Page 55 - From GMS to LTE

Global System for Mobile Communications (GSM) 41

4 × E-1
1 × E-1
BSC TRAU MSC

FR (ca. 13 kbit/s),
EFR (ca. 13 kbit/s),
HR (ca. 7 kbit/s) 64 kbit/s PCM

Figure 1.32 Speech compression with a 4:1 compression ratio in the TRAU.

the EFR codec and a PCM‐coded speech signal, the voice quality of the HR codec is
noticeably inferior. The advantage for the network operator of the HR codec is that the
number of simultaneous voice connections per carrier can be doubled. With the HR
codec, a single timeslot, which is used for a single EFR voice channel, can carry two
(HR) TCHs.
Another speech codec development is the Adaptive Multirate (AMR) algorithm
[22] that is used by most devices and networks today. Instead of using a single codec,
which is selected at the beginning of the call, AMR allows a change of the codec dur-
ing a call. The considerable advantage of this approach is the ability to switch to a
speech codec with a higher compression rate during bad radio signal conditions to
increase the number of error detection and correction bits. If signal conditions per-
mit, a lower rate codec can be used, which only uses every second burst of a frame for
the call. This in effect doubles the capacity of the cell as a single timeslot can be
shared by two calls in a similar manner to the HR codec. Unlike the HR codec, how-
ever, the AMR codecs, which only use every second burst and which are thus called
HR AMR codecs, still have a voice quality which is comparable to that of the EFR
codec. While AMR is optional for GSM, it has been chosen for the UMTS system as a
mandatory feature. In the United States, AMR is used by some network operators to
increase the capacity of their network, especially in very dense traffic areas like New
York, where it has become very difficult to increase the capacity of the network any
further, with over half a dozen carrier frequencies per sector already used. Further
information about AMR can also be found in Chapter 3.
The latest speech codec development used in practice is AMR‐Wideband (AMR‐WB)
as specified in ITU G.722.2 [23] and 3GPP TS 26.190 [24]. The algorithm allows, as its
name implies, digitization of a wider frequency spectrum than is possible with the PCM
algorithm that was described earlier. Instead of an upper limit of 3400 Hz, AMR‐WB
digitizes a voice signal up to a frequency of 7000 Hz. As a consequence, the caller’s voice
sounds much clearer and more natural on the other end of a connection. A high com-
pression rate is used in practice to reduce the datarate of a voice stream down to 12.65
kbit/s. This way, an AMR‐WB data stream can be transmitted in a single GSM timeslot,
and also requires no additional capacity in a UMTS network. As AMR‐WB is not com-
patible with the PCM codec used between the BSC and MSC, it is sent transparently
between the two nodes. This means that most of the bits in a 64 kbit/s PCM timeslot are