The night shift: when the EFHW80 catches up Kevin McKeown, 10/10/202610/10/2026 M7XOM • Antenna field notes • 9–10 October 2026 I had a feeling that my 20 m vertical did better towards the end of the day, while the 80 m EFHW was happier in the morning. This time, I let both listen together—and kept the logs. Complete receive experiment • 20:00 UTC, 9 October to 09:34 UTC, 10 October • recording interrupted • all times below are UTC +3 dBVertical’s evening median advantage−1 dBEarly recovery: EFHW80 leads−0.5 dBLater morning: almost level The Flex makes this rather clever. Two slices monitor 20 m FT8 simultaneously, one listening on ANT 1 and the other on ANT 2. Separate WSJT-X instances report under M7XOM and M7XOM/P. One receiving installation uses my EFHW20 vertical; the other uses the longer EFHW80 wire. Both are receiving on 14.074 MHz. PSK Reporter gives a useful view of who is being heard, but the local WSJT-X ALL.TXT files let us go further: compare the same message, transmitted at the same moment, and examine its received signal-to-noise ratio on both antennas. The results give some support to my hunch. The vertical has a substantial evening advantage. During the early morning recovery, the EFHW80 hears more stations and has a modest paired-SNR lead. Later in the morning, the two installations are almost level. It is an interesting result for this collection, rather than an established rule about which antenna wins at a particular hour.The three substantive phases: evening, early recovery and later morning. Compare message totals within each phase; phase lengths differ. Positive SNR difference favours the vertical. Each paired station contributes one balanced result.The laptop ends its shift earlyThe plan was to record from 20:00 until 08:00. The laptop had other ideas and shut down. Both logs stop at 06:15:15, then resume at 07:16:15 for a brief snapshot ending at 07:17:00. A separate morning collection covers 07:51:15–09:34:00.I showed the laptop pictures of possible replacements and reminded it that it could be replaced. Whether that improves its future attendance remains to be seen.Missing recording is not a dead band. The hour between the shutdown and restart slots is marked as a laptop gap. No supplied recording covers the interval between the restart snapshot and the later collection. Neither interval is interpreted as a propagation measurement. Extra rows outside each pair of files’ common endpoints are excluded. Watching the evening fade and the morning returnThe evening is busy, then logged activity dwindles sharply. Neither file contains decodes in the complete 00:45–02:00 blocks. The EFHW80 records a few signals from 02:15 onward. Both installations become much busier around 04:45, and the recovery is well under way by 05:00.In the complete 06:00–06:15 block, the vertical records 1,067 messages and the EFHW80 records 1,106. The later morning collection is busy throughout; both files contain decodes in every FT8 slot within its common window.This shows the changing reception I wanted to explore, with one limit: ALL.TXT is a decode log, not a continuous monitoring heartbeat. An empty period means no decodes were logged. By itself, it cannot prove that the decoder kept running or that propagation closed. Changes in transmitting activity also contribute.Complete 15-minute blocks only. Partial windows are omitted from the count lines, which break across missing recordings. Shading marks the laptop interruption and the later interval without supplied data. Zero means no logged decodes.What each phase saysPhaseMessages V / RStations V / RMedian ΔPaired stationsEvening9,064 / 6,525480 / 369+3 dB320Late evening573 / 34443 / 37+3 dB27Quiet night306 / 35433 / 40+2 dB27Morning recovery3,901 / 4,269219 / 244-1 dB197Restart snapshot112 / 12374 / 77-1.25 dB66Later morning11,425 / 11,528478 / 490-0.5 dB439V = EFHW20 vertical; R = EFHW80 reference. Evening 20:00–23:00; late evening 23:00–01:00; quiet night 01:00–05:00; early recovery 05:00–06:15:15; restart 07:16:15–07:17:00; later morning 07:51:15–09:34:00. Δ is vertical minus reference. Unique station counts cannot be added across phases. Between 20:00 and 23:00, the vertical records 38.9% more messages and hears 480 identifiable transmitters against 369. Among 320 stations with paired measurements, its station-balanced median advantage is 3 dB.During the early recovery, the EFHW80 records 9.4% more messages and hears 244 stations against 219. The paired station median favours it by 1 dB. In the later morning, its lead narrows to 0.9% more messages, 490 stations against 478, and a 0.5 dB paired station median advantage.Most 15-minute windows in the later morning have a paired station median of 0 dB; the 08:30 block favours the EFHW80 by 1 dB. That is near parity, rather than a decisive morning victory.The quiet-night results also show why counts and SNR belong together. The EFHW80 hears more stations, but the much smaller subset