Starship Super Heavy acoustics: Comparing launch noise from Flights 5 and 6 Open Access

SpaceX's Starship Super Heavy (i.e., “Starship”) is one of several heavy-lift rockets being developed or flown. With seven flights complete and rapid flight test schedule, Starship is on its way to becoming a fully reusable, multipurpose launch vehicle (SpaceX, 2024). Starship's possible community and environmental impacts are of interest because of its thrust, increasing launch cadence, and return-to-pad “flyback” maneuvers. Prior Starship studies include ground cloud prediction (Wongprapinkul , 2024) and the supersonic vehicle's shock wave and explosion impact on the ionosphere (Yasyukevich , 2024). Starship's noise is of considerable interest; the noise from development and testing is mentioned by Seedhouse (2022) as having generated local protests. Pilger and Hupe (2024) studied the long-range infrasound generated by Starship's Flight 2 in November 2023 and discussed multiple acoustical arrivals at the Caribbean Guadeloupe infrasound monitoring station (3800 km from Starbase, TX), including liftoff noise and explosions of both the Starship second stage and the Super Heavy booster.

Of direct relevance to this Letter are recent far-field acoustical measurements during Starship's fifth test flight. The resulting Letter (Gee , 2024b) discussed both the launch noise and flyback sonic boom associated with the Super Heavy booster catch in the context of environmental assessment (EA) predictions [Federal Aviation Administration (FAA, 2022)] that used the “RNoise” model (Sutherland, 1993; Plotkin , 1997; Plotkin, 2010). This Letter builds on the prior study in discussing the launch noise of Flight 6 relative to Flight 5 and an updated EA (FAA, 2024). Comparisons between Starship's launch noise with NASA's Space Launch System (SLS) (Gee , 2023; Kellison , 2023) and Falcon 9 (Mathews , 2021; Anderson , 2024) are updated. Finally, since Starship is expected to launch from NASA's Kennedy Space Center (KSC) and, eventually, nearby Cape Canaveral Space Force Station (CCSFS), launch noise estimates in and around KSC are discussed.

During liftoff of Flights 5 and 6 from Starbase's Orbital Launch Pad A (OLP-A), Starship Super Heavy's 33 Raptor 2 engines produced 74.4 MN (16.7 × 10⁶ lb) of thrust, with a combined-plume mechanical power of 127 GW. Relative to SLS (Kellison and Gee, 2023; Kellison , 2024), Starship produces 1.9 times greater thrust (39.1 MN) and 2.4 times greater mechanical power (∼52 GW). Compared to Falcon 9 Block 5, Starship generates nearly 10 times more thrust (7.6 MN) and almost 11 times more mechanical power (11.8 GW) (Mathews , 2021). For the two flights, the launch profiles were similar, but the result of the planned booster flyback and catch was different. For Flight 5 (Gee , 2024b), the Super Heavy booster returned to OLP-A about 7 min after liftoff, resulting in a flyback sonic boom, rocket noise during landing, and an additional sonic boom due to the hot-staging ring falling from high altitude into the ocean. However, a communications antenna was damaged during Flight-6 liftoff, resulting in an ocean Super Heavy booster landing about 35 km from OLP-A and a subsequent explosion. Although the flyback of Flight 6 and hot-staging ring booms were detected at several stations, this Letter focuses solely on Starship's launch noise.

The two launches' weather conditions also differed. Flight 5 lifted off at 07:25 am local on 13 October 2024 with nearly zero surface wind, but Flight 6 (04:00 pm local on 19 November 2024) experienced 6–7 m/s surface winds from the north-northeast. Radiosonde launches from Brownsville Airport (Durre , 2023) showed that after a near-ground temperature inversion during Flight 5 (+5 °C over the first 120 m), temperatures followed the same altitude trends for both flights, with the tropopause at ∼16 km. However, wind conditions were markedly different, resulting in the effective sound speed, $ceff$ (the sum of the thermodynamic sound speed and the heading-dependent wind velocity component), depending greatly on wind and observer directions. Figure 1 shows $ceff$ from Flights 5 and 6 for sound propagation to the north (N) and to the southwest (SW), spanning the measurement array's angular aperture beyond 10 km. As a key input to acoustic ray tracing [e.g., Salomons (2001)], $ceff$ provides insights into the acoustic propagation for a particular altitude and heading. As Fig. 1 shows, $ceff$ for Flight 6 varied much more with altitude and direction than Flight 5, with a much lower SW $ceff$⁠. In Sec. 4, this variation is used to explain data trends.

