The electric field changes in ELF to VLF were observed with a ball antenna in fair weather at Kochi (latitude 33.3°N, longitude 133.4°E) during 2003–2004. Some 376 Q bursts were obtained, seven examples of which are analyzed in the present study. The continuous frequency spectra of the Q bursts and the background noises from 1.0 Hz to 11 kHz are compared, and it was found that the Q bursts prevail over the background in the frequency range from 1 to 300 Hz. The surplus is 20 dB (in amplitude) near the fundamental mode frequency. The “W”-type changes found in the initial portion of the Q burst waveforms are interpreted as the combined electromagnetic waveform of direct and antipodal waves from the causative lightning strokes. From the time intervals between the two waves, the source-receiver distances are estimated as far as 19 Mm. The pulses to excite the Schumann resonances in the Q bursts are clearly identified.
 The Q burst is an isolated large transient damping oscillation lasting for 0.3–1.5 s. When the Q burst was identified and named in a narrow band observation in 1966 [Ogawa et al., 1966a, 1967], the Q burst oscillated at 8 Hz that is the fundamental mode frequency of the Schumann resonances in the Earth-ionosphere cavity. There occur also the Q bursts of 14 Hz, the second mode frequency [Ogawa, 2002]. In order to see mechanism of the Q bursts, it is necessary to observe more precise waveforms in wider frequency range.
 After 29 years from the first identification Boccippio et al.  found that the Q burst was an electromagnetic (EM) wave emitted from the large positive polarity lightning stroke related to the sprite that was a newly discovered upper atmospheric optical emission [Sentman and Wescott, 1993]. The observed largest charge moment changes are associated with the positive cloud to ground strokes that simultaneously produce sprites and Q bursts in the Earth-ionosphere cavity [Huang et al., 1999]. The negative polarity Q bursts also occur about 15% of all Q bursts [Jones and Kemp, 1971; Ogawa, 2002].
 It is interesting to see a relation of the Q bursts observed distant in a global scale with the causative lightning strokes and sprites. Sato et al.  observed the Q bursts at Syowa station in Antarctica relating with the sprites observed in Colorado. E. R. Williams et al. (Sprite lightning heard around the world by Schumann resonance, submitted to Radio Science, 2006, hereinafter referred to as Williams et al., submitted manuscript, 2006) observed the Q bursts in Rhode Island correlating with sprite-producing lightning flashes in northern Australia, at a distance of 16.6 Mm.
 The slow tail atmospherics have been explained to be associated with the continuing current in lightning flashes. Reising et al.  linked the sprites with the slow tails. Wait , however, argued that purely impulsive lightning without continuing current would be capable of creating a slow tail, simply on the basis of waveguide dispersion. It is necessary to make clear the relation between the slow tails and the Q bursts in a simultaneous observation. From a radio propagation point of view, the measurement of Q bursts with the direct and antipodal waves from the lightning strokes occurring near the antipode is interesting, although it is explained as only in the theoretical discussion [Nickolaenko and Hayakawa, 2002].
 The purpose of the present study is to measure the electric field changes from ELF to VLF over a wide frequency range to obtain precise Q burst waveforms in a global scale. From the analysis of these waveforms, the detailed features of Q bursts in both the time domain and the frequency domain are examined. It is investigated how the Q burst will be produced by the emitted waves from distant lightning strokes, especially near the antipode. The antipode of the present observation site locates in the southeastern part of South America (latitude 33.3°S, longitude 46.6°W). The source-receiver distances of causative lightning strokes of the Q bursts are estimated by measuring the time intervals between the direct and antipodal waves.
 The ball antenna was installed at Tochi station (latitude 33.3°N, longitude 133.4°E) in Kochi city (with the population of 0.33 million), Shikoku, in fair weather afternoon, and the ELF to VLF waves were observed for 1–2 hours on some ten days, during 2003–2004. The observed signals have been stored in a personal computer through the AD converter at 22 kHz.
 The frequency response of the present ball antenna is effective from 1.0 Hz (−20.6 dB) to 11 kHz (−1.09 dB) [Ogawa et al., 1966b]. When the observed signals are analyzed, a narrow band (4.0–30 Hz) filter is applied to discern the Q bursts from the background noises. The frequency responses of the system (wide band and narrow band) are shown in Figure 1. The antenna height was set at 3.8 m.
