# Knickebein VHF Beam Propagation: Globe vs Flat ![[ITU_Calc_knickebein_beam_map_telefunken_cycle_1s_mega 1.gif]] Everything that follows is an attempt to explain the picture above. Each yellow equisignal is a Knickebein beam fired over the North Sea at one of the five ranges the manufacturer logged in the July 1939 Telefunken range campaign. The solid pink segment is the region where the globe-model ground-wave calculator still clears the receiver noise floor. The hatched pink segment beyond it is the same beam on the same sphere with the signal now buried in galactic plus thermal noise. The primary source ([[1939_BArch_RL19-6-40_230Q8_App2_Telefunken_Range_Tests|BArch RL 19-6/40 ref. 230Q8 App. 2]]) records reliable reception at every one of those cyan-halo targets out to 1,000 km. A sphere of radius 6,371 km cannot reproduce the reception at the far three. That mismatch is the null hypothesis this document sets out to test. ## Background During the Second World War, Germany deployed a VHF radio navigation system called Knickebein ("crooked leg") to guide bomber aircraft to targets across England. The system used two intersecting Lorenz beams transmitted at 31.5 MHz from stations on the European continent. Bombers flew along one beam and dropped their payload when the second beam crossed the first. The equisignal (the narrow corridor where the dot and dash sub-beams overlap in equal strength) was measured by British monitoring flights at 400 to 500 yards wide at Spalding, 440 km from the transmitter at Kleve. The system was a scaled-up version of the Lorenz blind landing system, which is a line of sight VHF navigation aid designed for airport approaches. Knickebein used the same principle but with a much larger antenna array (99 m wide, 29 m tall) to produce a tighter beam over continental distances. ### How it was discovered None of the evidence that led to the discovery of Knickebein was scientific or mathematical. It was entirely circumstantial: - **March 1940:** A He 111 from Kampfgruppe 26 was shot down. A note recovered from it referred to a radio beacon called "Knickebein." - **Two months later:** Another He 111 from the same unit was shot down. A diary in the wreckage mentioned Knickebein. - **June 1940:** POWs provided details on a bombing apparatus called the X-Gerat. One prisoner suggested Knickebein was similar but had a very narrow beam. - **5 June 1940:** An Enigma signal was intercepted and decrypted, mentioning "Knickebein at Kleve." - **18 June 1940:** Documents recovered from a crashed aircraft in France provided coordinates of two Knickebein stations. - **21 June 1940:** Churchill ordered acceptance of the existence of Knickebein and prioritized countermeasures. - **21 to 22 June 1940:** The Wireless Intelligence and Development Unit (WIDU) detected Knickebein transmissions on 31.5 MHz, confirming the intersecting beam theory. Churchill was not convinced by the science. He was convinced by the weight of circumstantial evidence that R.V. Jones ([[1978_Jones_Most_Secret_War|*Most Secret War*]]) assembled. The scientific establishment was divided. T.L. Eckersley, Britain's leading expert on VHF propagation, initially told intelligence that Knickebein beams could not reach England from Germany at 31.5 MHz. His assessment was based on the expected diffraction loss over a curved Earth surface. [[1937_Eckersley_Ultra_Short_Wave_Refraction_Diffraction]] > [!quote] Eckersley (1937), p. 42 > "As a concession to the wave theory, it is generally admitted that some energy leaks into the shadow region, but the amount of spread is generally left to guesswork." Eckersley was right about the physics on a sphere. He was wrong about what happened in practice. The beams reached England with sufficient strength and precision to guide bombers to their targets. The British measured them, jammed them, and eventually defeated them through countermeasures (Aspirin and Starfish decoys), not through any failure of the beams themselves. [[1978_Jones_Most_Secret_War]] [[2017_Price_Instruments_of_Darkness]] ### The system | Parameter | Value | Source | |---|---|---| | Frequency | 31.5 MHz ($\lambda = 9.517$ m) | Knickebein operational spec, 30 to 33.3 MHz band ([[1943_DLuft_T4058_FuBl_2_Geraete_Handbuch\|D.(Luft) T.4058 §16]], [[1979_Trenkle_Deutsche_Funk_Navigation\|Trenkle 1979]]) | | Transmit power | 3,000 W (34.8 dBW) | Large Knickebein Telefunken spec ([[2024_Doerenberg_Knickebein_Reference\|Dörenberg]] citing [[1939_BArch_RL19-6-40_230Q8_App2_Telefunken_Range_Tests\|BArch RL 19-6/40 ref. 230Q8]]) | | Antenna aperture | 99 m × 29 m | [[2004_Bauer_German_Radio_Navigation_1907_1945\|Bauer (2004)]] p. 12; [[2024_Doerenberg_Knickebein_Reference\|Dörenberg]] from BArch RL 19-6/40 p. 12 | | Directivity | 26.0 dBi | $G = 4\pi A / \lambda^2$ ([[Balanis_Antenna_Theory\|Balanis]], *Antenna Theory*, Ch. 12) | | TX elevation | Kleve 111 m, Stollberg 72 m | christianch.ch Kink leg article citing [[1979_Trenkle_Deutsche_Funk_Navigation\|Trenkle 1979, p. 67]] | | Measured equisignal | $0.066°$ (at Spalding, 440 km from Kleve) | British monitoring flights ([[1978_Jones_Most_Secret_War\|Jones 1978, pp. 100-102, 181]]) | | Aircraft altitude | 6,000 m (19,685 ft) | He 111 operational altitude ([[1978_Jones_Most_Secret_War\|Jones 1978]]; [[2017_Price_Instruments_of_Darkness\|Price 2017]]) | | Audio keying tone | 1,150 Hz | [[1943_DLuft_T4058_FuBl_2_Geraete_Handbuch\|D.(Luft) T.4058]] §16; [[2024_Doerenberg_Knickebein_Reference\|Dörenberg]] | | Airborne receiver | FuBl 2 (EBL 3 + EBL 2) | [[1943_DLuft_T4058_FuBl_2_Geraete_Handbuch\|D.(Luft) T.4058]]; [[Radiomuseum_EBL3_Blindlandegeraet\|Radiomuseum EBL 3]] | | Max rated range | 500 km at 6,500 m altitude | [[2004_Bauer_German_Radio_Navigation_1907_1945\|Bauer (2004)]], p. 12, from German specs | These parameters are independently confirmed by [[2004_Bauer_German_Radio_Navigation_1907_1945|Bauer (2004)]] working from German Telefunken archives, and cross-checked against the Luftwaffe primary-source documentation at the Bundesarchiv-Militärarchiv Freiburg ([[1939_BArch_RL19-6-40_230Q8_App2_Telefunken_Range_Tests|RL 19-6/40]]) as compiled by [[2024_Doerenberg_Knickebein_Reference|Dörenberg]]. ### The paths | Path | Distance | Target area | |---|---|---| | Kleve to Spalding | 440 km | Where the beams were measured | | Stollberg to Beeston/Derby | 694 km | Where the beams guided bombers to targets | The Stollberg beam was detected at Beeston by British monitoring: > "That there is a second beam having similar characteristics but with dots to the north and dashes to the south synchronized with the southern beam, apparently passing through a point near Beeston on a bearing lying between 60° + and less than 104°." > — [[1978_Jones_Most_Secret_War|Jones (1978), *Most Secret War*]] The calculated beam width at Beeston from the measured Kleve divergence ($0.066°$) projected to 694 km: $W = 694{,}000 \times \tan(0.066°) \approx 799$ m $\approx 874$ yd. ### Knickebein Midlands towers-to-targets pairing table Every Knickebein raid into the Midlands required **two simultaneous equisignal beams** crossing over the target, one director and one cross beam. The two operational Knickebein stations during the winter Blitz (Oct 1940 to May 1941) were **Kleve (Kn-4)** and **Stollberg/Bredstedt (Kn-2)**. Kleve is positioned southwest and at slightly shorter range to the Midlands; Stollberg is further north-east and at longer range. Every Midlands target pair therefore reads "Kleve = director, Stollberg = cross" with Stollberg providing the second leg of the intersection. Great-circle distances computed from the station coordinates in `knickebein_paths.csv`: | Target | Kleve → Target (km) | Stollberg → Target (km) | Role | Source | |---|---|---|---|---| | **Derby** (Rolls-Royce Merlin works) | 529.6 | 710.1 | Confirmed operational target, principal aim-point for Kleve + Stollberg Midlands raids | [[1978_Jones_Most_Secret_War\|Jones 1978, p. 182]]; [[2017_Price_Instruments_of_Darkness\|Price 2017, p. 38]] | | **Birmingham** | 550.7 | 753.8 | Confirmed operational target (e.g. 26 Oct 1940 raid) | [[2017_Price_Instruments_of_Darkness\|Price 2017, p. 38 caption]] | | **Liverpool** | 639.6 | 790.6 | Confirmed operational target during the early Liverpool Blitz (Nov 1940 to Feb 1941, before Y-Gerät took over the May 1941 raids) | [[2017_Price_Instruments_of_Darkness\|Price 2017, p. 35]] | | **Spalding** | 439.6 | 632.7 | Bufton/Mackey Anson flight on 21 Jun 1940 measured the Kleve equisignal here (400 to 500 yd width). Spalding was not an operational target; it was chosen as the flight-detection site because the Kleve director beam was calculated to pass over it | [[1978_Jones_Most_Secret_War\|Jones 1978, pp. 181 to 182]] | | **Beeston** | 512.6 | 693.5 | Bufton's same 21 Jun 1940 flight detected the Stollberg cross beam near Beeston, flying north from Spalding along the Kleve director line. Beeston was not an operational target; it was a cross-beam detection event | [[1978_Jones_Most_Secret_War\|Jones 1978, p. 181]] | | **Retford** | 511.9 | 663.6 | Luftwaffe Enigma decrypts on 5 Jun 1940 named Retford as the pointing coordinate of a Knickebein beam. Retford is not a confirmed target or a measurement site; it is an Enigma-derived coordinate that helped British intelligence locate and confirm the Kleve beam before Bufton's flight | [[1978_Jones_Most_Secret_War\|Jones 1978, p. 163]] | **Reasoning on specifics:** 1. **Only Kleve and Stollberg were operational Large Knickebein stations during the winter Blitz.** [[2024_Doerenberg_Knickebein_Reference|Dörenberg's station database]] lists Kn-2 (Stollberg) and Kn-4 (Kleve) as the only two Large Knickebein units built before the Battle of Britain. Kn-6 through Kn-11 (French-coast sites) came online later and were never used for Midlands raids. Kn-12 (Maulburg) pointed at Italy and was irrelevant. The pair is therefore Kleve + Stollberg by elimination. 2. **Kleve is always the director, Stollberg the cross beam.** Kleve sits at 51.79°N, 6.10°E; Stollberg at 54.64°N, 8.94°E. For every Midlands target the Kleve bearing is shorter (by 170 to 180 km) and the Stollberg bearing is the longer cross. The director is always the shorter leg because the bomber flies along the director and drops when the cross intersects it. 3. **The three confirmed operational targets (Derby, Birmingham, Liverpool) are cited directly in Jones (1978) and Price (2017).** For Derby and Birmingham the raid histories and station pairings are explicit. For Liverpool, Price (2017) p. 35 lists it as a Knickebein-era target during the early Liverpool Blitz; by May 1941 Y-Gerät had taken over (see [[X_Y_Gerat_Propagation_Null#Y-Gerät]] and [[YG]]). 4. **Spalding and Beeston are measurement sites, not targets.** Bufton's 21 Jun 1940 Anson flight flew a search pattern following the expected Kleve beam path. Spalding is where the Kleve equisignal was physically measured (400 to 500 yd width at 5,800 m altitude), and Beeston is where the Stollberg cross beam was detected on the same flight. Neither site was bombed. 5. **Retford is an Enigma intercept coordinate.** A Luftwaffe signal decrypted on 5 Jun 1940 named Retford in connection with a Knickebein station assignment. This gave British intelligence a pre-Bufton hint on where Kleve was aimed. Retford was not a target and the flight did not visit it. It is included here for completeness because the distance from Kleve (511.9 km) is within 0.7 km of the distance to Beeston (512.6 km), meaning the Kleve-beam propagation graphs show both at essentially the same point. 6. **No Kleve → Liverpool or Stollberg → Spalding "pairing" is separately documented.** The operational record names Kleve + Stollberg as the pair; specific raid dates do not always list station assignments. The per-target distances above are computed from station and target coordinates. The Kleve → Liverpool leg at 639.6 km and the Stollberg → Spalding leg at 632.7 km are therefore **inferred from the two-beam requirement**: if Knickebein guided a bomber to Liverpool, both beams must have been present at Liverpool, and the Kleve director beam travelling 639.6 km is the only geometric option for the director leg. Same reasoning for Stollberg → Spalding as the inferred cross beam for the Bufton measurement flight. ### Prior X-beam trials: range and accuracy confirmed Before Knickebein, the X-beam programme (66 to 77 MHz) ran extensive trials. Bauer (2004, p. 11) reports two results: > "It was found that **ranges up to 500 km could be achieved at flying altitudes of 6000 m.