Introduction: Why Phase Noise Matters in Signal Generation

In modern electronics and telecommunications, signal generators serve as the bedrock of accurate testing and reliable system performance. Whether you are designing a radar system, characterizing an RF amplifier, or verifying a communication link, the quality of the test signal directly determines the validity of your results. One of the most subtle yet impactful characteristics of a signal generator is its phase noise. While often overlooked by novice engineers, phase noise can quietly degrade measurement accuracy, limit system sensitivity, and introduce errors that are difficult to diagnose. This article provides a comprehensive, technically grounded examination of phase noise: its origins, how it affects measurements and systems, methods for measuring and quantifying it, and practical steps to minimize its impact. By understanding phase noise, you can make informed decisions about equipment selection and test procedures, ensuring that your measurements reflect the true performance of the device under test rather than the imperfections of the stimulus.

Fundamentals of Phase Noise: The Frequency Domain View

Defining Phase Noise in the Time and Frequency Domains

Phase noise is the term used to describe the short-term, random fluctuations in the phase of a sinusoidal signal. In the time domain, these fluctuations appear as jitter—variations in the zero-crossing times of the waveform. However, engineers more often analyze phase noise in the frequency domain, where it manifests as sidebands around the carrier signal. A perfect oscillator would produce a single spectral line at the carrier frequency. In reality, phase noise broadens that line, creating a skirt of noise that extends symmetrically on either side.

The standard metric for phase noise is dBc/Hz (decibels relative to the carrier per hertz of bandwidth). For example, a phase noise specification of –120 dBc/Hz at a 10 kHz offset from a 1 GHz carrier means that, in a 1 Hz bandwidth 10 kHz away from the carrier, the noise power is 120 dB below the carrier power. Lower numbers (more negative) indicate cleaner signals. Phase noise is typically specified at one or more offset frequencies, often 1 kHz, 10 kHz, 100 kHz, and 1 MHz, to characterize the noise profile across different regions.

Close-In vs. Far-Out Phase Noise

The phase noise spectrum is not flat. It is conventionally divided into regions based on the offset from the carrier. Close-in phase noise (within a few kilohertz) is dominated by flicker noise, often referred to as 1/f noise. In the time domain, this corresponds to long-term jitter. Far-out phase noise (beyond about 100 kHz) is typically dominated by white noise (thermal and shot noise) and is flatter. The crossover point depends on the oscillator design—crystal oscillators have very low close-in noise, while VCOs (voltage-controlled oscillators) tend to have higher flicker noise. Understanding these regions is critical because different applications are sensitive to different offsets: a radar system may care more about close-in noise because of Doppler processing, while a high-speed digital link may be affected more by far-out noise.

The Relationship between Phase Noise and Jitter

While phase noise is a frequency-domain parameter, many digital applications specify jitter (in picoseconds or seconds). The two are mathematically related: jitter can be computed by integrating the phase noise spectral density over a given offset frequency range. This integration converts the noise power into a root-mean-square (RMS) time deviation. For a given phase noise profile, the jitter depends on the integration bandwidth—typically from a few kilohertz up to the clock frequency. Some signal generator datasheets provide both phase noise and jitter figures, but it is important to verify the integration limits used, as they can vary between manufacturers.

Sources and Types of Phase Noise in Signal Generators

Oscillator Circuit Topology

The primary source of phase noise in any signal generator is its reference oscillator. Crystal oscillators (XO, TCXO, OCXO) offer excellent phase noise performance, especially close to the carrier. Oven-controlled crystal oscillators (OCXOs) achieve the lowest phase noise by stabilizing the crystal temperature, reducing drift and noise. Voltage-controlled oscillators (VCOs) are more tunable but generally have higher phase noise, particularly at larger offsets, because they use varactor diodes and have higher impedance conversion noise. Many modern signal generators combine a low-noise crystal reference with a phase-locked loop to synthesize higher frequencies. The PLL introduces its own noise contributions from the phase detector, loop filter, and VCO, so careful loop bandwidth design is essential to minimize overall phase noise.