For Flight 5, acoustic pressure waveforms were recorded at eight sites 9.7–35.5 km from OLP-A (Gee , 2024b). As shown in Fig. 2, 21 stations from 1.00 to 35.5 km were deployed for Flight 6 to better characterize the propagation and compare against EA predictions. Station numbers mostly increase with distance and from north to south for similar distances. Six repeated measurement stations from Flight 5 are marked in blue. Stations 15 and 17 were in the same areas as during Flight 5 but were relocated to be farther from roads and crowds. Data acquisition was carried out as described by Gee (2024b) and references therein, using 6.35 mm (0.25 in.) GRAS 46BG, 46BD, and 46BE microphones for Stations 1–9 and 12.7 mm (0.50 in.) GRAS 146AE and 47AC microphones for the remainder. All microphones were located on a ground plate in a weather-robust windscreen. Each station was at ground level except for Station 11, which was located on the Margaritaville Hotel roof. Because of significant low-frequency energy in the radiated noise, a pole-shift digital filtering technique (Rasband , 2023; Marston, 2008) was used to compensate for instrumentation response roll-off. This correction was mostly important for the 6.35 mm microphones, with corner frequencies of ∼3 Hz (Anderson , 2024).

Our prior Flight-5 Letter described four launch noise metrics: the maximum unweighted (Z-weighted) level, $LZ,max$⁠, and maximum A-weighted level, $LA,max$⁠, both calculated over a 1s interval, as well as the cumulative Z (⁠ $SELZ$⁠) and A-weighted (⁠ $SELA$⁠) sound exposure levels. Figure 3 from Gee (2024b) showed these levels as a function of distance from OLP-A, as well as EA predictions (FAA, 2022) for $LZ,max$⁠, $LA,max$⁠, and $SELA$⁠. Figure 3 repeats the four Flight-5 plots but adds the Flight-6 data and least-squares fits to updated EA predictions (FAA, 2024). These updated predictions, made for a 103 MN-thrust version of the Super Heavy booster, increase $LZ,max$ and $LA,max$ by ∼1.5 dB at a given distance over the prior EA (FAA, 2022), which was for a 61.7 MN vehicle. The as-flown Super Heavy, with a thrust of 74.4 MN, falls between the two modeled scenarios. However, the predictions have not been adjusted because the updated EA also describes modeling improvements that particularly affect $SELA$⁠, an area of disagreement with Flight-5 levels. Differences in vehicle thrust represent a relatively minor correction (<2 dB).

For $LZ,max$ in Fig. 3(a), the combined-flight levels match the EA prediction to within 2.5 dB out to almost 20 km. However, for the three ∼10 km stations repeated for both flights (Stations 11–13 in Fig. 2), Flight 5 is 3–4 dB greater than Flight 6. Given that the Flight-5 and 6 launch trajectories were nominally the same, the cause must be meteorological. The maximum launch noise at 10 km is emitted at an altitude of ∼3.4 km, based on a peak noise directivity angle of 71° (re plume) that is predicted from the Raptor 2's plume velocity (Gee , 2024a). In Fig. 1, relatively large differences occur for $ceff$ beginning around 3–4 km. However, because its change for the N/SW headings is not consistent between the two launches, the 3–4 dB difference in $LZ,max$ at 10 km cannot be fully explained with these data. This difference in $LZ,max$ at 10 km also increases the uncertainty in the Flight-6 levels for lesser distances; the data of Flight 6 may not be representative of all Starship launches.

At distances greater than 10 km in Fig. 3(a), a split occurs in the $LZ,max$ trends, as $LZ,max$ of Flight 5 decays more slowly than much of the Flight-6 data. However, Station 14 (16.6 km) measured the same level for both launches and Station 18 (27.1 km) is 6 dB greater than the Flight-6 trend, making these data closer to the level decay of Flight 5. The $ceff$ of Fig. 1 may explain this split; this is done after discussion of all four metrics.

Figure 3(b) shows $LA,max$ decaying smoothly up to 10 km for Flight 6. At around 10 km, however, there is a ∼10 dBA difference between Station 11 and Station 13. Station 11 follows the Flight-5 decay, whereas Station 13 appears to match the more rapid roll-off of most of the other stations. By 35.5 km (22 miles), the difference between the Flight-5 and Flight-6 levels has grown to 34 dBA. In the Flight-5 analysis, we suggested that the measured $LA,max$ generally matched the EA. However, the Flight-6 data inside 10 km demonstrates a growing disagreement with the EA: at 1 km, the EA overpredicts the $LA,max$ $of Flight 6$ by ∼9 dBA. While there could be some uncertainty in digitization and fitting of the EA contour plots, the growing discrepancy is smooth for distances between 1 and 10 km. Even allowing for thrust differences, this result suggests the “RNoise” model (Sutherland, 1993; Plotkin , 1997; Plotkin, 2010) does not treat the frequency-dependent source levels or directivity appropriately.

