15-25 MHz Fractional-N Synthesizer

Design Schematics Embedded CPU JAM STAPL Loop Filter Spectra MASH Simulation Back to projects

19.876543 MHz

This phase noise peaking at an offset of 500 Hz agrees well with the SCILAB closed loop gain prediction and simulation. The noise shelf, which extends to beyond 100 KHz in the first graph, is the phase noise of the spectrum analyzer.

19.001000 MHz

Illustrating the removal of integer-N boundary spurs by slightly altering the comparison frequency:

Reference divisor M=100

Reference divisor M=104

Integer-N boundary spurs occur when harmonics of the comparison frequency mix with the VCO (or its harmonics) and the resultant beat frequency falls within the loop bandwidth e.g. when the fractional command is very small. Since the synthesizer can multiply by any fractional amount, the spurs can be avoided by slightly shifting the reference frequency, adjusting the N.F command by a reciprocal amount.

19.051000 MHz

Lower amplitude spurs also occur close to ½, ¼ ... 1/2x integer-N boundaries e.g:

Reference divisor M=100

Reference divisor M=104

When the fractional command is F=0.51, the spurs are 2 KHz from the carrier. The 381st harmonic of the reference frequency beats with the 2nd harmonic of the VCO! More screening / filtering between the VCO and the loop filter might be a good idea!

There is very little on the web about integer-N boundary spurs. According to Skyworks[2,3] the mechanism that precipitates them is not entirely understood! They should not be confused with fractional spurs, which occur at the smallest fractional step, and are well below the noise floor in this synthesizer.

Skyworks noted: "the spurs tend to lessen when the VCO drive level into the VCO divider port is lowered." Finite rise times may be a factor. Sometimes VCO and reference edges coincide e.g. 1/100e3 - 190/19.001e6 = 526 ps which is less than their rise times. Ground bounce from either will shift the zero crossings of the other. Deterministic jitter could produce the spur levels shown above: -47 dBc on 19.001 MHz is 150 ps peak-to-peak.

Update - April 2009

My new 38-76 MHz Fractional-N Synthesizer is free from integer-N boundary spurs.