GNSS RF bandwidth determines which part of the incoming spectrum reaches subsequent processing. Tracking-loop bandwidth determines how code or carrier estimates respond to changing errors after correlation. Correlation and integration introduce another frequency response between those stages. These quantities can all be expressed in hertz, but they are not interchangeable controls: narrowing a tracking loop does not narrow the antenna passband or protect an overloaded RF stage.
For an integration engineer, the first decision is therefore which layer a problem belongs to. Preserving the required signal spectrum, tolerating residual frequency error and following platform dynamics are different jobs. There is no general rule that the narrowest setting gives the best interference protection.
Put each bandwidth beside the operation it describes
The receiver chain below is a functional description, not a drawing of a ResiNav product. Analogue and digital filtering can be distributed differently in different implementations.
| Quantity | What it acts on | What its number does not mean |
|---|---|---|
| RF, IF or pre-correlation passband | Frequency content delivered through a particular input or sample-processing path | The rate at which a tracking loop can follow changing code or carrier errors |
| Correlation and coherent integration response | Agreement with a local code/carrier replica over an observation interval | A replacement antenna bandwidth or a universal loop-filter setting |
| DLL noise bandwidth | Fluctuations in the code-delay tracking feedback process | The width of the received GNSS signal spectrum |
| PLL noise bandwidth | Fluctuations in the carrier-phase tracking feedback process | An RF notch width, interference-rejection rating or navigation output rate |
ESA-hosted Navipedia's receiver-design description separates front-end conversion from channel processing and identifies code and carrier loops. That separation is the reason a megahertz input filter and a hertz-scale loop parameter must not be compared as competing specifications.
RF filtering decides which information is available
An input filter has a frequency response, including its useful passband, transition and rejection regions. Narrowing it can exclude unwanted energy outside the retained band, but it can also remove wanted spectral components. A shared centre frequency is not sufficient to establish that two GNSS modulations need the same processing bandwidth.
Navipedia's front-end explanation relates signal structure to front-end selection. Its correlator reference connects the signal spectrum with the correlation shape used for ranging. The objective is not simply to collect the least noise power: it is to retain the information required for the intended measurement while managing unwanted input energy.
A downstream algorithm cannot restore signal components that an earlier filter has removed. Nor can changing a DLL or PLL setting undo distortion already created by an overloaded amplifier or converter. The separate adjacent-band review explains where physical filtering belongs; this article explains why the later loop setting is a different control.
State the bandwidth convention before comparing numbers
A passband width measured between specified attenuation points is not automatically an equivalent noise bandwidth. The latter describes noise admitted through the complete response under its stated normalization. One-sided baseband cut-off, two-sided span and a measurement instrument's resolution bandwidth also need their own labels.
For an instrument-specific example, Keysight's resolution-bandwidth documentation distinguishes its VSA equivalent-noise-bandwidth convention from the 3 dB convention commonly used by swept analysers. This does not redefine GNSS loop bandwidth; it demonstrates why the word “bandwidth” alone is insufficient.
Sampling rate is another separate quantity. GNSS-SDR's input-filter documentation specifies sample rate, filter edges, transition width and possible decimation separately. A sample-rate entry is not evidence of a flat analogue passband of the same numerical width. Retain the real/complex sampling convention and the actual filter response when interpreting a recording or receiver description.
Integration changes tolerance to residual frequency error
Correlation first removes the hypothesized code and carrier. Coherent integration then accumulates what remains. A residual carrier-frequency error rotates the accumulated signal, so contributions can cancel rather than reinforce. The existing correlation guide explains accumulation gain; the bandwidth question is how that accumulation responds to an imperfect frequency match.
Consider an explicitly idealized calculation: perfect code alignment, constant amplitude, rectangular averaging for T seconds, and a constant residual frequency error Δf in hertz. Assume no data-bit or secondary-code sign changes and no other phase evolution. Direct integration of that rotating phasor gives the normalized amplitude response:
G(Δf) = |sin(π Δf T) / (π Δf T)|, with G(0) = 1.
The first zeros are at Δf = ±1/T. An invented 1 ms interval therefore has its first zeros at ±1000 Hz; a 20 ms interval has them at ±50 Hz. These are frequency-mismatch nulls of this rectangular calculation, not its 3 dB width, not a declared equivalent noise bandwidth and not a receiver's RF passband.
Longer accumulation thus demands a better frequency/phase match. It does not permit a designer to replace a signal's required RF spectrum with a 50 Hz input filter. This simplified relationship is consistent with the correlation model in GNSS-SDR's tracking documentation; it is not a recommended integration interval or a ResiNav operating limit.
Loop bandwidth trades noise against response
A DLL adjusts the local code delay; a PLL adjusts carrier phase. Their feedback filters govern how the estimates react to discriminator errors. A narrower loop can reject more rapid noise fluctuations, but may respond too slowly to the signal changes that the receiver must follow.
Navipedia's tracking-loop discussion treats bandwidth, filter order, integration interval and aiding together. Dynamics include motion projected along the satellite line of sight and oscillator behavior; a quiet platform does not make every relevant error constant. Changing one nominal bandwidth while changing order or aiding is not a clean one-variable comparison.
A limited model illustrates the noise side. In Navipedia's stated arctangent-PLL thermal-noise expression, hold C/N0, wavelength, integration time and the applicable model assumptions fixed. The thermal-noise standard deviation scales with the square root of loop noise bandwidth. Hypothetically changing 16 Hz to 4 Hz gives:
σ₄ / σ₁₆ = √(4/16) = 0.5.
That halves this modelled thermal-noise contribution, not total error. It says nothing by itself about dynamic tracking, oscillator disturbance, multipath or interference tolerance. The two bandwidths are teaching inputs, not default or recommended settings.
RF bandwidth also enters the code measurement
The earlier passband and the later loop can both influence a measurement without being the same parameter. Navipedia's DLL performance model explicitly distinguishes double-sided front-end bandwidth from loop noise bandwidth, and also includes signal spectrum, integration time and early–late spacing.
This explains why a smaller DLL bandwidth does not automatically compensate for a narrower input filter. The filter changes the information presented to the discriminator; the loop changes the response to its output. Correlator spacing, commonly specified in chips, is yet another design parameter—not a bandwidth in hertz. There is no justified universal conversion from one of these entries to another.
Choose the layer from the observed problem
For a blocker threatening an early RF stage, investigate input filtering, linearity and the affected stage. For a wanted signal being cut by a passband, compare the signal spectrum and filter response. Neither problem is solved merely by selecting a narrower PLL.
For cancellation caused by residual frequency error during accumulation, investigate synchronization and the integration interval. For noisy or lagging code/carrier estimates, compare the documented loop configuration, dynamics and oscillator conditions. A smoother trace alone cannot establish that tracking remains valid through the required changes.
Evaluate each change against both an appropriate quiet-input baseline and the relevant operating condition. Distinguish retained signal information, tracking error/continuity and the final navigation result. Use the receiver evaluation guide for the controlled comparison, not a universal narrow-band setting.
If a project independently needs added GNSS reception protection, the ResiNav catalogue is the published product-screening route. No adjustable loop, internal RF bandwidth or measurement performance is assigned to a ResiNav model here. The useful conclusion is to select the correct layer first, then compare a defined response—not to choose the smallest number labelled bandwidth.
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.