Figure 3(c) shows $SELZ$⁠, which does not have an EA comparison. For Flight 6, the difference between the $SELZ$ and $LZ,max$ grows from 11 dB at 1 km to 18 dB at 35.5 km. This difference means the launch cumulative sound exposure is equivalent to $LZ,max$ exposure for 13 s at 1 km and 63 s at 35.5 km. However, the Flight-5 differences were less than those of Flight 6: 12–13 dB from 10 to 20 km and 8 dB at 35.5 km. This lesser difference confirms the hypothesis in the Flight-5 Letter that the $LZ,max$ was driven by short-term, high-amplitude events. These events were less prominent during Flight 6, resulting in a greater offset between the $SELZ$ and $LZ,max$⁠. In fact, these high-amplitude transients during Flight 5 explain much of the 3–4 dB $LZ,max$ difference around 10 km, though not the meteorological cause.

Finally, $SELA$ is compared in Fig. 3(d) for Flights 5 and 6, along with the updated EA prediction. In the earlier Flight-5 analysis, we pointed out large discrepancies between the EA and data (18 dBA at 35.5 km). Although the updated model aligns more closely with the Flight-5 data, at 1 km the EA offset from the Flight-6  $SELA$ is larger than for $LA,max$⁠. These overpredictions for A-weighted levels again indicate a need to revisit EA models that presently overpredict both maximum and integrated A-weighted levels.

The decays in sound levels for distances greater than 10 km qualitatively match for all four metrics in Fig. 3. To the north of OLP-A, Flight 6's Stations 11, 14, and 18 follow the Flight-5 data trends, whereas at the stations ranging from northwest to west-southwest, Flight-6 levels roll off more quickly. The $ceff$ profiles in Fig. 1 point to a plausible qualitative explanation for the differences. From 10 to 35.5 km, and assuming a vertical trajectory (trajectory data are not presently available), the highest sound levels (maximum noise) are roughly generated with the rocket at a height of 3.4–12.2 km. However, as the rocket pitches east, the maximum noise emission altitude for a given observer range will be less. So, the maximum noise arriving at 10–35.5 km is likely generated at a vehicle altitude of ∼3–10 km. Figure 1 shows that in this altitude range, the Flight-6  $ceff$ gradient to the southwest is particularly negative, more so than to the north. Moreover, the differences between the north and southwest $ceff$ are much less for Flight 5. The negative $ceff$ gradient causes upward refraction of the radiated noise to the west and south of OLP-A more than to the north, qualitatively explaining the different trends for the Flight-6 north stations. If full trajectory data become available in the future, ray tracing should be carried out to study these phenomena further. However, regardless of the cause(s), the marked differences between Flight-5 and -6 levels demonstrate the potential variability and present uncertainty of Starship noise at distances important to estimating community impacts.

This far-field noise metric variability suggests that a key Flight-5 finding should be revisited. The Starship launch noise was previously compared to SLS (Gee , 2023; Kellison , 2023) and Falcon 9 (Mathews , 2021; Anderson , 2024) in terms of the number of equivalent launches. Curve fits were obtained for each dataset to provide the most consistent community noise comparison at distances of 10 and 20 km. A look at all four metrics and both distances resulted in the conclusion that a Starship launch was the acoustical equivalent of 4–6 SLS launches and at least 10 Falcon-9 launches. We stress that this comparison referred to equivalent acoustic energy, not a relative increase in human perceived loudness, i.e., Starship is not 10 times as loud as Falcon 9. Perceived loudness of noise from rockets is a topic that is not well understood and merits additional study.

Here, with Flight-6 noise data closer to OLP-A, the comparison is made using 1–10 km data for all three vehicles to produce a least-squares prediction at 1 km. For SLS, the closest measurement locations were 1.4–1.5 km, so an extrapolation was performed. Table 1 compares the vehicles' relative launch noise at 1 km for all four metrics in Fig. 3. Results are shown as vehicle equivalence ratios, e.g., Starship:SLS. For example, for $SELZ$⁠, the sound level from one Starship launch is equivalent to the level produced by 16.7 Falcon 9s launching from the same location at the same time. Also, shown in Table 1 are means of the metric-based equivalence ratios and ratios predicted solely from the vehicles' relative mechanical powers, which are linked to acoustic power through the acoustic efficiency. Although the acoustic efficiency for the Super Heavy booster has not yet been calculated, the acoustic efficiency of Falcon 9 has been estimated by Mathews (2021) as 0.34% and Kellison and Gee (2023) recently found an efficiency of 0.33% for SLS. At present, Starship's efficiency is assumed to be equal.

