Should you use averaging or reduce IF bandwidth to lower VNA noise?

Both averaging and IF bandwidth reduction improve the noise floor by 10*log10(N) dB for the same total measurement time. Reducing the IF bandwidth from 10 kHz to 100 Hz (a 100x reduction) improves the noise floor by 20 dB with a single sweep, taking the same time as 100 averages at 10 kHz IF bandwidth. The key difference is that IF bandwidth reduction applies to a single sweep, so the display updates continuously with improving data. Averaging requires completing N full sweeps before the final result is shown. IF bandwidth reduction is generally preferred because it also reduces the effect of interfering signals within the IF band. Averaging is preferred when measuring time-varying devices where you want to see each sweep's result and track trends, or when the DUT has long settling times that make slow sweeps impractical.

Test & Measurement

VNA Averaging

/vee-en-ay av-rij-ing/

A noise reduction technique in vector network analyzers that averages multiple sweep measurements of the complex S-parameter data (magnitude and phase) at each frequency point, reducing random trace noise by the square root of the number of averages while preserving the signal of interest. VNA averaging improves the effective dynamic range by 10×log₁₀(N) dB for N averages, enabling accurate measurement of high-isolation devices, low-level crosstalk, and filter stopband rejection that would otherwise be buried in the noise floor.

Category: Test & Measurement

Improvement: 10×log₁₀(N) dB

Types: Sweep averaging, Point averaging

Understanding VNA Averaging

A VNA measures S-parameters by injecting a stimulus signal and measuring the complex (magnitude and phase) response at each frequency point. The measurement includes both the desired signal and random noise from the VNA's receivers, source phase noise, and environmental interference. Averaging exploits the fact that the desired signal is deterministic (repeats identically each sweep) while noise is random. When N sweeps are averaged, the signal adds coherently (magnitude proportional to N) while noise adds incoherently (magnitude proportional to √N), improving the signal-to-noise ratio by √N, or 10×log₁₀(N) dB.

A critical distinction exists between VNA averaging and spectrum analyzer video averaging. VNA averaging operates on the complex (real + imaginary) data, which is a vector operation that preserves phase information. This means that even very weak signals buried below the noise floor in a single sweep can be recovered by sufficient averaging, as long as the phase is stable. Spectrum analyzer video averaging operates on the magnitude (scalar) data and cannot recover signals below the noise floor because the noise magnitude has a nonzero mean (Rayleigh distributed) that does not average to zero.

Noise Floor Improvement

Noise Floor Improvement:
ΔNF = 10 × log₁₀(N) dB

Effective Dynamic Range:
DR_eff = DR_raw + 10 × log₁₀(N)

Example: VNA with 110 dB raw dynamic range at IFBW = 10 Hz:
100 averages → DR_eff = 110 + 20 = 130 dB

Equivalent IFBW Reduction:
N averages at IFBW₁ ≡ 1 sweep at IFBW₁/N
10 averages at 1 kHz IFBW = 1 sweep at 100 Hz IFBW (same time, same improvement)

Trace Noise RMS:
σ_N = σ₁ / √N

Averaging vs. IFBW Reduction

Method	Noise Improvement	Display Update	Interferer Rejection	Best For
10x Averaging at 1 kHz	10 dB	After each sweep	No (full IFBW exposed)	Tracking drift, triggered measurements
1 sweep at 100 Hz	10 dB	Continuous per point	Yes (narrower IFBW)	General use, interfering environments
100x Averaging at 10 kHz	20 dB	After each sweep	No	High dynamic range, stable DUT
1 sweep at 100 Hz	20 dB	Continuous per point	Yes	Filter stopband, isolation measurement

Common Questions

Frequently Asked Questions

How much does averaging improve the VNA noise floor?

Averaging N sweeps improves the noise floor by 10*log10(N) dB. Ten sweeps gives 10 dB improvement; 100 sweeps gives 20 dB. For a VNA with 100 dB raw dynamic range at 10 kHz IFBW, 100x averaging extends effective dynamic range to 120 dB. The improvement follows the square root of N law: the signal adds coherently (proportional to N) while noise adds incoherently (proportional to sqrt(N)). However, 100 averages means 100x measurement time, which may be impractical. For most applications, reducing IF bandwidth is more time-efficient for the same noise reduction.

Should you use averaging or reduce IF bandwidth?

Both methods improve the noise floor by 10*log10(N) dB for the same total measurement time. Reducing IFBW from 10 kHz to 100 Hz (100x reduction) gives 20 dB improvement in a single sweep, taking the same time as 100 averages at 10 kHz. IFBW reduction is generally preferred because it also rejects interfering signals within the IF band and provides continuous display updates. Averaging is preferred when measuring time-varying devices (to see each sweep's result and track trends) or when the DUT has long settling times that make slow sweeps impractical.

What is the difference between sweep averaging and point averaging?

Sweep averaging performs N complete frequency sweeps and averages the complex data at each point across all sweeps. The display updates after each complete sweep with a progressively smoother trace. Point averaging measures each individual frequency point N times before moving to the next, completing the averaged measurement in a single pass. Both achieve the same noise reduction for the same total averages. Sweep averaging is better for tracking drift or detecting intermittent problems. Point averaging is better for triggered measurements where you need the entire averaged dataset in one acquisition.

Related Terms

← Averaging AWG (Optical) →