**" > "First practical trials in December 1936 proved that it was possible to get most of the dropped bombs within a square of **300 m by 300 m over a guidance range of 350 km**." 300 m targeting accuracy at 350 km requires a coherent, narrow beam maintaining its geometry over the full path. A diffracted signal smeared across a 47.8 km creeping wave ribbon does not produce 300 m precision. The earlier X-beam results at similar frequencies and ranges establish that VHF guidance beams were operationally proven at these distances before Knickebein was even built. Bauer also notes (p. 10) that the early 5 degree beam aperture produced an 8 km wide beam at 100 km. This is simple plane trigonometry: $W = d \times \tan(5°) \approx 8.7$ km. The beam width scaled linearly with distance, which is what rectilinear propagation predicts. If the beam were diffracting around a curved surface, the spreading would follow the Fock diffraction pattern (exponential decay), not linear trigonometric scaling. --- ## Null hypothesis **H₁:** The Knickebein signal at the equisignal crossover is above the receiver noise floor (galactic + thermal) on a spherical Earth of radius 6,371 km for every confirmed operational target. **H₀:** The Knickebein signal at the equisignal crossover is below the receiver noise floor on a spherical Earth of radius 6,371 km at one or more confirmed operational targets. Meaning the receiver cannot pick up the signal and the two-beam intersection cannot form. If H₁ holds, the beam works on a globe and the shape of the surface is not constrained by this observation. If H₀ holds, the beam cannot function on a globe because no signal reaches the antenna. Since the beams were physically detected, measured, and operationally used to guide aircraft to Derby, Birmingham, Liverpool, and Beeston across the 1940–41 Night Blitz (roughly September 1940 through May 1941, approximately eight months of mass-bomber night raids), the system demonstrably functioned. A surface geometry that prohibits the observed function is falsified by the observation. We compute the flat surface (rectilinear) SNR alongside the globe model at each path to show that the observed performance is quantitatively consistent with unobstructed propagation. **Critical framing:** Knickebein is a two-beam system. Both the director beam (Kleve) and the cross beam (Stollberg) must deliver signal at the target simultaneously for the intersection to exist. A director beam without a cross beam gives the pilot a line to fly along but no release point. The weakest link of each station pair determines whether the system works. **Why the receiver noise floor is the right threshold.** Once RF signal is present at the receiver antenna above the internal noise floor, the FuBl 2 / EBL 3 airborne set amplifies it through three IF stages and a two-stage AF amplifier, with full automatic gain control on the RF and first two IF stages, a manual sensitivity knob (Empfindlichkeitsregler) accessible to the operator, and an automatic volume control on top (see "Detection criterion" section below and [[1943_DLuft_T4058_FuBl_2_Geraete_Handbuch|D.(Luft) T.4058 §24]]). Primary-source German bomber crew testimony (Bundesarchiv ref. 230Q18) says the beam signals were *"easily heard"* in operational He 111 cockpits, even through British low-power jamming. The only question is whether the RF signal physically exists at the antenna above the noise floor at the target range and altitude. --- ## Units and conventions: why field strength in μV This document expresses all final received-signal results in **microvolts (μV)** at the receiver's 50 Ω input, not in dB. The reason is operational. The FuBl 2 receiver is sensitivity-limited: it has to discriminate the Knickebein carrier out of the same antenna voltage that contains galactic noise plus thermal noise. The relevant question at the antenna terminals is *"how many microvolts of Knickebein arrive, and how many microvolts of noise are arriving at the same instant?"* That question is answered in volts, not in decibels. The 31.5 MHz noise floor at the receiver input, computed from ITU-R P.372-16 Eq. 14 ($F_{am} = 52 - 23 \log_{10}(f_{MHz}) = 17.5$ dB) plus the 1940s vacuum-tube receiver noise figure (~10 dB) at a **500 Hz matched-filter detection bandwidth** for the MCW / A2 Knickebein emission (1,150 Hz audio tone with dot-dash keying), is: $V_{noise} \approx 0.079 \, \mu V \text{ at the 50 Ω input}$ This is the reference all field-strength tables and graphs in this document are drawn against. The horizontal noise-floor line on every per-station sweep and master bar chart in `make_p526_vs_p368_graphs.py` is at 0.079 μV. Anything sitting on or above that line is operationally detectable; anything below it is buried in galactic + thermal noise and cannot be recovered by amplification. dB still appears in this document where it is the natural unit: free-space path loss accumulates in dB along a path, Fock diffraction loss is given in dB, antenna directivity is in dBi. Where dB is used in the math walkthroughs, the corresponding μV value at the receiver input is shown alongside or stated immediately after the dB figure. For quick conversion between the two conventions: $V_{rx}(\mu V) = V_{noise} \times 10^{SNR(dB)/20} = 0.079 \times 10^{SNR/20}$ So an SNR of +20 dB means $V_{rx} = 0.079 \times 10 = 0.79 \, \mu V$. An SNR of −20 dB means $V_{rx} = 0.0079 \, \mu V$. The "USABLE / MARGINAL / DEAD" classification used throughout this document is just this conversion applied to the +10 dB and 0 dB SNR thresholds: | Classification | $V_{rx}$ at receiver input | Equivalent SNR | Meaning | |---|---|---|---| | **USABLE** | $\geq 0.25 \, \mu V$ | $\geq +10$ dB | Signal clearly above galactic noise, FuBl 2 AGC delivers an audible tone | | **MARGINAL** | $0.079$ to $0.25 \, \mu V$ | $0$ to $+10$ dB | Signal at or near the noise floor, detection uncertain | | **DEAD** | lt; 0.079 \, \mu V$ | lt; 0$ dB | Signal below the thermal + galactic noise floor; no amplification recovers it | Every per-path classification later in this document refers to these three field-strength bands. --- ## ITU standards and formulas used The analysis combines four propagation models plus the ITU-R noise-floor calculation to produce a field strength in μV at the 50 Ω receiver input for every Knickebein path. Derivations, variable definitions, and worked examples are in the [Amendments](#amendments) at the bottom of this document. The summary below names each model, states what it computes, and links to its Amendment. - **Sommerfeld-Norton flat-Earth** → the rigorous flat-surface solution from the ITU Handbook on Ground Wave Propagation (2014) Part 1 §3.2.1, summing the direct ray, the Fresnel-reflected ray, and the Norton surface-wave attenuation function. This is the flat-Earth comparison model plotted **green** on every graph in the body. - **ITU-R P.526-16 Fock smooth-Earth diffraction** → [Amendment B](#amendment-b-fock-smooth-earth-diffraction-itu-r-p526-16). The globe model. Computes the exponential field-strength decay past the geometric horizon caused by the signal having to bend around a 6,371 km sphere. Plotted **magenta**. [[2025_ITU-R_P526-16_Diffraction]] - **ITU-R P.368 GRWAVE (LFMF-SmoothEarth)** → ITU's reference Fortran implementation of the full multi-mode residue series. Independent cross-check of P.526; discussed in detail in [[GRWAVE_P368_BotB]]. Full methodology and per-parameter justification live there. - **Friis free-space path loss** → [Amendment A](#amendment-a-friis-free-space-path-loss). Rectilinear shape-agnostic link budget. Used in the Amendments only as the baseline for deriving Sommerfeld-Norton and Fock in dBW link-budget terms; not plotted in the body. - **Equisignal crossover loss** → [Amendment D](#amendment-d-equisignal-crossover-loss-5-squint). 19.87 dB below the sub-beam peak at the Telefunken 5° squint. Applied uniformly to every field-strength value in this document because the pilot flies the equisignal corridor, not the sub-beam peak. - **ITU-R P.372-16 noise floor** → [Amendment E](#amendment-e-noise-floor-calculation-itu-r-p372). Galactic + thermal receiver noise at 31.5 MHz over a 3 kHz audio detection bandwidth. 0.079 μV reference line on every graph. - **Surface electrical constants (K, β)** → [Amendment F](#amendment-f-surface-electrical-constants-at-315-mhz). The $\varepsilon$ and $\sigma$ values for land, sea, and medium ground that feed the P.526 β parameter and drive the Fock height-gain function. - **Formulas reference table** → [Amendment G](#amendment-g-formulas-reference-table). Every equation in the analysis, with source citations, in one table. All implementations live in `botb_itu_analysis.py` (the physics library) and `make_p526_vs_p368_graphs.py` (the graph generator). See the [Source Code](#source-code) section for the git link and file inventory. --- ## Geometry on a globe On a spherical Earth with effective radius $R_{eff} = \frac{4}{3} \times 6{,}371 = 8{,}495$ km (standard atmospheric refraction, ITU-R P.453), a horizontal beam from 200 m elevation rises above the curved surface: ### Kleve to Spalding (440 km) | Parameter | Ground (10 m) | Aircraft (6,000 m) | |---|---|---| | Curvature drop $d^2/(2R_{eff})$ | 11,372 m | 11,372 m | | TX radio horizon $\sqrt{2R_{eff}h_{tx}}$ | 58.3 km | 58.3 km | | RX radio horizon | 13.0 km | 319.3 km | | Total line of sight range | 71.3 km | 377.6 km | | **Shadow distance** (beyond LoS) | **368.2 km** | **62.0 km** | The beam centre is 11.6 km above the surface at 440 km. Even with an aircraft at 6 km, 62 km of the path is beyond radio line of sight. ### Stollberg to Beeston (694 km) | Parameter | Ground (10 m) | Aircraft (6,000 m) | | ------------------------------- | ------------- | ------------------ | | Curvature drop $d^2/(2R_{eff})$ | 28,312 m | 28,312 m | | Shadow distance | **622.2 km** | **316.0 km** | The beam centre is 28.5 km above the surface. With an aircraft at 6 km, 316 km of the 694 km path (46%) is beyond radio line of sight. --- ## Fock smooth-Earth diffraction (results) Full derivation of the residue series, the $F(X)$ distance term, the $G(Y)$ height gain term, and the $\beta$ polarisation parameter is in [Amendment B](#amendment-b-fock-smooth-earth-diffraction-itu-r-p526-16). The constants that depend only on frequency and Earth radius (and not on the path) are also in Amendment B. Two numbers from there flow directly into the results tables below: - **Creeping wave ribbon height**: 47.8 km at 31.5 MHz, $R_{eff} = 8{,}495$ km. Vertical elevation extent of the wave that hugs the surface after diffracting past the horizon. - **Surface attenuation rate**: 0.292 dB/km at 31.5 MHz. The exponential decay rate per kilometre of path past the geometric horizon, from the first Airy zero of the Fock residue series. Plugged into the ITU-R P.526-16 Eq. 13 ($20 \log_{10}(E/E_0) = F(X) + G(Y_1) + G(Y_2)$) for the two benchmark Knickebein paths at aircraft altitude: ### Kleve to Spalding (440 km) | Parameter | Value | |---|---| | $X$ (normalised distance) | 7.29 | | $Y_1$ (Kleve TX, 111 m) | 0.52 | | $Y_2$ (aircraft, 6,000 m) | 28.09 | | $F(X)$ | −108.6 dB | | $G(Y_1)$ | −5.5 dB | | $G(Y_2)$ | +76.3 dB | | **Total diffraction loss** | **37.8 dB** | ### Stollberg to Beeston (694 km) | Parameter | Value | |---|---| | $X$ (normalised distance) | 11.33 | | $Y_1$ (Stollberg TX, 72 m) | 0.34 | | $Y_2$ (aircraft, 6,000 m) | 28.09 | | $F(X)$ | −177.9 dB | | $G(Y_1)$ | −9.6 dB | | $G(Y_2)$ | +74.8 dB | | **Total diffraction loss** | **112.8 dB** | At ground-antenna heights the $G(Y_1)$ TX contribution is negative because the transmitter sits inside the 47.8 km creeping-wave ribbon rather than above it (Kleve loses 5.5 dB, Stollberg 9.6 dB). The entire positive height budget comes from the aircraft at 6,000 m, which earns +76.3 dB. The distance term $F(X)$ is dominated by the $-17.6 X$ residue-series slope in the deep shadow (see [Amendment B](#amendment-b-fock-smooth-earth-diffraction-itu-r-p526-16) for the closed form), which is the exponential decay that makes the Stollberg a non-viable path on a globe. --- ## Signal budget Free space spreading uses the ITU-R P.525 link equation. Noise at 31.5 MHz is dominated by galactic synchrotron radiation from the Milky Way (ITU-R P.372 curve E, $F_a = 18$ dB) above the 1940s receiver noise figure of 10 dB. At 500 Hz detection bandwidth the total noise floor is 0.079 μV at the 50 Ω receiver input. > [!note] ITU-R P.372 galactic noise, not man-made noise > An earlier draft used $F_a = 32$ dB, corresponding