Lab & Measurement Skills

A swept-tuned superheterodyne spectrum analyzer works by:

1. Input attenuator: Protects the input mixer and controls the signal level to avoid overdriving.
2. Mixer + swept LO: The input signal is mixed with a swept local oscillator. As the LO sweeps in frequency, different input frequencies are sequentially converted to the fixed IF.
3. IF filter (RBW filter): A bandpass filter at the IF frequency. Its bandwidth is the Resolution Bandwidth (RBW) — this determines the frequency selectivity.
4. Envelope detector: Extracts the amplitude of the IF signal.
5. Video filter (VBW): A low-pass filter that smooths the detected signal to reduce noise variance.
6. Display: Shows power (y-axis) vs. frequency (x-axis) synchronized to the LO sweep.

RBW is the bandwidth of the IF filter in the spectrum analyzer. It determines the ability to resolve (distinguish) two closely spaced signals.

• Narrower RBW: Better frequency resolution, lower displayed noise floor (noise power is proportional to BW), but slower sweep time. Use for measuring closely spaced signals or low-level spurs.
• Wider RBW: Faster sweep, higher displayed noise floor, but signals may overlap and appear as one.

The displayed noise floor changes with RBW:

Reducing RBW by 10x lowers the displayed noise floor by 10dB. Two signals can be resolved if their frequency separation is greater than the RBW.

VBW is the bandwidth of the low-pass filter after the envelope detector. It smooths the displayed trace by averaging the noise:

• VBW greater than or equal to RBW: No smoothing. The trace is noisy but represents the true signal + noise.
• VBW less than RBW: Smooths the noise, making the average noise floor more visible and stable. Does NOT change the measured noise floor — only reduces its variance (makes the trace cleaner).

When to use narrow VBW: Measuring CW signals near the noise floor — the smoothing makes weak signals easier to see. Typical setting: VBW = RBW/10 to RBW/100 for noise floor measurements.

When NOT to narrow VBW: Measuring pulsed or time-varying signals, as VBW filtering can distort the measurement.

For an analog (swept) spectrum analyzer, the minimum sweep time is constrained by the RBW filter settling time:

where k is a constant (~2-3 depending on filter type and acceptable amplitude error). This means:

• Narrower RBW: Sweep time increases as RBW squared — going from 10kHz to 1kHz RBW increases sweep time by 100x.
• Wider span: More frequency range to sweep = longer time.
• Sweeping too fast: Causes amplitude errors (signals appear lower than actual) and frequency errors (signals shift to the right).

Modern FFT-based spectrum analyzers bypass this limitation by capturing a block of time-domain data and computing the spectrum, offering much faster measurement speed.

In traditional spectrum analyzers, IF bandwidth and RBW are the same thing — the RBW filter IS the IF bandpass filter. The terms are used interchangeably.

However, in some modern instruments (especially VNAs and signal analyzers), IF bandwidth specifically refers to the bandwidth of the digital IF filter used in the final processing stage. In a VNA context, narrower IFBW reduces noise but increases measurement time:

Rule of thumb: Reducing IFBW by 10x lowers noise by 10dB but increases sweep time by 10x.

The input attenuator controls the signal level reaching the first mixer:

• More attenuation: Protects the mixer from overload, reduces internally generated distortion (better IMD performance), but raises the displayed noise floor by the same amount.
• Less attenuation: Lower displayed noise floor (better sensitivity), but risks overdriving the mixer, generating false intermodulation products inside the analyzer.

Key rule: If you change the input attenuation by X dB and the measured distortion products change by X dB, they are real (from the DUT). If they change by more than X dB (e.g., 3X dB for IM3), they are generated internally by the spectrum analyzer.

This is the standard method to verify whether observed spurs are from the DUT or the instrument.

Systematic verification method:

1. Attenuator test: Add 10dB external attenuation at the SA input. If IM3 products drop by 30dB (3:1 ratio), they were generated by the SA. If they drop by only 10dB, they are from the DUT.
2. Internal attenuation sweep: Change the SA internal attenuation in 5dB steps. DUT-generated products remain constant relative to the fundamental; SA-generated products change.
3. Check vs. specifications: Compare input signal levels against the SA’s published TOI (Third-Order Intercept) and 1dB compression point. Keep input levels at least 10dB below the mixer’s compression point.
4. Reduce mixer level: Increase attenuation until distortion products stop changing — this is the safe operating point.

