Application note SLWA066 explores the impact a switching power supply makes on an RF synthesizer’s phase noise.  The gold standard phase noise performance comes from use of a truly linear supply, such as an HP6213A. Even a clean switching supply such as the Agilent E3631A makes a notable dent in performance.

 In response, the TRF3765 EVM includes the TPS74201 ultra-low noise regulator to clean up even a dirty supply. It does a tremendous job at generating a clean 3.3V rail. But what about the customer who wants to test performance using a 5V VCC_Tank supply to achieve ultimate phase noise performance? Unfortunately, there is no equivalent ultra-low noise regulator for 5V.

 Up through revision D of the TRF3765 EVM board, a separate filter board with some big honkin’ inductors and capacitors shipped with every kit. The filter board cleaned up supplies but was inconvenient for customer and astonishingly expensive to produce for TI. The revision E board moves those filter components onto the board to eliminate the separate filter board. So how well does it work?

 The gold standard performance is seen below, with an unfiltered HP6213A supply. This plot is repeated on all of the comparison plots for easy viewing. All plots used default configuration EVM and GUI settings files.

Using an unfiltered Agilent E3631A for the supply degrades phase noise by 1.7dB, especially through the 10kHz offset region. Not too bad, overall, but some supplies and some conditions can be much worse.

Using the on-board filter restores the phase noise performance to linear supply levels.

The on-board filter actually outperforms the off-board filter by about 0.5dB.