 The observed signals were usually documented with the strong commercial power line noises at 60Hz, which were eliminated by mixing with artificially produced antiphase 60 Hz signals when the signals were analyzed. It is, however, difficult to eliminate whole 60 Hz noises. Some 60 Hz power remains by 10–20 dB.
3. Results and Discussions
 Some 376 Q bursts qualifying for detailed analysis have been obtained. Seven of these Q bursts were analyzed in the present study as follows.
3.1. Q Burst in the Frequency Domain
 In order to see the whole picture of the Q burst in the frequency domain and to see the effect of the chosen band-pass filter, the amplitude spectra for 1 s in the wide and narrow bands are calculated by FFT and compared. An example is shown in Figure 2.
 The amplitude spectrum (Figure 2a) is obtained throughout the frequency range from 1.0 Hz to 11 kHz in a single display. Although the antenna attenuation is −20.6 dB at 1.0 Hz (Figure 1), the presentation of the Q burst like Figure 2a is adequate to see the total picture of the Q burst, because the strong ULF component is properly limited. In the traditional estimation of ELF-VLF frequency spectra, measured values in several frequency bands were jointed together [e.g., Fraser-Smith et al., 1991].
 The Schumann resonance spectral peaks can clearly be detected at the first four mode frequencies, 7.9, 13.5, 20, and 26 Hz. The strong signal attenuation at 28 Hz is characteristic in the data on this day, 5 November 2004, but any reasonable explanation is not given at the moment.
 The bristle-type signals in 300 Hz to 3 kHz overlapping on natural signals may be higher harmonics of 60 Hz and other artificial signals of little interest in the present study. The amplitude spectral level of the Q burst and the associated sferics (Figure 2a) lies between −40 dB and −110 dB (in amplitude), the antenna height being 3.8 m. The spectrum gives the frequency roll-off with −40 dB/decade over the range 50–500 Hz.
 It is shown in the narrow band spectrum that the circuit electronic noise in the measurement system seems to generate at the maximum level of −115 dB in the frequencies higher than 200 Hz. As the minimum level of the frequency spectrum of the Q burst and the associated sferics given in the wide band is −106 dB at the lowest-level frequency of 1.2 kHz, the wide band spectrum represents the most realistic Q burst and associated background noises without any significant technical limitation. The attenuation of EM waves at the cutoff frequency in the Earth-ionosphere waveguide is not sharp at a single frequency but is rather smooth like pan bottom over 800 Hz to 3 kHz in this example.
3.2. Q Bursts in the Time Domain
 Three different types of Q bursts, “W,” “V,” and “NST” are discussed in the following. The “W”-type Q bursts are the most characteristic found in the present study. The “W” in the initial portion of Q burst has two minima. They are interpreted as the combined waveform of direct and antipodal waves from the causative lightning stroke referring to Nickolaenko and Hayakawa  and E. R. Williams and R. Boldi (private communication, 2006).
 Appearance of positive humps at the end of “W” is also characteristic. They repeat a few times. Such recurrence of positive humps indicates that each hump is arrival of the wave traveling around the globe. The wave velocity can be estimated from the circumference of the Earth, 40 Mm divided by the time interval between the first and second humps.
 The source-receiver distance D (Mm) can be estimated from the time intervals between the two minima of “W”:
where T is the time interval between the two minima, the arrivals of direct and antipodal waves, and V is the mean wave traveling velocity of 265 Mm/s estimated from the four Q bursts.
Burke and Jones  made the global radiolocation of the Q bursts. They demonstrated the ingenious method to estimate the source distance by calculating the wave impedance of the observed data. They estimated the source distance as 10.0 Mm by this method.
 They presented the electric field waveforms [Burke and Jones, 1995, Figure 2] in the frequency band of 5–45 Hz and showed that the initial pulse is the direct wave, and the second pulse is the antipodal wave “arriving some 72 ms later” [Burke and Jones, 1995, p. 26,268]. If we apply the formula (1) to estimate the source-receiver distance to this example, the distance will be 10.5 Mm. The present estimation value is over the Burke and Jones's estimation by 5%. The present method by measuring the wide band waveforms is simple compared with that calculating the wave impedance to estimate the source-receiver distance of the Q burst.
 The time interval between two minima of “W” in the Q burst, only the spectrum of which is given in Figure 2, is measured as 20.0 ms, then the source-receiver distance is estimated as 17.4 Mm. A possible lightning position on the globe may be located in the central to southeastern part of South America by the equidistance contour map centered at Kochi given in Figure 3.