decoded by both still has a positive median difference for the vertical. One-sided reception and paired SNR describe different parts of the experiment.Unique station overlap across the complete experiment, alongside the paired station median in each 15-minute block. Pale disconnected points have fewer than five paired stations. Partial blocks are included for SNR where available, not compared as full activity windows.A different morning sky—or the same paths changing?The evening and morning samples mostly contain different transmitters. That matters: a changing mix of paths can change the aggregate antenna result even without every individual path reversing.Only four stations supply paired measurements in both the evening and early recovery. All four shift towards the EFHW80:StationEvening ΔEarly recovery ΔChangePV8AJ+0 dB-4.5 dB-4.5 dBPY2DPM+5.5 dB+1.5 dB-4 dBPY5DC+4 dB+1 dB-3 dBZ62NS+8.5 dB-1 dB-9.5 dBThat is useful supporting evidence, but four stations are a small sample. PY2DPM and PY5DC still favour the vertical during the early recovery, just less strongly than in the evening.Seven stations have paired measurements in both the evening and later morning. Six move towards the EFHW80 and one is unchanged. The direction agrees with the earlier comparison, though these small, selected groups are not proof of a general daily pattern. Play the reception backThe combined sample contains 1,310 identifiable transmitters. Of these, 1,117 have consistent four-character grid locators in the logs, allowing a geographical playback in 15-minute windows. The 193 without usable grids remain in station counts but are absent from the maps.A symbol means a station was decoded within that window. A disappearing symbol can also mean that the station stopped sending. “Both” means both antennas heard it somewhere in that window; the separate paired-SNR calculation requires the same actual transmission.Animated reception throughout the supplied collection. Triangles: vertical only; squares: reference only; circles: both within the window. Missing recordings and partial frames are explicitly labelled.Whole-experiment signal origins. Teal favours the vertical in station-balanced paired SNR; brown favours the reference; grey has no attributable paired result. Positions are centres of logged Maidenhead grids, not exact station locations.Individual paths have their own preferencesThe heatmap follows the 28 most consistently paired stations. Persistent differences and changes are easier to see here than in one overall score. Blank cells contain no attributable matched measurement; they are not measured zeros.Median vertical-minus-reference SNR for each station in each 15-minute block. Colours saturate at ±10 dB. Blank cells include missing matches and recording gaps. These are observed receive differences, not a measured antenna radiation pattern.The complete experiment in numbersMeasureEFHW20 verticalEFHW80 referenceDecoded messages25,38123,143Unique identified transmitters1,1811,120Transmitters heard only by this antenna190129Transmitters heard by both991991Across 19,517 matched messages, 966 identifiable stations support a station-balanced result. The whole-experiment median is +1 dB for the vertical; the middle half runs from -2 to +4 dB. Of these stations, 550 favour the vertical, 48 are equal and 368 favour the EFHW80.Whole-experiment totals are useful context, but they conceal the timing. The +1 dB overall median does not mean a fixed 1 dB advantage throughout the night.Left: one balanced result per paired station. Right: all matched messages, with repeated transmissions contributing to density. Points above the equality line favour the vertical. Repeated decodes are not independent trials.What I make of itMy feeling about evening and morning performance has some support in this recording: a strong evening showing from the vertical, a modest early recovery advantage for the EFHW80, then almost equal performance later in the morning.I inspected the EFHW80 after worrying about a possible fault, and it looks good. Its morning reception gives no indication of broad underperformance. That does not exclude an intermittent issue or differences between the two receiving paths, but there is no basis here for declaring the antenna faulty.The next useful experiment is to swap the antennas between ANT 1 and ANT 2 while keeping receiver and decoder settings fixed. That would help reveal whether a persistent offset follows the antenna or the receiving chain. Several more nights would show whether this evening-to-morning pattern repeats.For now, having two different pieces of wire available seems worthwhile. They hear many of the same signals, but they do not always favour the same paths. How the comparison was made All six supplied receive-log uploads were combined by antenna. Repeated rows from the overlapping evening and overnight uploads were removed: 9,144 from the vertical set and 6,551 from the reference. No