The required Layer 1 software is supplied with the chipset. In
addition, an object code license for Layers 2 and 3 of the
protocol stack is available. This exact package of Layers 1,2,3 of
software, coupled with the AD20msp410 chipset, is today in
phones that have passed European GSM Final Type Approval.
A standard GSM Handset can be divided into five functional
• Analog and Digital Baseband Processing Subsystem
• (Voice to Radio)
• Layer 1 Software (Physical Layer)
• Protocol Stack Software (Layers 2 and 3)
• Radio Subsystem
• User Interface Software (MMI)
Analog Devices and The Technology Partnership (TTP)
provide a cost effective and proven method of attaining the
baseband processing subsystem and protocol stack software.
This data sheet includes functional descriptions of the baseband
processing subsystem and the Layer 1 software. The Technol-
ogy Partnership can provide licenses to software and reference
designs in all the other areas of a GSM hand-portable terminal.
For detailed information about the individual chipset compo-
nents, please refer to the ADSP-2178 (ASP), AD7015 (BBC)
and ADPLP01 (PLP) data sheets for electrical characteristics
and timing information.
Figure 1 is a functional block diagram of the GSM baseband
processing chipset. The chipset can be viewed as a functional
block that contains a number of discrete functional units. The
electrical and functional interfaces to the rest of the system are
briefly described at the end of this section and described in
detail in the individual data sheets for each component.
CHANNEL INTER- ENCRYPT
DECODE LEAVE DECRYPT
CONTROL + MMI + I/O
Figure 1. Functional Description
The uplink baseband processing functions include the following
Analog-to-Digital Voice Conversion (BBC)
A conventional microphone, connected directly to the BBC,
provides an analog input signal to the ADC. The voice ADC
function uses a sigma-delta converter to convert and noise shape
the input signal, achieving a Signal-to-Noise Ratio plus Total
Harmonic Distortion (SNR+THD) of greater than 62.5 dB.
The analog voice signal is sampled at 8 kHz, producing 13-bit
linear values corresponding to the magnitude of the input. The
resulting data is passed to the ASP through a dedicated serial
Speech Encoding (ASP)
The ASP receives the voice data stream from the BBC and
encodes the data from 104 kb/s to 13 kb/s. The algorithm used
is Regular Pulse Excitation, with Long Term Prediction (RPE-
LTP) as specified in the 06-series of GSM recommendations.
The algorithm is tested and proven to be bit-exact against the
GSM test vectors including all VAD/DTX functions. After
encoding the data is transferred to the PLP through a parallel
port in discrete blocks of 260 bits at 20 ms intervals.
Channel Coding (PLP)
The information received from the ASP contains data values
and filter coefficients that have different levels of priority. These
are subsequently protected to different levels within the channel
coding. The encode protection process incorporates block
coding and convolutional encoding. In addition to the normal
speech traffic channels, the channel coding function also
supports data transmission at full rate and half rate. After the
interleave process, if necessary, the data is encrypted using the
required A5/1 or A5/2 encryption algorithm. Data is then
formatted into bursts, with the required timing and training
sequences and sent to the BBC through a dedicated serial port.
GMSK Modulation and D/A Conversion (BBC)
The BBC receives data at 270 kb/s. The on-chip lookup-table
ROM modulates and spectrally shapes the data being sent. A
pair of 10-bit matched differential DACs convert the modulated
data from the digital domain to the analog domain and pass I
and Q data to the transmit section of the radio subsystem.
The downlink baseband processing functions include the
Analog-to-Digital Conversion (BBC)
The receiver I and Q signals are sampled by a pair of ADCs at
270 kHz. The resulting digital words are transferred to the ASP
through a dedicated receive path serial link and DMA control.
The equalizer recovers and demodulates the received signal and
establishes local timing and frequency references for the mobile
unit. The equalization algorithm is a version of the Maximum
Likelihood Sequence Estimation (MLSE) using the Viterbi
algorithm. Two confidence bits per symbol provide additional
information about the accuracy of each decision to the channel
codec’s convolutional decoder. The equalizer outputs a
sequence of bits including the confidence bits. This data is
transferred to the PLP through a dedicated parallel port on the
ASP. At this point, the training sequence and trailing bits,
contained within the burst, are discarded.
Channel Decoding (PLP)
The A5/1 or A5/2 decryption algorithm is used, as required, to
recover the data that is ready for the deinterleave process. The
deinterleave process is an exact inversion of the interleave
process used by the transmit section. Data can pass directly to
this function, without the A5/1 or the A5/2 decryption, con-
trolled by the Layer 1 processing. The decode function then
performs convolutional decoding and parity decoding. The
convolutional decoder uses a Viterbi algorithm, with two soft
–2– REV. 0