Abstract:
We report a novel design of a digitally controlled oscillator (DCO), its operation, and characteristics, whose performance is boosted with machine learning (ML) implementation. The DCO outputs 10–300 MHz at the stability of rubidium atomic clocks, and the finest possible frequency tunability resolution is greatly enhanced to 10 MHz with ML. The system described here is dual-channel and easily scalable to multichannel, where the interchannel output frequencies and phases are synchronized; however, they are tunable following end users’ control. The system shows interchannel frequency and phase drifts of 1.3 MHz and 3 mrad, respectively, over 24 h of continuous operation and a phase noise of <−140 dBc/Hz at 10 kHz. This makes it suitable for versatile time-sequenced precision radio frequency (RF) measurements that simultaneously require multiple outputs with shot-to-shot redundancies. The system can be fully controlled from an in-built touch panel and also from a remote PC, thus making it user-cum-time friendly.