An elevation mask selects observations by satellite direction. C/N0 screening selects them by a receiver-reported signal metric. Observation weighting retains measurements but changes their influence on an estimate. Exclusion and downweighting are therefore different operations. None, by itself, proves that a retained measurement is unbiased or that a position is correct. Before comparing settings, identify the processing stage they actually change.
Three controls, three different decisions
| Control | What it changes | What it does not establish |
|---|---|---|
| Elevation mask | Eligibility based on angle above the local horizon, in degrees, for the satellites or observations covered by the setting. | Whether a particular path is direct, reflected or unbiased. |
| C/N0 screening | Eligibility based on a reported carrier-to-noise-density ratio, usually expressed in dB-Hz. Check which signals and processing stage the rule covers. | Whether a sufficiently strong tracked signal gives an unbiased range. |
| Observation weighting | The influence of retained observations under a specified error model. Lower, nonzero weight is not exclusion. | Removal of a systematic measurement bias or validation of the error model. |
These controls can be combined. A processor may reject observations first and then assign weights to the surviving set. Its residual-based fault checks may make another decision later. NovAtel's Waypoint differential settings document elevation and C/N0 pre-filtering separately from subsequent automatic residual outlier detection. That is a particular processing workflow, not a universal sequence for every receiver.
First locate the control in the processing chain
The word “mask” does not tell you whether a setting changes signal tracking or only the measurements admitted to a solution.
For a receiver example, NovAtel's OEM7 GPS ECUTOFF command can prevent automatic satellite search below the cutoff when satellite positions are known. It also stops tracking below the cutoff unless a satellite is manually assigned. Below-cutoff satellites are excluded from the internal position and clock solution in either case. This describes that GPS command, including its exception; it is not a rule for every constellation or product.
By contrast, Waypoint's documented elevation and C/N0 rejection settings operate on observations during post-processing. A pre-filter there does not retrospectively change what the receiver searched for or tracked in the field.
Logging and export need their own check. Do not infer that all tracked observations were recorded, or that a solution exclusion removed the raw record. Conversely, offline processing cannot recreate observations that were never acquired or saved. This distinction determines which settings can genuinely be compared using the same recorded data.
Elevation and C/N0 provide partial evidence
Elevation describes direction, not the physical path that reached the antenna. Low-elevation observations can face longer atmospheric paths and difficult reflection conditions, but rejecting an angular band is not a direct-path detector. The tropospheric and ionospheric delay guide explains why atmospheric terms depend on propagation and modeling, rather than an angle alone.
C/N0 is a signal metric, not an error in metres. A reflected signal can be strong enough to remain tracked even when the direct path is blocked. NovAtel's multipath explanation describes that NLOS case. Passing a C/N0 threshold therefore cannot certify an unbiased range. The multipath versus NLOS guide develops the path distinction in more detail.
Also distinguish a satellite from an individual signal channel. OEM7's TRACKSTAT log can contain multiple channel entries for one satellite and reports C/N0, pseudorange residual, rejection code and pseudorange weight as separate fields. Those fields are different evidence, not interchangeable names for quality. Their definitions belong to that implementation; the log is not a claimed ResiNav output.
Lower weight changes influence, not the observation's bias
In a diagonal weighted least-squares model, the processor minimizes a sum of squared residuals multiplied by weights. A larger weight makes a residual more costly and gives that observation more influence. This is the diagonal case of the ESA weighted least-squares formulation.
Weights often represent assumed measurement uncertainty. That assumption must be stated: a number called “weight” in a vendor log need not use the convention of a teaching example. Correlated observations may require a covariance model with off-diagonal terms, rather than independent scalar weights. The receiver uncertainty guide explains why the model can report more confidence than the independent evidence supports.
Reducing an observation's weight reduces its influence. It does not subtract a bias from that observation. A biased measurement can still pull the estimate, and assigning a smaller influence does not replace fault detection or independent validation.
A hypothetical example: retained, downweighted or excluded
Consider three measurements of one scalar distance, not a GNSS position solution. Let the true value be 0 m and the observed values be 0 m, 0 m and +6 m. The third value has an imposed +6 m bias. Hold those observations fixed.
For this example only, use positive inverse-assumed-variance weights, in m⁻². The fitted scalar is:
x̂ = (w₁z₁ + w₂z₂ + w₃z₃) / (w₁ + w₂ + w₃).
With weights (1, 1, 1) m⁻², the estimate is 2 m. Retaining the third observation with weight 0.1 m⁻² gives 0.6 / 2.1 m = 2/7 m, approximately 0.286 m. Excluding it leaves an estimate of 0 m in this constructed case.
The smaller weight corresponds to an assumed variance of 10 m² instead of 1 m² for that observation. It has not changed the +6 m bias. The example supplies no elevation-to-variance or C/N0-to-variance mapping, and it predicts neither receiver accuracy nor satellite geometry. It simply isolates what influence reduction does. Exclusion happens to help here because the biased observation and reference value were specified in advance; a real processor must establish its evidence rather than know that answer beforehand.
Geometry and availability remain separate outcomes
Removing observations changes the usable set and can weaken geometry or leave no acceptable solution. Keeping biased observations can also worsen a result. Neither satellite count nor a geometry statistic resolves both effects.
The DOP versus position accuracy guide explains that propagation problem. For an implementation-specific caution, OEM7's RTKDOP documentation distinguishes satellites used in the RTK solution from those merely tracked, and states that different observation weights are not included in that log's DOP calculation. Do not assume that an unchanged displayed DOP means a weighting change had no effect.
There is consequently no single best elevation or C/N0 threshold to prescribe independently of the antenna environment, signal set, processor, dynamics and acceptance criteria.
Compare settings without changing the question
For controls that act only on saved observations, process the same data with a documented baseline, then change one control at a time. Keep corrections, time intervals, coordinate frame, antenna reference point and solution configuration consistent. Record the exact software and setting semantics. A replay of saved observations can compare solver selection and weighting; it cannot by itself reproduce the acquisition or tracking consequences of a live receiver command.
Use an appropriately independent reference, aligned in time and position reference point. Keep tuning data separate from assessment data. Then retain four complementary views:
- Position error: horizontal and vertical errors, distributions and difficult intervals against the reference. Describe reference uncertainty. Compare a common valid-epoch subset as well as the complete run's outcomes.
- Availability: use a predefined acceptable-solution criterion and the same planned-epoch denominator. Preserve outages and rejected intervals instead of silently removing them from the error comparison.
- Observation evidence: distinguish tracked and used signals, rejection reasons, C/N0 and the documented weighting model. Identify whether the reported geometry statistic includes the weighting convention.
- Solution behavior: retain mode changes, initialization or recovery periods and any relevant validation state. A favorable average must not conceal loss of the required operating state.
If the control changes live tracking, a paired field or RF-replay experiment needs a documented repeatability and reference strategy; it is not the same experiment as changing a post-processing filter. The receiver evaluation guide provides the broader context for defining the evidence and acceptance scope.
The useful decision is which layer to change, why its criterion is informative, and whether the resulting error and availability meet the application requirement. A cleaner-looking satellite list, higher C/N0 or lower reported uncertainty is evidence to interpret, not a substitute for that comparison.
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.