It is recommended that the EVS codec

• have mandatory support of high-quality and high-efficiency operation of NB, WB, and SWB audio;
• may support FB audio.

Stereo or multi-channel presentation is one way to realize significantly improved QoE. The codec may provide stereo/ multi-channel coding capability. Multiple monophonic coding, i.e. one monophonic coding per channel, can be realized by appropriate packetization techniques while stereo/multi-channel coding, i.e. joint coding of the channels, is part of the audio coder algorithm. The choice of whether using dedicated stereo/multi-channel coding or multiple monophonic codings depends on a trade-off between achievable quality, available bit rates, available delay, complexity and other implementation factors. If stereo/multi-channel coding capability is provided, it is necessary to specify how stereo or multi-channel capture/presentations could be achieved in mobile communications.

It is recommended that the codec shall span a large range of bit rates from low rates needed for high efficiency conversational speech services to high rates required for EVS with high quality operation. It is recommended that the offered span of bit rates shall be wide enough to allow for rate adaptation in response to available transmission resource.

The codec delay requirement for the EVS codec is recommended to be flexible within certain limits, allowing for overall optimizations of the system performance and considering that there is a trade-off between delay consumed by the speech codec and delay consumed by the PS transmission via the LTE air interface. The delay requirements should be the same regardless of the nature of the input content (e.g. speech, music and mixed content). The algorithmic delay of the EVS codec should be such that the overall end-to-end delay in an EVS-UE to EVS-UE connection meets or exceeds the preferred performance expectations in TS 22.105. According to TS 22.105 v9.0.0, conversational voice one way delay is <150 ms preferred and <400 ms limit where it is noted that the one way delay in the mobile network (from UE to PLMN border) is approximately 100 ms. In addition it is recommended to consult the relevant 3GPP working groups (SA1, SA2, RAN, CT) in order to get a technically relevant breakdown of the various possible delay contributing elements of the end-to-end EVS-UE to EVS-UE transmission chain, which in turn enables specifying the allowable limits for the algorithmic delay of the EVS codec.

The EVS Codec should be implementable on a mobile device using today's technology. The EVS codec should provide low computational complexity not significantly exceeding the design limits set during the AMR-WB codec standardization, and should have low memory usage. Increased computational complexity and memory usage should be commensurate with the gain in quality of user experience (e.g. higher audio bandwidth such as SWB or stereo if it is supported) or with increased efficiency (e.g. lower bit rate for same quality when compared to a reference codec).

Backward interoperability with existing 3GPP codecs aims to reduce the need for transcoding as much as possible. Backward interoperability with existing codecs can be achieved through

• the use or negotiation of existing 3GPP codecs previously defined for voice services, or
• bitstream interoperability with one or more of these codecs when a new codec is defined.

It is recommended that the EVS codec shall achieve backward interoperability to the existing 3GPP AMR-WB codec by supporting all AMR-WB codec formats used in 3GPP conversational speech telephony services including CS.

It is recommended that the EVS codec achieves the following quality requirements:

• For narrowband signals: the EVS codec shall offer significant improvements in trade-offs among capacity, delay, error robustness and speech quality over the state-of-the-art 3GPP narrowband codec (AMR).
• For wideband signals: the quality shall be significantly improved with respect to the state-of-the-art 3GPP wideband codec at equivalent operating points.
• For super-wideband signals: the quality shall be significantly better than the state-of-the-art 3GPP wideband codec with wideband input and be no worse than state-of-the-art conversational super-wideband codecs at equivalent operating points.

Besides, the codec performance should show an increase in quality with increased bit rate and audio bandwidth. For the AMR-WB codec modes that are included in the EVS codec (see section 4.3) it is additionally recommended that they should consistently provide significantly improved quality over the pre-Rel-10 AMR-WB codec when operating at the same bit rate

• in codec configurations with the AMR-WB encoder inside EVS and the AMR-WB decoder inside EVS,
• in codec configurations with the AMR-WB encoder inside EVS and the pre-Rel-10 AMR-WB decoder,
• in codec configurations with the pre-Rel-10 AMR-WB encoder and the AMR-WB decoder inside EVS, and
• in relevant 3GPP CS system configurations.

In order to retain music quality when interoperating with legacy CS networks (such as the PSTN), or when interoperating with networks that do not support signaling the start and stop of music, mixed content and music coded by the EVS codec should have quality that is significantly better than pre-Rel-10 3GPP speech codecs. The following is recommended:

• For narrowband signals: the quality shall be significantly improved with respect to the state-of-the-art 3GPP narrowband codec (AMR) at equivalent operating points.
• For wideband signals: the quality shall be significantly improved with respect to the state-of-the-art 3GPP conversational wideband codec at equivalent operating points.
• For super-wideband signals: the quality shall be significantly better than the state-of-the-art 3GPP conversational wideband codec with wideband input and be no worse than state-of-the-art conversational super-wideband codecs at equivalent operating points. Note: It is envisioned that the EVS codec will provide better performance than a specific state-of-the-art conversational super-wideband codec at equivalent operating points.

For the AMR-WB codec modes that are included in the EVS codec (see section 4.3) it is additionally recommended that they should consistently provide significantly improved quality over the pre-Rel-10 AMR-WB codec when operating at the same bit rate

• in codec configurations with the AMR-WB encoder inside EVS and the AMR-WB decoder inside EVS,
• in codec configurations with the AMR-WB encoder inside EVS and the pre-Rel-10 AMR-WB decoder,
• in codec configurations with the pre-Rel-10 AMR-WB encoder and the AMR-WB decoder inside EVS, and
• in relevant 3GPP CS system configurations.

The quality of the EVS codec should be evaluated using patterns of losses and delays based upon the expected behaviour of deployed EPS systems under a wide range of packet loss and delay/jitter conditions. As network conditions degrade, the degradation of EVS speech quality should be minimized. Furthermore, the performance of the codec should be evaluated in realistic scenarios including different delay and error profiles consistent with the objective requirements on JBM in TS 26.114. The following is recommended:

• For narrowband and wideband, speech quality in lossy conditions shall be significantly improved with respect to state-of-the-art 3GPP conversational codecs at equivalent operating points under the same loss conditions.
• For super-wideband, quality should be maintained to be significantly better than state-of-the-art super-wideband codecs under equivalent loss conditions.