repeated timestamp/audio-frequency/message key disagreed in logged SNR. The analysis begins at the requested 20:00 UTC on 9 October 2026. Three common windows are used, ending at 06:15:15, covering the 07:16:15–07:17:00 restart snapshot, and covering 07:51:15–09:34:00 on 10 October. Unmatched endpoint tails are excluded. Selection is receive-only FT8 on 20 m. All selected rows use 14.074 MHz. The earlier transmit/PSK Reporter antenna comparisons are a different experiment and are not mixed into these receive statistics. Matching requires identical timestamps and complete message text, with audio frequencies within 10 Hz. Each row is used once. Actual frequency separations in the matched data are no greater than 4 Hz. Standard messages identify the sender as the second callsign; CQ/QRZ messages use the callsign after qualifiers. Compound semicolon and unresolved ellipsis messages remain in decode counts and may be matched, but are not assigned to a single transmitter. Within each phase, take each station’s median paired SNR difference in each 15-minute block, then its median across those blocks, then the median across stations. The whole-experiment result uses the same procedure over all selected data. This prevents a prolific transmitter from dominating the headline. Paired SNR covers only transmissions decoded by both installations. One-sided decodes are reported separately, without inventing missing SNR values. Correlated transmissions and paths mean the decode count is not a count of independent experimental trials; no formal confidence interval is claimed. Map positions use consistent four-character Maidenhead grid centres. RR73 is excluded as a locator. No conflicting grid assignments were found. A locator learned later in this same experiment may locate an earlier decode retrospectively. Coastlines are Natural Earth outlines from World Atlas 2. Maps show observed signal origins, not ionospheric trajectories, propagation modes or exact station positions. Zero-decode intervals do not verify continuous monitoring. The known laptop gap and the later interval without supplied recording are separate from those quiet periods. Activity charts omit partial blocks. The setup is simultaneous Flex slices and separate WSJT-X instances, as described by the operator. The logs cannot verify identical gain, attenuation, processing, audio levels or decoder settings. SNR depends on signal and noise, so this compares complete receiving installations rather than isolating antenna gain. Antenna Experiments
M7XOM • Antenna field notes • 9–10 October 2026 I had a feeling that my 20 m vertical did better towards the end of the day, while the 80 m EFHW was happier in the morning. This time, I let both listen together—and kept the logs. Complete receive experiment • 20:00 UTC, 9 October to 09:34 UTC, 10 October • recording interrupted • all times below are UTC +3 dBVertical’s evening median advantage−1 dBEarly recovery: EFHW80 leads−0.5 dBLater morning: almost level The Flex makes this rather clever. Two slices monitor 20 m FT8 simultaneously, one listening on ANT 1 and the other on ANT 2. Separate WSJT-X instances report under M7XOM and M7XOM/P. One receiving installation uses my EFHW20 vertical; the other uses the longer EFHW80 wire. Both are receiving on 14.074 MHz. PSK Reporter gives a useful view of who is being heard, but the local WSJT-X ALL.TXT files let us go further: compare the same message, transmitted at the same moment, and examine its received signal-to-noise ratio on both antennas. The results give some support to my hunch. The vertical has a substantial evening advantage. During the early morning recovery, the EFHW80 hears more stations and has a modest paired-SNR lead. Later in the morning, the two installations are almost level. It is an interesting result for this collection, rather than an established rule about which antenna wins at a particular hour.The three substantive phases: evening, early recovery and later morning. Compare message totals within each phase; phase lengths differ. Positive SNR difference favours the vertical. Each paired station contributes one balanced result.The laptop ends its shift earlyThe plan was to record from 20:00 until 08:00. The laptop had other ideas and shut down. Both logs stop at 06:15:15, then resume at 07:16:15 for a brief snapshot ending at 07:17:00. A separate morning collection covers 07:51:15–09:34:00.I showed the laptop pictures of possible replacements and reminded it that it could be replaced. Whether that improves its future attendance remains to be seen.Missing recording is not a dead band. The hour between the shutdown and restart slots is marked as a laptop gap. No supplied recording covers the interval between the restart snapshot and the later collection. Neither interval is interpreted as a propagation measurement. Extra rows outside each pair of files’ common endpoints are excluded. Watching the evening fade and the morning