Anyone tried a UCD topology as a high level modulator for radio?

Hi all,
I am working in the design of an efficient shortwave transmitter and was wondering if anyone had tried a UCD style topology as a modulator for high level modulation of an RF final amplifier?

This would be single ended, as I do not need to drive the output negative, so really a self oscilating buck converter, and would need to swing the output from approximately 12V - ~48V into a 2.5 ohm load, at a rate sufficient to roughly track the envelope of a ssb transmisstion (Fine control is via a cartesian loop).

Does UCD work with more then the simple two pole output filter? I need to avoid switching frequency and harmonics getting through to generate noise sidebands on the final RF output, but the closed voltage loop of the UCD topology is very attractive.

Finally, has anyone tried to come up with a workable polyphase variant?

Regards, Dan.
Last edited:
I wouldn't use a UcD for this purpose but a triangular-wave based class-d amp.
Then you can use an Elliptic filter for instance to remove the residuals. Since you are connecting an antenna to the whole arrangement in the end you definitley have to make sure that carrier suppression is better than the one of an audio amplifier.


All modulation is by phasing method, which is straightforward for AM/SSB/PSK/CW/FSK and slightly less easy for FM.

The plan is to vary the final amplifier supply from approximately 12V to 48V to follow the modulation envelope (Cannot go below about 12V as the mosfet behaviour becomes stupid down there, hence cartesian rather then polar loop for this one), further by monitoring the load impedance I can ensure that the supply voltage suits the load presented by the aerial so that 1.5:1 hi Z (75 ohms) gets more volts then 1.5:1 low Z (33 ohms).
The bias is also variable and the plan is to run in class AB at low output power, transitioning to class C as the output comes up (With the feedback automatically ramping the drive power to compensate the gain changes).

For modes with high power backoff ratios like most voice modes, this should massively increase overall efficiency compared to a classical class AB amplifier.

A NCO and CORDIC core forms the modulator, with a second cordic core doing the demodulation from the ADC to close the cartesian loop and linearize the whole mess.

One of the nice things about starting off with an I/Q pair is that you can derive the envelope of the final rf simply as sqrt (I^2 + Q^2).

73 M0HCN