about 3 kHz, and higher audio would widen the FM signal, so the lowpass filter removes it before the modulation. A 300 Hz cutoff would remove most of the speech itself.

To the 12,8 MHz crystal oscillator (VCXO)

The filtered audio drives the VCXO, whose 12,8 MHz output is divided down to become the reference, so the reference carries the FM. Many synthesised rigs modulate the VCO instead, but not the manual’s design.

A varicap diode pulls its frequency slightly

The audio voltage changes the varicap’s capacitance, and with it the crystal’s frequency, by a small amount: pulling the crystal. Switching between two frequencies would be frequency-shift keying, used for data.

It is locked to a reference that carries the FM

The loop holds the VCO at N times the reference, so when the reference swings with the speech, the VCO swings with it, N times as far in hertz. Stray coupling would be a fault, not the design.

±1,4 kHz

The ÷1 024 divides the swing and the loop multiplies it by N: 350 × 4 200 ÷ 1 024 = 1 436 Hz, nearest ±1,4 kHz. It is not passed on unchanged; leaving out the ÷1 024 gives ±1,5 MHz.

260 Hz

The ÷1 024 passes on 64 ÷ 1 024 = 0,062 5 Hz of error, which the loop multiplies by N: 0,062 5 × 4 160 = 260 Hz. The error does not stay at 64 Hz.

It divides the VCXO’s output down to the reference

It divides the frequency-modulated 12,8 MHz VCXO signal by 1 024 to give the 12,5 kHz reference. Dividing the VCO’s output for the comparator is the job of the ÷N divider, which the microprocessor sets.

6,25 kHz

6 400 000 Hz ÷ 1 024 = 6 250 Hz, or 6,25 kHz, a channel step some radios use. Dividing by 102,4 instead, one decimal place out, gives 62,5 kHz.

512

The ratio is VCXO ÷ reference = 12 800 kHz ÷ 25 kHz = 512, half the manual’s 1 024. Keeping ÷1 024 would give the manual’s 12,5 kHz reference, half of what is wanted.

Phase comparator, lowpass filter, VCO and ÷N

The loop itself is boxed: phase comparator, lowpass filter, VCO and ÷N divider. The microprocessor that sets N, and the VCXO and ÷1 024 that make the reference, lie outside the box.

It sets the ÷N ratio for the chosen frequency

The microprocessor reads the user’s choice and loads the matching ratio into the ÷N divider, from 11 521 to 11 679 across the 2 m band. The ÷1 024 ratio is fixed.

17 200 to 17 600

N = VCO frequency ÷ reference: 430 000 kHz ÷ 25 kHz = 17 200, and 440 000 ÷ 25 = 17 600. Dividing by 12,5 kHz instead gives the wrong 34 400 to 35 200.

435,000 MHz

The VCO settles at N × reference = 34 800 × 12 500 Hz = 435 000 000 Hz, or 435,000 MHz. Taking the reference as 1,25 kHz, one decimal place out, gives the wrong 43,500 MHz.

35 092

N = VCO frequency ÷ reference = 438 650 kHz ÷ 12,5 kHz = 35 092. Taking the reference as 1,25 kHz, one decimal place out, gives the wrong 350 920.

37,5 kHz

Each step of N moves the VCO by one reference step, so 3 × 12,5 kHz = 37,5 kHz. Taking the whole change as a single 12,5 kHz step ignores how far N moved.

It smooths the comparator’s output to steer the VCO

The comparator’s output carries ripple at the 12,5 kHz reference besides the phase-error voltage; the filter removes the ripple, leaving a smooth voltage, speech swing included, to steer the VCO. It does not filter the RF.

On the output frequency, in the 2 m band

The VCO is locked at N times 12,5 kHz, from 144,012 5 to 145,987 5 MHz, so it runs on the output frequency itself. No tripler follows it; the PA amplifies its output directly.

A power amplifier, then a lowpass filter

The VCO already runs on the output frequency, so the PA simply amplifies it and the lowpass filter removes its harmonics. Older crystal-controlled FM rigs multiplied up, but this design needs no multiplier.

Class C

An FM signal’s amplitude stays constant, so the PA need not be linear and efficient class C suits it, as chapter 18 says; the lowpass filter removes its harmonics. Linear class AB is needed for SSB or low-level AM.