to ITU-R P.372 curve B (residential man-made noise). That was incorrect for an aircraft at 6,000 m altitude over rural England at night in 1940. Man-made noise from ground sources is heavily attenuated at operational altitude and was minimal in wartime blackout conditions. Galactic synchrotron radiation from the Milky Way is the correct dominant contribution, since it comes from above. The corrected galactic value is $F_a = 18$ dB (ITU-R P.372-16 Eq. 14 at 31.5 MHz), giving a noise floor of $-159.0$ dBW. This is 14 dB lower than the old value, which raises every SNR in this section by 14 dB. ### At beam peak Values below are the canonical output of the analysis library with `rx_gain = 0 dBi`, matching the headline master bar chart above and the per-station bar charts in the Full ITU certified graph set to within display rounding. | Quantity | Sommerfeld-Norton FE (Kleve → Spalding) | Fock globe (Kleve → Spalding) | Fock globe (Stollberg → Beeston) | |---|---|---|---| | Free-space spreading | 115.3 dB | 115.3 dB | 119.2 dB | | Ground / diffraction factor | +4.0 dB (SN surface coupling) | −38.1 dB (Fock first-mode residue) | −92.9 dB (Fock residue, β = 0.81 sea clamp) | | $V_{rx}$ at 50 Ω (peak) | **21,175 μV** | **165.9 μV** | **0.191 μV** | | Ratio to 0.079 μV noise floor | 268,000× above | 2,100× above | 2.4× above | On the flat-Earth Sommerfeld-Norton solution the Kleve beam peak arrives at 21,175 μV, more than five orders of magnitude over the 0.079 μV noise floor. On the globe Fock solution the Kleve peak is 165.9 μV because the 6,000 m aircraft height gain partially offsets the residue-series distance term. Stollberg → Beeston, 255 km further back, drops to 0.191 μV at globe-model beam peak (only 2.4× above the noise floor at peak); the β = 0.81 over-sea correction prevents it from falling further at the peak, but the equisignal crossover loss in the next section puts it firmly below noise. > [!note] What if the transmitter was more powerful? > Bauer (2004, p. 12) documents an 80 kW VHF transmitter (type S. 561a) in German archives, operating in the 30 to 34 MHz band. Substituting it for the 3 kW Large Knickebein multiplies every receiver voltage in this section by 5.16×. On the globe Fock model, the Stollberg → Beeston beam-peak voltage rises from 0.191 μV to 0.985 μV, and the equisignal voltage rises from 0.0194 μV to 0.1 μV, which just squeaks over the 0.079 μV noise floor. The three longer Stollberg paths do not: Derby equisignal rises to 0.062 μV (still 1.3× below noise), Birmingham to 0.019 μV (4.2× below), Liverpool to 0.0067 μV (12× below). The exponential shadow-zone decay outpaces any linear power boost. The flat-Earth Sommerfeld-Norton solution has headroom to spare at any power level. The globe model remains broken for the longer Stollberg equisignal paths even at 80 kW. [[2004_Bauer_German_Radio_Navigation_1907_1945]] ### At the equisignal crossover The pilot does not fly at the beam peak. The pilot flies at the equisignal, where each sub-beam (dot or dash) is well below its own peak. The crossover depth depends on the squint angle between the two sub-beams. The squint angle is an operator-adjustable parameter. This creates a trade-off: a deeper crossover gives a narrower equisignal but a weaker signal. The Sommerfeld-Norton flat-Earth solution keeps tens of thousands of microvolts of headroom at Kleve → Spalding across every sensible squint choice. The Fock globe solution has only 16.85 μV of signal at the same 5° Knickebein squint, which is enough for Kleve but cannot survive another station's worth of additional path loss. | Squint | Crossover loss | Sommerfeld-Norton $V_{rx}$ (Kleve → Spalding) | Fock globe $V_{rx}$ (Kleve → Spalding) | Equisignal width | |---|---|---|---|---| | 1.0° | −0.5 dB | 20,000 μV | 157 μV | 8,712 yd | | 2.0° | −2.0 dB | 16,820 μV | 132 μV | 4,052 yd | | 3.0° | −4.8 dB | 12,170 μV | 95.4 μV | 2,334 yd | | 3.5° | −6.8 dB | 9,660 μV | 75.7 μV | 1,775 yd | | 4.0° | −9.6 dB | 7,010 μV | 54.9 μV | 1,304 yd | | 4.5° | −13.5 dB | 4,470 μV | 35.0 μV | 876 yd | | **5.0°** | **−19.87 dB** | **2,150 μV** | **16.85 μV** | **~500 yd** | For Kleve → Spalding, the globe model clears the noise floor at every squint angle. The aircraft height gain is big enough to rescue the director beam at this specific 440 km distance. Bufton's Anson flight in June 1940 physically measured the 400 to 500 yd equisignal here; the globe-model equisignal at the 5° squint (16.85 μV, 213× above the 0.079 μV noise floor) is consistent with a clean detection, which matches the historical record. The falsification does not come from Kleve → Spalding. It comes from every other Midlands path. For Stollberg → Beeston (694 km), the globe beam-peak signal is already only 2.4× above the noise floor at peak. At the 5° squint that produces a 500 yd equisignal, the signal drops to 0.0194 μV, which is 4× below the 0.079 μV noise floor. The receiver hears galactic hiss, not a tone. The Stollberg cross beam cannot form an intersection with the Kleve director beam over Derby, Birmingham, Liverpool, or Beeston on a sphere. Its equisignal signal simply does not exist at those distances. --- ## Detection criterion: does the signal exist at the antenna? The only question that matters for the null hypothesis is whether the Knickebein signal is physically present at the receiver antenna above the thermal + galactic noise floor. Everything else is solved by the receiver electronics once the signal is present. ### The FuBl 2 / EBL 3 has full AGC and operator gain control From [[1943_DLuft_T4058_FuBl_2_Geraete_Handbuch|D.(Luft) T.4058]] §24 *"Schaltung des EBL 3"* (circuit of the EBL 3 receiver), the airborne Knickebein receiver is not a passive envelope detector. Its gain architecture is: - **1 RF amplifier stage** (Hochfrequenz-Verstärkerstufe) with image-frequency trap - **1 mixer stage** driven by a local oscillator - **3 IF amplifier stages** at 6 MHz intermediate frequency (Zwischenfrequenz-Verstärkerstufen) - **1 detector stage** (Gleichrichterstufe) - **2-stage audio-frequency amplifier** in the EBL 2 unit, with switchable gain (single-stage for the EZ glow-lamp marker, two-stage for audio reception of the AFF director beam) - **Automatic gain control** acting on the RF and first two IF stages via grid-bias regulation - **Manual sensitivity control** (Empfindlichkeitsregler) adjusting the screen-grid voltage of the RF and first IF stages through the W42 potentiometer at the Pegelregler - **Automatic volume control** (*"selbsttätige Lautstärkeregelung"*) on top, explicitly added to compensate for level variations as the aircraft range changes - **Separate manual volume knob** (Lautstärkeregler) for the operator From the manual (§24, page 14 of the OCR): > *"Regelanordnung zur selbsttätigen Regelung der Verstärkung der Hochfrequenz- sowie ersten und zweiten Zwischenfrequenz-Verstärkerstufe... Ein Empfindlichkeitsregler wirkt durch Verstärkungsänderung der Hochfrequenz- und Ersten Zwischenfrequenz-Verstärkerstufe... Zum Ausgleich von Lautstärkeänderungen ist zusätzlich eine selbsttätige Lautstärkeregelung vorgesehen."* **What this means operationally.** Once the RF signal exceeds the receiver's bare detection threshold (signal above the internal thermal + galactic noise floor), the AGC and the manual sensitivity knob can push the gain chain to whatever level is needed to drive the 2 kΩ Dfh.b magnetic earphones to audible SPL at the pilot's ear. The cockpit acoustic environment is not a bottleneck as long as the signal is physically present at the antenna, because the receiver electronics have enough gain in the chain (RF + 3× IF + 2× AF, with AGC and ALC trimming the response) to deliver full audio output across a wide range of RF input levels. ### Primary-source testimony: German bomber crews said the signals were "easily heard" Bundesarchiv ref. 230Q18 is the primary file; the interviews and reasoning are published in [[2003_Price_Wizard_War_RAFHS28|Price (2003) *A New Look at the Wizard War*]] pp. 17 to 20. The two key passages: > "The 'official' British line on the jamming of Knickebein is that the German system soon fell into disuse, ASPIRIN was so effective in disrupting the beams. Having spoken to several German bomber crewmen who flew over Britain at that time, I would assert that while Knickebein did indeed fall into disuse, it was not because ASPIRIN was particularly effective in disrupting the radio beams. In fact, crews have told me that **usually it was quite easy to hear the Knickebein signals through the jamming**." > — Price (2003), p. 17 > "The modified hospital diathermy sets gave a jamming output of 150 watts, while ASPIRIN gave an output of 500 watts. In each case, the jamming was radiated from an omnidirectional aerial. As many people in this audience will readily appreciate, such levels of power were inadequate to provide effective jamming against Knickebein signals radiated at 500 watts and boosted by a high gain directional aerial array." > — Price (2003), p. 17 The Pfadfinder (KGr 100) testimony about X-Gerät is on p. 20: > "I asked one radio operator how much trouble he had had from the British jamming of X-Gerät. He said he did not remember any jamming... None of them remembered hearing the British Morse dashes." > — Price (2003), p. 20 Three things follow for the audibility question the null hypothesis is testing. **First, the crews discriminated the real beam against a much stronger local interferer.** The Aspirin jammer radiated omnidirectionally from a ground station located inside Britain near the target area, at a few kilometres to tens of kilometres from the bomber during its run-in. The Knickebein transmitter was at Kleve or Stollberg, 500 to 700 km away. The raw inverse-square signal-strength geometry at the bomber favours the nearby jammer by roughly 40 dB over the distant beam before any directivity is applied. Despite that, the operators heard the Knickebein signal "easily" through the jamming. That is direct operational evidence that the Knickebein field strength at the FuBl 2 antenna was high enough to dominate a substantially closer omnidirectional interferer. **Second, the engineering reason Price gives is the directivity ratio at the receiver.** Knickebein's 500 W (or 3,000 W for the Large variant at Kleve and Stollberg) was radiated from a high-gain directional array with 26 dBi peak directivity. Aspirin's 500 W came from an omnidirectional aerial, spreading the power over $4\pi$ steradians. The ratio of directed-beam power density at the receiver to omnidirectional-jammer power density determines whether the jamming masks the beam or leaves it intact. The fact that the crews report the Knickebein tone as easily heard is empirical confirmation that this directivity ratio left the beam intact at the receiver, which is only possible if the beam's RF field strength at the bomber's antenna was a real, detectable signal above the combined jammer + cockpit + receiver noise environment. **Third, the crews could discriminate fine tonal differences in the audio.** The Aspirin dashes were not synchronised to the real Knickebein keying, so they produced wailing beat tones that a trained operator could distinguish from the steady 1,150 Hz equisignal note. If cockpit noise had masked audio discrimination, neither the real beam reception nor the detection of unsynchronised jamming would have worked. The operators did both, routinely, across the Night Blitz operational campaign. No primary Luftwaffe source complains about cockpit noise preventing Knickebein reception. The operational problems that ultimately degraded Knickebein were British jamming plus the crews' concern that the jammers revealed the beam axes to RAF night fighters, not any acoustic limit inherent to listening to the audio tones in a He 111 cockpit. ### Therefore the threshold is the RF noise floor The operational criterion for the BotB null hypothesis is: **is the globe-model equisignal field strength at the receiver input above the 0.079 μV ITU-R P.372 galactic + thermal noise floor?