TDMA Third Generation Wireless - Adaptive Multi-Rate Speech Codec Minimum Performance Requirements
|Publication Date:||1 April 2001|
This document specifies the procedures to ensure that non-bit-exact implementations of TIA/EIA-136-440 or GSM 06.90-compatible speech codecs operating in CTX or DTX-high and CNG [Ref. 11, 16] operating in DTX-low meet the recommended minimum performance specifications under channel error free conditions. The speech codec is used to digitally encode the speech signal for transmission at one of eight data rates: 12200, 10200, 7950, 7400, 6700, 5900, 5150, or 4750 bits persecond.
The GSM 06.73 standard describes a bit-exact specification of the speech codec using a fixed point ANSI C code [Ref 15]. A number of test sequences are provided with standard GSM 06.74 [Ref 17] for maximum coverage of the fixed point ANSI C code. There are separate test sequences for testing the encoder and for testing the decoder portions of the standard speech codec [Ref 11, 14]. A bit-exact implementation of this standard on a fixed point device would have to pass all of the test sequences for the encoder and for the decoder to comply with the standard [Ref 11, 14].
At the Base Station end, non-bit-exact implementations of the standard AMR speech codec are permitted. For such implementations, it may not be possible to generate test sequences that can guarantee compatibility with the standard by using them as input to the speech encoder or decoder and by examining their output for exact replication of the corresponding reference sequence. This document describes a series of tests which are used to test conformance to the specification. These tests do not necessarily ensure that the codec operates satisfactorily under all possible input signals and it is still required that the manufacturer ensure that the implementation operates in a consistent manner. The purpose of these tests is to examine minimum performance levels. The manufacturer should attempt to provide the highest performance possible.
The testing of any non-bit-exact implementation of the codec operating in CTX or DTX-high is based upon two classes of procedures:
Objective tests, which are based upon bench measurements of the speech codec function.
Subjective tests, which are based upon listening tests to judge overall speech quality.
Minimum performance specification for CNG is based on objective tests only. These tests are defined by the test database, a method for generating individual test data files, the test requirements on candidate CNGs, a method to collect performance metrics, and the means to test those metrics so as to validate conformance to the specification.
The main purpose of the testing is to ensure adequate performance between any Base Station s encoder and any Mobile Station s decoder and vice-versa. Since the MS is restricted to have bit-exact implementations of the AMR standard only, it suffices to demonstrate that minimum performance standards have been met when inter-operating with any bit-exact implementation. For this purpose, the fixed point ANSI C code provided as a part of the GSM 06.73 standard [Ref 15] can be treated as the reference or master codec.
The restriction that MSs can only have bit-exact implementations of the AMR standard [Ref 11, 14] is the primary reason that allows the present standard to ensure minimum performance standards between any pairing of base stations and MSs; otherwise, this could only be guaranteed by testing all combinations of encoded/decoder pairings that would be found in either base stations and mobile stations. With the number of equipment manufacturers expected to supply equipment this could become a prohibitive task.
By convention in this document, the Courier font is used to indicate C language and other software constructs such as file and variable names. The section mark, 0, is used to indicate section numbers.