DANL (Displayed Average Noise Level) is the spectrum analyzer’s own noise floor, measured with the input terminated in 50 ohms and 0dB input attenuation. It represents the minimum detectable signal level.

DANL is specified at a particular RBW (typically 1Hz). The actual displayed noise floor at your measurement RBW is:

Example: SA with DANL of -165 dBm/Hz at RBW = 1kHz: Noise floor = -165 + 30 = -135 dBm. To measure a -130 dBm signal, you need the noise floor at least 5-10dB below, so RBW must be narrow enough.

A VNA measures both magnitude and phase of S-parameters (reflection and transmission coefficients) as a function of frequency. It is a stimulus-response instrument — it generates a known signal and measures what comes back/through.

• vs. Scalar Network Analyzer: SNA measures only magnitude, not phase. VNA provides complex impedance, group delay, and time-domain reflectometry.
• vs. Spectrum Analyzer: SA is a receiver only — it measures unknown signals. VNA is both a source and receiver, measuring the DUT’s response to a known stimulus.

• S11 (Input Reflection): Ratio of reflected wave to incident wave at port 1 (with port 2 terminated). Represents input return loss/VSWR/impedance match.
• S21 (Forward Transmission): Ratio of transmitted wave at port 2 to incident wave at port 1. Represents gain (amplifier) or insertion loss (passive device).
• S12 (Reverse Transmission): Ratio of transmitted wave at port 1 to incident wave at port 2. Represents reverse isolation or reverse gain.
• S22 (Output Reflection): Ratio of reflected wave to incident wave at port 2 (with port 1 terminated). Represents output return loss.

S-parameters are measured at the reference impedance Z0 (usually 50 ohms) and are frequency-dependent complex numbers (magnitude and phase).

VNA calibration removes systematic errors from cables, connectors, and the VNA’s internal hardware, moving the measurement reference plane to the DUT. Without calibration, measurements can have errors of several dB and tens of degrees.

Common calibration types:

• SOLT (Short-Open-Load-Thru): Most common. Uses four known standards. Requires accurate standard definitions.
• TRL (Thru-Reflect-Line): Uses a thru connection, a reflect (short or open), and a transmission line. Self-calibrating — doesn’t need precisely known standards. Preferred for on-wafer and fixture measurements.
• ECal (Electronic Calibration): Automated module with multiple internal impedance states. Fast, repeatable, but expensive.

Key points: Always calibrate at the measurement reference plane (at the DUT connectors, not at the VNA ports). Recalibrate when temperature changes significantly or cables are moved.

1-port calibration corrects for three error terms (directivity, source match, reflection tracking) and is sufficient for measuring reflection only (S11 or S22). Uses short, open, and load standards.

Full 2-port calibration corrects for all 12 error terms (forward and reverse) and is required when measuring transmission (S21, S12) accurately, especially for:

• High-isolation measurements (filters, amplifiers with high reverse isolation)
• Measurements where port match affects the DUT (low-return-loss devices)
• Group delay measurements requiring phase accuracy

Port extension electrically shifts the calibration reference plane by adding or removing a length of ideal transmission line in the VNA’s error correction. Use it when:

• You can’t physically calibrate at the DUT (e.g., DUT is inside a fixture and you calibrated at the fixture input).
• There’s an additional cable or adapter between the calibration plane and the DUT.
• You need to de-embed a known length of trace on a PCB to see the component’s true S-parameters.

Port extension only compensates for electrical delay (phase rotation) and does not correct for loss or impedance mismatch in the extended section. For full correction, use proper de-embedding techniques.

TDR transforms the frequency-domain S11 data into the time domain using an inverse FFT, showing impedance vs. distance along a transmission line. It reveals:

• Impedance discontinuities: Connector transitions, trace width changes, via transitions.
• Fault location: Opens, shorts, or damage along a cable or PCB trace.
• Characteristic impedance: Z0 of each section of a transmission path.
• Connector quality: Small impedance bumps from poor connectors or solder joints.

Spatial resolution depends on the VNA’s frequency range: Resolution = v_p / (2 x f_max). A 20GHz VNA gives ~7.5mm resolution in free space.