Table 1 represents an update to the Flight 5-based vehicle comparison (Gee , 2024b) that used data more subject to atmospheric variability. While the prior comparison was valid for that launch and those data, close-range data should represent more accurately and consistently the source characteristics for a given metric. For Starship vs SLS, the prior comparison that one Starship launch was equivalent to 4–6 SLS launches has been updated to 2.0–2.5 SLS launches, with a metric average of 2.2. Note that these values are closely approximated by the mechanical power ratio, which suggests that one Starship launch is the equivalent of 2.4 SLS launches. Current 1-km results are in line with what is expected from relative mechanical powers.

While metric-to-metric differences for Starship and SLS's equivalence comparison are relatively small, the Starship:Falcon 9 comparisons in Table 1 are more metric-dependent. For the A-weighted metrics, the vehicle equivalence ratios are smaller; one Starship launch is 4.8 and 9.2 Falcon-9 launches for $LA,max$ and $SELA$⁠, respectively. For the unweighted metrics, the differences are larger: one Starship to 12.7 and 16.7 Falcon 9s for $LZ,max$ and $SELZ$⁠, respectively. While the average, 1:10.8, happens to match the mechanical power equivalence ratio, the results highlight the differences with unweighted vs A-weighted metrics. The equivalence ratios based on A-weighted metrics are less than their unweighted counterparts because Falcon 9 has a smaller effective diameter across its nine engines relative to Super Heavy and generates more sound power at higher frequencies weighted more strongly by A-weighting. The large differences in equivalence ratios indicate the importance of determining which weighting curve is most appropriate to assess human and wildlife responses to both the rumble and crackle present in rocket noise. Nonetheless, this analysis slightly updates the Flight-5 analysis: one Starship launch's noise radiation is approximately that of 11 simultaneous Falcon-9 launches, though this is metric-dependent. Again, these refer to metric-averaged acoustic energy and not to a perceived increase in loudness, a potential point of confusion with the prior Letter (Gee , 2024b).

This Letter's analysis concludes by estimating Starship's $LZ,max$ in eastern Florida, because of ongoing discussions to potentially launch from KSC's Launch Complex (LC) –39A and later from CCSFS's LC-37 at a combined rate of 120 launches per year. This limited analysis is done with caveats. First, the prediction ignores effects of wind or temperature gradients, which can greatly influence long-range sound levels [like during the 2022 Artemis-I launch from LC-39B (Kellison , 2023)]. The KSC Starship EA (NASA, 2019) acknowledges that a “prevailing onshore or offshore breeze may…strongly influence noise levels in…communities” but also does not address detailed weather effects. Here, data from both launches in Fig. 3(a) have been combined and a curve fit produced using a double-exponential function. As seen in Fig. 4(a), the function roll-off beyond 10 km falls between the Flight-5 and -6 data. Second, ground surface differences are ignored. However, a recent analysis (Hart and Gee, 2023) suggests a large rocket's $LZ,max$ at the ground is virtually unchanged for all but the softest ground impedances.

Figure 4(b) shows the projection of the Flight-5/Flight-6 curve fit onto a partial Florida map, from about 15 km east of LC-39A to Orlando. Identified are several cities: Orlando, Titusville, Merritt Island, Cape Canaveral, and Cocoa Beach. Per Fig. 4(b), the area around LC-39A will be exposed to $LZ,max$ > 130 dB inside a few kilometers and 120 dB within 12 km. Outside KSC, the 110 dB contour reaches beyond Titusville to the west and Cape Canaveral to the south. These three contours have historical significance related to different risk criteria for structure damage from rocket noise. Although the most recent draft Starship EA (FAA, 2024) advocates for using impulse noise criteria [i.e., Fenton and Methold (2016)], it is unclear if these are appropriate for continuous, shock-containing noise during launch that could excite structural resonances. Regarding structure damage from continuous noise, one report from the National Research Council (NRC, 1977) suggested no damage is expected for $LZ,max$ < 130 dB. Per this criterion, no structure damage risk exists within residential areas. Conversely, Guest and Sloane (1972) studied resident complaints and structure damage claims generated from 1960s Saturn-V first-stage static fires at Marshall Space Flight Center in Huntsville, Alabama. They found that 111 ± 3 dB (95% confidence) correlated with residential structure damage claims at the rate of 1 for every 1000 households. By $LZ,max$ = 120 ± 3 dB, the rate increased to 1 claim for every 100 households.