Active and Passive Component Noise

Amplifiers, buffers, and dividers in the signal path all contribute to phase noise. Bipolar junction transistors (BJTs) tend to have lower 1/f noise than FETs, which is why high-end RF oscillators often use discrete BJTs. Resistors generate thermal noise that is upconverted to phase noise via nonlinearities in active devices. Power supply noise—ripple and switching transients—modulates the oscillator’s phase, creating spurious sidebands that are sometimes mistaken for phase noise but are actually deterministic (spurs). Proper and robust power supply decoupling and regulation are critical. Even passive components like capacitors can contribute if they are microphonic (sensitive to vibration), inducing frequency modulation.

Environmental Factors

Temperature changes, mechanical vibration, and electromagnetic interference can all induce phase noise. For example, a signal generator placed near a cooling fan may experience vibration-induced noise sidebands at the fan’s rotational frequency. Many high-performance generators use shock mounts and isolators to reduce mechanical coupling. Electromagnetic fields from nearby switching power supplies can also couple into sensitive oscillator nodes, creating unwanted frequency modulation. Shielding and careful layout in the generator’s design mitigate these effects, but the user’s test environment matters too—avoiding placing the generator near motors, fans, or radio transmitters is good practice.

How Phase Noise Affects System Performance and Measurements

Impact on Receiver Sensitivity in Communication Systems

In modern communication systems—whether cellular (4G/5G), Wi-Fi, or satellite—phase noise directly degrades the receiver’s ability to detect weak signals in the presence of strong adjacent ones. This is called the reciprocal mixing phenomenon. When a strong blocker signal is present, the receiver’s local oscillator (LO) has phase noise sidebands that mix the blocker down to the intermediate frequency (IF) along with the desired signal. The resulting noise floor rise reduces the signal-to-noise ratio (SNR) and can desensitize the receiver. For a given LO phase noise, the degradation is more severe when the blocker is stronger and closer in frequency. This is why 5G NR specifications place stringent phase noise requirements on oscillators used in base stations and user equipment.

Furthermore, in digital modulation schemes like QAM (quadrature amplitude modulation), phase noise causes constellation points to rotate, increasing the error vector magnitude (EVM). High phase noise makes it impossible to achieve high-order modulation (e.g., 256-QAM or 1024-QAM) and the associated data rates. Engineers testing such systems must ensure their signal generator’s phase noise is well below the device under test’s specification, or they will incorrectly attribute poor EVM to the DUT.

Radar and Doppler Processing

Radar systems are particularly sensitive to close-in phase noise. In Doppler radar, the signal reflected from a moving target is shifted in frequency by the Doppler effect. This tiny frequency shift (often less than 1 kHz) must be detected in the presence of large clutter returns from stationary objects. If the transmitter oscillator has high close-in phase noise, the clutter sidebands can mask the Doppler shift. This is known as the radar clutter rejection problem. Similarly, in imaging radar (SAR), phase noise limits the coherent integration time and degrades image resolution. For these reasons, military radar often uses cavity oscillators or DROs (dielectric resonator oscillators) with phase noise specifications approaching –160 dBc/Hz at 10 kHz offsets.

Testing RF Components and Systems

When using a signal generator as a test stimulus, the generator’s phase noise becomes an integral part of the measurement accuracy. For example, when measuring the phase noise of a device under test (DUT) with a spectrum analyzer, the analyzer’s own LO phase noise may dominate if not properly managed. This is why dedicated phase noise test sets use cross-correlation techniques. More generally, if you are testing an amplifier’s noise figure, the signal generator’s phase noise can increase the effective input noise, giving an artificially high noise figure reading. Similarly, in filter response measurements, phase noise can broaden the apparent bandwidth of a sharp filter or obscure its stopband rejection. The key lesson: always verify that the test signal’s phase noise is at least 10 dB lower than the expected measurement floor, or use a calibrated phase noise subtraction method.

Digital Clock Jitter in High-Speed Data Systems

While the original article focused on analog RF, phase noise is equally critical in digital systems. High-speed serial links (PCIe, USB, Ethernet, etc.) rely on clock signals to sample data. Jitter from the clock source reduces the timing margin, causing bit errors. The phase noise of the clock oscillator directly translates to deterministic and random jitter. In many test setups, the signal generator is used as a clock source for digitizers or pattern generators. Using a generator with poor phase noise will add jitter that is indistinguishable from the DUT’s jitter, leading to false failures or pass/fail ambiguity. Selecting a signal generator with low phase noise (or a dedicated low-jitter clock source) is essential for high-speed digital validation.