Use a 10x probe. A 1x probe presents ~50-100pF of capacitive load, which at 100MHz creates a low impedance (less than 30 ohms) that loads the circuit and distorts the signal. A 10x probe reduces the capacitive loading to ~10-15pF and presents higher input impedance (10M ohms), minimally disturbing the circuit.

The trade-off is reduced signal amplitude (divided by 10), but the scope compensates for this in its gain setting.

Oscilloscope bandwidth is the frequency at which the displayed amplitude drops to -3dB (70.7%) of the actual signal. The rule of thumb for bandwidth selection:

The relationship between bandwidth and rise time (for a Gaussian response):

The measured rise time combines the scope and signal rise times:

A 1GHz scope has a rise time of 350ps — it cannot accurately measure signals with rise times faster than about 1ns.

• Signal Generator (RF/Analog): Generates CW, modulated (AM/FM/PM), or pulse signals with precise frequency and amplitude. Uses a synthesizer (PLL/DDS) for clean, low-phase-noise output. Best for: RF measurements, receiver testing, phase noise analysis.
• AWG (Arbitrary Waveform Generator): Generates any user-defined waveform from stored digital samples via a DAC. Can create complex modulated signals (OFDM, QAM), multi-tone, chirps, or custom waveforms. Best for: baseband/IF signal generation, radar waveforms, protocol testing.
• Vector Signal Generator (VSG): Combines both — a signal generator with an IQ modulator fed by an internal AWG. Best for: generating standard-compliant modulated RF signals (LTE, 5G, Wi-Fi).

Techniques to measure signals near the noise floor:

1. Reduce RBW: Every 10x reduction in RBW lowers the noise floor by 10dB. Trade-off: slower sweep.
2. Reduce VBW: Set VBW to RBW/10 or RBW/100 to smooth the noise and make weak signals more visible.
3. Use averaging: Trace averaging reduces noise variance by sqrt(N_averages). 100 averages give 10dB improvement.
4. Minimize input attenuation: Use 0dB attenuation (if input levels are safe) to get closest to the SA’s DANL.
5. Use a preamplifier: An external low-noise preamp before the SA improves system NF. Effective noise floor = SA_DANL – Preamp_Gain + Preamp_NF.
6. Noise correction: Some SAs can subtract their own noise floor mathematically for a few dB of additional sensitivity.

Three primary methods:

• Y-factor method: Uses a calibrated noise source (ENR). Measure output noise with source ON (hot) and OFF (cold). Y = P_hot/P_cold. Then: F = (ENR – Y + 1) / (Y – 1) or NF = ENR(dB) – 10 log(Y-1). Most common method, used by dedicated NF analyzers.
• Cold-source (direct) method: Measures output noise with a terminated input (no noise source). Requires accurate gain measurement from a VNA. More accurate for high-NF or frequency-converting devices (mixers).
• Signal generator method: Adjust input signal until output SNR = 1 (signal equals noise). Input power at that point relates to NF. Simple but less accurate.

Key consideration: The NF meter/analyzer itself adds noise. Use cascaded NF formula to de-embed: NF_DUT = NF_measured – NF_instrument/G_DUT.

Phase noise measurement methods:

• Direct spectrum method: Use a spectrum analyzer with very low phase noise LO. Measure the noise sidebands around the carrier. Limited by the SA’s own phase noise. Simple but not the most sensitive.
• Phase detector method: Mix the DUT with a reference oscillator of equal frequency (phase-locked or free-running). The mixer output is the phase difference, which is analyzed by a baseband FFT analyzer. Most sensitive method.
• Delay line discriminator: Split the DUT output, delay one path, and mix. Converts phase noise to amplitude noise. No reference oscillator needed, but sensitivity limited by delay length.
• Cross-correlation method: Uses two independent reference oscillators and cross-correlates to reject their noise. Can measure phase noise below the reference oscillator’s own noise. Used in modern dedicated phase noise analyzers.

1. Short (or terminate with known source impedance) the LNA input.
2. Connect the LNA output to a low-noise spectrum analyzer or dynamic signal analyzer.
3. Measure the output noise spectral density V_n,out(f) in V per root-Hz.
4. Divide by the measured gain A_v(f) to get input-referred noise: V_n,in = V_n,out / A_v.