3.2.1. “W”-Type Q Burst
 A typical example of the “W”-type Q burst observed at 1816.2 s after 0545:00 UT on 5 November 2004 is shown in Figure 4. The electric field is defined as positive when directed vertically downward. The initial negative going excursion of the Q burst electric field is interpreted as a positive charge brought down from the cloud to the ground. The “W” is clearly seen in the wideband waveform given in Figures 4a and 4b but not clearly seen in the narrow band given in Figure 4c.
 A small non-Q burst producing “W” sferic occurred at 1815.7 s, about 0.6 s earlier than the Q burst producing “W” sferic (Figure 4a). Such small “W” sferics occur often in plural numbers near the Q burst. Such behavior of “W” sferics seems like the sprite appearance, but the problem is not discussed in the present study.
 An additional damping oscillation of about 60 Hz was produced right after the “W.” It would be caused by the finite band pass width of the frequency band of the observation system [Nickolaenko and Hayakawa, 2002], because the ball antenna system functions as a kind of high-pass filter as shown in Figure 1, although it has a wide frequency band.
 The Schumann resonance oscillation continued for about 0.8 s as seen in the narrow band (Figure 4c). The Schumann resonance peak frequencies are estimated as 7.9 and 13.8 Hz for the first two resonance modes (Figure 4d).
 The frequency spectrum for 1 s of the Q burst was compared with the background spectrum (Figure 4d). Data for 1 s just before the occurrence of Q burst were used for the background spectrum. The comparison demonstrates that the Q burst power prevails over the background in the frequency range from about 1–300 Hz. The spectral level difference is 20 dB near the Schumann resonance fundamental mode peak frequency. It is interesting to see the identical spectra in the frequencies higher than 300 Hz for the Q burst and the background. The wave “cutoff” frequency in the Earth-ionosphere waveguide is seen at 2 kHz.
 The total spectral level extends from −45 dB at 8 Hz down to −105 dB at 2 kHz. The amplitude of the Q burst electric field change is 30 mV/m (Figure 4b). The deep spectral missing at about 30 Hz of the Q burst is of a special feature on 5 November 2004, and is not properly explained yet.
 The positive hump after “W” and the second hump shown in Figure 4b are used to calculate the wave traveling velocity. Similar estimation was made for the Q bursts in Figures 5, 6, and 7, and the mean velocity was obtained as 265 Mm/s as already explained in section 3.2.
 The time interval between the two minima of “W” is measured as 20.0 ms. Estimation of the source-receiver distance for the causative lightning stroke by the formula (1) results in the distance of 17.4 Mm. Then the antipodal wave was observed after propagating along the path of 22.6 Mm. A possible lightning position may be located in the central to southeastern part of South America in Figure 3.
3.2.2. “W”-Type Q Burst: Comparison With a Normally Observed Positive Polarity Slow Tail
 In Figure 5, the “W”-type Q burst observed at 1000.99 s after 0610:00 UT on 23 October 2004 is compared with a normally (not very often) observed positive polarity slow tail, that happened at 1001.16 s within the duration of Q burst (Figure 5b). It is obvious that this slow tail is much smaller than the Q burst and not related with the occurrence of the Q burst.
 Multiple occurrence of small five or six “W” sferics was observed (Figure 5a). Only the “W” with high right shoulder produced the Q burst. There was no VLF sferic at the starting point of “W” in the Q burst (Figure 5b). The amplitude of the Q burst electric field change is 20 mV/m. The additional about 60 Hz damping oscillation is seen right after the “W” similar to Figure 4b.
 The first mode Schumann resonance peak is seen at 7.9 Hz in the spectrum (Figure 5d). The wave “cutoff” attenuation in the Earth-ionosphere waveguide is not sharp but continuous from 600 Hz to 3 kHz.
 The time interval between the first and second minima in “W” is 13.0 ms (Figure 5b), and the source-receiver distance is estimated as 18.3 Mm. A possible lightning position may be located in the southeastern part of South America in Figure 3.
