Skip to content

VCO Selection: Frequency, Phase Noise, and Tuning Range

Phase noise is the spec that separates transmitters that work from transmitters that pass certification. And it starts with the VCO.

A voltage-controlled oscillator sets the carrier frequency for your transmitter. Get it wrong, and no amount of filtering or amplification downstream will fix the modulation quality, spectral purity, or adjacent-channel performance.

What a VCO actually does in a transmitter

The VCO generates the local oscillator (LO) signal that drives the mixer. In a superheterodyne transmitter, the mixer combines the baseband signal with the LO to produce the RF output at the target frequency. In a direct-conversion architecture, the VCO runs at the output frequency itself.

The VCO's output frequency is controlled by an analog tuning voltage — typically 0.5 V to 4.5 V for a CMOS VCO, or 1 V to 15 V for a discrete varactor-tuned design. The relationship between tuning voltage and output frequency defines the tuning sensitivity (KV), measured in MHz/V.

In almost all modern transmitter designs, the VCO is embedded inside a phase-locked loop (PLL) that locks it to a stable reference oscillator — typically a 10-100 MHz crystal or TCXO. The PLL corrects the VCO's frequency drift and reduces its close-in phase noise. Outside the PLL bandwidth, the VCO's inherent phase noise dominates. Inside the bandwidth, the reference oscillator's phase noise (multiplied to the output frequency) dominates.

This means VCO selection and PLL design are coupled problems. A VCO with excellent phase noise at 1 MHz offset but poor performance at 10 kHz offset might work with a narrow-bandwidth PLL. A VCO with mediocre far-out noise but clean close-in performance needs a wider PLL bandwidth to let the reference clean up the near-carrier noise.

Phase noise: the spec that matters most

Phase noise quantifies the spectral purity of the oscillator. It's measured in dBc/Hz at a given offset from the carrier — for example, -110 dBc/Hz at 100 kHz offset from a 5 GHz carrier.

Why it matters for transmitters:

EVM degradation. Error vector magnitude measures how accurately the transmitted signal matches the ideal constellation. Phase noise directly rotates constellation points, increasing EVM. A 5G NR 256-QAM signal requires EVM below ~3.5%. To achieve that, the integrated phase noise contribution from the LO typically needs to be below -35 dBc (integrated over the signal bandwidth). For 64-QAM, the requirement relaxes to about -30 dBc.

Adjacent channel leakage. Phase noise creates a noise skirt around the carrier that extends into adjacent channels. Cellular standards specify adjacent channel leakage ratio (ACLR) limits — typically -45 dBc for LTE and -28 to -45 dBc for 5G NR depending on the band. The VCO's phase noise at offsets corresponding to the channel spacing directly affects whether these limits are met.

Reciprocal mixing. In a frequency-hopping or multi-carrier system, phase noise from one carrier can mix with adjacent carriers, degrading the signal-to-noise ratio of neighboring channels.

Real phase noise specs to benchmark against:

At 5 GHz, a good integrated VCO/PLL achieves -100 to -110 dBc/Hz at 100 kHz offset. An excellent one reaches -115 dBc/Hz. At 28 GHz, the same design achieves roughly 15 dB worse due to the 20log(N) frequency multiplication penalty — so -95 dBc/Hz at 100 kHz offset is a strong result.

Discrete VCOs with external resonators can achieve -120 dBc/Hz or better at 100 kHz offset at 5 GHz, but at the cost of board space, power consumption, and design complexity.

Tuning range and frequency planning

VCO tuning range determines which output frequencies the transmitter can reach. This sounds straightforward — pick a VCO that covers your band. In practice, it's more nuanced.

Continuous tuning range. A VCO might specify 4.5-5.5 GHz tuning range. But is that range continuous, or does it have dead zones or mode jumps? Some wideband VCOs use switched capacitor banks to cover a wide range, with each bank covering a sub-range. The transitions between banks can create brief frequency discontinuities that matter for fast-hopping applications.

