Electronics Guide

The Tuning Characteristic and K_VCO

The relationship between output frequency and control voltage is called the tuning characteristic, or tuning curve. For a control voltage V_tune, the instantaneous output frequency is approximated as:

fout = f0 + KVCO × Vtune

where f₀ is the free-running frequency at the chosen reference point and K_VCO is the slope of the curve at that point. Because real tuning curves bend, K_VCO is properly the local derivative df/dV rather than a single constant, and it generally changes across the tuning range. In control-loop work the gain is often expressed in angular terms, K_VCO in radians per second per volt, because the loop responds to phase, which is the time integral of frequency.

The value of K_VCO is a design compromise. A large gain allows a wide tuning range from a limited control-voltage span, which is convenient, but it also magnifies the effect of any noise on the tuning line: a given microvolt of noise on V_tune produces frequency modulation proportional to K_VCO. A small gain rejects tuning-line noise but narrows the achievable range and may force the control voltage to swing close to the supply rails. Much of VCO design is the management of this trade-off.

Frequency and Phase Modulation

Because the output frequency follows the control voltage, a VCO is by construction a frequency modulator: applying an information-bearing signal to the tuning input produces frequency modulation of the carrier, the basis of FM transmitters and of the modulation features in laboratory function generators. The same mechanism, viewed through the integral relationship between frequency and phase, makes the VCO the element a phase-locked loop adjusts to track the phase of a reference. Any unwanted voltage on the tuning line, whether supply ripple, thermal noise in the tuning network, or coupling from nearby circuitry, becomes unwanted frequency modulation, which is why the cleanliness of the control voltage is as important to spectral purity as the oscillator core itself.

Tuning Range and the Tuning Ratio

The tuning range is the span of frequencies the VCO covers between the minimum and maximum control voltages. It is often quoted as a tuning ratio, f_max divided by f_min, or as a percentage of the center frequency. A narrowband VCO for a fixed-channel radio may tune only a few percent, while a wideband VCO for a broadband synthesizer or a test instrument may cover an octave (a 2:1 ratio) or more. Wide tuning range conflicts with low phase noise and with tuning linearity, because achieving a broad range usually requires a large tuning-element variation and a high K_VCO, both of which work against spectral purity and a constant tuning slope.

The Varactor Diode

A reverse-biased p-n junction has a depletion region whose width grows as the reverse voltage increases. Because the depletion region separates two conductive regions, the junction behaves as a parallel-plate capacitor whose plate spacing widens with voltage; the junction capacitance therefore falls as the reverse bias rises. Varactor diodes (also called varicaps or tuning diodes) are junctions designed and characterized specifically for this behavior. Their capacitance follows approximately:

Cj = Cj0 / (1 + VR/φ)γ

where C_j0 is the zero-bias junction capacitance, V_R is the applied reverse voltage, φ is the built-in junction potential (roughly 0.7 V for silicon), and γ is the grading coefficient set by the doping profile. For an abrupt junction γ is about 0.5; hyperabrupt junctions are doped to give γ near 1 or higher, producing a larger and more nearly linear capacitance change for a given voltage swing, which is valuable for wide-range and linearized tuning.

From Capacitance Change to Frequency Change

In an LC oscillator the resonant frequency depends on the total tank capacitance, so the varactor's voltage-dependent capacitance translates directly into voltage-dependent frequency. For a tank inductance L and total capacitance C_tank:

f = 1 / (2π√(L × Ctank))

The varactor rarely sets the whole tank capacitance. A fixed capacitor is usually placed in series or parallel with it, partly to set the center frequency and partly to dilute the varactor's influence so that K_VCO and the tuning range come out at the intended values. This combination also limits how far parasitic and fixed capacitances can be swamped, which caps the achievable tuning ratio. Because frequency depends on the square root of capacitance, even a hyperabrupt varactor yields a tuning curve that is only approximately linear, and the residual curvature is one source of the K_VCO variation discussed earlier.

