1、 Report ITU-R BS.2161(11/2009)Low delay audio coding for broadcasting applicationsBS SeriesBroadcasting service (sound)ii Rep. ITU-R BS.2161 Foreword The role of the Radiocommunication Sector is to ensure the rational, equitable, efficient and economical use of the radio-frequency spectrum by all ra
2、diocommunication services, including satellite services, and carry out studies without limit of frequency range on the basis of which Recommendations are adopted. The regulatory and policy functions of the Radiocommunication Sector are performed by World and Regional Radiocommunication Conferences a
3、nd Radiocommunication Assemblies supported by Study Groups. Policy on Intellectual Property Right (IPR) ITU-R policy on IPR is described in the Common Patent Policy for ITU-T/ITU-R/ISO/IEC referenced in Annex 1 of Resolution ITU-R 1. Forms to be used for the submission of patent statements and licen
4、sing declarations by patent holders are available from http:/www.itu.int/ITU-R/go/patents/en where the Guidelines for Implementation of the Common Patent Policy for ITU-T/ITU-R/ISO/IEC and the ITU-R patent information database can also be found. Series of ITU-R Reports (Also available online at http
5、:/www.itu.int/publ/R-REP/en) Series Title BO Satellite delivery BR Recording for production, archival and play-out; film for television BS Broadcasting service (sound) BT Broadcasting service (television) F Fixed service M Mobile, radiodetermination, amateur and related satellite services P Radiowav
6、e propagation RA Radio astronomy RS Remote sensing systems S Fixed-satellite service SA Space applications and meteorology SF Frequency sharing and coordination between fixed-satellite and fixed service systems SM Spectrum management Note: This ITU-R Report was approved in English by the Study Group
7、 under the procedure detailed in Resolution ITU-R 1. Electronic Publication Geneva, 2010 ITU 2010 All rights reserved. No part of this publication may be reproduced, by any means whatsoever, without written permission of ITU. Rep. ITU-R BS.2161 1 REPORT ITU-R BS.2161 Low delay audio coding for broad
8、casting applications (2009) 1 Operational requirements of a low delay audio coding 1.1 Requirements for digital wireless microphones For wireless microphones, it is essential to reduce latency so that the sound of voice such as speech or vocals is reproduced from loudspeakers simultaneously. In this
9、 section, the required latency in each operation of a wireless microphone is described. The requirements were formulated by broadcast audio experts in Japan. Table 1 presents a list of requirements for digital wireless microphones. 1.1.1 Studio In the studio, particularly for live broadcasting, the
10、maximum acceptable delay time is approximately 1 ms to ensure smooth conversation over talk-back. In other cases, for programme production in a studio for recording, the relative delay between picture and audio should be minimized and not vary although it can be adjusted later through the editing pr
11、ocess. In addition, there are multiple audio sources in most cases, and differences of delays among the sources should be minimized. As some of the sources are from wired microphones, the delay time of wireless microphones should be less than approximately 1 ms. It is also desirable for operators to
12、 know the actual delay time. 1.1.2 ENG and outside broadcasting In outside broadcasting, the acceptable delay time is the same as in the studio case. In live sports programmes, the maximum acceptable delay time is approximately 5 ms for the sound of a ball and other sounds made in a game to achieve
13、good synchronization with the picture. The delay time should not vary. On the other hand, when it is not live broadcasting, the acceptable delay could be relaxed to approximately 25 ms as a trade-off to get robustness against interference. It is also desirable for operators to know the actual delay
14、time. 1.1.3 Talk-back Speakers or singers find it difficult to speak or sing if their talk-back voice has significant latency, so very strict delay time management is required. The delay of both a wireless microphone and a talk-back circuit should be taken into consideration. A delay of less than 1
15、ms for studio use and less than 5 ms for outside broadcasting are required. The delay time should not vary. It is also desirable for operators to know the actual delay time. 1.1.4 Concerts On the stage, various delays are generated depending on the allocation of speakers and microphones. For example
16、, a 3-ms delay corresponds to a distance of 1 m. It is considered that professional players can detect a 2-ms delay. 2 Rep. ITU-R BS.2161 The relative delay between wireless microphone and wired microphone or other electronic musical instruments should be minimized. The maximum acceptable delay is a
17、pproximately 2 ms, or 1 ms if possible. 1.1.5 Musicals and plays To express fine vocals and music performance, the delay time should be as small as possible. A good singer would not wish to use a microphone if the delay exceeds 3 ms, and more than 5 ms is unacceptable. As this value is the total del
