IKS | Lab – Surrounded by Sound
The aim of immersive 3D sound is to reproduce or synthesize a complete acoustic scene just like in real life. In many applications, such as virtual reality or immersive cinema, 3D audio offers even more expansive experience than its visual counterpart. For example, spatial sound enables us to hear sounds from all directions and is not limited to the field of view of the display.
For the subjective assessment of 3D audio quality, a new listening lab, the so-called IKS|Lab, has been installed step-by-step at the Institute. The acoustic characteristics of the listening room as well as the reproduction devices follow the base reference ITU-R BS.1116-2 to enable a rigorous control of the experimental conditions.
The IKS|Lab is equipped with 36 Neumann KH120D loudspeakers and a couple of subwoofers. Mounted on a rig system the loudspeakers are approximately located on a spherical surface and the positions were chosen such that almost all proprietary and standardized multi-channel audio formats from mono to 22.2 can be played under reference conditions. The new IKS|Lab opens up new research activities in the area of audio signal processing for immersive audio systems, e.g., 3D audio formats, binaural signal processing, and spatial audio playback. One example is our new cooperation with the Aachen Symphony Orchestra, where we created professional 3D audio and 360° video content.
Reference Listening Conditions
Recommendation ITU-R BS.1116-2 specifies among other three main room acoustic characteristics to be met within certain tolerance ranges: the reverberation time, the noise rating and the operational room response curve. All characteristics are measured in one-third octave bands over frequency. In the following we describe how we met all three criteria by an iterative optimization of the room characteristics.
At first the status of the empty room was documented. For this purpose, the reverberation time (T60) of the empty room was calculated based on impulse response measurements at different locations in the room. The T60 curve revealed the necessity of additional acoustic absorption over a broad frequency range. To quantify the influence of additional acoustic absorbers, a 3D computer model of the room was created where the materials and absorption coefficients can be adjusted. The model was fed into a room acoustic simulation to calculate the resulting reverberation time including different elements of acoustic treatment, such as acoustic panels, sound-control acoustic curtains, or diffusors. For best results, we introduced the different elements of acoustic treatment step-by-step, followed by corresponding measurements of the reverberation time.
Early measurements induced a carpet on the floor and curtains around the walls of the room. The resulting reduction in T60 was neither sufficient to meet the constraints nor was the reduction as large as promised by simulation. Still absorption over a wide frequency range was needed.
In a next step, broadband absorbers were installed at the ceiling. They achieve high absorption of acoustic energy over a wide frequency range by combining passive, porous absorbers and active, resonant elements. Measurements confirmed a drastic reduction of reverberation time. However, in the low frequency range, T60 was still too high whereas a slight overdamping at high frequencies could be observed.
Now in the fine-tuning stage as countermeasure for the underdamped low frequencies, compound panel absorbers were mounted at the ceiling. They are optimized for absorption around 125 Hz, which had been identified as the most critical range. To tackle the overdamping at high frequencies, it helped to rearrange the curtains.
Finally, we equipped the room with additional acoustic foam pads in front of the windows, diffusors and bass traps in the corners of the room, such that the reverberation time lies inside the tolerance range specified by the recommendation.
For the calculation of the noise rating (NR) we carried out long-term measurements lasting 12 hours and more. To be able to observe the low noise levels specified in ITU-R BS.1116-2, a pre-calibrated microphone with low self-noise was used. As expected, the noise rating requirements are not permanently fulfilled during the day. The NR curves over time reveal the hours of normal working days where the noise rating is inferior to the recommended thresholds of NR10 and NR15. In addition, they assisted to identify possible acoustical incidents. This information can now be used to schedule very sensitive measurements or critical listening tests
Operational room response curve
A remaining task is the analysis of the operational room response curve which is the transfer function between each loudspeaker and the reference listening position. According to ITU-R BS.1116-2, they shall fall within certain tolerance limits. This can be achieved by minor acoustic modification at or near specific speakers and by digital equalization of the loudspeaker signals.
Loudspeaker reproduction has a great advantage over headphone playback as listeners can turn their heads while the scene remains static. This is highly desirable since slight head-movements are an important psycho-acoustical cue for sound localization.
For the audio reproduction in our new IKS|Lab a loudspeaker array of 36 professional studio monitors plus sub-woofers has been installed. For the lab we chose Neumann KH120D loudspeakers with digital input due to their flat frequency response (free-field) in the range of 54Hz-20kHz (± 2dB) and their excellent spatial resolution. Official test reports and a subjective listening test in an anechoic chamber led to this decision. In addition, these studio monitors have been designed with rigid chassis and professional mounting systems for the traverses.
The loudspeakers can be combined in any way to enable flexible reproduction setups: either standard setups, for instance, mono, stereo, 5.1, 7.1, 7.1+4, up to 22.2, or non-standard setups individually defined by the user, e.g., for the use of Higher Order Ambisonics reproduction.
As we normally do not experience our senses individually, the IKS|Lab is also equipped with video reproduction devices. Besides a retractable and acoustic transparent projection screen (with video projector), an HTC Vive headset is available as virtual reality platform. The combination of the headset with the loudspeaker rig provides full immersive 3D audio and 360° video scenes.
In order to work with adequate professional content (audio and video), we took the great opportunity to join the Aachen Symphony Orchestra for recordings in the midst of the musicians. For these recordings special audio equipment such as the em32 Eigenmike®, a spatial microphone array with 32 electret microphones, and HMS II.3 dummy heads from HEAD acoustics is available at the institute. For the video recordings we used the Kodak SP360 4K BK5 "Dual Pro Pack" camera system as well as GoPro HERO 4 cameras to document the fields of view from the dummy head positions.
Signal Sources, Routing and Control
Besides the professional and flexible reproduction setups, the IKS|Lab supports multiple audio interface formats. The signal source may originate from different PCs, an A/V receiver, a Blue-ray player, and a variety of mobile playback devices. Different kinds of digital and analog audio signal interfaces comprising balanced and unbalanced analog audio with microphone and line level, digital AES/EBU, SPDIF, ADAT, MADI, and others are supported.
Inevitably, a sophisticated concept of how to connect the audio sources with the different playback setups is necessary. This is achieved using a MADI/AES10 based multichannel audio transmission backbone. The topology is that all source signals are aggregated into multiple MADI/AES10 streams, which are all connected to a single router. From the router, the audio signals are distributed to the desired reproduction devices. The router consists of a proprietary MADI switch and a special control and watchdog PC.
The control and watchdog PC's functionality is twofold: Firstly, a web front-end allows to select and create presets for a certain MADI switch configuration and to rearrange the channels within a multichannel audio stream to achieve a certain loudspeaker mapping. Thereby, the MADI switch is remote operated over MIDI. Secondly, the watchdog serves as an emergency shut-down for ear protection, e.g. in the unlikely case of software issues along the signal processing chain. The shut-down is triggered automatically whenever certain level thresholds are exceeded or can be triggered manually. The design of the watchdog is redundant. An additional protection circuit monitors the watchdog PC. In case that the watchdog PC does not react anymore, the loudspeakers are turned off automatically.
The IKS|Lab comprises two powerful PCs which are used as audio sources and for signal processing purposes. One is equipped with a high-end GPU and can be used with our virtual reality headset. The other one is exclusively reserved for students and is available for the Real-Time Audio Processing lab course as well as for master/bachelor theses.