A “1 claim for 1000 households” for $LZ,max$ > 110 dB was adopted as a significance threshold in past rocket environmental documents [e.g., U.S. Department of the Air Force (1998), NASA (2013), and FAA (2016)]. This criterion is worth examining in Fig. 4(b) while noting three caveats. First, Guest and Sloane (1972) did not characterize damage severity. Second, they stated that only 20% of claims were judged valid, i.e., a claim only rarely indicated actual damage. Third, we do not know differences in construction between 1960s homes around Huntsville, Alabama and present-day structures in eastern Florida. Thus, applied to Starship or other heavy-lift rockets, the Guest and Sloane (1972) study may be more indicative of resident concerns about structural vibrations than actual structure damage.

Depending on how representative the curve fit of Fig. 4(a) is of local sound propagation, the 110 dB contour in Fig. 4(b) suggests likely structural vibrations of nearby Florida residences during Starship launches. From U.S. census and GIS data, the 110 dB contour includes a population of ∼135 000 residents and ∼55 000 households, assuming 2.43 residents per household in Brevard County, FL. To contrast Starship with Falcon 9 launches that occur from both KSC and CCSFS, the data used to create Table 1 indicate Falcon 9's $LZ,max$ = 110 dB contour is limited to 4–5 km. This result, which is well away from homes, agrees with predictions (FAA, 2016). If Starship launches from CCSFS's LC-37 (∼9 km south of LC-39A), sound levels to the south would increase such that Cocoa Beach and Merritt Island would fall within the 110 dB contour. However, noise levels in environmentally sensitive areas to the north (e.g., the Merritt Island National Wildlife Refuge) would be reduced.

The limits of Fig. 4(b) include a portion of Orlando. Although weather could cause actual levels to vary by at least 10 dB from those shown, the two-launch curve fit in Fig. 4(a) predicts a maximum level of 90 dB at the eastern edge of the greater Orlando area and 80 dB near the center. Although A-weighted levels will likely be below ambient levels, considerable low-frequency energy (i.e., infrasound) may persist in Orlando. Whether these levels will induce noticeable structural vibrations or rattle may be answered by additional long-range measurements and resident surveys around the Boca Chica region during Starship launches.

This Letter's comparison of Flights 5 and 6 expands understanding of Starship Super Heavy's launch noise beyond the prior Flight-5 analysis (Gee , 2024b). Additionally, the comparative analysis with other vehicles expands our knowledge of heavy-lift rocket acoustics in general. Although many questions and analyses remain, this Letter communicates the following findings. First, both unweighted and A-weighted maximum and sound exposure levels inside 10 km follow a relatively smooth trend, regardless of station heading. Beyond 10 km, Flight-6 levels decay more rapidly than Flight 5, except for stations north of the launch pad. Radiosonde data suggest that, for noise generated in the first several kilometers of Starship's Flight-6 ascent, stronger upward atmospheric refraction occurred to the west than to the north. Second, environmental assessment predictions closely match measured Flight-6 unweighted maximum levels (⁠ $LZ,max$⁠) out to 10 km, but beyond that, the EA underpredicts Flight-5 sound levels and overpredicts Flight-6 levels. For A-weighted metrics, the EA model significantly overpredicts Starship's actual noise; the conclusion from the prior analysis (Gee , 2024b) that the EA adequately matched $LA,max$ was based on the range-limited data of Flight 5. Third, the comparative analysis between Starship, Space Launch System (SLS), and Falcon 9 has been updated by focusing on levels at 1 km. One Starship launch is the acoustic-energy equivalent of 2.2 SLS launches and nearly 11 Falcon 9 launches, though this latter equivalence depends on whether A-weighting is applied. Finally, an estimate of $LZ,max$ around Kennedy Space Center in Florida suggests that notable structural vibrations may be generated by Starship launches in nearby communities and infrasound will persist at greater distances. As Starship and other heavy-lift rockets launch increasingly often, this Letter contributes to understanding their potential effects on surrounding structures, people, and the environment.

No conflicts to disclose.

Some data that support this study's findings are available upon written request to the corresponding author.