Measuring Phase Noise: Techniques and Best Practices

The Direct Spectrum Method

The simplest method for measuring phase noise is direct spectrum analysis using a spectrum analyzer. The signal is fed into the analyzer, and the noise sidebands are measured relative to the carrier at various offset frequencies. This technique works well only if the analyzer’s own phase noise is significantly lower than the DUT’s. For many modern signal generators with phase noise below –130 dBc/Hz at 10 kHz, the analyzer may limit the measurement. The direct method also cannot distinguish between amplitude noise and phase noise; the spectrum analyzer measures total AM+PM noise. For pure oscillators, AM noise is usually negligible, but it can become significant in certain cases (e.g., when testing DUTs that have AM noise components).

The Phase-Locked Loop (PLL) Method

The PLL method uses a reference oscillator (usually with better phase noise than the DUT) to downconvert the DUT’s signal to a baseband voltage that is proportional to phase difference. The resulting voltage is then analyzed with a low-frequency spectrum analyzer. This approach separates phase noise from amplitude noise and allows much lower noise floors because the reference is only compared in phase. However, it requires the DUT and reference to be at the same frequency, and the loop bandwidth must be carefully set to avoid suppressing the phase noise being measured. Manual PLL methods are becoming less common with the availability of dedicated phase noise analyzers.

Cross-Correlation Phase Noise Measurement

The gold standard for high-accuracy phase noise measurement is the cross-correlation technique. Two independent but nominally identical measurement channels are used, each with its own mixer and reference oscillator. The outputs are cross-correlated in a digital processor. Since the noise from the two channels is uncorrelated, it averages to zero, while the DUT’s phase noise is correlated and accumulates. This method can achieve measurement floors below –180 dBc/Hz, far below what any single-channel system can achieve. Cross-correlation analyzers are expensive but indispensable for characterizing the best signal generators and oscillators in metrology labs. Many references and application notes from Rohde & Schwarz detail this technique.

Practical Measurement Setup Tips

Regardless of method, common pitfalls include:

  • Impedance mismatches: Use high-quality cables and attenuators to ensure proper termination and minimize reflections that can add spurious noise.
  • Power level and saturation: Overdriving a mixer or spectrum analyzer can create harmonics and increase noise floor. Always operate within manufacturer-specified input ranges.
  • Ground loops: Use coaxial isolators or balanced connections to break ground loops that introduce 50/60 Hz hum and harmonics.
  • Temperature stability: Allow the DUT and measurement equipment to warm up to stable operating temperature; thermal drift can look like low-frequency phase noise.

For a thorough treatment, Keysight’s phase noise application note remains a classic reference.

Strategies to Minimize Phase Noise in Signal Generators

Selecting the Right Oscillator

The most direct way to reduce phase noise is to choose a signal generator that uses a high-quality internal oscillator. For benchtop RF generators, look for models that employ an oven-controlled crystal oscillator (OCXO) with specified phase noise below –140 dBc/Hz at 10 kHz (for a 1 GHz carrier). Some generators allow an external reference input, so you can supply an even lower-noise atomic or crystal reference. In generator-level design, using a low-noise oscillator topology like a Clapp or Colpitts oscillator with high-Q resonators and careful impedance matching reduces close-in noise.

Power Supply Conditioning and Shielding

Designers must pay attention to power supply filtering. Use multiple stages of low-dropout regulators (LDOs) with high ripple rejection. A noisy DC-DC converter can be kept away from the oscillator module with proper shielding and ferrite beads. The RF output port should be well isolated from the oscillator circuit to prevent load pulling and injection of external noise. For the user, adding an external low-noise amplifier after the generator is rarely advisable because the amplifier’s own phase noise will degrade performance. Instead, use a generator with sufficient output power directly.

Filtering and Spur Reduction

Some signal generators include built-in tracking filters that reduce harmonic content and spurious signals. Lower spurs mean less chance of near-carrier noise contamination. While filters do not reduce phase noise of the fundamental, they can remove unwanted mixing products that might be misinterpreted as phase noise in a measurement. In digital generators (DDS or direct digital synthesis), the output often includes quantization noise and spurs. These can be mitigated by choosing a DDS-based generator with high resolution and low spurious specifications, but careful comparison to PLL-based generators is necessary.