3.2.3. “W”-Type Q Burst: Comparison With a Normally Observed Negative Polarity Slow Tail
 In Figure 6, the “W”-type Q burst is compared with a normally very often observed negative polarity slow tail that happened within the Q burst. The multiple “W” sferics occurred but only the second “W” at 3886.7 s after 0610:00 UT on 23 October 2004 is shown in (Figure 6b). The two large “W” sferics jointly produced the Q burst (Figure 6c). A normally very often observed negative polarity slow tail occurred at 3886.79 s within the Q burst (Figure 6b). It is clear that the “W” sferics to make the Q burst are much larger than the normally observed slow tail. No VLF sferic occurred at the starting point of “W” (Figure 6b). The amplitude of the Q burst electric field change is 21 mV/m. The additional about 60 Hz damping oscillation was generated right after the “W” similar to Figures 4b and 5b. The first two remarkable Schumann resonance peaks are seen at 7.8 Hz and 14 Hz in the spectrum (Figure 6d).
 The time interval between the first and second minima of “W” in the Q burst is 15.0 ms. The source-receiver distance is estimated as 18.0 Mm. A possible causative lightning position may be located in the southeastern part of South America in Figure 3.
3.2.4. “W”-Type Q Burst Occurring Near the Antipode
 If the two minima of the “W” are very close, the lightning source would be near the antipode. Such a Q burst was observed at 2731.3 s after 0610:00 UT on 23 November 2003 and is illustrated in Figure 7.
 The VLF sferic is not seen at the initial portion of the Q burst (Figure 7b). The amplitude of the electric field change is 29 mV/m. The Schumann resonance oscillation continues for about 0.7 s (Figure 7c).
 The resonance peak frequencies are estimated as 7.8 and 20 Hz, the first and third mode frequencies, respectively, with a lack of the second mode peak frequency (Figure 7d). The time interval between the arrivals of direct and antipodal waves given by the two minima of “W” is measured as only 6.0 ms. The source-receiver distance is estimated as 19.2 Mm. The spectral missing at the second mode Schumann resonances can be interpreted by the theory with this lightning source distance.
 The parent lightning of the Q burst would occur at only 0.8 Mm from the antipode. The corresponding lightning position may be located in the southeastern part of South America in Figure 3.
3.2.5. “V”-Type Q Burst
 The Q burst occurring relatively near the observation point was obtained at 3129.0 s after 0911:10 UT on 8 May 2004 and is shown in Figure 8. Active VLF noises occurred in the background (Figure 8a).
 Two “V” sferics occurred (Figures 8a and 8b) and jointly made the Q burst (Figure 8c). The Schumann resonance oscillation continues for about 0.3 s (Figure 8c). The VLF sferic is seen at the initial portion of the Q burst (Figure 8b). The amplitude of electric field change is 15 mV/m. The Schumann resonance peak frequencies are seen for the second and fourth mode frequencies at 13 Hz and 26 Hz, respectively, with a lack of the first and third mode peaks (Figure 8d).
 The first “V” at 3129.01 s is interpreted as the direct wave and the second “V” at 3129.08 s the antipodal wave from the causative lightning stroke (Figure 8b). The time interval between the two waves is 78.0 ms, and the source-receiver distance is estimated as 9.7 Mm. The corresponding lightning position may be located in the western part of Africa or in the middle of the Pacific Ocean in Figure 3.
 The two pulses of the first and second oscillations of the Q burst caused the prominent second mode frequency of the Schumann resonances. This is consistent with the theory that the Q burst that occurs at 10 Mm distant has the second mode with a lack of first mode. This is a 14 Hz Q burst reported by Ogawa .
3.2.6. Negative Polarity Slow Tail-Type Q Burst
 The negative polarity Q burst is known to occur about 15% of all Q bursts [Jones and Kemp, 1971; Ogawa, 2002]. Such a Q burst produced by the large negative polarity slow tail-type pulse (NST) was observed at 1287.4 s after 0825:00 UT on 17 May 2004, and is shown in Figure 9. The background VLF sferics dominates, suggesting the lightning occurring close to the observation site (Figure 9a).
 A large VLF sferic at the start of the Q burst is characteristic. The time width of the first positive pulse is very short, only 3.5 ms (Figure 9b). The amplitude of the electric field change is 48 mV/m.
 The Schumann resonance can be seen relatively large in the second mode peak at 13 Hz (Figure 9d). The first resonance mode peak at 8.2 Hz is lower than the second mode peak by about 5 dB.