Tuning linearity. The relationship between tuning voltage and frequency is rarely linear. A VCO with 1 GHz of tuning range might have 400 MHz/V sensitivity at the low end and 100 MHz/V at the high end. Nonlinear tuning sensitivity affects PLL loop dynamics — the loop gain varies across the band, which changes the phase margin, lock time, and noise transfer function.

Frequency planning for superheterodyne. If your transmitter uses an intermediate frequency (IF), the VCO frequency equals the output frequency plus or minus the IF. For a transmitter targeting 3.5 GHz with a 1 GHz IF, the VCO needs to operate at either 4.5 GHz (high-side LO) or 2.5 GHz (low-side LO). High-side injection avoids the image frequency falling in-band but requires a higher-frequency VCO.

Harmonic and sub-harmonic operation. Some designs use a lower-frequency VCO with a frequency multiplier (doubler or tripler) to reach mmWave frequencies. A 14 GHz VCO doubled to 28 GHz avoids the challenges of building a stable oscillator at 28 GHz directly. But the doubler adds 7-10 dB of conversion loss and produces unwanted harmonics that need filtering. The trade-off between VCO complexity and multiplier chain complexity depends on the target frequency and required phase noise.

Power consumption and output level

VCO power consumption ranges from under 20 mW for low-power IoT applications to over 500 mW for high-performance synthesizers. The power directly trades against phase noise — a VCO running at higher bias current generally achieves better phase noise because the carrier-to-noise ratio improves.

Output power matters for driving the mixer. Most passive double-balanced mixers want +7 to +13 dBm of LO drive. An active mixer might need only 0 dBm. If the VCO's output is -5 dBm, you need a buffer amplifier between the VCO and mixer — adding components, power, and potential noise.

Integrated VCO/PLL modules (like those from Analog Devices' ADF series or Texas Instruments' LMX series) include output buffers with programmable power levels. A typical integrated solution provides +2 to +5 dBm output, which is sufficient for most active mixers but may need amplification for passive mixers.

Integrated vs. discrete VCO selection

Integrated VCO/PLL modules. Components like the ADF4372 (62.5 MHz to 16 GHz), LMX2595 (10 MHz to 19.6 GHz), or MAX2871 (23.5 MHz to 6 GHz) combine the VCO, PLL, dividers, and output buffers in a single package. Advantages: reduced board space, pre-characterized loop performance, simplified design. Disadvantages: phase noise floor is fixed by the integrated VCO design, and you can't independently optimize the oscillator.

Discrete VCOs. Separate VCO modules from companies like Mini-Circuits, Synergy Microwave, or Z-Communications offer the widest range of frequencies and often better phase noise than integrated solutions. But they require external PLL design — loop filter, charge pump, reference distribution — which adds design time and board space.

MMIC VCOs. Gallium arsenide or SiGe MMIC oscillators for mmWave frequencies above 20 GHz. These are often the only option at the highest frequencies but require careful attention to supply bypassing, output matching, and thermal management.

The trend is strongly toward integration. For most designs below 20 GHz, an integrated VCO/PLL saves enough design time and board space to outweigh the phase noise penalty. Above 20 GHz, discrete or MMIC solutions are still common because integrated options are limited.

Selection criteria in practice

When selecting a VCO for a transmitter design, the decision matrix typically includes:

  1. Does the tuning range cover the required LO frequency (including margin for PLL pull-in range)?
  2. Does the phase noise at relevant offsets meet the system EVM and ACLR requirements?
  3. Is the output power sufficient to drive the chosen mixer, or is a buffer needed?
  4. Does the power consumption fit the thermal budget?
  5. Is the package compatible with the PCB layout and grounding strategy?
  6. Is the part available, in production, and not on a single-source end-of-life trajectory?

Failing any one of these often means restarting the search. And because VCO selection affects the PLL design, which affects the mixer LO drive, which affects the gain chain — changing the VCO late in the design process cascades through the entire transmitter.

What's your approach to the integrated vs. discrete VCO decision? I've found that integrated modules save significant design time below 10 GHz, but above that the phase noise compromise starts to hurt.

This is Part 2 of an 8-part series on RF transmitter design. Previously: "The Anatomy of an RF Transmitter Chain." Next: "Mixer and Upconversion: Getting Your Signal to RF."