** If yes, the receiver electronics amplify the signal, the crew hears the tone, the system functions. If no, no amount of gain creates signal out of thermal noise. The USABLE / MARGINAL / DEAD bands are defined in the "Units and conventions" section above and reproduced here in field-strength terms: | Classification | $V_{rx}$ at 50 Ω input | Equivalent SNR | Meaning | |---|---|---|---| | **USABLE** | $\geq 0.25 \, \mu V$ | $\geq +10$ dB | Clearly above galactic noise, FuBl 2 AGC delivers audible tone | | **MARGINAL** | $0.079$ to $0.25 \, \mu V$ | $0$ to $+10$ dB | At or near noise floor, detection uncertain | | **DEAD** | lt; 0.079 \, \mu V$ | lt; 0$ dB | Below galactic + thermal noise floor, no amplification recovers it | Applying this to the Kleve + Stollberg operational paths. The field strength $V_{rx}$ in the table below is the **equisignal** value at the receiver input, the voltage the pilot's FuBl 2 actually sees while riding the steady-tone corridor (19.87 dB below the sub-beam peak for the 5° Knickebein squint geometry). Values are the post-β-fix output of the canonical analysis library with `rx_gain = 0 dBi`, matching the per-station equisignal sweeps in the [Full ITU certified graph set](#full-itu-certified-graph-set). | Path | Equisignal $V_{rx}$ | Noise floor | Equisignal SNR | Status | |---|---|---|---|---| | Kleve → Spalding (440 km) | **16.85 μV** | 0.079 μV | +46.5 dB | **USABLE** (Bufton's physical measurement confirms) | | Kleve → Retford (512 km) | **1.39 μV** | 0.079 μV | +24.8 dB | **USABLE** | | Kleve → Derby (529 km) | **0.77 μV** | 0.079 μV | +19.7 dB | **USABLE** | | Kleve → Birmingham (550 km) | **0.37 μV** | 0.079 μV | +13.4 dB | **USABLE** | | Stollberg → Beeston (694 km) | **0.0194 μV** | 0.079 μV | −12.3 dB | **DEAD**, 4× below noise floor | | Stollberg → Derby (711 km) | **0.0121 μV** | 0.079 μV | −16.4 dB | **DEAD**, 6.5× below noise floor | | Stollberg → Birmingham (754 km) | **0.00363 μV** | 0.079 μV | −26.8 dB | **DEAD**, 22× below noise floor | | Stollberg → Liverpool (791 km) | **0.00129 μV** | 0.079 μV | −35.8 dB | **DEAD**, 61× below noise floor | All four Stollberg Midlands paths sit below the 0.079 μV galactic + thermal noise floor at the equisignal corridor, by factors of 4× to 61×. All four Kleve paths clear the +10 dB detection threshold. Cross-beam intersection requires both stations to deliver signal at the same target; the Stollberg column alone vetoes the system on a sphere. > [!critical] The falsification rests on Stollberg > Every confirmed operational intersection target in the Midlands required **both** the Kleve director beam and the Stollberg cross beam. The Kleve beam delivers 0.37 to 16.85 μV at every Midlands target on the globe, all above the +10 dB detection threshold. The Stollberg cross beam delivers between 0.00129 μV and 0.0194 μV at the equisignal crossover at those same targets on the globe, which is 4× to 61× below the 0.079 μV galactic + thermal noise floor. No AGC chain, no volume knob, no sensitivity setting creates a signal where none exists. No 1,150 Hz tone can be amplified out of pure galactic hiss. Without the Stollberg cross beam, no two-beam intersection forms, and no precision bombing is possible. On a globe, the Knickebein two-beam system has no operational intersection anywhere in the Midlands. --- ## Sommerfeld-Norton flat-Earth vs ITU Fock globe diffraction The two propagation models plotted green and magenta on every graph in the body of this document are the Sommerfeld-Norton rigorous flat-Earth solution (ITU Handbook on Ground Wave Propagation 2014 Part 1 §3.2.1, direct plus ground-reflected plus surface-wave terms) and the ITU-R P.526-16 Fock smooth-sphere diffraction (residue series past the radio horizon). Both run on the same ground electrical parameters (seawater $\sigma = 5$ S/m, $\varepsilon_r = 70$ for Stollberg; average land $\sigma = 5 \times 10^{-3}$ S/m, $\varepsilon_r = 15$ for Kleve), the same 26 dBi directional transmitter, the same 6,000 m aircraft receiver, and the same 0.079 μV receiver noise floor. The only physical difference between the green curve and the magenta curve is the shape of the surface between the transmitter and the receiver: an infinite flat plane on one side, a sphere of radius 6,371 km on the other. The equisignal voltages at the 50 Ω receiver input for the two benchmark paths (all values in microvolts): | Model | Kleve → Spalding peak | Kleve → Spalding equisignal | Stollberg → Beeston peak | Stollberg → Beeston equisignal | | ---- | ---- | ---- | ---- | ---- | | **Sommerfeld-Norton FE** (green) | 21,175 μV | 2,150 μV | 3,958 μV | 402 μV | | **ITU-R P.526 Fock globe** (magenta) | 165.9 μV | 16.85 μV | 0.191 μV | 0.0194 μV | The Sommerfeld-Norton flat-Earth equisignal holds at 2,150 μV on the Kleve director leg and 402 μV on the Stollberg cross leg, which is between 5,000× and 27,000× above the 0.079 μV noise floor. The Fock globe equisignal clears the noise floor at Kleve → Spalding by a factor of 213× (16.85 μV versus 0.079 μV, USABLE) and sits 4× below the noise floor at Stollberg → Beeston (0.0194 μV versus 0.079 μV, DEAD). The gap between the two rows is 128× on the Kleve path and 20,700× on the Stollberg path. The Stollberg gap is the falsification: the flat-Earth solution delivers an audible signal at the antenna input, the Fock sphere does not. --- ## Equisignal width: flat model consistency check The British measured the beam divergence at Spalding as $0.066°$. On a flat surface, this angle projects to a width at range $d$: $W_{flat} = d \times \tan(0.066°) = 439{,}541 \times \tan(0.066°) \approx 506 \text{ m} \approx 554 \text{ yd}$ The measured equisignal was 400 to 500 yd. The measured divergence angle, projected rectilinearly over the measured distance, gives the measured width. This is a consistency check, not an independent prediction: the divergence angle came from the same measurement. But the consistency confirms that no bending, diffraction, or atmospheric correction is needed to connect the transmitter geometry to the received beam width. A straight line from the antenna to the measurement point, at the measured angle, gives the measured width. On a globe, this straight line path does not exist. The beam must diffract around the curved surface, and the diffraction mechanism (creeping waves) smears the signal in the elevation dimension by 47.8 km. The horizontal beam pattern is approximately preserved through diffraction (the horizon presents no obstruction in the azimuthal direction), but the signal strength in the shadow zone is far below the noise floor (see signal budget above). --- ## Eckersley was right about the physics Eckersley's 1937 paper calculated the diffraction loss for VHF signals beyond the radio horizon. His curves (Figs. 6 to 9) show field strength dropping steeply for wavelengths of 2 to 10 m, consistent with the Fock analysis computed here. His initial assessment to intelligence that Knickebein beams could not reach England from Germany was a correct application of spherical Earth diffraction theory. [[1937_Eckersley_Ultra_Short_Wave_Refraction_Diffraction]] The establishment overruled him, not because his physics was wrong, but because the beams were physically detected. The beams existed. The physics on a sphere said they should not. That tension was never resolved during the war. The beams were jammed, and everyone moved on. --- ## Independent validation: US Air Force SEKE propagation code The Fock residue series is not an obscure theoretical exercise. It is the equation that the US Air Force uses to predict radar propagation beyond the horizon. MIT Lincoln Laboratory implemented Fock's series in their SEKE (Spherical Earth Knife Edge) radar propagation code under Air Force contract F19628-85-C-0002. [[1988_Shatz_Spherical_Earth_Diffraction_SEKE]] Shatz and Polychronopoulos (1988) present their implementation (Eq. 1, p.5): $F(x,y,z) = 2\sqrt{\pi x} \sum_{n=1}^{\infty} f_n(y) f_n(z) \exp\left[\frac{1}{2}(\sqrt{3}+i)a_n x\right]$ This is the same equation as Fock (1965), Eq. (6.10), p. 209, in different notation. Same Airy zeros ($a_n = \tau_s$), same exponential decay, same residue series. Their Figure 4 computes F$^2$ at 167 MHz (VHF) with antenna height 10 m and target height 6000 m (same as He 111 operational altitude). The corrected Fock curve shows: | Range | F$^2$ (propagation factor) | | --------------------------- | -------------------------- | | 200 km | ~0 dB (near free space) | | 350 km (at optical horizon) | ~-25 dB | | 425 km | ~-80 dB | | 500 km | ~-110 dB | At 167 MHz (higher frequency = less diffraction loss than 31.5 MHz), the signal is already -80 dB at 425 km. Our analysis at 31.5 MHz gives -77 dB diffraction loss at 440 km. The numbers are consistent: lower frequency means slightly more loss at the same distance. > [!note] The US Air Force could not get the computation right on the first try > MIT's original subroutine (SPH35) used polynomial fits to approximate the Airy function and produced errors up to 28.65% at 5000m target height. The code would "diverge" near the optical horizon and fall back to incorrect values. They had to rewrite the entire Airy function evaluation (SPH35N) using Taylor series and Gaussian quadrature integration. The math is hard enough that MIT Lincoln Lab needed a dedicated technical report to fix it. The physics, however, is unambiguous: VHF signals decay exponentially past the optical horizon on a sphere. --- ## Independent validation: ITU-R P.368 GRWAVE ground-wave calculator The ITU-R diffraction standard (P.526) is not the only piece of international ITU physics that gives an answer for this problem. ITU-R Recommendation **P.368** ("Ground-wave propagation curves for frequencies between 10 kHz and 30 MHz") is the companion standard that computes the full ground-wave field, including the geometric-optics / two-ray region, the extended flat-Earth region, and the residue-series shadow region. It is published as Fortran source code by the ITU radiocommunication sector and is maintained in the open-source Python wrapper at [space-physics/grwave](https://github.com/space-physics/grwave). P.368 is the physics DC Dan invoked by name in his 13 Feb 2025 response when he argued that vertical-polarisation ground-wave propagation with "grazing-ray reflection" and "Sommerfeld" mechanics should rescue the beams on a globe. Running ITU's own ground-wave Fortran on the Knickebein geometry is therefore a direct test of his claim, using his preferred physics, with ITU's reference implementation. Full methodology and per-parameter justification is in [[GRWAVE_P368_BotB]]. Driver scripts and commit history live in the BotB GitHub repo at [AlanSpaceAudits/battle-of-the-beams](https://github.com/AlanSpaceAudits/battle-of-the-beams), specifically `run_grwave_knickebein.py` (single-point queries) and `make_p526_vs_p368_graphs.py` (the figures below). ### Master comparison bar chart ![[ITU_Calc_sn_vs_grwave_master_bargraph.png]] Thirteen-path headline comparison: the Sommerfeld-Norton flat-Earth equisignal (green) against the ITU-R P.368 GRWAVE globe equisignal (magenta) at every confirmed Knickebein and Telefunken path. The y axis is the peak SNR at the 5° equisignal corridor, expressed in dB above the 31.5 MHz galactic plus thermal receiver noise floor (0 dB = 0.079 μV at the 50 Ω receiver input per the noise calculation in [Amendment E](#amendment-e-noise-floor-calculation-itu-r-p372)). The four Kleve Midlands targets and the two shorter Telefunken over-sea tests clear the noise floor on both models. Every Stollberg Midlands target and every Telefunken test at 700 km or further collapses below the noise floor on the globe model, while the flat-Earth line continues sitting 70 to 90 dB above it. The full per-path μV breakdown, the station-specific bar charts, and the continuous distance sweeps are gathered in the [Full ITU certified graph set](#full-itu-certified-graph-set) section that precedes the conclusion. ### Agreement summary Eleven of thirteen confirmed paths give the same USABLE / UNUSABLE verdict under both ITU standards (P.526-16 Fock and P.368 GRWAVE), with the two borderline paths flipping due to margin compression where the curves transition from the line-of-sight region into the residue-series shadow decay. The load-bearing falsification (Stollberg → Beeston / Derby / Birmingham / Liverpool at 694 to 791 km) is agreed by both standards at 9 to 46 dB below the detection floor. Ground-wave propagation does not rescue the Knickebein system at 31.5 MHz on a globe, and the ITU ground-wave calculator running ITU's own Fortran says so directly. --- ## The refraction rescue and why it fails The standard globe rescue for VHF signals reaching beyond line of sight is atmospheric refraction. The Fock analysis already accounts for standard refraction by using the effective Earth radius $R_{eff} = \frac{4}{3}R = 8{,}495$ km (the ITU-R P.453 standard atmosphere model). This makes the Earth appear 33% less curved from the radio wave's perspective. Even with this concession built in, the