Intermediate Reference System

Description

The Intermediate Reference System (IRS) is a critical concept in 3GPP specifications for audio and speech quality assessment. It is defined as a standardized electro-acoustic chain that simulates the characteristics of a typical telephone handset, including the microphone, receiver, and associated frequency responses. The IRS is not a physical device but a mathematical model or a set of filter curves specified in ITU-T Recommendations (e.g., P.48) and adopted by 3GPP. Its primary role is to serve as a reference point during subjective and objective testing of speech codecs, ensuring that evaluations are consistent and reproducible across different laboratories and manufacturers. The term 'No IRS' or 'rather flat' refers to a condition where the IRS filter is not applied, meaning the audio signal is processed without the typical telephone band-limiting, resulting in a flatter frequency response that is closer to the original wideband audio.

Architecturally, the IRS is implemented within test setups defined in specifications like 26.975 (Speech and video telephony tester) and 26.978 (Voice over IP test). In these setups, the IRS filter is applied to the send (speaking) and receive (listening) paths to emulate the acoustic interface of a traditional telephone. The send-side IRS modifies the input speech signal to mimic the frequency response of a typical handset microphone, which typically has a bandpass characteristic (e.g., 100 Hz to 8 kHz) with specific attenuations. The receive-side IRS does the same for the output, simulating the handset's receiver. This ensures that codec performance is evaluated under conditions that represent real-world usage, rather than ideal laboratory conditions with perfect transducers.

How it works involves integrating the IRS filter into the test sequences for codec qualification. During subjective tests like Mean Opinion Score (MOS) evaluations, listeners assess speech quality after the signal has passed through the IRS, providing ratings that reflect typical user experience. For objective tests, algorithms like POLQA or PESQ may use the IRS as a reference to compare degraded speech against a pristine reference. The IRS is key to ensuring that codec performance metrics, such as those for AMR, EVS, or VoLTE/VoNR codecs, are comparable across different implementations and network conditions. Its inclusion in 3GPP specs like 29.163 (Interworking between SIP-I and BICC) and 46.008 (MS speech codec) underscores its importance in interoperability and quality assurance for voice services.

Purpose & Motivation

The purpose of the Intermediate Reference System (IRS) is to provide a standardized and reproducible acoustic reference for testing speech codecs and audio quality in telecommunications. Before its adoption, codec evaluations were often inconsistent because different test setups used varied microphone and receiver characteristics, leading to incomparable results. The IRS solves this by defining a unified electro-acoustic chain that mimics a typical telephone handset, ensuring that all tests are conducted under the same conditions. This allows for fair comparison between different codecs (e.g., GSM FR vs. AMR-WB) and ensures that performance metrics like speech quality, noise robustness, and delay are measured in a way that reflects real user experience.

Historically, the need for IRS arose from the proliferation of digital speech codecs in 2G/3G networks, where quality assurance became critical for service acceptance. The ITU-T developed the IRS concept (e.g., in Recommendation P.48) to standardize testing, and 3GPP incorporated it starting in Release 8 to support voice services over GSM, UMTS, and later LTE/5G. The 'rather flat' condition (No IRS) was introduced to accommodate wideband and super-wideband codecs like AMR-WB and EVS, which benefit from a broader frequency response. This addresses the limitation of narrowband IRS, which was tailored for traditional telephone bandwidth (300-3400 Hz), by allowing tests that better represent enhanced voice services.

The motivation for IRS inclusion in 3GPP was driven by the need for interoperability and quality consistency across global networks. By using IRS, manufacturers and operators can verify that codecs meet minimum quality standards, facilitate roaming, and ensure backward compatibility. It also supports the evolution from circuit-switched voice to packet-switched VoLTE/VoNR, where maintaining speech quality across diverse networks and devices is paramount. Thus, IRS remains a foundational element in audio testing, enabling reliable deployment of voice services from 2G to 5G.