The pandemic has changed attitudes towards offices, companies have revised business processes and transferred part of their employees to remote work. Many have decided to keep this format as more efficient. We at BIOCAD have also transferred more than 50% of our employees to remote work. From a technical point of view, we had all the tools necessary for this – corporate files in the cloud, application delivery infrastructure, virtual machines, remote desktops and video conferencing software with colleagues left in the office.
Employees quite easily adapted to work in the new realities, and the company continued to move towards its goals. Before the pandemic, we often received guests in our office, these were international delegations, foreign partners and representatives of state authorities, now all such meetings have moved to an online format. As it turned out, our meeting rooms were not fully prepared for regular video conferences. In the meeting room, there was a Samsung QM65N TV on a stand, a Cisco Roomkit Plus system with complete video cameras, a CTS-MIC-TABL20 external microphone, as well as Polycom Studio and Polycom Trio complexes.
The microphones did not provide good sound quality – some conference participants were hard to hear. There were also problems with the video. Usually, a Cisco camera with a neural network automatically detects the speaker and points the camera at him, but its capabilities were not physically enough to “zoom in” the speaker at the far end of the room, and three to five more people fell into the frame. Particular difficulties for the system were moving around the conference room.
At the same time, Cisco and Polycom cameras and microphones duplicated each other, but we used them to work with various services and protocols – for example, WebRTC, Teams, Zoom and other software. As a result, employees on calls were confused about which hardware was being used at a particular moment. With the increased frequency of video conferencing, the room’s acoustic deficiencies have also become apparent. Three walls in the room are made of glass at once, which creates problems with reverberation and echo.
To remedy the situation, we decided to re-equip the meeting room and set ourselves several key tasks:
Improve the quality of audio and video, as well as set up auto-targeting to the speaker throughout the meeting room.
Get rid of duplicate equipment.
Reduce the effect of echo in the room and do it exclusively in software and hardware. We could not resort to re-planning or changing architectural solutions, as it is expensive and time-consuming. At the same time, we would have to modify the rest of the conference rooms, since from an aesthetic point of view they should have the same style.
Together with partners from Cisco, we worked out a hardware configuration of three codecs – the main Cisco RoomKit Pro and two auxiliary Cisco RoomKit Plus – taking into account the characteristics of our room. Four Shure MX202WP-A\C positioning microphones determine which part of the room the sound is coming from, while Cisco QuadCamera microphones provide speaker pointing.
We installed three Sennheiser SL Ceiling Mic II microphone arrays for speech capture, and eight JBL 64P/T-WH ceiling speakers for sound reproduction.
The Crown XLS1002 two-channel amplifier is now responsible for amplifying the audio signals. Their frequency correction and switching is handled by the BiAmp TesiraFORTE DAN CI audio processor. It broadcasts sound on the local network according to the standard Dante, sharpened for the transmission of uncompressed audio with almost zero delay. Such a composition of the audio equipment path made it possible to achieve perfect speech intelligibility of all people in the meeting room, even if they are standing against the wall or actively moving around the room. We also managed to get rid of the “echo” effect that was observed earlier.
To broadcast images from cameras, we use an ATEN UC3021 video capture card connected to an Intel NUC computer – it is installed in a separate rack with a media complex. The media complex that we have built allows you to work with any connection options – for example, SIP, H323, WebRTC protocols and Zoom, Webex software, including using the full functionality of Teams. As for video switching, we use Aten VM0808HB and Aten VM0801HB matrix switchers that support 4K resolution. The latter automatically switches the video stream when a new device is connected – for example, a laptop. To do this, conference participants can use one of the eight HDMI points located in the hatches of the table.
To switch Ethernet signals, we configured a Huawei S1720-28GWR-PWR-4P-E switch with the ability to power devices using PoE technology.
The system based on the Crestron CP3 central processor is responsible for managing the entire hardware complex, and the Cisco TOUCH10+ tablet acts as a user interface.
Interestingly, multimedia can be controlled remotely from a laptop or tablet connected to the same local network.
How exactly all our equipment is installed can be seen on the plan below.
It is marked with numbers:
1 – RoomKit Pro + QuadCamera and Samsung QE82T TV”;
2 – RoomKit Plus + QuadCamera (2 pcs.) and Samsung QM65R” (2 pcs.);
7 – Eight-port HDMI True 4K switch with auto-switching function;
8 – HDMI transmitter VE801T HDBaseT-Lite (7 pcs.). They made it possible to place server equipment in a separate room, located 40 meters from the meeting room.
12 – Quad Port USB 2.0 Daisy Chain Extension Cable;
13 – Ceiling speakers JBL 64P / T-WH (8 pcs.);
15 – Ceiling microphones Sennheiser SL CEILING MIC II (3 pcs.);
18 – Additional positioning microphones Shure MX202 (4 pcs.);
19 – Shure MXA310B desktop mute buttons (4 pcs.).
Also in a separate room, a server rack is installed, in which all the central equipment is placed. A UPS is installed in its lower part, which protects the operation of the complex from sudden power surges.
How it all works
The script we developed provides the ability to automatically switch the camera when a certain minimum value of the incoming sound is reached. We can assume that until the sound from the microphone has reached a predetermined level, it is only background noise. In this case, it is not perceived by the camera and switching to the speaker does not occur. This is very convenient, in this way all extraneous noises in the meeting room are filtered out, which prevents involuntary switching of attention of the meeting participants.
Additional functionality was also provided. For each of the positioning microphones, a priority level was created with an initial value. This level can be adjusted dynamically, which allows you to use this functionality even during meetings.
The yellow zone microphone is initially set to level -1, and the red zone microphone is set to level 1. If the speaker speaks at a volume of 5 units, then the yellow zone microphone will have a level of 4, and the red zone microphone will have a level of 6, respectively. Priority will be given to the camera from the red zone.
In total, with a script, we can adjust more than 14 different parameters for more accurate positioning (switch delay time, minimum frame hold time, individual priority setting for each microphone, minimum response threshold, etc.).
Such settings help the system determine the camera and further activate the built-in capabilities of Cisco neural networks to select the best frame and track the speaker.
What is the result
This is what our meeting room looks like after the upgrade. We have already successfully held several meetings there.
We implemented Cisco SpeakerTrack based on 3 codecs and 3 camera complexes using custom scripts that we wrote from scratch ourselves. At the same time, we can work with any protocols and video services: SIP, H323, WebRTC, Zoom, Teams, and even non-standard software – because all functions (including speaker tracking) are implemented at the hardware level.
At the time of launch, this was the first case for the implementation of a videoconferencing complex using 3 Cisco codecs connected into one logical videoconferencing system.
The team of BIOCAD Infrastructure Solutions Department worked on the project. The article was prepared for you by the head of BIOCAD technical support Denis Sitnikov.
What else we write about in the BIOCAD blog: