---
Every radar system, regardless of its architecture or mission profile, depends on one fundamental building block: a stable reference frequency source. Whether the radar is tracking aircraft at long range, resolving the micro-Doppler signature of a walking pedestrian, or imaging the surface of a distant planet, the purity of its local oscillator (LO) signal directly governs the quality of the information it can extract. The primary imperfection that limits this purity is phase noise — random, short-term fluctuations in the phase of a periodic signal that spread a nominally pure spectral line into one with skirts of unwanted energy.
Phase noise is not merely an academic concern for oscillator designers. It has cascading consequences throughout a radar's signal chain, affecting minimum detectable velocity, clutter rejection, range accuracy, Doppler resolution, and even the ability to detect targets that hide beneath the spectral skirts of nearby clutter. In modern radar environments, where electronic countermeasures are sophisticated and target cross-sections are increasingly small, understanding and mitigating phase noise is not optional — it is mission-critical.
This article provides a comprehensive treatment of phase noise in the context of radar systems. We begin with the mathematical definition of single-sideband phase noise spectral density, explore Leeson's foundational oscillator phase noise model, examine how the LO chain propagates and compounds phase noise, analyze its impact on range and Doppler accuracy, and conclude with practical mitigation strategies.
---
A mathematically ideal oscillator produces a perfectly sinusoidal output:
$$v(t) = V_0 \cos(2\pi f_0 t + \phi_0)$$
where $V_0$ is the amplitude, $f_0$ is the carrier frequency, and $\phi_0$ is a constant initial phase. The power spectrum of such a signal is a Dirac delta function — an infinitely narrow spectral line at $f_0$. In practice, however, every real oscillator is perturbed by thermal noise, flicker noise, supply ripple, mechanical vibrations, and other stochastic processes. The result is a signal with both amplitude and phase fluctuations:
$$v(t) = V_0[1 + \alpha(t)]\cos[2\pi f_0 t + \phi(t)]$$
where $\alpha(t)$ represents amplitude noise and $\phi(t)$ represents phase noise. In most well-designed oscillators, amplitude fluctuations are suppressed by gain-limiting mechanisms (such as automatic gain control or saturation in the amplifier), so the dominant degradation comes from $\phi(t)$. Single-sideband (SSB) phase noise, denoted $\mathcal{L}(f_m)$, is the standard metric used by oscillator manufacturers and radar engineers. It is defined as the ratio of noise power in a 1 Hz bandwidth at an offset frequency $f_m$ from the carrier to the total carrier power:
$$\mathcal{L}(f_m) = \frac{P_{\text{sideband}}(f_0 + f_m, \text{1 Hz BW})}{P_{\text{carrier}}} \quad [\text{dBc/Hz}]$$
The unit, dBc/Hz (decibels relative to the carrier per hertz of bandwidth), describes the noise power density normalized to the carrier. A high-quality crystal oscillator (XCO) might exhibit $\mathcal{L}(f_m) = -160$ dBc/Hz at $f_m = 10$ kHz offset, while a free-running dielectric resonator oscillator (DRO) at microwave frequencies might show only $-90$ dBc/Hz at the same offset.
The spectral shape of $\mathcal{L}(f_m)$ is not flat. It typically exhibits distinct regions with different slopes, each dominated by a different physical noise mechanism. At offsets very close to the carrier, $\mathcal{L}(f_m)$ may fall off as $f^{-3}$ (the flicker-frequency noise region), transitioning to $f^{-2}$ (the white-frequency noise region), then flattening at far offsets where the noise floor is set by the amplifier's additive white noise. This characteristic shape is precisely what Leeson's model captures.
---
In 1966, David B. Leeson published a remarkably intuitive and practically useful model for predicting the phase noise of a feedback oscillator. The model treats the oscillator as a high-Q resonant circuit inside a feedback loop with an amplifier that contributes broadband noise. The result is the celebrated Leeson equation:
$$\mathcal{L}(f_m) = 10 \log_{10}\left[\frac{2FkT}{P_s} \cdot \frac{1}{4Q_L^2} \cdot \left(\frac{f_0}{f_m}\right)^2 \cdot \left(1 + \frac{f_c}{f_m}\right)\right] \quad [\text{dBc/Hz}]$$
where:
The model reveals several powerful design insights:
The Leeson model, while semi-empirical (it requires knowledge of $F$ and $f_c$, which are not always easy to determine a priori), remains one of the most widely used tools in oscillator design. More rigorous analyses based on Hajimiri and Lee's time-varying phase noise model or nonlinear perturbation methods have extended the theory, but Leeson's equation provides an indispensable starting intuition.