18、ay time in the audio system from microphone to loudspeaker, the maximum acceptable delay of a wireless microphone is 2 ms. 1.1.6 In-ear monitor Music players play the music picking up the beat through this monitor. The maximum acceptable delay is 1 ms. TABLE 1 Requirements for digital wireless micro
19、phones Application Studio ENG and outside broadcasting Talk-back Concerts Musicals and plays In-ear monitor Content Voice Voice Voice and broadcast programme Voice and musical instruments Voice and musical instruments Voice and musical instruments in stereo Audio frequency 20 Hz-20 kHz 20 Hz-20 kHz
20、(50 Hz-10 kHz by trade-off with interference) 100 Hz-10 kHz(100 Hz-7 kHz by trade-off with interference or latency) 20 Hz-over 20 kHz 20 Hz-over 20 kHz 20 Hz-15 kHz Audio dynamic range More than 100 dB (preferably 20-bit linear PCM and more than 120 dB) More than 100 dB More than 70 dB More than 100
21、 dB 90 dB 95-100 dB Maximum sound pressure level of microphone More than 130 dBSPL More than 140 dBSPL 140 dBSPL 130 dBSPL Maximum acceptable latency 1 ms 5 ms (25 ms by trade-off with interference) 5 ms 2 ms 2 ms 1 ms Audio interface AES/EBU output at receiver AES/EBU input at transmitter AES/EBU o
22、utput at receiver AES/EBU input at transmitter Rep. ITU-R BS.2161 3 2 Delay calculation for audio codecs It is essential to consider the delay introduced by the codecs when comparing the achieved audio quality. Therefore, the underlying assumptions of the delay calculation are described before the p
23、articular overall delays are defined in the corresponding codec 2.1 to 2.4. For the codec described in 2.5 the end-to-end delay was measured. There are three main categories of delay in communication systems: Algorithmic delay the part of the latency introduced by the algorithm which is independent
24、from the properties of the transmission channel and the speed of the digital signal processor. Transmission delay the part of the latency introduced by sending the bit reduced audio data from the encoder to the decoder. Processing delay the part of the latency dependent on the processing speed of th
25、e digital signal processor. As the focus of this document is a broadcasting environment, the transmission delay is important. We assume a restricted channel capacity (constant bit rate) for the data transfer, which equals the bit rate of the signal or exceeds it slightly. The additional delay in mil
26、liseconds caused by this so-called “continuous transmission” is equal to the average bit-stream frame-length divided by the network transmission clock 0. The “continuous transmission” mode (with a fixed upper limit for the bit rate) implies furthermore that the bit reservoir, which is implemented in
27、 several audio codecs, have to be incorporated into the delay calculation of the algorithmic delay. Besides the aforementioned bit reservoir, other potential sources for algorithmic delay are the framing delay, the filter bank delay and the look-ahead delay for a potential block switching decision.
28、Assuming an unlimited hardware computation capacity for the encoding/decoding process, any additional delays caused by time for computation is therefore ignored. This is reasonable because the processing delay in general is small compared to the other two factors and is getting smaller with technica
29、l progress in microelectronics. The algorithmic delay and the transmission delay sum up to the total delay of a codec. 2.1 ISO/IEC 11172-3 (MPEG-1 Audio) Layer II and MPEG-1 Layer III 2.1.1 Overview The international standard ISO/IEC 11172 was introduced November 1992. The MPEG/Audio-Group developed
30、 the audio part of the standard, which is described in ISO/IEC 11172-3. This part consists of three perceptual coders, called Layers, with their complexity and performance increasing from Layers I to III. Layer I can be thought of as a coding scheme for applications that do not require very low bit
31、rates, e.g. for home recording on Digital Compact Cassette. Layer II is used for broadcasting purposes. More precisely, Layer II is the coding scheme for contribution, distribution and emission applications in the broadcasting domain. Layer III is more complex than the other layers, but achieves a h
32、igher compression performance. In the ISO standard, only the decoder structure and the bit stream are exactly defined. The encoding structure given in the standard can be seen as a minimum requirements version. Proprietary enhancements in the encoder, e.g. a better psychoacoustic model can be implem
33、ented. Both Layers II and III are referenced in Recommendation ITU-R BS.1115. 4 Rep. ITU-R BS.2161 2.1.2 Structure and delay of Layer II MPEG-1 Layer II uses a 32-channel polyphase filter bank (PQMF) with a length of 512 taps to map the time domain audio signal samples into the time-frequency domain