Precautions: Ensure the instrument noise floor is at least 10dB below the DUT output noise. Use averaging to reduce measurement uncertainty. Shield the setup from external interference. Measure with both short and known resistance to separate voltage and current noise contributions.

Cause: Mains power line interference (50Hz Europe/Asia, 60Hz Americas) coupling through:

• Ground loops between equipment
• Magnetic field coupling from power transformers
• Capacitive coupling from AC power wiring
• Insufficient PSRR in the power supply

Solutions: Use shielded cables, eliminate ground loops (single-point grounding, isolation transformers), use battery-powered equipment, add common-mode chokes, improve PCB power filtering, and use differential signaling.

Measurement uncertainty is the range of values within which the true value is expected to lie. Sources in RF measurements include:

• Mismatch uncertainty: Reflections between source, DUT, and load. Reduced by using well-matched components and adapters.
• Cable/connector repeatability: Variations from connecting/disconnecting. Use torque wrenches, high-quality connectors, minimize reconnections.
• Instrument accuracy: Amplitude, frequency, and phase accuracy specifications.
• Temperature drift: Instruments and DUTs drift with temperature. Allow warm-up time and control ambient temperature.
• Calibration age: Calibration degrades over time. Recalibrate regularly.

Total uncertainty is calculated by RSS (Root Sum of Squares) of individual contributions for uncorrelated sources.

• Peak power: Maximum instantaneous power. Critical for component stress (compression, breakdown). Measured with peak power meters or fast oscilloscopes.
• Average power: Power averaged over time. What thermal power meters measure. Determines total energy and heating.
• RMS power: For CW signals, RMS = average. For modulated signals, RMS captures the “true heating power” including the modulation envelope.

Peak-to-Average Power Ratio (PAPR): The difference between peak and average power in dB. OFDM signals (LTE, 5G, Wi-Fi) have high PAPR (8-13dB), meaning the PA must handle peaks much higher than the average power without compressing.

• Power meter + sensor: Measures true RMS power (thermal or diode sensor). Very accurate (within 0.1dB), broadband, and captures total power regardless of modulation. Does NOT provide frequency information — it measures total power across its entire bandwidth.
• Spectrum analyzer: Shows power vs. frequency. Can measure power in specific frequency bands (channel power). Less accurate for absolute power (within 0.5-1dB typically) but provides spectral information.

Use power meter: For accurate absolute power measurements, calibration verification, transmitter output power. Use spectrum analyzer: For spectral analysis, spurious measurements, channel power, occupied bandwidth.

1. Characterize the failure mode: Is it INL, DNL, or missing codes? Which codes are affected? Consistent or localized?
2. Lot analysis: Is it specific to certain wafer lots, wafer positions (edge vs. center), or assembly lots?
3. Correlation analysis: Compare failing vs. passing units on parametric data (offset, gain, reference voltage).
4. Process monitoring: Check for fab process shifts (threshold voltage, capacitor matching, oxide thickness).
5. Test setup validation: Verify equipment calibration, input signal quality, and reference voltage accuracy.
6. Failure analysis: Cross-section and inspect suspect units for physical defects.

Systematic verification process:

1. Attenuator test: Add 10dB external attenuation. Real harmonics drop by 10dB. SA-generated harmonics drop by 20dB (2nd) or 30dB (3rd).
2. Change SA settings: Increase internal attenuation. If harmonics change relative to fundamental, they are SA artifacts.
3. Use a bandpass filter: Place a filter at the fundamental frequency before the SA. If harmonics disappear, they were from the SA or from the DUT being overdriven.
4. Reduce DUT output power: Real harmonics track the DUT power level. SA artifacts track the power at the SA input mixer.
5. Check with a different instrument: If harmonics appear on both instruments at the same levels, they are real.

Automated RF testing typically involves: instrument control via GPIB/LAN/USB using SCPI commands, scripting in Python (with PyVISA), MATLAB, or LabVIEW. Key elements include:

• Calibration routines: Automated cal verification at test start.
• DUT control interfaces: Serial, SPI, I2C for DUT configuration.
• Data logging and analysis: Automated pass/fail with statistical process control.
• Temperature integration: Chamber control for over-temperature testing.
• Report generation: Automated data export and report creation.
• Test time optimization: Minimize settle times, parallelize independent measurements, use list-mode sweeps on instruments.