Varactor Nonidealities

Real varactors depart from the ideal voltage-variable capacitor in ways that affect VCO performance. The diode has a finite series resistance that lowers the quality factor of the tank and therefore worsens phase noise, so high-Q varactors are preferred for low-noise designs. The capacitance is sensitive to temperature, contributing to frequency drift. And because the junction is a nonlinear element, large radio-frequency voltage swings across the varactor modulate its own capacitance over each cycle, which can rectify signal into a bias shift, generate harmonics, and degrade the spectrum; keeping the radio-frequency swing across the varactor modest, or splitting the tuning across an anti-series varactor pair, mitigates this effect.

LC VCOs

An LC VCO sustains oscillation across a resonant tank of an inductor and capacitance, with a varactor providing the tuning. Because the tank is a high-Q frequency-selective element, the LC VCO offers far lower phase noise than the alternatives, which is why it dominates demanding radio-frequency applications such as wireless transceivers and frequency synthesizers. The cross-coupled differential pair, in which two transistors supply the negative resistance that cancels the tank loss, is the workhorse integrated LC VCO. The price of the LC topology is a comparatively narrow tuning range, set by how much the varactor can move the tank, and the silicon area or board space consumed by the inductor, whose on-chip quality factor ultimately limits the achievable noise.

Ring VCOs

A ring VCO is a ring oscillator, an odd number of inverting delay stages connected in a loop, whose stage delay is made controllable by the tuning voltage. The control voltage typically adjusts the bias current that charges each stage's load capacitance, so a higher current shortens the delay and raises the frequency. Ring VCOs are attractive because they require no inductor, occupy very little area, integrate readily in any digital process, and tune over a wide range, often several octaves. Their weakness is phase noise: lacking a high-Q resonator, a ring oscillator accumulates timing jitter from the noise of its own switching devices, and its spectral purity is far inferior to that of an LC VCO. Ring VCOs are therefore the natural choice for on-chip clock generation, clock-and-data recovery, and other applications where wide range and easy integration matter more than the lowest possible noise.

Relaxation VCOs

A relaxation VCO derives its frequency not from a resonance but from the time a capacitor takes to charge between two thresholds, with the control voltage setting the charging current. Because a constant current charging a capacitor produces a linear ramp, the frequency of a current-controlled relaxation oscillator is directly proportional to the charging current, giving these circuits an inherently linear and very wide tuning characteristic. Their phase noise is the poorest of the three families, since the threshold-crossing instants are exposed to comparator noise and the timing has no resonant element to stabilize it. Relaxation VCOs appear where waveform flexibility and an enormous tuning range outweigh spectral purity, as in the cores of analog function generators and in low-cost voltage-to-frequency conversion. The mechanism is treated in detail under the relaxation-oscillator topic linked below.

Crystal and Other Resonator VCOs

When frequency stability matters more than tuning range, the resonator can be a quartz crystal, giving a voltage-controlled crystal oscillator (VCXO). Here a varactor pulls the crystal slightly off its natural frequency, providing only a very small tuning range, typically tens to a few hundred parts per million, but inheriting the crystal's exceptional stability and low noise. VCXOs are used where a stable reference must be trimmed or disciplined, as in clock recovery and synchronization. The same idea applied to a surface-acoustic-wave or dielectric resonator yields narrowband, low-noise VCOs at higher frequencies.

Defining Phase Noise

Phase noise is quantified as the ratio of the noise power in a one-hertz bandwidth at a specified offset frequency from the carrier to the total carrier power. It is denoted L(f_m), where f_m is the offset, and is expressed in decibels relative to the carrier per hertz (dBc/Hz). A figure such as −110 dBc/Hz at a 100 kHz offset means that in a one-hertz slice 100 kHz away from the carrier the noise sits 110 dB below the carrier. Phase noise grows rapidly closer to the carrier, so the offset must always be stated for the number to have meaning.

