Cave 1 is an incomplete Vihara (Dwelling Place) [4]. Although this is a traditionally accepted opinion, to some researchers it is an incomplete chaitya.

Cave 3 It is the main chaitya (Prayer Hall). It was initiated by two merchant brothers Gajasena & Gajamitra. Subsequently it was completed by merchant Aparenu. Thera Saumala was overseer of all the work, indicates an inscription in this cave. While entering in verandah, on the jamb there is inscription of Satavahana ruler Yajna Sri Satakarni (2^nd century CE). This cave measures 86.5 ft. × 39 ft. 10 in (from wall to wall). Dagoba, where sacred relic of a venerable person were kept at the time of construction, is 16 ft. in diameter [4] (see Figs. 1a & 1b).

Cave 11 This cave is known as “Darbar Cave”. It served as Vihara (Resting or Dwelling Place) as well as an assembly hall. According to an estimate it was able to house around 500 persons. It measures 73 ft. × 32 ft. [4] (see Fig. 1c).

In a survey of 41 caves reverberant environment were noted in Caves 1, 3, 4, 7, 8, 11, 12, 17, 18, 33, 34, 75 (see Fig. 2).

It can handle sound pressure input up to 120 dB SPL. It is able to record 360° audio. Highest resolution supported is up to 24 bit/96 kHz. Supported recording formats are Ambisonic A (Raw format), Ambisonic B (FuMa and AmbiX). Microphone position is detected automatically when recording starts. This recorder has built-in motion sensor to generate the playback sound for the desired direction from the data recorded in every direction. It also supports two 3D methods: Ambisonics and Binaural. In order to minimize reflections, it is recommended that recorder must be placed as far from walls and the floor as possible during recording. Binaural recording is supported only up to “48 k/8 bit” or “48 k/16 bit” [7]. H3-VR is noisier, but it is lot better than H2n. “H3-VR is a little bright, even with foam + fur, which with a touch of eq manages its tone and noise too.” After decoding with the Zoom Software much self-noise is generated and “spatiality” decreases a bit. A full-sphere capture (360°) in this instrument allows rotations on three axes in post-production of a sound field while in the case of 2D capture, only one axis is allowed. The ambix signal out of the H2n is a totally “legit” ambisonic signal, except that it lacks vertical information. The H3-VR is natively A-format with cardioid capsules. So each capsule is both pressure and velocity. Once converted to B-format (FuMa or ambiX), channel 1 is pressure, channels 2, 3, and 4 are velocity. Analog audio output of H3-VR is unbalanced stereo. So there is a risk of RFI/EMI noise. Harpex software and Array2SH (one of the plug-ins in SPARTA) are more spatially accurate than the Zoom converters. They are also less noisy; Harpex has least noise while converting to other 3D formats [8].

“We can imagine the result: some more or less significant roll-off below 100 Hz, some mild boost in the 2000 to 5000 Hz region and a somewhat irregular response above 10 kHz” [8].

In FOA four signals which are recorded, referred as A-format (raw format without any editing, post processing, conversion to other formats). They are labeled as based on their location as shown in Fig. 3a. A-format do not represent full 3D environment [9]. A to B-format conversion can be done with help of mathematical equations. They represent four monophonic recordings. With this equation four independent signals can be converted into a single four-channel audio file. B-format represent full 3D space using four audio streams: W containing the Sound Pressure Level (SPL) and functions as an omnidirectional recording X (front-to-back), Y (left-to-right), and Z (top-to-bottom) [9]. The equation to reach these values follows:

(1) W=FLU+FRD+BLD+BRU

(2) X=FLU+FRD−BLD−BRU

(3) Y=FLD−FRU+BLD−BRU

(4) Z=FLU−FRD−BLD+BRU

EDT: It is based on 0 to 10 dB of initial decay. Influenced by early reflections, it depends upon measuring position of source-receiver and room geometry [10]. This parameter is measured from a linear regression line fit to the initial portion of the impulse response and computed separately for different frequency of octave band [11].

(5) EDT=RT0,−10(inSec)

D50: It is evaluated at 500 Hz, 1 KHz, 2 KHz, 4 KHz and should have values >50% for good speech intelligibility [10]. D will be 100% if the impulse response does not contain any components with delays in excess of 50 ms [12].

(6) D50=∫050⁡h2(t)dt/∫0∞⁡h2(t)dt

where h²(t)_{0-50 ms} = RIR measured at some distance preferably 1 m, h²(t)_0-∞ is a RIR measured with omnidirectional source.

C80: It is a measure of transparency of musical structures [13]. In practice an unweight average of 125 Hz–4 KHz is used [14]. Optimal value for C80 are –4 dB to 0 dB (orchestral music), 0 dB to 4 dB (singers), –4 dB to 4 dB (general purpose) [15]. By substituting 0 to 50 limit in Eq. (7) as 0 to 80 this parameter can be obtained.

C50: It is based on the Haas Effect for speech, i.e., when an acoustical reflection reaches within 50 ms of the direct sound. It improves when strong early reflections are present in a room. Optimal values for C50 are ≥3 dB [11]. A weighted average for 500 Hz to 4 KHz is often used for assessment of a room [14].

(7) C50=10log⁡(∫050⁡h2(t)dt/∫50∞⁡h2(t)dt)

where h(t) = power in first 50 ms and thereafter.

Reverberation Time (RT): According to ISO 3382 measured RT is obtained by extrapolation of a 60 dB line fitted to a decay curve. In reality, it is difficult to achieve Signal to Noise Ratio (SNR) which allows 60 dB decay range, therefore –5 dB to –35 dB decay is used, i.e., T30 [16]. T10 is calculated for range –5 dB to –15 db of decay. T20 is an estimation of the 60 dB decay time by extrapolation. SNR of minimum 35 dB is preferred for measuring T20 [11]. For greater speech intelligibility RT should be in range of 0.8 s – 1.0 s [14]. It is determined by Sabine equation as follows

(8) T60=0.1611×(V/Sα)

where V = Volume of room in m³, S = Total surface area in m², α = Absorption coefficient.

Fig. 4 shows many resonance peaks on low frequency side. Table 1 lists peaks at low frequency (less than 1 kHz), with corresponding SPL and musical note. Frequency corresponding to lowest dB level is in Italics.

Tables 2–4 show room acoustic parameters like EDT (sec), T10 (sec), C50 (dB), D50 (%). Here CH = Channel number.

Table 5 gives the values in condensed form. For this we first took average of all channel values for each frequency. These values were then added and averaged for octave band. Here we have values of Clarity (C50, C80), Definition (D50), Early Decay Time (EDT), Reverberation Time (T10, T20, T30), Speech Transmission Index (Male, Female). C80 (for Music) value was evaluated only for chaitya.

We then calculated IACC (Inter-aural Cross Correlation) as shown in Table 6 below.

Inter-aural Cross Correlation (IACC) value denotes the correlation between the two signals arriving at two ears. +1 denotes two identical signals with perfect correlation, 0 for no correlation, −1 perfect correlation with out of phase signals (see Figs. 5 & 6).