Calibration and Maintenance

Phase noise performance of a signal generator can degrade over time due to component aging, crystal drift, or contamination of connectors. Regular calibration (typically every 12–24 months) ensures the internal oscillator remains within specification. Recalibration of the PLL loop parameters may be needed after repairs. For critical measurements, a quick sanity check of phase noise using a known stable source can catch problems before they affect results.

Practical Considerations for Engineers: Trade-offs and Decision Making

Phase Noise vs. Other Performance Parameters

Signal generators must often balance phase noise against other specifications such as frequency range, switching speed, output power, and spurious performance. For example, a wideband VCO that can sweep from 2 to 18 GHz will have higher phase noise than a narrowband VCO covering only 10–11 GHz. Fast switching (< 1 µs) typically requires a wider PLL loop bandwidth, which can increase close-in noise. In modular or benchtop generators, you can choose optimized configurations: use the low-noise mode for narrow-band applications and the fast-switching mode for swept tests. Understanding your application’s priorities allows you to select the appropriate instrument or mode.

Interpreting Datasheet Phase Noise Specs

Datasheet phase noise specifications are often given at a single offset (e.g., –110 dBc/Hz @ 100 kHz) but may omit the carrier frequency and test conditions. The same oscillator at a lower carrier frequency will have better phase noise relative to the carrier because the sideband offset is a larger fraction of the carrier. Always check the phase noise plot (usually provided in the manual) to see the full curve. Beware of “typical” versus “guaranteed” specs. For critical applications, request tested data or buy a generator with guaranteed phase noise. Some manufacturers, like Analog Devices, publish detailed application notes that help interpret oscillator phase noise.

When to Use External References

If your measurement demands extremely low phase noise that your signal generator cannot meet, consider using an external low-noise reference oscillator (e.g., a Rubidium atomic clock or a premium OCXO) as the generator’s reference input. This can improve the close-in phase noise by 10–20 dB because the generator’s internal PLL divides the reference noise upward but often filters out the distant noise. However, the external reference must have low phase noise itself and be stable—a poor external reference can degrade performance. Many high-end signal generators have a 10 MHz reference input; ensure the input impedance is terminated properly (usually 50 Ω) and use a high-quality coaxial cable.

Cost vs. Performance

Low phase noise signal generators are expensive. Prices can range from a few thousand dollars for basic models to over $100,000 for ultra-low-noise instruments used in metrology. Before investing, evaluate the phase noise requirements of your most critical test. If your test only requires –100 dBc/Hz at 10 kHz (suitable for many basic RF tests), a mid-range generator is sufficient. For high-end receiver testing or radar work that demands –150 dBc/Hz, you likely need a dedicated analog signal generator with a premium OCXO. Always consider whether the generator’s phase noise will be the limiting factor in your measurement uncertainty budget.

Conclusion: Taking Control of Phase Noise for Reliable Measurements

Phase noise is a fundamental property of signal generators that can quietly undermine the accuracy of even the most carefully designed test setup. By understanding its origins—from oscillator design and component noise to environmental factors—you gain the ability to anticipate and mitigate its effects. Recognizing how phase noise impacts different applications, such as receiver sensitivity, radar Doppler processing, and high-speed digital jitter, allows you to prioritize your requirements when selecting test equipment. Proper measurement techniques, whether using direct spectrum analysis or cross-correlation, help you characterize phase noise objectively and verify that your generator is meeting its specifications.

The steps to minimize phase noise are well established: choose a generator with a low-noise internal oscillator, ensure clean power and good shielding, use proper measurement practices, and maintain regular calibration. But perhaps the most important takeaway is to treat phase noise as a test parameter in its own right—just as you would check power output, frequency accuracy, and harmonic distortion. By making phase noise a standard part of your evaluation process, you will produce more trustworthy results and reduce costly design iterations. For further reading, the Wikipedia article on phase noise provides a solid overview, and Microwave Journal’s article on phase noise and jitter measurement covers practical methods in depth.