All,

If the IETF codec WG is formed, Broadcom Corporation would like to submit its family of BroadVoice®16 (BV16) and BroadVoice32 (BV32) standard codecs to the IETF codec WG for consideration as a candidate codec or a candidate building block for the proposed Internet Wideband Audio Codec (IWAC) standard.  Currently Broadcom is already offering its patent rights, floating-point C source codes, and fixed-point C source codes of BV16/BV32 on a royalty-free basis, and Broadcom is even willing to make BV16/BV32 open source.

A little background on BV16/BV32.

BV16 and BV32 have been standardized by multiple standard organizations for VoIP applications in cable telephony.  BV16 is a 16 kb/s codec for 8 kHz narrowband signals; it is a standard codec in the following standards: PacketCableT 1.5, PacketCable 2.0, ANSI/SCTE 24-21 2006, and ITU-T Recommendation J.161.   Similarly, BV32 is a 32 kb/s codec for 16 kHz wideband signals, and it is a standard codec in PacketCable 2.0, ANSI/SCTE 24-23 2007, and ITU-T Recommendation J.361.  The RTP payload formats for BV16 and BV32 are specified in RFC4298.  BV16 and BV32 belong to the same family of codecs.  They have very similar codec structures and share most of the algorithm modules, so if the two are implemented together, substantial code sharing and memory reduction can be achieved.

Broadcom developed BV16 and BV32 from the ground up with a goal of avoiding known third-party intellectual property rights (IPR).  Although it was correctly pointed out earlier in the IETF codec WG email discussion that no one can be absolutely sure that a given codec is completely free of third-party IPR, we took two steps to help increase our confidence.  First, to help avoid the codec IPR minefield, we purposely avoided the popular and heavily-patented CELP coding paradigm, and instead took an "ancient art" (circa. 1954) technique of noise feedback coding (NFC) that nobody seemed to be interested in (i.e., it was the subject of very few recent publications) and improved it to get BV16/BV32.  Any patent on the fundamental NFC presumably expired long ago.  Second, several extensive patent searches were performed during the development of BV16/BV32 to help avoid third-party IPR.  To the best of our knowledge, BV16 and BV32 do not infringe on any valid unexpired third-party patent claim.  

BV16 and BV32 were also designed from the ground up to be optimized for voice transmission over IP networks, with high speech quality, low latency, low complexity, and high robustness to packet loss as primary design goals.  Thus, BV16 and BV32 are ideally suited for low-latency, high-quality voice transmission over the Internet.  The following summarizes the attributes of BV16 and BV32:

.	Low Delay (Latency): algorithmic buffering delay of merely 5 ms (compared with 15 to 40 ms of competing codecs)
.	Low Complexity: much lower MIPS requirements than most competing codecs (typically ½ to 1/3 of comparable ITU-T G.72x codecs), also lower memory requirement than most other competing codecs
.	High Quality: equivalent or better speech quality than competing codecs in extensive formal subjective MOS listening tests conducted by AT&T Labs, COMSAT Labs, and Dynastat, Inc.
.	Availability: Broadcom is offering its patent rights, floating-point and fixed-point C source codes of BV16/BV32 on a royalty-free basis; BV16/BV32 to become open source soon.

For comprehensive comparisons between BV16/BV32 and other codecs, please see the codec comparison tables in Annexes C and D (pages 73 - 81) of the PacketCable 2.0 Codec and Media Specification, available at:

http://www.packetcable.com/downloads/specs/PKT-SP-CODEC-MEDIA-I02-061013.pdf

Broadcom believes that as compared to the other nearly two dozen dominant codecs listed there, BV16 and BV32 offer the most compelling combinations of low delay, low complexity, high quality, moderate bit-rate, and royalty-free patent rights and C source codes.  The algorithm descriptions of BV16 and BV32 can be found in the two ANSI/SCTE standard documents below:
http://www.scte.org/documents/pdf/Standards/ANSISCTE24212006.pdf
http://www.scte.org/documents/pdf/Standards/ANSI_SCTE24-232007.pdf

It seems clear from the previous email discussion in the IETF codec WG reflector that voice transmission over the Internet is the primary application of the proposed IWAC, although transmitting general audio (music, etc.) at high fidelity is also desirable/necessary. It is also clear from the emails that royalty-free, low latency, and high quality are all highly desirable.  Given what was described above about BV16/BV32, it would seem that BV16/BV32 is an ideal candidate and is sufficient to cover the narrowband (8 kHz) and wideband (16 kHz) voice transmission aspects of the applications of IWAC.  To get higher sampling rates and higher audio bandwidths, it is possible to use BV16 and BV32 as the base layers and add additional bits on top of the BV16/BV32 bit-streams to encode an appropriately formulated difference signal to get a scalable, embedded codec that goes from 8 kHz sampling at 16 kb/s all the way to 48 kHz sampling at a much higher bit-rate.  One advantage of doing so is that the resulting IWAC will be interoperable with the BV16/BV32 in the PacketCable, ANSI/SCTE, and ITU-T J.161/J.361 standards. This is similar to the relationship between ITU-T G.729.1 and G.729.

Juin-Hwey (Raymond) Chen
Broadcom Corporation