diffraction losses are 77 dB and 149 dB. To eliminate the remaining diffraction loss entirely, you would need super-refraction or tropospheric ducting: an atmospheric inversion layer that bends the beam downward enough to follow the curved surface. This rescue has three problems: ### 1. No seasonal variation observed The Telefunken engineers who built and operated the Knickebein system noticed no real difference in beam performance between summer and winter. If the beams depended on atmospheric refraction to reach England, the signal should vary significantly with season. Summer atmospheres produce stronger temperature inversions (warm air over cool sea) and more frequent ducting conditions. Winter atmospheres are more well mixed and produce weaker refraction. The observation of consistent year-round performance is inconsistent with a refraction dependent propagation path. It is consistent with a direct, unobstructed path that does not depend on atmospheric conditions. ### 2. Two beams, two paths, same intersection Knickebein requires two beams from two different stations (at different distances and different angles) to intersect over a single target with sub-kilometre precision. If both beams depended on atmospheric ducting, the refraction conditions along each path would need to produce the correct bending independently, at the same time, to maintain intersection geometry over the target. The atmosphere along a 440 km path from Kleve and a 694 km path from Stollberg would need to produce two separate extreme refraction gradients that happen to converge at the same point. The probability of this occurring reliably, night after night, across varying weather conditions, is negligible. ### 3. Ducting destroys beam geometry Tropospheric ducting traps the signal in a layer, typically 100 to 500 m thick. The signal bounces between the top and bottom of the duct, spreading vertically within the layer. This is incompatible with a 500 yd equisignal. The duct layer is itself wider than the measured beam. A ducted signal does not arrive as a coherent, narrow beam with a well defined equisignal crossover. It arrives as a smeared, multipath signal with fading and distortion. > [!note] The refraction coefficient argument > Explaining observations that exceed the geometric horizon by increasing the effective Earth radius (refraction coefficient $k > 4/3$) is standard practice. But the required coefficient to eliminate the diffraction loss entirely at 440 km would imply an effective Earth radius roughly 5 times the actual radius. At that point the "sphere" is so large relative to the path that the surface is, for all practical purposes, flat. The refraction rescue converges toward the flat model it is trying to avoid. --- ## Telefunken range table comparison (4,000 m altitude) The Telefunken company documented operational ranges for Knickebein at 4,000 m flight altitude over sea, based on measurements from July 1939 onwards. The primary source is Appendix 2 to Luftwaffe OKL/Ob.d.L. reference 2/14/39, secret command matter, LN Verb. 47F 68, dated 10 September 1939. Filed at the Bundesarchiv-Militärarchiv Freiburg under RL 19-6/40 (formerly RL 19/537). The document header literally reads: *"Reichweite von UKW-Leitstrahlen über See bei 4000 m Flughöhe"* ("Range of VHF Guide Beams over Sea at 4000 m Flight Altitude"). [[1939_BArch_RL19-6-40_230Q8_App2_Telefunken_Range_Tests]] [[2024_Doerenberg_Knickebein_Reference]] The radio line-of-sight range at 4,000 m from a 72 m TX (Stollberg) is **296 km**. Every documented Telefunken range exceeds this. Canonical post-β-fix output at `rx_gain = 0 dBi` (matching the Telefunken over-sea graph in [[GRWAVE_P368_BotB]]): | Receiver config | Telefunken typical (km) | Flat equisignal $V_{rx}$ / SNR | Globe equisignal $V_{rx}$ / SNR | Status (vs 0.079 μV noise floor) | |---|---|---|---|---| | FuBl 1, rod antenna | 400 | 1,642 μV / +78.5 dB | 18.06 μV / +39.4 dB | **USABLE** (95× above noise floor) | | FuBl 1, trailing wire | 500 | 1,314 μV / +76.6 dB | 1.06 μV / +14.7 dB | **USABLE** (5.6× above noise floor) | | FuBl 1 + increased selectivity, rod | 700 | 938 μV / +81.5 dB | 0.0038 μV / −34.1 dB | **DEAD**, 50× below noise floor | | FuBl 1 + increased selectivity, trailing wire | 800 | 821 μV / +72.5 dB | 0.00023 μV / −58.3 dB | **DEAD**, 820× below noise floor | | Special RX (EBL 3?), cross dipole | 1,000 | 657 μV / +70.6 dB | ~10⁻⁵ μV / −106.7 dB | **DEAD**, ~200,000× below noise floor | > [!critical] The exponential wall > The flat model sits at 657 μV to 1,642 μV equisignal across every single row. The globe model passes the noise floor comfortably at 400 km and 500 km and collapses beyond. Going from 500 km to 700 km costs 49 dB on the globe (from +14.7 to −34.1). Going from 700 km to 1,000 km costs another 73 dB (down to −106.7). The Fock exponential decay dominates every linear improvement. Adding receiver sensitivity does not rescue paths 50× or more below the noise floor, because no amplification creates signal out of below-noise hiss. The Germans measured these ranges themselves, at their own transmitters, with their own engineers, and filed the numbers in their own classified internal archive. The globe model cannot reproduce Telefunken's own measured range data beyond the shortest test. The flat model reproduces them trivially. --- ## Required TX height for line of sight The falsification turns on transmitter geometry. Here is the TX height each path would need to sit **inside** line of sight on a globe (aircraft at 6,000 m, or 4,000 m for the Telefunken rows), which is the height at which the globe model would produce the same answer as the flat model: | Path | Distance | RX altitude | Actual TX height | **Required TX height** | |---|---|---|---|---| | Kleve → Spalding | 440 km | 6,000 m | 111 m | 858 m | | Kleve → Retford | 512 km | 6,000 m | 111 m | 2,186 m | | Kleve → Derby | 529 km | 6,000 m | 111 m | 2,589 m | | Kleve → Birmingham | 550 km | 6,000 m | 111 m | 3,133 m | | Stollberg → Beeston | 694 km | 6,000 m | 72 m | 8,265 m | | Stollberg → Derby | 711 km | 6,000 m | 72 m | 9,032 m | | Stollberg → Birmingham | 754 km | 6,000 m | 72 m | 11,124 m | | Stollberg → Liverpool | 791 km | 6,000 m | 72 m | 13,098 m | | Telefunken 500 km | 500 km | 4,000 m | 72 m | 3,371 m | | Telefunken 700 km | 700 km | 4,000 m | 72 m | 11,360 m | | Telefunken 800 km | 800 km | 4,000 m | 72 m | 17,120 m | | Telefunken 1000 km | 1000 km | 4,000 m | 72 m | 32,172 m | Put Knickebein on Mount Everest (8,849 m) and it works on a globe for most operational targets. The height gain $G(Y_1)$ extends the TX horizon from 43 km (at 111 m) to 388 km (at 8,849 m), giving a total LoS of 707 km when combined with the aircraft at 6,000 m. Every Kleve operational target falls inside that LoS, and Stollberg → Beeston (694 km) is inside by a hair. Only the very longest Stollberg path (Liverpool at 791 km) and the Telefunken 800/1000 km configurations would still sit in the shadow zone from Everest. But the actual Kleve and Stollberg antennas were **111 m and 72 m** above the terrain, not 8,849 m. The TX horizons from real heights are 43 km and 35 km respectively, not 388 km. The geometry that allows the system to work on a globe doesn't exist at the actual transmitter sites. The Germans didn't have mountain peaks. They had steel antenna frames in flat German farmland. Kleve sits on a 83 m hill in Nordrhein-Westfalen; Stollberg sits on a 44 m ridge in Schleswig-Holstein. There is no elevation in either region that gets you even halfway to the 858 m minimum needed for the shortest operational path, let alone the 13 km needed for Stollberg → Liverpool. --- ## Full ITU certified graph set Everything the argument rests on now belongs on the same canvas. The figures in this section are regenerated from the same library that produced the headline master bar chart, at the same 31.5 MHz, 3 kW, 26 dBi geometry, and against the same 0.079 μV galactic plus thermal noise floor. All are drawn in absolute receiver-input microvolts on a logarithmic axis so the viewer can read the signal voltage straight off the y axis and compare it to whatever receiver sensitivity they like. The dB SNR variants used for the derivation upstream are omitted here on purpose: the physics is already settled by that point, and the operational question is how many microvolts of Knickebein arrive at the 50 Ω antenna terminals against how many microvolts of noise. ### Static beam map: Kleve and Stollberg over the Midlands ![[ITU_Calc_knickebein_beam_map 1.png]] The operational two-beam geometry for a 1940-41 Midlands raid. Kleve (111 m, overland path to Derby via Spalding) fires the director beam in green. Stollberg (72 m, sea path to Derby via Beeston) fires the cross beam with a magenta wedge collapsing into the hatched shadow segment past roughly 560 km, because the shorter TX and the sea path's β clamp at 31.5 MHz pull the globe-model signal into the noise well before the beam reaches any confirmed operational target. The same beam on Sommerfeld-Norton flat-Earth physics stays saturated green across the whole length. ### Telefunken July 1939 over-sea snapshots The five nominal range tests logged in [[1939_BArch_RL19-6-40_230Q8_App2_Telefunken_Range_Tests|BArch RL 19-6/40 ref. 230Q8 App. 2]], rendered as static frames. Cyan halos mark the target under test. Each solid pink wedge is the part of the beam where the globe ground-wave calculator still clears the noise floor; each hatched pink wedge is the part where the globe signal is buried in noise. The green Sommerfeld-Norton wedge stays saturated across every frame. ![[ITU_Calc_knickebein_beam_map_telefunken_400km 1.png]] 400 km target, closest range. Sommerfeld-Norton equisignal 1,024 μV, GRWAVE equisignal 20.6 μV. Both well above 0.079 μV, USABLE on both models. ![[ITU_Calc_knickebein_beam_map_telefunken_500km 1.png]] 500 km target. Sommerfeld-Norton 827 μV, GRWAVE 3.95 μV. The last range at which both ITU standards agree the signal is USABLE. ![[ITU_Calc_knickebein_beam_map_telefunken_700km 1.png]] 700 km target. Sommerfeld-Norton 507 μV, GRWAVE 0.047 μV. GRWAVE crosses the noise floor by a factor of 1.7. The hatched pink band visualises precisely this gap. ![[ITU_Calc_knickebein_beam_map_telefunken_800km 1.png]] 800 km target. Sommerfeld-Norton 401 μV, GRWAVE 0.0039 μV, roughly 26 dB into the shadow on the globe. The 1939 logbook records reception at this distance. The globe model cannot reproduce it. ![[ITU_Calc_knickebein_beam_map_telefunken_1000km 1.png]] 1,000 km target, farthest range. Sommerfeld-Norton 269 μV, GRWAVE 2.7 × 10⁻⁵ μV, about 69 dB below the noise floor. The 1939 report explicitly records successful reception at this distance. The globe prediction misses the observation by more than 60 dB. ### Per-station master bar charts in microvolts Each bar pair plots the Sommerfeld-Norton flat-Earth equisignal (green) alongside the ITU-R P.526-16 Fock globe equisignal (magenta) at the receiver input in μV on a log axis. The white 0.079 μV line is the operational threshold: a bar at or above it is audible; a bar below it is buried in galactic plus thermal noise. ![[ITU_Calc_master_bargraph_kleve_operational 1.png]] Kleve at 111 m firing overland to the four Midlands targets (Spalding, Retford, Derby, Birmingham). Both models clear the noise floor at every target. The globe model tracks within an order of magnitude of the flat-Earth line because the Kleve paths are still inside the usable diffraction regime at these ranges. ![[ITU_Calc_master_bargraph_kleve_telefunken 1.png]] Same Kleve 111 m transmitter, this time over sea to the five Telefunken test targets. Because Kleve sits south-west of Stollberg, the haversine distances from Kleve are shorter than the Stollberg-labelled nominals. The globe equisignal crosses the noise floor between the 500 km test (3.95 μV) and the 700 km test (0.047 μV). ![[ITU_Calc_master_bargraph_stollberg_operational 1.png]] Stollberg at 72 m over sea to the four Midlands operational targets. Every globe prediction sits below 0.079 μV: Beeston 0.019 μV, Derby 0.012 μV, Birmingham 0.0036 μV, Liverpool 0.0013 μV. The flat-Earth line runs between 326 μV and 402 μV on the same four paths. This is the bar chart that carries the falsification. ![[ITU_Calc_master_bargraph_stollberg_telefunken 1.png]] Stollberg 72 m to the historical test geometry at 400, 500, 700, 800, and 1,000 km. The flat-Earth column stays usable across the entire set. The globe column is already below the noise floor by the 500 km test and roughly 10⁷ times too weak by 1,000 