Although most HF transceivers will allow you to listen on 472 kHz, only a few can transmit here. And even if they can, the output power is often limited. So, unless you buy a JUMA TX500 or a TVTR1 630 m transverter , homebrew is the only way to get on the air.
Basically there are 4 options:
A 472 kHz transmitter. This is the simplest solution, but will limit the transmit modes to CW (including QRSS and Opera).
A 472 kHz transverter. This is a bit more complicated but will allow you to use all popular modes.
A 472 kHz upconvertor. The audio signal is upconverted to 472-479 kHz. This will allow all modes. A special design is needed in order to suppress all unwanted mixing products and carriers.
Modification of an existing HF transceiver (see "Getting started").
472 kHz CW transmitter
The design of a CW transmitter is straightforward: oscillator - driver - PA.
The oscillator can be a well designed and built "free running" VFO. The stability will be sufficient for CW, but probably not for QRSS or Opera.
A crystal oscillator will offer excellent stability, but you are tied to one frequency. As crystals in the 472-479 kHz range are rare (and expensive) it is common practice to use crystals at a multiple of 472-479 kHz and divide the signal down.
A ceramic resonator VXO is a low cost solution that offers good stability and a tuning range of some kHz. Here also it is common practice to have the VXO running at a higher frequency that is divided down to 472-479 kHz.
Other options are a PLL or DSS oscillator. Both offer excellent stability and access to the complete frequency range.
The PA is almost always a class D or E power amplifier using low cost MOSFETs. These amplifiers have an efficiency in excess of 80%. In both designs the MOSFET is used as a switch. As the input capacitance of power MOSFETs can be rather high (several nF) the driver must be able to deliver high peak currents, in order to switch the MOSFET properly. Very convenient are high-speed power MOSFET drivers IC's such as TC4426/28, IR2304, ...
Class D amplifiers, and to a lesser extend class E amplifiers, require a good low pass filter between the PA and antenna.
Transverter
A transverter that downconverts a HF signal to 472 kHz is more versatile. If all stages, including the PA are linear any mode that your HF transceiver can handle can be used on 472 kHz. But, as mentioned before, for most popular modes used on 472 kHz (CW, QRSS, DFCW, JT9, WSQ, WSPR, Opera) don't require a linear PA.
Unless the HF transceiver has a (low power) transverter output an attenuator will be needed to avoid overload or even destruction of the mixer. At the mixer output a lowpass filter, or eventually a bandpass filter, will keep all unwanted signals from the following stages. If linear amplification is not required the driver and PA can be the same as for a CW transmitter.
Upconverter
With an upconverter the audiosignal is directly converted to the desired RF frequency. This sounds simple, but a special technique is required to guarantee sufficient suppression of all unwanted signals.
An example: assume we have a 1000 Hz audiosignal that we want to convert to 475 kHz. This can be done by mixing the audiosignal with a 474 kHz carrier, the sum product will be 475 kHz. But there will be unwanted signals at 473 kHz (474-1 kHz) and 474 kHz (carrier). A very narrow bandpass filter at RF would be needed in order to suppress this unwanted signals.
With a special technique called Third Method Sideband Generation and using the fact in practice only narrow bandwidth modes (< 100 Hz) are used, a clean RF signal can be produced without the need of a narrow RF bandpass filter.
The audio signal (from a PC soundcard) is downconverted to a "baseband" of ≈ 20 - 200 Hz. This is done twice where both mixers are fed with a 90° phaseshifted signal. Both baseband signals are then upconverted to the desired RF frequency, again using 2 signal that are 90° phaseshifted. After summing all RF signals, except for the wanted one, will be suppressed. Finally a 500 kHz low pass will suppress all harmonics.
Class D power amplifier
A class-D power amplifier utilises the active device(s) as a switch. This requires proper on/off switching of the active devices. Fast switching Power MOSFETs are very suitable as active device, and they are cheap.
Assuming the devices act as a perfect switch efficiency can be 100%, as either the voltage over the device or the current through the device is zero. In reality efficiencies of 80% - 90% can be achieved.
The main causes of loss (for power MOSFETs) are:
"ON resistance" of the MOSFET: in the ON state the MOSFET will act as a resistor. Typical values are from a few mΩ to several 100 mΩ. The lower the ON resistance the less loss.
Rise and fall time of the MOSFET: MOSFETs do not switch instantly, there are transition periods from the ON to the OFF state and vice versa. During these periods both the MOSFET current and voltage will not be zero and the MOSFET will dissipate a significant amount of power. The shorter the rise and fall times the less loss.
Due to the on/off switching of the active device(s) the output signal will contain a lot of harmonics, a good lowpass filter at the TX output is an absolute must!
Class D power amplifiers are often designed as a "push-pull" amplifier. As the MOSFETs must be driven by square wave signals the input transformer is replaced by a driver circuit that produces 2 opposed square wave signals.
One of the advantages of a push-pull amplifier is a good suppression of all even harmonics. In particular the suppression of the 2nd harmonic will make the design of the output lowpass filter less critical.
Class E power amplifier
In a class D amplifier the main losses occur when the MOSFET is switched (on or off). These losses can be minimised if:
The rise of voltage (MOSFET switched off) is delayed until after the current has reduced to zero.
The voltage returns to zero (MOSFET switched on) before the current begins to rise.
The MOSFET voltage at switch on time is nominally zero.
The slope of the MOSFET voltage waveform is nominally zero at turn on time.
These timing requirements are fulfilled by a suitable network between the MOSFET and the load (antenna). As a result the waveforms never have high voltage and high current simultaneously, reducing the power dissipation in the MOSFET.
With class E power amplifiers an efficiency of well over 90% can be achieved on 472 kHz.
But there are also some disadvantages:
In contradiction to class D has a class E amplifier a narrow bandwidth.
The efficiency of a class E amplifier will drop rapidly if the load (antenna) is not properly matched.