---
A radar rarely uses the output of a single oscillator directly. Instead, a carefully designed LO chain generates the final transmit and receive frequencies through a cascade of frequency multipliers, dividers, mixers, amplifiers, and filters. Each element in this chain modifies the phase noise characteristics of the signal in specific and predictable ways.
A frequency multiplier by a factor $N$ multiplies the carrier frequency by $N$ but also multiplies the phase excursions by $N$. Since phase noise power is proportional to $\phi^2$, the phase noise spectral density increases by $N^2$, or equivalently, $20 \log_{10}(N)$ dB. For example, a frequency doubler ($N = 2$) degrades phase noise by 6 dB; a tenfold multiplier ($N = 10$) degrades it by 20 dB.
This is a critical consideration in radar design. A radar operating at 10 GHz that uses a 100 MHz reference oscillator multiplied by 100 will carry a $20 \log_{10}(100) = 40$ dB penalty. If the reference oscillator has $\mathcal{L} = -155$ dBc/Hz at 10 kHz offset, the LO at 10 GHz will have $\mathcal{L} = -115$ dBc/Hz at the same offset — before accounting for any additional noise from the multiplier itself.
Conversely, a frequency divider by $M$ reduces phase noise by $M^2$, or $20 \log_{10}(M)$ dB. Dividers are commonly used to generate lower-frequency reference signals from a high-stability source. However, they also add their own noise floor, typically from the digital logic circuitry.
Many modern radar LO chains employ phase-locked loops to synthesize stable, programmable frequencies. A PLL locks a voltage-controlled oscillator (VCO) to a stable reference. The PLL's phase noise profile is shaped by the loop bandwidth:
The designer must carefully choose the loop bandwidth to optimize the trade-off. A wider loop bandwidth suppresses VCO close-in noise (beneficial if the VCO is noisy), while a narrower bandwidth protects against reference noise spurs. The loop filter design also determines how well the PLL rejects reference spurs, which appear as discrete spectral lines in the LO spectrum and can create false targets in the radar return.
In a realistic radar transmitter/receiver, multiple stages — a reference oscillator, PLL synthesizer, frequency multipliers, power amplifiers, and distribution networks — form a chain. The total phase noise at the output is the power sum of the phase noise contributions from each stage, each transformed according to its multiplication or division factor and its own noise figure. Modern system-level simulation tools model each component's additive phase noise and combine them coherently, providing a complete picture of the system-level LO phase noise.
---
Phase noise affects virtually every aspect of radar performance, but its impact is most directly felt in two domains: range measurement and Doppler processing.
In a pulse-Doppler radar, targets are separated from clutter (ground, sea, weather) based on their Doppler shift. The radar's ability to detect a slowly moving target near strong clutter depends on the clutter cancellation ratio, which is limited by the spectral purity of the transmitted signal.
Phase noise from the LO directly modulates the transmitted waveform. When the radar return is mixed with the LO in the receiver, the phase noise of the transmitted signal and the received signal correlate (since they share the same LO during the coherent processing interval). For targets at zero Doppler, this correlation provides perfect cancellation. However, for targets with a nonzero Doppler shift or for clutter returns arriving from different ranges (and thus different round-trip times), the correlation is imperfect.
The decorrelation of phase noise between the transmit and receive paths creates residual noise power that limits the clutter-to-noise improvement factor. For clutter at range $R$, the time delay is $\tau = 2R/c$. The phase noise decorrelation for offset frequencies below $f_m = 1/\tau$ is small (nearly perfect cancellation), but for offset frequencies above $1/\tau$, the cancellation degrades. The effect is that close-in phase noise (small $f_m$) limits the detection of targets at long range, while far-out phase noise (large $f_m$) limits the detection of targets at short range against nearby clutter.
This insight explains why ground-based radars watching slow-moving targets (e.g., MTI radars for ground traffic or helicopters) demand extraordinarily low close-in phase noise — often $-120$ dBc/Hz or better at 1 kHz offset — while maritime surveillance radars combating sea clutter at short ranges benefit from good far-out phase noise performance.
Range in a pulsed radar is determined by measuring the time delay of the echo. Phase noise introduces timing jitter into the received signal, which translates directly into range measurement errors. The relationship can be expressed as:
$$\sigma_R = \frac{c}{2} \cdot \sigma_\tau = \frac{c}{2} \cdot \frac{\sigma_\phi}{2\pi f_0}$$
where $\sigma_R$ is the range error, $\sigma_\tau$ is the timing error, and $\sigma_\phi$ is the RMS phase error integrated over the signal bandwidth. For a radar operating at $f_0 = 10$ GHz with a phase noise of $-100$ dBc/Hz integrated over a 10 MHz bandwidth, the resulting range jitter is on the order of millimeters — negligible for most applications. However, for high-resolution synthetic aperture radar (SAR) or inverse SAR (ISAR) imaging, where centimeter-level accuracy is required, even small phase noise contributions must be carefully managed.