km. ### Continuous distance sweeps The continuous curves fill in the structure behind the per-target bars. Distance runs from 50 km out to just past the longest nominal target for each station. Blue dotted verticals mark the confirmed operational or test targets. The dashed grey line is the 4/3-Earth radio horizon computed from the transmit and receive heights. ![[ITU_Calc_sweep_kleve_operational 1.png]] Kleve → Midlands sweep, TX 111 m, RX 6,000 m, overland. Sommerfeld-Norton stays near 1 kμV across the whole 900 km span. The globe ground-wave curve tracks close to the flat-Earth line until the 354 km radio horizon, then falls off a cliff and crosses the noise floor near 600 km. Retford, Derby, and Birmingham labels are staggered across rows because they sit within 40 km of each other. ![[ITU_Calc_sweep_stollberg_operational 1.png]] Stollberg → Midlands sweep, TX 72 m, RX 6,000 m, sea. Same cliff-edge pattern as Kleve but the globe crossover arrives earlier, near 560 km, because of the lower transmit height and the sea path's β clamp below the 300 MHz cutoff for vertical polarisation over seawater. By the time the beam reaches Beeston at 694 km the globe model predicts 0.019 μV while Sommerfeld-Norton predicts 402 μV. ![[ITU_Calc_sweep_telefunken 1.png]] Telefunken July 1939 sweep. Same 72 m transmitter but with the aircraft at 4,000 m per the manufacturer's test altitude in the 1939 range memo. Radio horizon drops to 296 km. The globe curve crosses the noise floor just past 500 km; the flat-Earth curve stays usable out to 1,200 km. If the 1939 report logs a signal at 700, 800, or 1,000 km, the globe prediction cannot reproduce it and the flat-Earth prediction can. ### Sommerfeld-Norton against GRWAVE alone The four sweeps below show the same Sommerfeld-Norton flat-Earth line (green) against ITU's own ground-wave globe calculator, P.368 GRWAVE (magenta). GRWAVE runs Rotheram's full multi-mode residue-series solver, which agrees with the P.526 first-term Fock curve within a few dB across every path (see [[GRWAVE_P368_BotB]] for the full cross-check). The headline master bar chart higher up in this document uses the same two-model pairing; the three station sweeps and the fubl2 Dan-comparison variant are gathered here. ![[ITU_Calc_sn_vs_grwave_uv_master_bargraph.png]] Thirteen-path two-model master, Sommerfeld-Norton against GRWAVE. The green column and the magenta column are the same two quantities plotted in the headline dB chart, re-expressed at the 50 Ω antenna terminals so the reader can read the equisignal voltage directly. The 0.079 μV white line is the 500 Hz matched-filter galactic plus thermal noise floor derived in [Amendment E](#amendment-e-noise-floor-calculation-itu-r-p372). ![[ITU_Calc_sn_vs_grwave_uv_kleve.png]] Kleve sweep. Both curves clear the noise floor at every Midlands target. The Kleve director leg is the non-discriminating side of the test. ![[ITU_Calc_sn_vs_grwave_uv_stollberg.png]] Stollberg sweep. The globe GRWAVE curve crosses the noise floor well short of Beeston and remains there for every confirmed operational target. The Sommerfeld-Norton curve stays saturated above 300 μV across the same paths. ![[ITU_Calc_sn_vs_grwave_uv_telefunken.png]] Telefunken sweep at the 1939 test altitude. GRWAVE drops below the noise floor around 550 km; Sommerfeld-Norton holds usable past 1,200 km. The manufacturer logged reception at 700, 800, and 1,000 km. Only the flat-Earth line is consistent with the logbook. ![[ITU_Calc_sn_vs_grwave_fubl2_master_bargraph.png]] Master bar chart with the 2 μV reference line DC Dan entered into his link-budget tool during the 13 February 2025 *Drunk Debunk Show* presentation. Every Stollberg → Midlands path and every long Telefunken path sits well below both the 2 μV line and the 0.079 μV physics noise floor on the globe model. The noise floor is the load-bearing bound, not the chosen receiver sensitivity. --- ## Summary Threshold: signal above the galactic + thermal receiver noise floor. Noise floor: −159.0 dBW (ITU-R P.372 galactic synchrotron at 31.5 MHz, Eq. 14, curve E, plus 290 K thermal over 500 Hz detection bandwidth). > [!note] Values refreshed against the current library output (April 2026) > Equisignal voltages and SNRs in the tables that follow have been regenerated from the live output of the canonical analysis library. Two corrections propagate from there: the squint crossover loss is computed dynamically as $20\log_{10}|\text{sinc}(5°)| = -19.87$ dB at the 5° Telefunken geometry rather than the earlier rounded −19 dB, and the physics noise floor converts to 0.079 μV at the 50 Ω receiver input on the current 500 Hz MCW matched-filter bandwidth rather than the earlier 0.078 μV rounding. Both changes shift individual equisignal values by tenths of a dB; no operational verdict changes. All values below match the master bar chart above and the canonical Google sheet output. Noise floor reference is 0.079 μV. | Quantity | Flat $V_{rx}$ / SNR | Globe $V_{rx}$ / SNR | Observed | |---|---|---|---| | Equisignal width (Kleve, 440 km) | 554 yd (consistent) | same corridor geometry | 400 to 500 yd (detected) | | Beam peak (Kleve → Spalding) | 13,320 μV / +104.5 dB | 165.9 μV / +66.4 dB | Detected by WIDU | | Equisignal (Kleve → Spalding) | 1,353 μV / +84.7 dB | **16.85 μV / +46.5 dB USABLE** | Detected by Bufton's Anson | | Equisignal (Kleve → Retford) | 1,162 μV / +83.4 dB | **1.39 μV / +24.8 dB USABLE** | Enigma 5 Jun 1940 | | Equisignal (Kleve → Derby) | 1,125 μV / +83.1 dB | **0.77 μV / +19.7 dB USABLE** | Detected, bombers guided | | Equisignal (Kleve → Birmingham) | 1,082 μV / +82.8 dB | **0.37 μV / +13.4 dB USABLE** | Detected, bombers guided | | Beam peak (Stollberg → Beeston) | 8,445 μV / +100.6 dB | **0.191 μV / +7.6 dB MARGINAL** | Bufton cross beam detected | | Equisignal (Stollberg → Beeston) | 858 μV / +80.7 dB | **0.0194 μV / −12.3 dB DEAD** | Bufton cross beam detected | | Equisignal (Stollberg → Derby) | 837 μV / +80.5 dB | **0.0121 μV / −16.4 dB DEAD** | Detected, bombers guided | | Equisignal (Stollberg → Birmingham) | 789 μV / +80.0 dB | **0.00363 μV / −26.8 dB DEAD** | Detected, bombers guided | | Equisignal (Stollberg → Liverpool) | 752 μV / +79.6 dB | **0.00129 μV / −35.8 dB DEAD** | Detected | | Continuous beam tracking | Yes | No (shadow zone) | Yes (continent to target) | **The four Kleve paths are all detectable on the globe (equisignal $V_{rx}$ from 0.37 μV to 16.85 μV, SNR from +13.4 to +46.5 dB above the 0.079 μV noise floor), consistent with Bufton's physical detection at Spalding. All four Stollberg paths are 12 to 36 dB below the noise floor at the equisignal crossover on the globe (0.00129 μV to 0.0194 μV versus 0.079 μV), i.e. 4× to 61× below. No amplification recovers signal from below the noise floor, so the Stollberg cross beam does not exist at any confirmed Midlands operational target on a sphere.** --- ## Conclusion The [Full ITU certified graph set](#full-itu-certified-graph-set) carries the visual argument and the Summary table above carries the numeric one; both agree on the same verdict. **Failed to falsify H₀.** The flat model produces a detectable, audible signal at every documented range, with equisignal $V_{rx}$ from 752 μV to 1,353 μV (SNR from +79.6 to +84.7 dB above the 0.079 μV noise floor) regardless of path. The measured beam divergence, projected rectilinearly over the path distance, is consistent with the measured 400 to 500 yard equisignal at Spalding. No bending, diffraction, or atmospheric correction is needed. **H₁ is falsified.** The globe model fails at every confirmed Stollberg path: 1. **Kleve → Spalding and all Kleve Midlands targets (440 to 550 km):** detectable on the globe. Spalding equisignal is 16.85 μV (+46.5 dB), Retford 1.39 μV (+24.8 dB), Derby 0.77 μV (+19.7 dB), Birmingham 0.37 μV (+13.4 dB). All are above the 0.079 μV noise floor. These values are consistent with Bufton physically measuring the equisignal at Spalding and with the operational record of Kleve-guided raids into the Midlands. The Kleve director beam alone is not the falsification; it passes. 2. **Stollberg → all four Midlands operational targets (694 to 791 km):** Post-β-fix Fock diffraction loss of 93 to 145 dB on top of free-space path loss. The aircraft height-gain term cannot rescue any of them. Globe equisignal $V_{rx}$ is 0.0194 μV at Beeston (694 km), 0.0121 μV at Derby (711 km), 0.00363 μV at Birmingham (754 km), 0.00129 μV at Liverpool (791 km). **All DEAD by 4× to 61× below the 0.079 μV thermal + galactic noise floor.** The Stollberg cross beam does not exist at any confirmed Midlands target on a sphere. No receiver gain amplifies signal out of below-noise hiss. Without the cross beam, no intersection exists, and no precision bombing is possible. This is the falsification. 3. **Continuous tracking:** Aircraft followed the beam continuously from Germany to England. On a sphere, the Kleve beam enters the diffraction shadow zone after 378 km (aircraft at 6 km). The signal decays exponentially from that point at 0.292 dB/km first-mode. There is no physical mechanism for continuous tracking through 300+ km of shadow zone on a globe. 4. **Telefunken's own range data:** The globe model cannot reproduce the operational ranges documented in September 1939 ([[1939_BArch_RL19-6-40_230Q8_App2_Telefunken_Range_Tests|Bundesarchiv-MA Freiburg RL 19-6/40 ref. 230Q8 App. 2]], as compiled by [[2024_Doerenberg_Knickebein_Reference|Dörenberg]]). Of five measured configurations, the 400 km and 500 km paths pass on the globe after the β = 0.81 over-sea correction (20.56 μV and 3.95 μV respectively, both above the 0.079 μV noise floor), but every path at 700 km and beyond collapses: 0.047 μV at 700 km, 0.0039 μV at 800 km, 2.7 × 10⁻⁵ μV at 1,000 km, all 1.7× to 290,000× below the noise floor. The Fock exponential decay overwhelms any receiver sensitivity improvement. Adding receiver sensitivity does not rescue paths that far below noise. 5. **Required TX height:** To put Stollberg → Liverpool (791 km) inside line of sight on a globe, the transmitter would need to be at 13,098 m above the terrain. Stollberg's actual antenna frame sits at 72 m. The ratio is **182 to 1**. There is no terrain anywhere in Schleswig-Holstein that gets above 200 m. Even Mount Everest (8,849 m) would not cover the full Knickebein operational envelope. 6. **Primary-source crew testimony confirms the receiver, not the cockpit, is the issue.** Bundesarchiv ref. 230Q18 reports that interviewed German bomber crews said the real Knickebein signals were *"easily heard"* in the cockpit even through British Aspirin jamming. The FuBl 2 / EBL 3 airborne receiver has full AGC on its RF and first two IF stages, a manual sensitivity knob (Empfindlichkeitsregler), automatic volume control, and a two-stage AF amplifier ([[1943_DLuft_T4058_FuBl_2_Geraete_Handbuch|D.