VA Virtual Acoustic Space
As a core part of the generation of virtual acoustics comes from the geometric structure the sound sources propagate within, this session will run iterations on three models with varying size to determine what distinctions may be apparent between their VAS.
Model Resources :
Large “Mayan Pyramid” :
https://3dwarehouse.sketchup.com/model/f08ea0c8177071bd19a839891524a150/Mayan-Pyramid
Medium “Mediaeval tower-keep” :
https://3dwarehouse.sketchup.com/model/u05b3db35-4c97-4a18-a642-96030329585d/Medieval-tower-keep
Small “Spiral Minaret” :
https://3dwarehouse.sketchup.com/model/dc0b9713583712669f39c57cba71f74/Spiral-Minaret-Construction
Material will remain consistent across the structures, utilising the mirror material from the previous session to maximise ray transmissions. Size / scales / skews will be under consideration for adaptation of the spaces, source and receiver arrangement will remain
consistent. Sources will consist of 8 tracks of acoustic folk consisting primarily of guitar and vocals, centred around the receiver in plotted using the coordinate plane plotter in resources. Using a parent gameobject
to centre an anchor coordinate (0,0) the surrounding sources can be fit to those given here in the source configuration to the right. Sources maintain consistent volume on render from original stems, meaning any change is only due to the VA encoding, where distance rolloff has been set to linear. Recording 8 tracks on initialisation results in tearing, this issue could be due to sample rate or internal clock processing containing timing discrepancies. Until a fix is found, renders will be in two parts for the left and right groups of 4 sources and then layered in Logic. Although this increases render time, the results should be higher quality as virtual voice count (binaural ray cues) can be maximised for each side of sources.
Upon completion of the split LR recording of iteration 1, lining the two sides within Logic to correctly layer is a tenuous task to ensure the original phase is maintained between any bleed present in these multitracks. Upon the next iteration to remedy this, a click impulse will be enabled alongside audio sources similar to how a clapper box is used to align audio / video. During monitoring, minimal artefacts and tears were observed during the recording similarly to when all sources were attempted to be rendered at once. This recurring issue will need resolving to ensure consistent quality throughout renders.
Iteration 1 recordings :
Maya Temple -
Scar-TheLonelyWild-MayaTemple-VASTest.wav
https://drive.google.com/file/d/1_XI5PA0cy3k-RuzZD7LK4DtWs9wXDeFN/view?usp=share_link
Medieval Tower Keep -
Scar-TheLonelyWild-MedievalTowerKeep-VASTest.wav
https://drive.google.com/file/d/1VDeyqSAc402WSB8EmQpqim3LzDWmznrV/view?usp=share_link
Spiral Minaret -
Scar-TheLonelyWild-SpiralMinaret-VASTest.wav
https://drive.google.com/file/d/17OcotLylv4Efg13aQ_bf4DRpcXZcXv7_/view?usp=share_link
Dry Stems -
Scar-TheLonelyWild-DryStems-VASTest.wav
https://drive.google.com/file/d/171UApb0RZh5m7bzNQI_CUYBpFnotoKMt/view?usp=share_link
Through the differing geometric models, the size difference can be apparent when regarding the “air” present, with the smaller VAS having tighter rays resulting in a denser sound image in comparison to the large Maya Temple. Amplitude also remains consistent with previous tests in that it increases more in small spaces over larger spaces. There is present, within the spiral minaret model render, some timing issues where instrumentation within the track loses cohesion. This is likely due to extended ILD / ITD from the height of the structure leading to noticeable delay on rays bouncing off the ceiling and returning, this does negatively impact the recording as these timing differences are too significant to maintain. Along with this, as the high count of incoming rays from the ceiling results in a large portion of the stereo field centre receiver position to be muddied with these cues. Distinction between all VAS encoded renders across the three models when compared to the original dry stems is apparent, with the sound adapting to the virtual acoustic environment when filling the space with reflections along with the exocentric positioning of signals. As these models contained more of a ratio of larger height to width, the next iteration will focus on one model in a ratio where width is larger than height. Specifically, the spiral minaret model will be scaled from the original to 50% height, and width increased to 300%.
Iteration 2 -
Scaled Spiral Minaret
Scar-TheLonelyWild-ScaledSpiralMinaret-VASTest.wav
https://drive.google.com/file/d/1-wZCBItVR7Cyaesa1bekVmJPSXdO-t7I/view?usp=share_link
On completion of iteration 2 with the scaled spiral minaret model now with larger width and less height, difference to the mix in comparison is that it has lost some height information which previously muddied the stereo field centre.
With the reflected rays now predominantly coming in from horizontal angles, the exocentric acoustic effect is enhanced with a dense sound that is not compromised by timing issues which dominate the final recording. This is beneficial to be aware of, as adding height information through excessive reflected rays down vertically, leads to none exocentric audio which causes perceptual timing cues for sound sources which are inconsistent with intended sources. For this iteration, the clapper impulse response utilised on activation of sources in the rendering environment scene lead to a quick and more precise alignment of the two L and R renders.
