Catto / Topic Map · Electronic Measurements II Level 5
Electronic Measurements II · 120 h · Topic 3 of 10

Synthesized signal generators

For 1 MHz channels at 2.4 GHz, an integer-N synthesizer multiplies the phase noise of its comparator by 2440. A fractional one, by 49: its phase noise ends up 17 dB lower.

PLL N divider Phase noise Fractional-N Spurs

01From the DDS to the phase-locked loop

Direct digital synthesis in the signal sources of Measurements I gives very fine resolution, but its output cannot exceed a fraction of the clock. To generate signals from hundreds of megahertz to several gigahertz with the accuracy of a crystal, RF generators use a phase-locked loop (PLL): a voltage-controlled oscillator (VCO) whose frequency, divided by \( N \), is compared with a stable reference. The loop corrects the VCO until the two coincide:

\[ f_{out} = N \cdot f_{PFD} \] \[ f_{PFD} = \frac{f_{ref}}{R} \] \( f_{PFD} \) is the frequency at which the phase-frequency detector operates. In an integer-N synthesizer, the output frequency step is \( f_{PFD} \): for 25 kHz channels, the comparison has to be at 25 kHz.

02The building blocks of the synthesizer

BlockFunction
ReferenceA temperature-compensated (TCXO) or oven-controlled (OCXO) crystal oscillator, sometimes locked to GPS
R dividerLowers the reference to the comparison frequency
Phase-frequency detector with charge pumpDelivers current pulses proportional to the phase error
Loop filterIntegrates the pulses and sets the loop bandwidth and its stability
VCOGenerates the output; its frequency depends on the control voltage
N dividerBrings the output down to the comparison frequency

The PLL is a feedback system, and its bandwidth and phase margin are designed like those of a feedback amplifier. For stability, the loop bandwidth is kept below one tenth of \( f_{PFD} \): comparing at a low frequency forces a slow loop, which takes a long time to lock when the channel changes.

03Phase noise

A real generator does not produce a perfect spectral line: its phase fluctuates, and the spectrum has "skirts" on both sides of the carrier. Phase noise is specified in dBc/Hz at an offset from the carrier. Within the loop bandwidth, the noise of the reference and of the detector is multiplied by \( N \), that is, it rises \( 20 \log N \); outside it, the VCO noise dominates. The in-band noise is estimated with the detector’s figure of merit, about −220 dBc/Hz in current circuits:

\[ \mathcal{L}_{band} \approx FOM + 10\log f_{PFD} + 20 \log N \] For the same output frequency, \( N f_{PFD} \) is fixed: doubling \( f_{PFD} \) adds 3 dB through the second term and subtracts 6 dB through the third. It pays to compare at the highest possible frequency.

That is the reason for fractional-N synthesizers: the divider alternates between \( N \) and \( N+1 \) with a sigma–delta modulator, so that the average is fractional. The comparison is made at tens of megahertz and a step of a few hertz is obtained. The price is the fractional spurs, lines close to the carrier that the modulator and the loop filter have to keep low.

Lab · integer-N or fractional-N synthesizer

A synthesizer for the 2.4 GHz band with a figure of merit of −220 dBc/Hz. In integer-N, the comparison frequency is the channel step; in fractional-N, 50 MHz. The bars compare the in-band phase noise of the two architectures.

04How to read the datasheet of a synthesized generator

  • Frequency range and resolution, and the accuracy, which is that of the internal or external reference.
  • Phase noise at 10 kHz or 20 kHz from the carrier, the figure that separates a laboratory generator from a high-purity one.
  • Spurs and harmonics, in dBc.
  • Switching time between frequencies, which depends on the loop bandwidth.
  • Output level and its accuracy, and the available modulations: AM, FM, pulse, I/Q.

05In the laboratory

Activity 1 · Phase noise

Use a spectrum analyzer to measure the phase noise of a synthesized generator at 1 kHz, 10 kHz and 100 kHz from the carrier, correcting for the resolution bandwidth.

Activity 2 · An integrated PLL

Program an integrated PLL synthesizer with a microcontroller, in integer-N and in fractional-N, and compare the output spectrum and the spurs.

Activity 3 · External reference

Lock the laboratory’s generator and counter to the same 10 MHz reference and observe how the reading difference disappears.

06Common mistakes

  • Comparing at the channel step frequency without need. It raises the phase noise and makes the loop slow.
  • Forgetting that the reference noise is multiplied. A noisy reference ruins a GHz synthesizer.
  • Ignoring the fractional-N spurs. They can fall inside the adjacent channel.
  • Reading the analyzer’s phase noise without correcting for its bandwidth. It is expressed per hertz.

07Self-assessment

What N is needed to synthesize 915 MHz comparing at 200 kHz?

4575.

How much does the in-band phase noise rise when multiplying by 1000?

\( 20 \log 1000 = 60 \) dB.

Why can a fractional synthesizer have a faster loop?

Because it compares at a much higher frequency, and the loop bandwidth can be proportionally greater.

What is the disadvantage of fractional-N?

The fractional spurs close to the carrier.

08Further reading

  • Dean Banerjee. PLL Performance, Simulation, and Design. 5th ed., Dog Ear Publishing, 2017. The design of charge-pump synthesizers, with phase noise and spurs.
  • Ulrich L. Rohde. Microwave and Wireless Synthesizers: Theory and Design. Wiley, 1997. The classic reference on radio-frequency synthesizers.
  • Roland E. Best. Phase-Locked Loops: Design, Simulation, and Applications. 6th ed., McGraw-Hill, 2007. Analog and digital PLLs from the standpoint of control theory.
Development of the topic “Synthesized signal generators” of Electronic Measurements II (Level 5), based on the curriculum of the UTN Electronic Engineering program, 2023 curriculum — Ordinance No. 1849 of the UTN Higher Council. Back to the Topic Map · catto.ar