The Leeson Model

The most widely used description of oscillator phase noise is the model published by D. B. Leeson in 1966. It is a semi-empirical expression that captures, with one compact formula, how the resonator quality factor, the carrier power, and the amplifier noise combine to shape the noise spectrum. In one common form the single-sideband phase noise is written:

L(fm) = 10 log { (FkT / 2Ps) × [1 + (f0 / 2Q fm)2] × [1 + fc / fm] }

where f_m is the offset from the carrier, f₀ is the carrier frequency, Q is the loaded quality factor of the resonator, F is the device noise factor, k is Boltzmann's constant, T is absolute temperature, P_s is the signal power at the active device, and f_c is the flicker (1/f) noise corner of the device. The model is not a first-principles derivation, but it organizes the dominant dependencies correctly and remains the standard engineering tool for reasoning about oscillator noise.

Reading the Leeson Regions

The bracketed terms in the Leeson expression divide the spectrum into regions of characteristic slope as the offset moves away from the carrier. Far from the carrier, beyond the resonator half-bandwidth, the noise flattens into a constant floor set by FkT/2P_s. Closer in, the resonator term dominates and the noise rises at 20 dB per decade as the offset shrinks. Closest to the carrier, the flicker term adds a further 10 dB per decade, producing the steep 1/f³ region. The lessons for design follow directly: raise the loaded Q, because phase noise improves as 1/Q²; raise the signal power P_s, because noise falls inversely with it; and choose active devices with low noise factor and a low flicker corner. The 1/Q² dependence is precisely why an LC VCO, with its high-Q tank, so decisively outperforms a ring or relaxation VCO that has no resonator at all.

Phase Noise, Jitter, and Tuning-Line Noise

Phase noise integrated over a band of offsets corresponds to a root-mean-square jitter in the time domain, the quantity that matters for clocking and data recovery. Beyond the intrinsic noise captured by Leeson, a VCO has an extrinsic noise path that a fixed oscillator does not: the tuning line. Any noise voltage on the control input is converted to frequency noise by the factor K_VCO, so a high tuning gain that eases the range requirement simultaneously amplifies tuning-line noise into the output spectrum. Low-noise VCO design therefore extends beyond the core to the quiet generation and filtering of the control voltage.

Sources of Nonlinearity

The dominant source of tuning nonlinearity in an LC VCO is the varactor. Its capacitance-versus-voltage law is nonlinear, and the square-root relationship between tank capacitance and frequency adds further curvature, so the frequency-versus-voltage characteristic bends across the range. The curvature is usually worst near the extremes of the tuning voltage, where the varactor capacitance changes most slowly with bias. Hyperabrupt varactors, chosen for a more nearly constant slope, and series or parallel fixed capacitors, chosen to linearize the combination, are the usual remedies. Relaxation and current-controlled VCOs, by contrast, are inherently linear because frequency tracks charging current directly.

Why K_VCO Variation Matters

When a VCO sits inside a phase-locked loop, K_VCO is one of the factors in the loop gain. A tuning curve whose slope changes by, say, a factor of three from one end of the range to the other changes the loop gain by the same factor, which shifts the loop bandwidth, the damping, and the settling behavior as the synthesizer is tuned across its band. A loop designed for adequate phase margin at one frequency may be sluggish or under-damped at another. Designers therefore specify not only the average K_VCO but the ratio of its maximum to its minimum across the range, and they may compensate the loop, or shape the tuning network, to hold the product of K_VCO and the other gain terms roughly constant.

Switched Tuning and Coarse-Fine Architectures

A direct way to reconcile a wide overall range with a low, well-controlled K_VCO is to split the tuning into coarse and fine paths. A bank of switched capacitors (or switched sub-bands) sets the operating region in discrete steps, while the varactor provides continuous fine tuning within each step. Because each sub-band need cover only a small fraction of the total range, the varactor contribution and hence K_VCO can be kept small, which suppresses tuning-line noise, while the switched bank delivers the wide range. This coarse-fine, or switched-capacitor, architecture is standard in modern integrated LC VCOs for wideband synthesizers.