34、. This filter bank provides a frequency resolution of 500 Hz at a sampling rate of 32 kHz and a time resolution of 1 ms. The Layer II encoder is shown in Fig. 1 and the decoder is shown in Fig. 2. FIGURE 1 The system structure of the Layer II encoder Report BS.2161-01PQMFfilterbank32 subbandsLinearq
35、uantizerFFT1024 pointsPsychoacousticmodelCoding ofside informationBitstreamformattingCRC-CheckExternalcontrolSubband 31PCM768 kbit/sAncillary dataCoded audio signal32-384 kbit/s0FIGURE 2 The system structure of the Layer II decoder Report BS.2161-02InversePQMFfilterbanksubbandsDequantizationofsubban
36、dsamplesDecoding ofside informationDemultiplexinganderrorcheckSubband 310PCM768 kbit/sAncillary dataCoded audio signal32-384 kbit/sRep. ITU-R BS.2161 5 A psychoacoustic model is used to determine the masking threshold which is used to find the just noticeable noise level of each band in the filter b
37、ank. The actual quantizer level in each frequency sub-band of every time block is calculated by allocating the available bits depending on the difference between the maximum signal level and the masking threshold. Before the psychoacoustic model is computed, the time domain signal has to be transfor
38、med, in parallel to the PQMF, by a 1 024 point FFT which causes no additional delay. Those two transforms work independently of each other. The psychoacoustic model in Layer II delivers an accurate approximation of the masking threshold. It controls the quantization of the 32-channel PQMF sub-bands
39、through its blockwise adaptive bit allocation by adjusting the quantization step size. The bit allocation is done in an iterative process and is based on the principle of minimizing the “total noise-to-mask ratio over the frame with the constraint that the number of bits used does not exceed the num
40、ber of bits available for that frame” 0. This process has to be repeated until the needed bits for coding the samples, the scale factors and the bit allocation information are as close as possible to the number of available bits for the whole frame without exceeding the available bits. The block len
41、gth of one time domain frame is 1 152 samples which corresponds to 36 ms at 32 kHz. With 32 sub-bands this corresponds to 36 samples per sub-band in the time-frequency domain. The quantized sub-band samples and the side information have to be multiplexed and transmitted. The Layer II codec is operat
42、ing at bit rates between 32 kbit/s and 384 kbit/s. In Layer II the overall delay adds up to 2 815 samples. 511 samples for the filter bank delay plus 1 152 samples for the transmission of the sub-band samples plus 1 152 samples for the framing. This corresponds to an algorithmic delay of 59 ms at a
43、sampling rate of 48 kHz, and an algorithmic delay of 88 ms at a sampling rate of 32 kHz. 2.1.3 Structure and delay of Layer III In Layer III, there is a 32-channel PQMF filter bank cascaded with an 18-point MDCT. The cascading results in 32*18 = 576 spectral lines with a frequency resolution of 27.7
44、7 Hz and a time resolution of 6 ms at 32 kHz. The Layer III encoder is shown in Fig. 3 and the decoder is shown in Fig. 4. The psychoacoustic model processes the MDCT frequency domain output samples in order to provide the masking threshold and to estimate the tonality of the signal by using a magni
45、tude-phase prediction scheme. The calculation causes no additional delay. The window switching decision for the MDCT filter bank is done in parallel to the PQMF filtering in the time domain and produces a delay of 144 samples. The Layer III frame length is equal to the Layer II frame length, which i
46、s 1 152 samples or 36 ms at a sampling rate of 32 kHz. 6 Rep. ITU-R BS.2161 FIGURE 3 The system structure of the Layer III encoder Report BS.2161-03ExternalcontrolPQMFfilterbank32 subbandsNonuniformquantizationRate/distortioncontrolWindowswitchingPsychoacousticmodelCoding ofside informationBitstream
47、formattingCRC-check310PCM768 kbit/sAncillary dataCoded audio signal32-320 kbit/s0575HuffmanencodingMDCTFIGURE 4 The system structure of the Layer III decoder Report BS.2161-04InversePQMFFilterbank32 subbandsDescalingDecoding ofsideinformationDemultiplexinganderrorcheckPCM768 kbit/sAncillary dataCode
48、d audio signal32-320 kbit/sHuffmandecodingInverseMDCT0575 310Layer III uses a non-uniform quantizer and a Huffman entropy coder with static tables to compress the sub-band samples. For sequences of zeroes a run-length-coder reduces the bit demand further. A bit reservoir technique allows for saving
49、bits from frames which did not exploit the target bit rate. Those bits can be used to code subsequent frames of a higher complexity which would otherwise exceed the maximum channel bit rate. Rep. ITU-R BS.2161 7 The Layer III codec is operating at bit rates between 32 and 320 kbit/s and reaches excellent audio quality at 96 kbit/s/channel. In Layer III the overall delay sums up to 5 809 samples. 144 samples delay are caused by the window switching look-ahead. 481 samples delay are caused due to the PQMF filter bank plus an MDCT delay of 576 samples. Furthermore, there
