Do Receive-Antenna Revisions Make a Measurable Difference?
Do Receive-Antenna Revisions Make a Measurable Difference?
A constructive paired A/B method for separating a real design benefit from propagation, site noise, installation, common mode, receiver state and time.
A receive-antenna revision can improve symmetry, common-mode control, filtering, linearity, transfer response, mechanical repeatability or weather resistance. Those mechanisms can be useful even when the difference is not dramatic. The engineering question is narrower: does the benefit appear in a controlled comparison, follow the candidate through cross-swaps and remain larger than the unresolved variation?
Start by Defining What “Better” Means
An A/B test needs a defined measurand before the first record is collected. “Louder,” “more spots” and “heard farther” combine several effects and invite a favourable result to be selected after the fact. State the comparison, frequency range, installation and primary outcome in advance.
Compare calibrated signal level or transfer response by band, orientation and polarization at a declared reference plane.
Record wanted signal and noise separately so a change in SNR can be traced to signal transfer, coupled noise or both.
Count paired successes, one-sided decodes and outages under like mode, bandwidth, software and receiver settings.
Decide whether the object under test is the bare antenna, the active head, or the complete installed receiving system. Feedline routing, bias injection, isolation, mast bonding and common-mode paths belong inside the measurand when the claim concerns installed SNR. They must be controlled or characterized when the claim concerns the antenna element alone.
Write the decision rule first. Name the primary metric, useful secondary metrics, allowed exclusions, test bands, operating states and the smallest difference that would matter in practice. A long run cannot rescue a comparison whose question changed after the data were seen.
Choose the Comparison Cadence
HF propagation and local noise can change within seconds. Small revision effects therefore need simultaneous reception or rapid switching. A comparison made on different days is valuable for operational history, but it is weak evidence for a small A/B difference unless an independent reference and a repeated randomized design remove the time bias.
Both candidates see the same transmission and nearly the same propagation. Channel gain, filtering, clocking, overload and cable differences must be calibrated and cross-swapped.
A characterized RF switch lets both candidates share one receiver path. Record switch loss, isolation, repeatability, settling time and the time between A and B samples.
Long runs reveal reliability and seasonal behaviour. They do not by themselves separate a small revision effect from different propagation, occupancy, weather or site noise.
Nearby antennas can couple to each other, and a switch can change termination or common-mode current. Check whether disconnecting, detuning or moving the unused candidate changes the active candidate’s response. Record the switch state and unused-port termination as part of the fixture.
A Six-Step Paired Test
Photograph and measure height, orientation, spacing, mast, nearby conductors, feedlines, grounding, bias feeds and choke placement for both candidates.
Inject a known signal at declared reference planes, measure relative gain and noise response, and verify that neither path is compressed or internally generating products.
Pair the same transmitter, transmission interval, mode, frequency and decoder state. Preserve raw timestamps, levels, noise estimates and status flags.
Move A and B between physical positions, then independently reverse their cables or receiver channels. The candidate effect should follow the candidate.
Repeat the A/B and swap sequence across bands, day and night, quiet and noisy periods, and relevant weather or operating conditions.
Show the paired distribution, block-to-block variation, calibration limits, missing data and uncertainty interval—not only a favourable mean or total.
Hold Receiver State Still
A receiver can manufacture an apparent antenna difference. Automatic gain changes, preamp or attenuator transitions, different filters, ADC overload, decoder revisions and reporting downtime all alter the observation. Log the state with the data rather than relying on memory.
| Confounder | Control | Evidence to retain |
|---|---|---|
| RF gain, preamp, attenuator and AGC | Use fixed settings where practical; if automation is unavoidable, record every state change for both paths. | Receiver configuration, state log and calibration checks before and after each block. |
| Bandwidth, filter shape, mode and decoder | Use the same mode, passband, sample rate, decoder build and reporting rules. | Software version, mode/submode, filter settings and reference bandwidth. |
| Clock, frequency and time alignment | Use a common reference where possible and pair only observations that belong to the same transmission interval. | Clock source, measured offset, timestamps and pairing tolerance. |
| Compression, clipping and intermodulation | Check each path with strong-signal states and reject intervals in which either receiver is outside its linear region. | Input levels, overload indicators, reference traces and exclusion reason. |
| Feedline, bias and common-mode path | Match or characterize cable loss and routing; repeat with cable and channel cross-swaps. | Cable identity, route, loss, shield-current or route-sensitivity checks, grounding and bias arrangement. |
| Uptime and report handling | Use matched availability windows and retain one-sided decodes instead of silently discarding them. | Heartbeat logs, missing-data flags, deduplication rules and complete report export. |
Pair Signal and Noise, Not Just Spot Counts
For a matched observation, define the difference in the same direction every time, for example ΔSNR = SNRB − SNRA. Also retain calibrated signal and noise estimates because ΔSNR = ΔSignal − ΔNoise in decibels at matched reference conditions. A positive SNR delta can result from greater wanted-signal transfer, less coupled noise, or both; those are different engineering outcomes.
Decoder reports need extra care. The PSKReporter developer specification accepts an integer SNR, mode and decoder-software field, while stating that data gathering depends on the client. The WSJT-X user guide defines its signal reports against a 2500 Hz reference noise bandwidth. That definition is useful inside a controlled WSJT-X test, but it should not be assumed for every client, mode or software version. Compare like with like and archive the decoder identity.
Decode-only data are threshold-censored. If the analysis keeps only transmissions decoded by both candidates, it removes the cases nearest the detection boundary. Track paired decodes, A-only decodes, B-only decodes and neither-decoded intervals. Where possible, retain synchronized recordings or calibrated level/noise measurements so non-decodes are not treated as absent transmissions.
Coverage Maps Are Context, Not an A/B Result
The public spots.rf.guru maps show decoder reports collected by operational receiving systems. They are useful for checking coverage, uptime, band occupancy and unusual paths. The companion sdr.rf.guru receivers can provide listening context. Neither view isolates a revision unless the underlying records come from the controlled paired protocol.
Numerical SNR filters are not mode conversions
A submitted-SNR threshold can create a useful view of the archive, but it does not turn decoder data into a CW or SSB intelligibility test. Reference bandwidth, mode, decoder behaviour, fading, interference and operator criteria still matter.
Propagation and Site Noise Still Need Their Own Controls
Simultaneous reception greatly reduces propagation-time bias, but the two antennas do not occupy the same electromagnetic point. A local noise source can couple differently by position, orientation, feedline route or common-mode impedance. The current ITU-R P.372 radio-noise recommendation treats atmospheric, galactic and man-made noise as distinct contributions with substantial variability. That variability belongs in the test record, not in a catch-all explanation after the result.
Use position swaps, route-sensitivity checks and repeated blocks. If the advantage remains at one location, it is a site effect. If it remains with one receiver path, it is a chain effect. If it follows the candidate through both swaps, the case for a candidate-specific effect becomes stronger. If it changes by band or noise state, report that dependence instead of averaging it away.
Build an Uncertainty Budget
An uncertainty statement is not an admission that nothing can be known. It identifies how accurately the stated difference is known. The NIST Technical Note 1297 framework separates statistically evaluated contributions from contributions estimated through calibration records, specifications and other evidence, then requires the result and uncertainty to be reported together.
| Contribution | How to estimate it | What it limits |
|---|---|---|
| Relative gain and noise calibration | Repeated injected-signal and terminated-input checks at the declared reference planes. | Small signal-level and SNR deltas. |
| Switch loss, isolation and repeatability | Characterize every path over frequency and repeat the switching sequence. | Rapid-switch comparisons and possible interaction between candidates. |
| Receiver drift and state | Pre/post block verification, shared references and logged configuration changes. | Long runs and comparisons across temperature or time. |
| Position and installation sensitivity | Physical cross-swaps, repeated mounting and feedline-route trials. | How far the result can be generalized beyond the tested installation. |
| Fading, propagation and local-noise variation | Matched pairs grouped into independent time, band and station blocks. | Statistical precision and repeatability across operating conditions. |
| Decoder and reporting process | Fixed software, reference bandwidth and deduplication; retain one-sided and missing observations. | Interpretation of SNR distributions and decode counts. |
Thousands of reports are not automatically thousands of independent samples. Repeated transmissions from the same station, one band opening and one local-noise event can be strongly correlated. Summarize complete blocks by session, band, time or transmitter, then show whether the direction and size of the result repeat across those blocks.
A publishable quantitative result identifies: the two candidates and installed configurations; reference planes; bands and modes; receiver and decoder states; pairing and exclusion rules; calibration results; position and chain swaps; number of independent blocks; effect estimate; uncertainty interval; limitations; and a route to the underlying dataset.
What a Constructive Revision Claim Can Say
A revision may be worthwhile because its engineering mechanism is sound and its benefit is observable under stated conditions. Better symmetry or common-mode control can reduce installation-dependent pickup. More linear headroom or filtering can reduce overload products. A changed transfer response can improve a target band. Improved sealing, strain relief or mechanical repeatability can make field behaviour more consistent.
Each mechanism needs the right evidence. Network, balance and route-sensitivity tests support a common-mode claim; blocking or two-tone tests support a linearity claim; calibrated sweeps support transfer response; environmental and production records support durability and consistency. A paired field test connects those mechanisms to installed reception, but only for the systems, bands, site and states that were actually tested.
Practical Release Checklist
- Define candidate A, candidate B and the complete installed configurations.
- Choose simultaneous reception or a characterized rapid-switch path.
- Fix and log receiver gain, filtering, bandwidth, clock, decoder and reporting state.
- Calibrate relative signal and noise response at declared reference planes.
- Collect synchronized signal, noise, decode and status records.
- Perform independent physical-position and receiver-chain cross-swaps.
- Repeat complete blocks across the intended bands and operating conditions.
- Account for coupling, site noise, common mode, missing reports and correlated samples.
- Report the effect estimate with uncertainty, limitations and the supporting dataset.
The Engineering Answer
Yes, a receive-antenna revision can make a real difference. A credible result appears in matched observations, survives calibration, follows the candidate through position and receiver-chain swaps, and remains meaningful after uncertainty and installation dependence are stated. Until that record exists, describe the expected mechanism and the validation plan rather than assigning a product-specific dB or SNR improvement.
Mini-FAQ
- What is the fastest credible A/B test? A characterized rapid switch through one receiver chain can be strong evidence when switching is much faster than the channel variation and switch loss, isolation, settling and termination are known.
- Is simultaneous reception always better? It removes most time separation, but two channels add relative-gain, filtering, clock and overload differences. Calibrate and cross-swap both paths.
- Why swap positions and receiver channels? Position swaps expose site bias; independent cable or channel swaps expose signal-chain bias. A candidate-specific effect should follow the candidate through both.
- Can spot counts prove that a revision is better? Not alone. Counts also depend on propagation, transmitting activity, decoder thresholds, uptime, settings and reporting rules. Use matched pairs, one-sided decodes, signal and noise records.
- Should signal and noise be recorded separately? Yes. The same SNR change can come from more wanted signal, less coupled noise or a combination, and those outcomes imply different mechanisms.
- When can a product-specific dB claim be made? When the paired dataset, calibration, cross-swaps, analysis, uncertainty interval and tested conditions are available for review and the claim stays within those boundaries.