 Two sharp positive pulses are seen in Figure 9b. The first pulse is at 1287.39 s and the second pulse at 1287.46 s. The time interval between the two pulses is measured as 77.0 ms, causing the second mode Schumann resonance (Figure 9d). The first pulse is interpreted as the direct wave and the second pulse is the antipodal wave from the causative lightning stroke. The source-receiver distance is estimated as 9.8 Mm from the time interval between the two pulses. A possible lightning position may be located in the western part of Africa or in the middle of the Pacific Ocean.
 It is clear that the first large positive pulse excites the Schumann resonance, and the second mode frequency was observed at 9.8 Mm distant. This is a different type of 14 Hz Q burst other than the one in Figure 8.
4. Discussions and Conclusions
 The wide band ELF to VLF electric fields were observed and precise amplitude spectra in the frequency domain from 1 Hz through 11 kHz were obtained in a single display. The cutoff frequency in the Earth-ionosphere waveguide was not sharp at a single frequency but smooth over about 2 kHz range from 1 to 3 kHz in many cases.
 The precise waveforms of some 376 Q bursts were obtained, seven of which are analyzed in the present study. The “W”-type field changes were found in the initial portion of the Q bursts. The two minima of “W” are interpreted as the direct and antipodal waves from the causative lightning stroke. As we had no magnetic field data, the positioning of lightning strokes with the traditional wave impedance method [Burke and Jones, 1995; Williams et al., submitted manuscript, 2006] is not possible, but by using the time intervals between the two arrivals of direct and antipodal waves of the Q bursts, the source-receiver distances are estimated for the causative lightning strokes. In this case the wave traveling velocity is estimated. Two methods to estimate the velocity are considered.
 The first method is from the Schumann resonance frequencies. The resonance eigenfrequencies are given by the relation
where c is the light velocity, a the Earth's radius, and n the mode number. The theoretical resonance frequencies are 10.6, 18.4, 26.6, and 33.5 Hz for the first four modes, respectively. On the other hand the experimental resonance frequencies are 7.8, 14, 20, and 26 Hz, respectively. The experimental values of the Schumann resonance peak frequencies satisfy the √(n(n + 1)) dependence in equation (2), but the resonance frequencies are 25% less than the theoretical values. Therefore, if the coefficient (c/2πa) in (2) were 25% smaller than the real value, the theory will fit the observation. In this case, the effective wave velocity must be smaller than the light velocity by 25%, that is, 225 Mm/s.
 The second method to estimate the effective wave velocity is to use the time interval between the first and second wave arrivals, that is, the first and second positive humps of the Q burst waveform are used as shown in Figures 4, 5, 6, and 7. The time intervals are measured with the mean of 151 ms for the time to travel around the globe. Then the mean wave velocity is estimated as 265 Mm/s. This value is 15% larger than the first method.
 In the present paper we tried to estimate the source distance of the Q bursts using the waveforms measured at Kochi. Then the second method value, 265 Mm/s was used in the text. The source-receiver distances were estimated as 17.4, 17.4, 18.3, 18.0, 19.2, 9.7, and 9.8 Mm for the seven Q bursts presented in this study. The pulses to excite the Schumann resonances are clearly identified for the Q bursts at 10 Mm distant shown in Figures 8 and 9. The Schumann resonance peak frequencies of the Q bursts were measured at about 10 Mm and 19 Mm from the sources by using 1 s data, and the second mode and the first mode resonances, respectively, are predominant, consistent with theory.
 From the comparison of the Q burst spectrum with the background noise spectrum for 1 s, it is shown that the Q burst amplitude exceeds the background noise from 1 to 300 Hz with 20 dB excess near the fundamental mode frequency. As a result it is concluded that the Q burst is the phenomenon prevailing over the frequencies from 1 to 300 Hz.
 The EM waves emitted from the sprites have not been recognized in the analyzed data at the distances of 10 Mm and 19 Mm. The small “W” sferics often appear in plural numbers near the Q burst. Behavior of occurrence of such small “W” sferics seems like that of sprites. The problem to make clear the difference between the Q burst producing “W” sferics and the non-Q burst producing “W sferics remains for future work.
 T. O. thanks Earle Williams at MIT for his detailed discussions and kind advice to improve the text. He also thanks Robert Boldi at MIT Lincoln Laboratory for providing us the various sferic waveforms calculated with the computer. The author also thanks Paul Krehbiel at NMT for his help for submitting the paper to AGU.