GNSS receivers can detect satellite signals whose power is below the noise integrated across the receiver's input bandwidth because they look for a known signal pattern over time. They correlate the samples with a local ranging-code replica, compensate for frequency offset and accumulate the result. A matching signal builds a distinguishable correlation peak; random noise does not build the same consistent pattern. This is signal processing, not an increase in the power arriving at the antenna.
For integrators, the useful distinction is between an RF observation, a detected satellite signal and a navigation solution. Understanding that chain helps explain why a noise-like input can be usable, why longer observation has limits, and why a processing-gain calculation is not an anti-jamming product rating.
A known code makes a weak signal searchable
A pseudo-random noise code, or PRN code, looks irregular but is reproducible. The receiver can generate the relevant ranging code locally. Its individual elements are called chips; they are not the same as the slower navigation-message bits.
Correlation compares the incoming samples with that replica at a trial delay. The receiver multiplies corresponding values and adds the products. With the correct code and alignment, the desired contributions reinforce one another. At a different delay or with a different code, the reinforcement is generally much smaller. GNSS codes are designed for useful autocorrelation and cross-correlation properties, not perfect separation under every condition. ESA-hosted Navipedia's correlator explanation describes this operation.
The receiver must also account for frequency offset, including Doppler. A code match alone is insufficient if an uncompensated carrier rotates substantially during accumulation. The GNSS-SDR acquisition documentation makes the two-dimensional task explicit: search code delay and Doppler, then use the detected peak to initialize tracking. This is an implementation reference, not a statement about a ResiNav processing algorithm.
What accumulation changes: a simplified calculation
Consider a hypothetical, perfectly aligned correlator, not a GPS performance simulation. After code and carrier compensation, let each desired sample contribute the same amplitude A. Let the additive noise samples be independent, zero-mean, with variance σ². Assume the signal remains coherent throughout N samples.
The signal amplitude in the sum is N × A. The noise variance is N × σ², so its standard deviation grows as √N × σ. Consequently, the squared signal amplitude divided by noise variance improves by a factor N relative to one sample: an ideal gain of 10 log10(N) dB under these assumptions.
For N = 100, the signal sum is 100A while noise standard deviation is 10σ. The ideal ratio improves by 100, or 20 dB. If the results are divided by N to form an average, the signal stays at A and the noise standard deviation becomes σ/10. Summing and averaging describe the same ratio improvement with different scaling.
This example explains why a matched pattern can emerge without every input sample being visibly above noise. It does not predict a receiver's sensitivity: real samples can be correlated by filtering, and code error, frequency error, data transitions and implementation losses violate the ideal assumptions. Sample count is therefore not a shortcut to a product's usable integration gain.
Why “43 dB processing gain” needs a definition
One familiar GPS L1 C/A calculation uses the spreading chip rate divided by navigation-data bit rate:
10 log10(1,023,000 chips/s ÷ 50 bits/s) ≈ 43.1 dB.
Analog Devices' historical GPS receiver design note uses this chip-rate/data-rate convention. The number describes the stated L1 C/A signal relationship. It is not a universal GNSS value, an antenna gain, or a measured 43.1 dB suppression result for a protection device. The note's older product and constellation descriptions are not used here as current specifications.
The 43.1 dB calculation and the hypothetical 100-sample example use different reference quantities. They must not be added as independent improvements without defining the complete processing chain and avoiding double counting. When a comparison simply says “processing gain,” identify the signal, rate or bandwidth convention and integration interval before comparing the number with another report.
The existing C/N0 and SNR guide covers the separate bandwidth conversion between dB-Hz and dB. Correlation explains how the receiver uses signal structure; it does not change those unit definitions.
Longer integration helps only while the model remains valid
Coherent integration preserves phase while accumulating the signal. Non-coherent integration combines quantities such as the powers of successive correlation results. These are different operations, with different losses and constraints.
Navipedia's baseband-processing explanation identifies navigation-data sign changes and frequency-estimation accuracy as constraints on long coherent integration. It also distinguishes pilot components without navigation data from data-bearing components. A pilot removes that particular data-bit limitation; it does not remove every synchronization limit.
Thus “wait longer” is not an unlimited remedy. Accumulation helps while the local model follows the signal. A longer interval with an unresolved sign change or frequency error can lose the reinforcement being sought. Acquisition also has to search and make a detection decision, so sensitivity, observation time and false-alarm control belong together.
From a correlation peak to a usable result
Acquisition supplies an initial estimate. Tracking continuously refines it. In a simple early/prompt/late arrangement, the receiver compares replicas on either side of the expected alignment to determine which way to adjust the delay. Navipedia's multicorrelator description explains this feedback process. A prompt peak is not a static setting that remains correct indefinitely.
| Stage | What is being established | What that alone does not establish |
|---|---|---|
| RF sampling | A digital representation of the selected input band | Detection of a particular satellite signal |
| Acquisition and tracking | Signal presence and estimates that follow its code and carrier | A complete or authentic navigation solution |
| Navigation processing | A solution from suitable measurements, navigation data and receiver models | Independently verified accuracy or suitability for a platform's operating decision |
The generic receiver description separates signal processing from subsequent use of measurements and navigation data. That separation matters when reading a demonstration: a detected peak, a tracked signal and a position output are different observations.
For what those measurements contain, compare pseudorange and carrier phase before interpreting a tracked signal as a geometric range or a navigation solution.
Why correlation does not make interference protection unnecessary
Correlation operates on the samples supplied to it. The receiver's front end must first filter, amplify, convert and digitize the input within its operating range. Navipedia's front-end explanation describes these distinct stages and the importance of quantization and dynamic range. An ideal accumulation calculation cannot establish behavior when the earlier RF path is overloaded or its usable signal information is degraded.
To identify which setting to compare, distinguish RF passband, coherent-integration response and tracking-loop bandwidth before treating a narrower value as better protection.
Nor does a code match authenticate the source. Correlation tests a signal pattern, not the identity of the transmitter. The meaconing guide explains why even a replay of genuine content can create a different timing problem. Weak-signal extraction, interference mitigation and authenticity are separate questions, although a receiver may implement functions addressing more than one.
For an equipment comparison, separate three claims: the signal-processing principle, documented receiver operating behavior, and protection performance under specified conditions. Use the data-sheet review guide for suppression and J/S comparisons rather than treating a theoretical processing-gain figure as a substitute. If the project needs additional reception protection, continue to the ResiNav product catalogue and the selected model's existing enquiry route. This explanation does not assign a correlator design, sensitivity or processing-gain specification to any ResiNav model.
NEXT STEP
Turn available platform information into an engineering review.
Use the Technology centre to frame the discussion, review relevant application scenarios, then send the available platform and receiver details for confirmation.