October 10, 2005
Call quality testing
Factors affecting call quality include noise, echo, variations in the signal volume, etc. Voice quality is tested for listening quality, conversational quality, and transmission quality. Listening quality is a subjective assessment by listeners of what they hear. Delay, echo, ease of two-way communication including listening quality is rated when conversational quality is tested. Network service quality and the quality of the network connection are measured while checking the transmission quality.
The Absolute Category Rating (ACR) is a popular subjective test for testing voice quality. The test uses a scale of 1 to 5. The Mean Opinion Score (MOS) is calculated from the individual scores and should ideally be taken from a pool of at least 16 participants. MOS scores offered by companies for their codecs are subjective scores that are influenced by a number of variables. Laboratory testing of voice is done using phonetically balanced text such as the Harvard Sentences. This helps in obtaining a subject’s reaction to a voice that covers the whole range of sounds found normally in speech.
Degradation Category Rating (DCR) and Comparison Category Rating (CCR) are other examples of subjective tests. The amount of degradation that occurs with the damaged files is measured by the DCR and a DMOS score is given. Pairs of files are compared by the CCR and a CMOS score provides the results. The ITU distinguishes the scores as Subjective, Objective, and Estimated.
P.861 and P.862 are objective measurement techniques developed by the ITU. ITU developed P.861 (PSQM) and the newer P.862. Transmission systems and codecs can introduce a distortion into the system. These measurement techniques contrast a reference file with the weakened signals. The reference and the actual signals are divided into small segments and the Fourier Transform coefficients for each segment are calculated and compared.
The amount of distortion can be measured using these algorithms only if both the source file and the output files are accessible to the algorithms. It needs to be understood that this particular type of algorithm needs a high processing speed, i.e. processing capabilities for 8000 samples per second for narrowband voice and 16000 samples for wideband voice. The processing and memory capabilities required are quite high and in such cases a packet-based network is preferably used.
The E Model for testing VoIP quality was developed by the ETSI. VQmon® is a voice testing technology that requires far less processing power than the PSQM approach. The E Model is used to rate the transmission quality, denoted as “R”. It is a measure of what are commonly known as the “mouth to ear” factors of a conversation. The R-value has a nominal range of 0-120. For broadband telephony, the range is 50-110.
The E Model assumes that the impairments have an additive effect and is represented by the following equation:
R = Ro - Is - Id - Ie + A
Ro is a base factor. Is stands for the signal impairments. Id stands for the delayed impairments. Ie stands for the “equipment impairment factor”. A stands for the “advantage factor”.
An ACR is a subjective test and when it is performed on a wideband CODEC, the score may not be representative of the actual performance if the reference conditions are set for a narrowband CODEC.
--
Did you enjoy this post?
« Setting up a VoIP network | Main | Delay issues in VoIP calls »