returnThe evening is busy, then logged activity dwindles sharply. Neither file contains decodes in the complete 00:45–02:00 blocks. The EFHW80 records a few signals from 02:15 onward. Both installations become much busier around 04:45, and the recovery is well under way by 05:00.In the complete 06:00–06:15 block, the vertical records 1,067 messages and the EFHW80 records 1,106. The later morning collection is busy throughout; both files contain decodes in every FT8 slot within its common window.This shows the changing reception I wanted to explore, with one limit: ALL.TXT is a decode log, not a continuous monitoring heartbeat. An empty period means no decodes were logged. By itself, it cannot prove that the decoder kept running or that propagation closed. Changes in transmitting activity also contribute.Complete 15-minute blocks only. Partial windows are omitted from the count lines, which break across missing recordings. Shading marks the laptop interruption and the later interval without supplied data. Zero means no logged decodes.What each phase saysPhaseMessages V / RStations V / RMedian ΔPaired stationsEvening9,064 / 6,525480 / 369+3 dB320Late evening573 / 34443 / 37+3 dB27Quiet night306 / 35433 / 40+2 dB27Morning recovery3,901 / 4,269219 / 244-1 dB197Restart snapshot112 / 12374 / 77-1.25 dB66Later morning11,425 / 11,528478 / 490-0.5 dB439V = EFHW20 vertical; R = EFHW80 reference. Evening 20:00–23:00; late evening 23:00–01:00; quiet night 01:00–05:00; early recovery 05:00–06:15:15; restart 07:16:15–07:17:00; later morning 07:51:15–09:34:00. Δ is vertical minus reference. Unique station counts cannot be added across phases. Between 20:00 and 23:00, the vertical records 38.9% more messages and hears 480 identifiable transmitters against 369. Among 320 stations with paired measurements, its station-balanced median advantage is 3 dB.During the early recovery, the EFHW80 records 9.4% more messages and hears 244 stations against 219. The paired station median favours it by 1 dB. In the later morning, its lead narrows to 0.9% more messages, 490 stations against 478, and a 0.5 dB paired station median advantage.Most 15-minute windows in the later morning have a paired station median of 0 dB; the 08:30 block favours the EFHW80 by 1 dB. That is near parity, rather than a decisive morning victory.The quiet-night results also show why counts and SNR belong together. The EFHW80 hears more stations, but the much smaller subset decoded by both still has a positive median difference for the vertical. One-sided reception and paired SNR describe different parts of the experiment.Unique station overlap across the complete experiment, alongside the paired station median in each 15-minute block. Pale disconnected points have fewer than five paired stations. Partial blocks are included for SNR where available, not compared as full activity windows.A different morning sky—or the same paths changing?The evening and morning samples mostly contain different transmitters. That matters: a changing mix of paths can change the aggregate antenna result even without every individual path reversing.Only four stations supply paired measurements in both the evening and early recovery. All four shift towards the EFHW80:StationEvening ΔEarly recovery ΔChangePV8AJ+0 dB-4.5 dB-4.5 dBPY2DPM+5.5 dB+1.5 dB-4 dBPY5DC+4 dB+1 dB-3 dBZ62NS+8.5 dB-1 dB-9.5 dBThat is useful supporting evidence, but four stations are a small sample. PY2DPM and PY5DC still favour the vertical during the early recovery, just less strongly than in the evening.Seven stations have paired measurements in both the evening and later morning. Six move towards the EFHW80 and one is unchanged. The direction agrees with the earlier comparison, though these small, selected groups are not proof of a general daily pattern. Play the reception backThe combined sample contains 1,310 identifiable transmitters. Of these, 1,117 have consistent four-character grid locators in the logs, allowing a geographical playback in 15-minute windows. The 193 without usable grids remain in station counts but are absent from the maps.A symbol means a station was decoded within that window. A disappearing symbol can also mean that the station stopped sending. “Both” means both antennas heard it somewhere in that window; the separate paired-SNR calculation requires the same actual transmission.Animated reception throughout the supplied collection. Triangles: vertical only; squares: reference only; circles: both within the window. Missing recordings and partial frames are explicitly labelled.Whole-experiment signal origins. Teal favours the vertical in station-balanced paired SNR; brown favours the reference; grey has no attributable paired result. Positions are centres of logged Maidenhead grids, not exact station locations.Individual paths have their own preferencesThe heatmap follows the 28 most consistently paired stations. Persistent differences and changes are easier to see here than in one overall score. Blank cells contain no attributable matched measurement; they are not measured zeros.Median vertical-minus-reference SNR for each station in each 15-minute block. Colours saturate at ±10 dB. Blank cells include missing matches and recording gaps. These are observed receive differences, not a measured antenna radiation pattern.The complete experiment in numbersMeasureEFHW20 verticalEFHW80 referenceDecoded messages25,38123,143Unique identified transmitters1,1811,120Transmitters heard only by this antenna190129Transmitters heard by both991991Across 19,517 matched messages, 966 identifiable stations support a station-balanced result. The whole-experiment median is +1 dB for the vertical; the middle half runs from -2 to +4 dB. Of these stations, 550 favour the vertical, 48 are equal and 368 favour the EFHW80.Whole-experiment totals are useful context, but they conceal the timing. The +1 dB overall median does not mean a fixed 1 dB advantage throughout the night.Left: one balanced result per paired station. Right: all matched messages, with repeated transmissions contributing to density. Points above the equality line favour the vertical. Repeated decodes are not independent trials.What I make of itMy feeling about evening and morning performance has some support in this recording: a strong evening showing from the vertical, a modest early recovery advantage for the EFHW80, then almost equal performance later in the morning.I inspected the EFHW80 after worrying about a possible fault, and it looks good. Its morning reception gives no indication of broad underperformance. That does not exclude an intermittent issue or differences between the two receiving paths, but there is no basis here for declaring the antenna faulty.The next useful experiment is to swap the antennas between ANT 1 and ANT 2 while keeping receiver and decoder settings fixed. That would help reveal whether a persistent offset follows the antenna or the receiving chain. Several more nights would show whether this evening-to-morning pattern repeats.For now, having two different pieces of wire available seems worthwhile. They hear many of the same signals, but they do not always favour the same paths. How the comparison was made All six supplied receive-log uploads were combined by antenna. Repeated rows from the overlapping evening and overnight uploads were removed: 9,144 from the vertical set and 6,551 from the reference. No repeated timestamp/audio-frequency/message key disagreed in logged SNR. The analysis begins at the requested 20:00 UTC on 9 October 2026. Three common windows are used, ending at 06:15:15, covering the 07:16:15–07:17:00 restart snapshot, and covering 07:51:15–09:34:00 on 10 October. Unmatched endpoint tails are excluded. Selection is receive-only FT8 on 20 m. All selected rows use 14.074 MHz. The earlier transmit/PSK Reporter antenna comparisons are a different experiment and are not mixed into these receive statistics. Matching requires identical timestamps and complete message text, with audio frequencies within 10 Hz. Each row is used once. Actual frequency separations in the matched data are no greater than 4 Hz. Standard messages identify the sender as the second callsign; CQ/QRZ messages use the callsign after qualifiers. Compound semicolon and unresolved ellipsis messages remain in decode counts and may be matched, but are not assigned to a single transmitter. Within each phase, take each station’s median paired SNR difference in each 15-minute block, then its median across those blocks, then the median across stations. The whole-experiment result uses the same procedure over all selected data. This prevents a prolific transmitter from dominating the headline. Paired SNR covers only transmissions decoded by both installations. One-sided decodes are reported separately, without inventing missing SNR values. Correlated transmissions and paths mean the decode count is not a count of independent experimental trials; no formal confidence interval is claimed. Map positions use consistent four-character Maidenhead grid centres. RR73 is excluded as a locator. No conflicting grid assignments were found. A locator learned later in this same experiment may locate an earlier decode retrospectively. Coastlines are Natural Earth outlines from World Atlas 2. Maps show observed signal origins, not ionospheric trajectories, propagation modes or exact station positions. Zero-decode intervals do not verify continuous monitoring. The known laptop gap and the later interval without supplied recording are separate from those quiet periods. Activity charts omit partial blocks. The setup is simultaneous Flex slices and separate WSJT-X instances, as described by the operator. The logs cannot verify identical gain, attenuation, processing, audio levels or decoder settings. SNR depends on signal and noise, so this compares complete receiving installations rather than isolating antenna gain.