(Luft) T.4058 §24]]). As long as the RF signal is physically present at the antenna above the receiver noise floor, the receiver electronics deliver it to the earphones at audible SPL. This is why the falsification rests on the RF noise floor: the receiver chain handles everything downstream of that. The Stollberg cross beam fails not because the pilot cannot hear it but because there is no signal at the antenna at 694 km on a sphere at all. The Knickebein system worked exactly as a line-of-sight VHF beam would work over a flat surface. The measured beam divergence, signal strength, Telefunken range table, and continuous tracking from the continent to the target are all consistent with rectilinear propagation. The same observations are quantitatively inconsistent with propagation over a sphere of radius 6,371 km, as computed from the Fock smooth-Earth diffraction theory that the radio propagation community itself uses for this exact class of problem. The international standard ITU-R P.526-16 smooth-Earth diffraction algorithm, the ITU-R P.368 GRWAVE ground-wave Fortran calculator, the US Air Force SEKE production radar code (Shatz & Polychronopoulos 1988), and Fock's own 1945 residue series all agree within 0.2 dB on what the globe model predicts at these paths. They all say the same thing: the Stollberg cross beam cannot reach the Midlands on a sphere. These findings are independently corroborated by Bauer (2004), who documents from German Telefunken archives that the system was rated for 500 km at 6,500 m altitude, and that the preceding X-beam programme achieved 300 m bombing accuracy at 350 km guidance range in December 1936 trials. Both figures are quantitatively inconsistent with smooth-Earth diffraction at 31.5 MHz on a sphere of radius 6,371 km. Both are consistent with rectilinear propagation. [[2004_Bauer_German_Radio_Navigation_1907_1945]] Eckersley knew this. He told British intelligence the beams could not reach England. The physics on a sphere agreed with him. The beams reached England anyway. --- # Amendments Heavy-math reference sections. Every formula the body of the document uses is derived here with variable definitions, source citations, and worked examples. ## Amendment A: Friis free-space path loss The Friis equation computes the received power from a transmitter at distance $d$ in unobstructed free space. Published by Harald T. Friis of Bell Labs in 1946 and the standard starting point for any radio link budget. [[1946_Friis_Simple_Transmission_Formula]] $P_{rx} = P_{tx} \cdot G_{tx} \cdot G_{rx} \cdot \left(\frac{\lambda}{4\pi d}\right)^2$ In decibels (where multiplication becomes addition and division becomes subtraction): $P_{rx}(\text{dBW}) = P_{tx}(\text{dBW}) + G_{tx}(\text{dBi}) + G_{rx}(\text{dBi}) - \text{FSPL}(\text{dB})$ where FSPL (Free-Space Path Loss) is: $\text{FSPL} = 20\log_{10}\left(\frac{4\pi d}{\lambda}\right)$ ### Variable definitions | Variable | What it is | Formula / value | Knickebein value | |---|---|---|---| | $P_{tx}$ | Transmit power. The electrical power fed to the antenna. | Given | 3,000 W = 34.8 dBW | | $G_{tx}$ | Transmitter antenna gain (directivity). How much the antenna concentrates energy in one direction compared to radiating equally in all directions (isotropic). | $G = \frac{4\pi A}{\lambda^2}$ | 398 = 26.0 dBi | | $A$ | Physical area of the antenna aperture (width $\times$ height). | $L_H \times H_V$ | 99 m $\times$ 29 m = 2,871 m$^2$ | | $G_{rx}$ | Receiver antenna gain. 0 dBi (isotropic) throughout for the conservative case matching the graph generator. | Given | 0.0 dBi | | $\lambda$ | Wavelength. | $\lambda = c / f$ | 9.517 m | | $f$ | Frequency. | Given | 31.5 MHz | | $c$ | Speed of light. | Constant | 299,792,458 m/s | | $d$ | Distance from transmitter to receiver. Great-circle distance along the surface. | Given | 439,541 m (Kleve) or 693,547 m (Stollberg) | | FSPL | Free-Space Path Loss. Spreading loss as the signal radiates outward in all directions. Power decreases as $1/d^2$ (inverse-square law). | $20\log_{10}(4\pi d/\lambda)$ | 115.3 dB (Kleve), 119.1 dB (Stollberg) | ### Worked example: Kleve to Spalding (440 km) $P_{tx} = 10\log_{10}(3000) = 34.8 \text{ dBW}$ $G_{tx} = 10\log_{10}\left(\frac{4\pi \times 2871}{9.517^2}\right) = 10\log_{10}(398) = 26.0 \text{ dBi}$ $G_{rx} = 0.0 \text{ dBi (isotropic)}$ $\text{FSPL} = 20\log_{10}\left(\frac{4\pi \times 439{,}541}{9.517}\right) = 115.3 \text{ dB}$ $P_{rx} = 34.8 + 26.0 + 0.0 - 115.3 = -54.5 \text{ dBW}$ $\text{SNR} = P_{rx} - N = -54.5 - (-159.0) = +96.7 \text{ dB}$ Converted to a voltage at the 50 Ω receiver input at the sub-beam peak: $V_{rx} = 0.079 \times 10^{104.5/20} = 13{,}320 \, \mu V$. Applying the 19.87 dB equisignal crossover loss (Amendment D) gives 1,353 μV at the equisignal corridor. Enormous headroom either way on the flat model. --- ## Amendment B: Fock smooth-Earth diffraction (ITU-R P.526-16) The Fock diffraction formula gives the field strength relative to free space for a source on a conducting sphere. The general attenuation factor is (Fock 1965, Ch. 10, p. 209, Eq. 6.10): $V(x, y, q) = e^{i\pi/4} \cdot 2\sqrt{\pi x} \sum_{s=1}^{\infty} \frac{e^{ix t_s}}{t_s - q^2} \cdot \frac{w(t_s - y)}{w(t_s)}$ where $x = (ka/2)^{1/3}\theta$ is the normalised distance, $y$ is the normalised source height, $q$ is the surface impedance parameter, $t_s$ are the roots of $w'(t) - qw(t) = 0$ with phase $\pi/3$, and $w(t)$ is the Fock-Airy function. Root magnitudes $|t^0_s|$ for the perfectly conducting case are tabulated (Fock 1965, p. 209): $|t^0_1| = 2.33811$, $|t^0_2| = 4.08795$, $|t^0_3| = 5.52056$, $|t^0_4| = 6.78671$, $|t^0_5| = 7.99417$. [[1945_Fock_Diffraction_Radio_Waves_Earth]] For surface-to-surface ($y = 0$) on a perfectly conducting sphere ($q \to \infty$) the general form reduces to the standard radio propagation residue series (cf. Neubauer et al. 1969, Eq. 20 to 26): $\frac{E_{diff}}{E_0} = V(\xi) \times G(Y_1) \times G(Y_2)$ $|V(\xi)| = 2\sqrt{\pi\xi} \left| \sum_{s=1}^{N} \frac{\exp\left(i \cdot e^{i\pi/3} \cdot \tau_s \cdot \xi\right)}{[\operatorname{Ai}'(-\tau_s)]^2} \right|$ where $\xi = m\theta$ is the Fock normalised distance and $\tau_s = |t^0_s|$ are the Airy-function zeros. Since $t_s = \tau_s \cdot e^{i\pi/3}$, the exponential $e^{i\xi t_s} = \exp(i \cdot e^{i\pi/3} \cdot \tau_s \cdot \xi)$ has real part $-(\sqrt{3}/2) \cdot \tau_s \cdot \xi$. That negative real part produces exponential decay in the shadow region, which is the creeping-wave attenuation at 0.292 dB per kilometre at 31.5 MHz. ### Frequency-and-radius constants (path-independent) | Parameter | Formula | Value at 31.5 MHz, $R_{eff} = 8495$ km | |---|---|---| | $kR_{eff}$ | $\frac{2\pi}{\lambda} \times R_{eff}$ | 5,608,104 | | Fock parameter $m$ | $(kR_{eff}/2)^{1/3}$ | 141.0 | | Creeping ribbon height $W_{creep}$ | $R(kR)^{-1/3}$ | 47.8 km | | Surface attenuation $\alpha_1$ | $\frac{8.686 \times \tau_1 \sin(60°) \times m}{R}$ | 0.292 dB/km | The 47.8 km creeping-wave ribbon height is the vertical extent of the wave that hugs the surface after diffracting past the horizon. It is the height of the layer within which any diffracted signal exists. The 0.292 dB/km surface attenuation is the per-kilometre decay rate of the creeping wave mode past the horizon. ### Height gain function $G(Y)$ $G(Y)$ is the closed-form fit to the magnitude of Fock's first-mode height factor, standardised by the ITU in [[2025_ITU-R_P526-16_Diffraction|ITU-R P.526-16]] Eq. 17 to 18. With $B = \beta Y$: $G(Y) = \begin{cases} 17.6\sqrt{B - 1.1} - 5\log_{10}(B - 1.1) - 8 & B > 2 \\[4pt] 20 \log_{10}(B + 0.1\,B^3) & B \leq 2 \end{cases}$ $Y = 2\beta \left(\frac{\pi^2}{\lambda^2 \, R_{eff}}\right)^{1/3} h \;\;=\;\; \frac{2 m^2 h}{R_{eff}}$ At 31.5 MHz, $Y/h = 0.004682$ m⁻¹. The $B > 2$ branch is the far-field regime where the antenna is well above the creeping ribbon; the $B \leq 2$ branch is the near-field regime where the antenna sits inside the ribbon. Plugged in for the relevant Knickebein antennas: | Antenna | $h$ (m) | $Y$ | regime | $G(Y)$ | |---|---|---|---|---| | Kleve TX (83 m terrain + 28 m frame) | 111 | 0.520 | $B \leq 2$ | −5.5 dB | | Stollberg TX (44 m terrain + 28 m frame) | 72 | 0.337 | $B \leq 2$ | −9.4 dB | | He 111 aircraft (operational) | 6,000 | 28.09 | $B > 2$ | +76.3 dB | | Telefunken test aircraft | 4,000 | 18.73 | $B > 2$ | +59.7 dB | At Knickebein ground-antenna heights the TX contribution is negative because the transmitter sits inside the 47.8 km creeping-wave ribbon. The entire positive height budget comes from the aircraft at 6,000 m (+76.3 dB). > [!note] Fock and ITU are the same physics in two forms > The piecewise closed-form above is a curve-fit to Fock's residue series in the first-mode-dominated regime. The direct Fock residue series via the Shatz & Polychronopoulos (1988) 35-mode MIT Lincoln Laboratory implementation and the ITU closed form agree within 0.1-0.2 dB. [[1988_Shatz_Spherical_Earth_Diffraction_SEKE]] ### ITU-R P.368 GRWAVE: independent cross-check The ITU maintains a separate Fortran reference implementation of the full multi-mode residue series in **ITU-R P.368** ("Ground-wave propagation curves for frequencies between 10 kHz and 30 MHz"). The BotB code wraps this calculator in `make_p526_vs_p368_graphs.py` via the `p368_snr_peak()` function. Both P.526 Fock and P.368 GRWAVE are plotted on every graph in the main body and agree within 0.5 dB on every path. --- ## Amendment C: Fock flat-Earth (Weyl-van der Pol) Fock (1965), Ch. 10, p. 200-201, Eq. (3.22) to (3.23) derives the complete electromagnetic field at a receiver over a flat conducting surface, starting from Maxwell's equations (Helmholtz equation, Eq. 1.02). Not an add-on to Friis; an independent self-contained solution to the wave equation for a dipole source above a flat conductor. Full derivation chain from Maxwell's equations through the Hertz potential, Sommerfeld-Watson transformation, and stationary phase evaluation is in [[1945_Fock_Diffraction_Radio_Waves_Earth#Chapter 10, Sections 1-3: From Maxwell's Equations to the Flat Earth Formula]]. The solution gives the total received power as: $P_{rx} = P_{tx} \cdot G_{tx} \cdot G_{rx} \cdot \left(\frac{\lambda}{4\pi d}\right)^2 \cdot |W|^2$ The first four terms (free-space propagation) emerge naturally from solving the wave equation for an unobstructed path. The factor $|W|^2$ is the ground-interaction term the wave equation produces when a conducting surface is present below the propagation path. The attenuation factor $W$ in the illuminated (no-curvature-obstruction) region: $W = 1 + R_s \cdot e^{i\Delta\phi}$ | Variable | What it is | Value for Knickebein | |---|---|---| | $W$ | Attenuation factor. The ratio of the total field (direct + reflected) to the direct field alone. | 2 (for perfect conductor) | | $R_s$ | Surface reflection coefficient. For a perfect electrical conductor $R_s = 1$. Seawater at 31 MHz is close to a perfect conductor. | 1 (perfect conductor limit) | | $\Delta\phi$ | Phase difference between direct and reflected paths. At low grazing angles over a flat surface, this approaches 0. | $\approx 0$ at long range | For a perfectly conducting flat surface ($R_s = 1$, $\Delta\phi \to 0$): $W = 1 + 1 = 2 \quad\Rightarrow\quad |W|^2 = 4 = +6.0 \text{ dB}$ This is the standard ground-reflection result for a source over a conducting plane: Fock (1965), p. 201, Eq. (3.23). ### Worked examples **Kleve to Spalding (440 km):** $P_{rx}(\text{Fock flat}) = P_{rx}(\text{Friis}) + |W|^2 = -54.5 + 6.0 = -48.5 \text{ dBW}$ $\text{SNR}(\text{Fock flat, peak}) = -48.5 - (-159.0) = +102.7 \text{ dB}$ $\text{SNR}(\text{Fock flat, equisignal}) = 102.7 - 19.0 = +83.7 \text{ dB}$ **Stollberg to Beeston (694 km):** $\text{FSPL}(\text{Stollberg}) = 20\log_{10}\left(\frac{4\pi \times 694{,}000}{9.517}\right) = 119.2 \text{ dB}$ $P_{rx}(\text{Friis}) = 34.8 + 26.0 + 0.0 - 119.2 = -58.4 \text{ dBW}$ $\text{SNR}(\text{Fock flat, peak}) = +98.7 \text{ dB}; \quad \text{SNR}(\text{Fock flat, equisignal}) = +79.7 \text{ dB}$ The body of the document uses the conservative Friis figure throughout rather than the +6 dB Fock flat-Earth figure, in order to give the globe model every generosity. --- ## Amendment D: Equisignal crossover loss (5° squint) Knickebein radiates two squinted sub-beams (dot and dash) from the same 99 m × 29 m Telefunken rectangular aperture. The two sub-beams are offset by a total of 5° from the array centreline. The equisignal is the narrow corridor at the centreline where dot and dash sub-beam amplitudes are equal and the pilot hears a steady tone instead of dots or dashes. The equisignal sits at the crossover between two squinted lobes, below either lobe's peak. For the 5° squint geometry and the Telefunken aperture, the crossover loss is **19.87 dB**, computed in the canonical analysis library dynamically from the sinc-pattern array-factor evaluation at the 5° squint (earlier drafts of this document rounded the value to 19 dB; the live library output is 19.87 dB). Every field-strength value in the body of this document (master bar chart, per-station sweeps, all numeric tables) has this 19.87 dB crossover loss applied uniformly, because the pilot flies the equisignal corridor, not the sub-beam peak. The sub-beam peak value is 19.87 dB (9.83×) higher than the equisignal value shown in the body. For the derivation of the 0.066° equisignal corridor width from the 5° squint + 99 m × 29 m aperture geometry see [[DC_Dan]] Argument 4. ### Equisignal width consistency check The British measured the beam divergence at Spalding as 0.066°. On a flat surface this angle projects to a width at range $d$: $W_{flat} = d \times \tan(0.066°) = 439{,}541 \times \tan(0.066°) \approx 506 \text{ m} \approx 554 \text{ yd}$ The measured equisignal was 400 to 500 yd. The measured divergence angle projected rectilinearly over the measured distance gives the measured width. This is a consistency check that confirms no bending, diffraction, or atmospheric correction is needed to connect the transmitter geometry to the received beam width. --- ## Amendment E: Noise floor calculation (ITU-R P.372) The noise power at the receiver input: $N = k_B \cdot T \cdot B \cdot F_a$ In decibels: $N(\text{dBW}) = 10\log_{10}(k_B T B) + F_a$ | Variable | What it is | Value | |---|---|---| | $k_B$ | Boltzmann constant. | $1.381 \times 10^{-23}$ J/K | | $T$ | System reference temperature. | 290 K (standard, IEEE/ITU convention) | | $B$ | Detection bandwidth. The frequency range the receiver listens to. | 500 Hz (matched to Lorenz dot-dash keying rate) | | $F_a$ | External noise figure. At 31 MHz over rural England at night in 1940, dominant noise is galactic synchrotron radiation from the Milky Way (ITU-R P.372 curve E). Man-made noise is heavily suppressed at 6,000 m altitude. | 18 dB (ITU-R P.372-16 Eq. 14 at 31.5 MHz) | | $N$ | Total noise power. | $-177.0 + 18 = -159.0$ dBW | Converted to voltage at the 50 Ω receiver input: $V_{noise} = \sqrt{N \times 50 \, \Omega} = \sqrt{10^{-159.0/10} \times 50} \approx 0.079 \, \mu V$ This 0.079 μV value is the horizontal white reference line on every graph and the denominator of the USABLE / MARGINAL / DEAD classification. Earlier drafts of this document quoted 0.195 μV here (matched to the wider 3 kHz AM-voice bandwidth used before the MCW bandwidth correction); the refreshed 500 Hz matched-filter bandwidth gives 0.079 μV, which is what every current figure and table is drawn against. --- ## Amendment F: Surface electrical constants at 31.5 MHz The P.526-16 surface admittance $K$ and polarisation parameter $\beta$ depend on the ground's electrical characteristics along the path. Three ground types matter for Knickebein: | Ground type | $\sigma$ (S/m) | $\varepsilon_r$ | Used for | |---|---|---|---| | Average land | 0.005 | 15 | Kleve → Midlands (Dutch lowlands, UK rural) | | Seawater | 5.0 | 70 | Stollberg → Midlands (North Sea) | | Medium dry ground | 0.001 | 15 | (not used in Knickebein paths) | From [[2025_ITU-R_P526-16_Diffraction|P.526-16 Eq. 11a / 12a]] the normalised surface admittance at 31.5 MHz vertical polarisation: | Ground type | $K_V$ at 31.5 MHz | Regime | $\beta$ (Eq. 16) | |---|---|---|---| | Average land | 0.003 | $K < 1$, §3.1.1.2 valid | $\beta \approx 1$ (vert pol above 20 MHz over land) | | Seawater | 0.054 | $K < 1$, §3.1.1.2 valid | $\beta \approx 0.81$ (vert pol sea below 300 MHz; post-β fix) | The β = 0.81 value for Stollberg's over-sea paths was added to the BotB library in the April 2026 β-fix. Before that, β was hardcoded to 1.0, which overestimated Fock loss on Stollberg paths by ~8-10 dB. Every graph and number in the current document is post-β-fix. --- ## Amendment G: Formulas reference table Every equation used anywhere in the document, with source citations, in one table. | Formula | Expression | Source | |---|---|---| | Curvature drop | $h = d^2/(2R)$ | Standard geodesy | | Radio horizon | $d_{hor} = \sqrt{2R_{eff}h}$ | ITU-R P.453 | | Antenna directivity | $G = 4\pi A/\lambda^2$ | Friis (1946) Eq. 1, gain-form rearrangement; [[Balanis_Antenna_Theory\|Balanis, Ch. 12]] | | HPBW | $\theta_{3dB} \approx 0.886\lambda/L$ | Born and Wolf, Sec. 8.4.3 | | FSPL (Friis) | $20\log_{10}(4\pi d/\lambda)$ | Friis (1946); Amendment A | | Creeping ribbon width | $W_{creep} = R(kR)^{-1/3}$ | Fock (1945); Keller GTD | | Fock diffraction (general) | $V(x,y,q) = e^{i\pi/4} \cdot 2\sqrt{\pi x}\sum_s e^{ixt_s}/(t_s - q^2) \cdot w(t_s-y)/w(t_s)$ | Fock (1965), Ch. 10, p. 209, Eq. 6.10; Amendment B | | Fock diffraction (surface, $q \to \infty$) | $\lvert V(\xi)\rvert = 2\sqrt{\pi\xi}\bigl\lvert\sum_s \exp(i e^{i\pi/3}\tau_s\xi)/[\mathrm{Ai}'(-\tau_s)]^2\bigr\rvert$ | Specialisation of Eq. 6.10; Neubauer et al. (1969), Eq. 20-26 | | Height gain ($B > 2$) | $G(Y) = 17.6\sqrt{B-1.1} - 5\log_{10}(B-1.1) - 8$ | ITU-R P.526-16, Eq. 17 | | Height gain ($B \leq 2$) | $G(Y) = 20\log_{10}(B + 0.1B^3)$ | ITU-R P.526-16, Eq. 18 | | Normalised height | $Y = 2\beta(\pi^2/(\lambda^2 R_{eff}))^{1/3} h = 2m^2h/R_{eff}$ | ITU-R P.526-16, Eq. 14 | | Surface admittance $K_V$ | $K_V = K_H[\varepsilon^2 + (18000\sigma/f)^2]^{1/2}$ | ITU-R P.526-16, Eq. 12a; Amendment F | | Polarisation parameter $\beta$ | Eq. 16 piecewise fit in $K$ | ITU-R P.526-16, Eq. 16 | | Surface attenuation | $\alpha_1 = 8.686\tau_1\sin(60°)m/R$ dB/m | Fock (1945), first mode | | Flat equisignal width | $W = d\tan(\theta_{eq})$ | Plane trigonometry | | Weyl-van der Pol (flat Earth) | $W = 2$ (perfect conductor) | Fock (1965), Ch. 10, p. 201, Eq. 3.23; Amendment C | | Noise floor (dBW) | $N = 10\log_{10}(k_B T B) + F_a$ | ITU-R P.372-16 Eq. 14; Amendment E | | SNR → V_rx | $V_{rx} = V_{noise} \times 10^{SNR/20}$ | Ohm's law at 50 Ω; Amendment E | --- ## Sources Sorted by year (primary sources first, standing references at the end). - **Eckersley, T.L.** (1937). "Ultra-Short-Wave Refraction and Diffraction." *J.I.E.E.* 80, p. 286. The 1937 spherical-Earth diffraction calculation that correctly predicted Knickebein beams could not reach England on a globe. [[1937_Eckersley_Ultra_Short_Wave_Refraction_Diffraction]] - **Bundesarchiv-Militärarchiv Freiburg, RL 19-6/40, ref. 230Q7** (1939). Luftwaffen-Führungsstab Ic Nutzbereich (usable-region) chart for the Large Knickebein. Defines the usable altitude/range region as *"based on beacon signals audibility"*. [[1939_BArch_RL19-6-40_230Q7_Nutzbereich]] - **Bundesarchiv-Militärarchiv Freiburg, RL 19-6/40, ref. 230Q8 Appendix 2** (10 September 1939). *"Reichweite von UKW-Leitstrahlen über See bei 4000 m Flughöhe"*. Telefunken range measurements from July 1939, over open water, at 4,000 m aircraft altitude. Primary source for the Telefunken range table in this document. [[1939_BArch_RL19-6-40_230Q8_App2_Telefunken_Range_Tests]] - **Fletcher, H.** (1940). "Auditory Patterns." *Reviews of Modern Physics* 12(1), 47-65. Classical critical-band masking theory. [[1940_Fletcher_Auditory_Patterns]] - **Cameron, D. and Annand, W.J.D.** (1942). *Measurements of Cabin Noise in Bomber Aircraft.* A.&A.E.E. Report No. Res./162, January 1942; ARC R&M 2296, HMSO 1948. Calibrated octave-band SPL measurements at wireless-operator positions on eight WWII bombers. [[1942_Cameron_Annand_ARC_RM_2296_Bomber_Cabin_Noise]] - **D.(Luft) T.4058 Geräte-Handbuch FuBl 2** (February 1943). Luftwaffen-Führungsstab Ic. *Funk-Landegerät FuBl 2 Geräte-Handbuch*. 120-page service manual. §21 defines FuBl 2 sensitivity operationally. Confirms 1150 Hz audio tone, 30-33.3 MHz band. [[1943_DLuft_T4058_FuBl_2_Geraete_Handbuch]] - **Fock, V.** (1945). "Diffraction of Radio Waves Around the Earth's Surface." *Zhur Eksp Teor Fiz* 15, 479-496. DRIC Transl. No. 2747, April 1972. Reprinted as Ch. 10 in Fock (1965) *Electromagnetic Diffraction and Propagation Problems* (Pergamon, IMEM Vol. 1). Residue series at Eq. (6.10) p. 209; Weyl-van der Pol flat-Earth at Eq. (3.23) p. 201. [[1945_Fock_Diffraction_Radio_Waves_Earth]] - **Friis, H.T.** (1946). "A Note on a Simple Transmission Formula." *Proc. IRE* 34(5), 254-256. Flat-Earth link-budget baseline. [[1946_Friis_Simple_Transmission_Formula]] - **Vogler, L.E.** (1961). "Smooth Earth Diffraction Calculations for Horizontal Polarization." *NBS J. Res.* 65D(4), 397-399. - **Neubauer, W.G., Ugincius, P., and Uberall, H.** (1969). "Theory of Creeping Waves in Acoustics and Their Experimental Demonstration." *Z. Naturforsch.* 24a, 691-700. Standard specialisation of Fock Eq. 6.10 for a perfectly conducting surface. - **Jones, R.V.** (1978). *Most Secret War*. Hamish Hamilton. Discovery, Bufton's Anson flight, Enigma intercepts, Derby target. pp. 100-102 Bufton at Spalding; p. 163 Enigma; pp. 175-176 frequencies; pp. 181-182 Bufton details. [[1978_Jones_Most_Secret_War]] - **Trenkle, Fritz** (1979). *Die deutschen Funk-Navigations- und Funk-Führungsverfahren bis 1945*. AEG-Telefunken. German primary history of Luftwaffe radio navigation. p. 67 for 28 m antenna frame dimension. [[1979_Trenkle_Deutsche_Funk_Navigation]] - **Shatz, M.P. and Polychronopoulos, G.H.** (1988). "An Improved Spherical Earth Diffraction Algorithm for SEKE." Project Report CMT-111, ESD-TR-88-122. MIT Lincoln Laboratory. ADA195847. US Air Force production radar propagation code, independent Fock residue-series implementation. [[1988_Shatz_Spherical_Earth_Diffraction_SEKE]] - **Price, Alfred** (2003). "A New Look at 'The Wizard War'." *RAF Historical Society Journal* #28, pp. 15-23. German bomber-crew testimony that Knickebein signals were *"easily heard through the low-power jamming"*. [[2003_Price_Wizard_War_RAFHS28]] - **Bauer, Arthur O.** (2004). "Some historical and technical aspects of radio navigation, in Germany, over the period 1907 to 1945." cdvandt.org *Navigati.pdf*. Primary-source data on antenna dimensions, transmitter power, 1936 X-beam trials, rated max range. [[2004_Bauer_German_Radio_Navigation_1907_1945]] - **Price, Alfred** (2017). *Instruments of Darkness: The History of Electronic Warfare, 1939-1945*. Frontline Books. Station-target pairings (pp. 35-38), antenna dimensions (pp. 24-25), Liverpool operational raids (p. 35). [[2017_Price_Instruments_of_Darkness]] - **ITU-R P.833-10** (2021). "Attenuation in vegetation." International Telecommunication Union. Ground-level vegetation-attenuation models, not applicable to the 6,000 m air-to-ground Knickebein geometry. [[2021_ITU-R_P833-10_Vegetation]] - **Dörenberg, F.** (2004-2024). *Hellschreiber pages: Knickebein*. nonstopsystems.com. Wayback-snapshot compilation of Luftwaffe primary-source documentation with direct citations to the Bundesarchiv files. [[2024_Doerenberg_Knickebein_Reference]] - **ITU-R P.526-16** (2025). "Propagation by diffraction." International Telecommunication Union. Smooth-Earth diffraction standard; the piecewise height gain G(Y) in this document is Eq. 17-18. [[2025_ITU-R_P526-16_Diffraction]] **Standing references** (undated ITU Recommendations, textbooks, and supporting materials): - **ITU-R P.372-16**. "Radio noise." Galactic synchrotron noise at 31 MHz (Eq. 14, curve E). - **ITU-R P.453**. "The radio refractive index: its formulae and refractivity data." Effective Earth radius $k = 4/3$ model. - **ITU-R P.368**. "Ground-wave propagation curves for frequencies between 10 kHz and 30 MHz." Reference Fortran implementation of the full Fock residue series (LFMF-SmoothEarth). - **Balanis, C.A.** *Antenna Theory: Analysis and Design*, 4th ed. Aperture-antenna directivity formula $G = 4\pi A / \lambda^2$. [[Balanis_Antenna_Theory]] - **Born, M. and Wolf, E.** *Principles of Optics*, 7th ed. Half-power beamwidth of a uniformly illuminated rectangular aperture. - **Goodman, J.** *Introduction to Fourier Optics*, 3rd ed. - **NATO STO AGARD-AG-300 Vol. 10** *Historical Review of Aviation Hearing Conservation*. WWII aircraft headsets had 0-8 dB passive attenuation, no real hearing protection. [[NATO_STO_AGARD_AG300_Vol10_Aviation_Hearing]] - **Radiomuseum.org EBL 3** (Blindlandegerät E.Bl.3). Technical specs: 7 × RV12P2000 tubes, superhet, 34 channels at 100 kHz spacing, 30-33.3 MHz band. [[Radiomuseum_EBL3_Blindlandegeraet]] A complete extraction of numerical claims and direct quotes from downloaded primary sources is in **`BotB/sources/README_primary_sources_digest.md`**. --- ## Source Code The Python scripts that compute every number in this analysis are open source: [github.com/AlanSpaceAudits/battle-of-the-beams](https://github.com/AlanSpaceAudits/battle-of-the-beams) - `botb_propagation.py`: primary analysis of geometry, Fock diffraction, link budgets, equisignal crossover - `equisignal_prediction.py`: equisignal corridor width from first principles - `equisignal_globe_honest.py`: step by step walkthrough of what happens to the beam on a sphere - `telefunken_4000m_comparison.py`: Telefunken range table comparison at 4,000 m altitude --- ## See Also - [[GRWAVE_P368_BotB]] - [[DC_Dan]] - [[1937_Eckersley_Ultra_Short_Wave_Refraction_Diffraction]] - [[1939_BArch_RL19-6-40_230Q7_Nutzbereich]] - [[1939_BArch_RL19-6-40_230Q8_App2_Telefunken_Range_Tests]] - [[1940_Fletcher_Auditory_Patterns]] - [[1942_Cameron_Annand_ARC_RM_2296_Bomber_Cabin_Noise]] - [[1943_DLuft_T4058_FuBl_2_Geraete_Handbuch]] - [[1945_Fock_Diffraction_Radio_Waves_Earth]] - [[1946_Friis_Simple_Transmission_Formula]] - [[1978_Jones_Most_Secret_War]] - [[1979_Trenkle_Deutsche_Funk_Navigation]] - [[1988_Shatz_Spherical_Earth_Diffraction_SEKE]] - [[2003_Price_Wizard_War_RAFHS28]] - [[2004_Bauer_German_Radio_Navigation_1907_1945]] - [[2017_Price_Instruments_of_Darkness]] - [[2021_ITU-R_P833-10_Vegetation]] - [[2024_Doerenberg_Knickebein_Reference]] - [[2025_ITU-R_P526-16_Diffraction]] - [[Balanis_Antenna_Theory]] - [[NATO_STO_AGARD_AG300_Vol10_Aviation_Hearing]] - [[Radiomuseum_EBL3_Blindlandegeraet]] - [[X_Y_Gerat_Propagation_Null]] - [[00_Null_Hypothesis_Index]]