Frequency Pushing

Frequency pushing is the change in oscillation frequency caused by a change in the supply voltage, with the tuning voltage held fixed. It is expressed in hertz per volt of supply (or megahertz per volt). Pushing arises because the supply voltage modulates the bias conditions of the active device and, through them, the device capacitances and the effective tank, shifting the frequency. Pushing matters because supply noise and ripple, such as residue from a switching regulator, are thereby converted directly into phase noise and spurious tones on the carrier. The defenses are a clean, well-regulated and well-decoupled supply, and circuit topologies whose frequency is relatively insensitive to bias. Pushing is one reason a VCO is often given its own low-noise regulator rather than sharing a noisy digital supply.

Frequency Pulling

Frequency pulling is the change in oscillation frequency caused by a change in the load impedance presented to the oscillator output, with supply and tuning held fixed. It is specified for a stated load mismatch, commonly a given voltage standing-wave ratio swept through all phases. Pulling occurs because a reactive load reflects energy back into the resonator and shifts the phase condition that sets the frequency. The classic remedy is a buffer amplifier between the oscillator core and the variable load: by isolating the resonator from whatever the output drives, the buffer presents the core a stable load and absorbs the variation. Good reverse isolation in that buffer is what keeps a downstream mixer, switch, or antenna mismatch from pulling the VCO.

Load Isolation in Practice

Because both pushing and pulling convert environmental disturbances into frequency error, a well-engineered VCO is rarely a bare oscillator core. It is the core plus a regulated, decoupled supply that attacks pushing and an output buffer that attacks pulling. In a frequency synthesizer these defenses also protect the phase-locked loop, since frequency disturbances that the loop cannot track within its bandwidth appear directly on the output as spurs and elevated noise.

The VCO Inside the Loop

In a phase-locked loop a phase detector compares the phase of the VCO output (usually after frequency division) with the phase of a reference, a loop filter smooths the detector's error output, and the resulting control voltage drives the VCO. If the VCO drifts high, the detector senses the growing phase error and the loop lowers the tuning voltage to bring it back; the loop holds the VCO exactly on frequency. Within this loop the VCO acts as a perfect integrator from control voltage to output phase, because phase is the integral of frequency, a fact central to the loop's dynamics and one developed further in the dedicated phase-locked-loop topic linked below.

Frequency Synthesis

Placing a programmable frequency divider between the VCO and the phase detector turns the loop into a frequency synthesizer. When the loop is locked, the divided VCO frequency equals the reference frequency, so the VCO output frequency is the reference multiplied by the divider value N. Stepping N steps the output frequency in increments of the reference (or, in fractional-N synthesis, in far finer steps), so a single stable crystal reference and a tunable VCO together generate a dense grid of accurate output frequencies. This is the architecture behind the local oscillators of virtually every modern radio and the clock generators of countless digital systems; the VCO supplies the agility and the reference supplies the accuracy.

How the Loop Shapes VCO Noise

The loop does not merely set the VCO's frequency; it also shapes its phase noise. Within the loop bandwidth the feedback forces the VCO to track the reference, so close-in noise is governed by the (multiplied) reference and the loop components, and the VCO's own close-in noise is suppressed. Outside the loop bandwidth the feedback can no longer correct the VCO fast enough, and the output noise reverts to the free-running VCO. The loop bandwidth is therefore chosen as a crossover: wide enough to clean up the VCO where the VCO is noisy, yet narrow enough to reject reference noise and spurs where the reference is noisy. Understanding this division of labor is the key to specifying a VCO for a synthesizer, since only the noise outside the loop bandwidth, plus the tuning-line contribution, ultimately falls to the VCO itself.