Mary C. Gruszka / 06.09.2003 12:00AM
CBS, Neural Audio Capture 5.1 Madness
NCAA tourney brings HD with Surround Sound
CBS is on a roll with live HDTV broadcasts with 5.1 Surround Sound.
First it was the Grammy Awards in February with live 5.1 music mixing. Then it hit the courts at NCAA March Madness for 12 games in HD and 5.1 including the Final Four from New Orleans.
It was CBS's fourth year of HDTV at the basketball tournament, but the first time with 5.1 audio. As CBS discovered, creating a 5.1 Surround mix for this kind of sporting event, with action rapidly shifting from one side of the court to the other, posed a different set of challenges than the stage-based Grammys.
Seattle-based Neural Audio presented a solution in the form of the SECA 2-Goal microphone array system, which made its TV debut at the 12 CBS NCAA HD broadcasts.
"We wanted to get ahead of the curve ourselves and learn how to produce 5.1 audio," said Ken Aagaard, senior VP of Sports, Operations, Engineering and Production services at CBS. "The mic pushed us a long way in our development."
AN ARRAY OF SOUND
SECA stands for Sports Event Capture Arrays and is "presently effective for any sport involving scoring goals at either side of the playing surface with the surface surrounded by the audience," said Robert Reams, CTO and co-founder of Neural Audio. "There are also applications of this type of array for tennis, and there is a SECA version being prepared specifically for baseball."
The SECA 2-Goal consists of two arrays of four microphones each. The individual arrays are suspended on tubular supports clamped to the goal stands.
Two hyper-directional elements, comprising an intensity array, point forward and are spaced approximately 11 inches apart, Reams explained. Behind them are cardioid and figure-eight pattern elements comprising an M-S (mid-side) array, which is acoustically centered within the hyper-directional elements. Each intensity array is aimed at the court and each M-S array is aimed at the crowd.
An intensity array, as Reams explained, is a simple direction pair (one left, one right), while the M-S pair (cardioid and figure-eight) is matrixed to create a left and right signal.
The four outputs from each of the two arrays connect to a Neural Audio Expert Systems Rack, where internal neural processing manages volume level and frequency response to maintain a pre-determined spectral and level balance for the resulting output signals. This means that the output remains consistent, regardless of the sound source distance to the mic array or whether the sound source is on-axis or off-axis of the mic array.
Reams explained that as a sound source moves closer to an array, the received level increases 6dB for each doubling of distance. Also, as a sound source moves off-axis from the pickup pattern of an element of the array, the spectral characteristics change, with the highs rolling off faster than the lows.
Especially in a fast moving game like basketball, with sound sources constantly and rapidly moving, unprocessed sound pickup will exhibit varying audio levels and tonal characteristics that Reams said cause listener fatigue. Even a good audio operator would not be able to make enough fine adjustments in real time to compensate.
NO FUZZ HERE
But the Neural Audio processor takes care of these variations by using a form of artificial intelligence known as neural networks. "Neural networks are the next step up from fuzzy logic," Reams said. "They blend fuzzy logic with back propagation or feedback, and learn from what they did previously to narrow [subsequent] choices."
The Neural Audio processor is also an expert system-an expanded application of neural networks, Reams said, that can be "trained to conform to a set of rules and/or behavior targets as determined by expert guided training."
The expert is a human one-in this case, Paul Hubert, co-founder of Neural Audio. Hubert made the "first-pass" adjustments to the system using a special GUI that allowed him to change equalizer and level settings until the resulting sound, including how wide or narrow the coverage, was to his liking. These settings were later refined during a couple of trial basketball games CBS covered before the tournament and resulted in the target level and frequency response that the processor was to meet regardless of the position of the sound sources. Once set, the system was ready to go for any of the subsequent games.
Through the processing, different perspectives can be chosen and triggered by a video tally. Points-of-view can be created for the two goals and for the two long sides of the court with the front stereo image changing appropriately. The stereo rear-Surround image is derived from the audience behind the goals.
CBS took the resulting four main channels (front left and right and rear left and right), plus a low-frequency channel from the Neural Audio processor, and mixed them with other sources, including the announcers' mics, for a separate 5.1 mix on a Yamaha PM3000 in a trailer separate from the main production truck.
The low-frequency (LFE) or 0.1 channel "is a level, managed, coherent sum of the low-frequency information of all the capsules combined," Reams explained. The processor can also supply a left/right downmix, and there are allocations for including the announcer within the control of SECA, providing a full 5.1 experience.
CBS experimented with a number of different perspectives. "The telecast had a small audience, and we tried different things during the same broadcast to see which was the best way to go," Aagaard said. One early point of agreement was that the announcers should be placed in the center. "CBS found ways ... of using the system that produced better results than we did," Reams said.
Aagaard acknowledged that early forays into live HDTV broadcasts focused more on getting the video right. But now it's audio's turn.
"We were excited about what we got out of it," he said. "It puts the listeners right there at the scene. If you are a sports fan, you won't want to leave your seat. You feel like you can almost order a hot dog."