The minimum detectable velocity (MDV) of a pulse-Doppler radar — the slowest radial speed at which a target can be distinguished from stationary clutter — is directly governed by the spectral width of the clutter return. Phase noise broadens this spectral line, increasing the MDV. A radar with a first blind speed of 100 m/s but poor phase noise may have an effective MDV of 5 m/s, while a radar with excellent phase noise can push the MDV below 1 m/s.
This is why airborne early warning (AEW) radars and space-based moving target indication (MTI) systems invest heavily in ultra-low-noise oscillators and often employ pulse-to-pulse phase coherence techniques that actively correct for LO instabilities.
---
Given the pervasive impact of phase noise, radar engineers employ a multi-layered approach to mitigation:
The foundation of any high-performance radar LO chain is the reference oscillator. Oven-controlled crystal oscillators (OCXOs) provide phase noise as low as $-170$ dBc/Hz at 10 kHz offset. For the most demanding applications, sapphire-loaded cavity oscillators (SLCOs) and cryogenic sapphire oscillators (CSOs) achieve phase noise approaching $-190$ dBc/Hz, though at the cost of size, power consumption, and expense.
Careful selection of PLL loop bandwidth, using high-performance phase-frequency detectors with low noise floors, and employing fractional-N synthesizers with sigma-delta noise shaping can significantly reduce close-in phase noise. Advanced PLL chips from manufacturers like Analog Devices, Texas Instruments, and Hittite now offer integrated phase noise floors below $-160$ dBc/Hz.
Since each multiplication stage adds $20 \log_{10}(N)$ dB of phase noise, designers minimize the multiplication factor by starting with a higher-frequency reference or using direct digital synthesis (DDS) to generate signals closer to the final frequency.
When the transmit and receive paths share the same LO, the common phase noise partially cancels. This correlation cancellation can be enhanced by using short, matched cable lengths, minimizing temperature gradients between the two paths, and employing correlation loops — auxiliary receivers that measure and subtract the correlated phase noise component.
In modern digitally-generated radar waveforms (e.g., using DDS), the phase noise of the DAC clock can be measured and compensated in the digital domain before conversion. Similarly, at the receiver, digital signal processing can estimate and remove residual phase noise using pilot tones, autofocus algorithms, or entropy-minimization techniques — particularly in SAR/ISAR imaging.
Microphonic phase noise, caused by vibration-induced frequency modulation of the oscillator, is a significant concern in airborne and shipborne radars. Vibration isolation mounts, accelerometric feedback compensation, and vibration-tolerant oscillator designs (e.g., SC-cut crystals) mitigate this effect.
---
Phase noise is one of the most fundamental yet often underappreciated factors governing radar system performance. From the mathematical definition of $\mathcal{L}(f_m)$ through Leeson's insightful oscillator model to the complex interplay of multipliers, dividers, and PLLs in the LO chain, phase noise propagates through the radar signal chain and ultimately manifests as degraded clutter rejection, reduced Doppler sensitivity, increased minimum detectable velocity, and compromised range accuracy.
The Leeson equation — with its elegant dependence on resonator Q, signal power, oscillation frequency, and device noise figure — provides the conceptual framework for understanding and optimizing oscillator design. The LO chain's frequency transformations (multiplication by $N$ adding $20\log_{10}(N)$ dB of phase noise) demand careful architectural choices to keep the cumulative phase noise within budget.
Mitigation is an engineering discipline unto itself, spanning ultra-low-noise oscillator selection, optimized PLL design, correlation-based cancellation, digital correction, and environmental isolation. As radar systems push to higher frequencies, wider bandwidths, and more demanding detection requirements — detecting stealth targets, resolving micro-Doppler signatures, imaging at sub-meter resolution — the demand for spectral purity will only intensify.
In the end, the pursuit of lower phase noise is the pursuit of seeing more clearly through the noise of the physical world. Every decibel of improvement in $\mathcal{L}(f_m)$ translates into a radar that can see farther, measure more precisely, and distinguish the faintest whisper of a target from the roar of clutter. In radar engineering, purity of the signal source is not a luxury — it is the foundation upon which all performance is built.
--- Word count: approximately 2,500 words.
Need precision timing solutions? Get a quote from BRIDZA