Module 3
Module 3
📌 GSM's success came from being the first to offer a complete, standardized, interoperable
digital cellular solution. Every aspect — from air interface to network switching to subscriber
identity — was standardized, allowing any compliant handset to work on any compliant network.
Telephone Services Standard voice and emergency calls Voice calls, emergency calling,
over the GSM network facsimile (Videotex/Teletex
supported but not integral)
Bearer / Data Services Data transfer using OSI layers 1, 2, Packet-switched protocols at 300
and 3. Supports transparent mode bps – 9.6 kbps
(standard channel coding) and non-
transparent mode (GSM-specific
coding efficiency)
Supplementary ISDN Digital features unavailable in Call diversion, closed user groups,
Services analog networks caller ID, SMS, cell broadcast
Two features of GSM stand out from a user perspective:
• SIM Card (Subscriber Identity Module): This is arguably GSM's most innovative feature. The
SIM is a removable smart card that stores the subscriber's International Mobile Subscriber
Identity (IMSI), authentication keys, network access permissions, and user-specific data.
Without a SIM inserted, a GSM handset is completely non-functional and has no identity. The
subscriber activates service using a 4-digit PIN. Crucially, a subscriber can insert their SIM into
any compatible GSM handset anywhere in the world — hotel phone, rented phone, borrowed
phone — and all calls will be billed to their home account. This portability of identity was
revolutionary.
• On-Air Privacy: Unlike analog FM cellular (AMPS) which could be trivially monitored with a
radio scanner, GSM encrypts the digital bit stream transmitted over the air. The encryption uses
a secret cryptographic key that is known only to the cellular carrier and changes with each user
and each call. This makes eavesdropping on a GSM radio link virtually impossible. All GSM
equipment manufacturers must sign a Memorandum of Understanding (MoU) before receiving
the cryptographic algorithms.
• Short Message Service (SMS): GSM introduced SMS as part of its supplementary services.
Messages up to 160 characters (7-bit ASCII) can be sent while the phone simultaneously
carries a voice call. The cell broadcast variant allows the base station to send the same
message to all subscribers in a coverage area — used for highway alerts, weather warnings,
etc.
• Discontinuous Transmission (DTX): In a typical conversation, each party speaks less than
40% of the time. GSM exploits this by incorporating a Voice Activity Detector (VAD) in the
speech coder. When no voice activity is detected, the transmitter is switched off, saving battery
power and reducing interference to other users. At the receiving end, a Comfort Noise
Subsystem (CNS) generates low-level background noise to avoid the unnatural silence that
would otherwise occur.
Base Station Subsystem Radio Subsystem Manages all radio transmission between mobile
(BSS) stations and the MSC. Contains BSCs and
BTSs.
• HLR (Home Location Register): A permanent database that stores the service profile and
current location of every subscriber who calls this network home. When a subscriber is assigned
an IMSI number, a record is created in the HLR. The HLR knows the current VLR address for
each roaming subscriber, allowing incoming calls to be correctly routed.
• VLR (Visitor Location Register): A temporary database that stores subscriber information for
every roaming mobile currently in the coverage area of a particular MSC. When a mobile enters
a new MSC area, the VLR requests the subscriber's profile from the HLR and stores it locally.
This avoids the need to query the HLR (which may be geographically distant) for every call.
• AUC (Authentication Center): A highly secured database that stores and manages the
encryption and authentication keys (Ki) for every subscriber in the HLR/VLR. It generates
authentication triplets used to verify a subscriber's identity and produce the encryption key for
that session. It also contains the Equipment Identity Register (EIR), which identifies stolen or
fraudulently modified handsets by checking their IMEI.
📌 Key insight: Handovers between two BTSs controlled by the same BSC are handled by the
BSC alone — the MSC is not involved. This is called intra-BSC handover and greatly reduces
MSC load. Only inter-BSC handovers (between BTSs of different BSCs) require MSC
involvement.
GSM Interfaces
Interface Connects Standard / Protocol Key Point
📌 Think of an ARFCN as a highway lane (200 kHz wide). TDMA divides time on that lane into 8
turns. Each subscriber gets one turn per frame. This means each user has a raw channel bit rate
of 270.833 / 8 = 33.854 kbps, of which only 24.7 kbps is actual user data (the rest is overhead).
Parameter Value
Total ARFCNs in 25 MHz band 124 (practical, with 100 kHz guard bands at
edges)
Parameter Value
• Full-Rate TCH: User data occupies one time slot per frame. A full-rate speech channel
(TCH/FS) carries digitized speech at 13 kbps raw, which becomes 22.8 kbps after GSM channel
coding. This is the standard mode.
• Half-Rate TCH: User data occupies one time slot but only in every other frame (alternating
frames). Two half-rate users share the same time slot. A half-rate speech channel (TCH/HS) is
designed to carry ~6.5 kbps raw speech (using a more advanced codec), becoming 11.4 kbps
after coding. This effectively doubles capacity at some voice quality cost.
In the speech multiframe (26 TDMA frames = 120 ms), TCH data occupies 24 of the 26 frames. The
13th frame carries SACCH data and the 26th is idle (for full-rate) or carries SACCH for half-rate.
• FCCH (Frequency Correction Channel): Transmits a special data burst of all zeros, which
appears as a pure unmodulated tone. Mobiles use this to synchronize their internal local
oscillator to the exact frequency of the base station. It occupies TS0 of frame 0 and repeats
every 10 frames.
• SCH (Synchronization Channel): Transmitted in TS0 of the frame immediately following the
FCCH frame. It carries the Frame Number (FN, ranging 0 to 2,715,647) and the Base Station
Identity Code (BSIC). The mobile uses the FN to synchronize its TDMA frame counter with the
base station. Since a mobile can be up to 30 km from the BTS, timing advance commands are
also sent via SCH.
• BCCH (Broadcast Control Channel): Broadcasts cell identity, network identity, current control
channel structure, channel availability, and congestion information. Frames 2–5 of every 51-
frame control multiframe contain BCCH data.
• PCH (Paging Channel — Forward Link): Used by the base station to alert a specific mobile of
an incoming call (originated from PSTN). The base station broadcasts the IMSI of the target
mobile. All mobiles in the cell monitor the PCH during their assigned paging slot.
• RACH (Random Access Channel — Reverse Link): Used by a mobile to request service
(make a call) or to acknowledge a page. Since multiple mobiles may attempt to access the
RACH simultaneously, it uses a slotted ALOHA access scheme. All mobiles must transmit
RACH within TS0 of any GSM frame.
• AGCH (Access Grant Channel — Forward Link): The base station's response to a RACH
transmission. It assigns the mobile to a specific ARFCN and time slot for SDCCH signaling. This
is the final CCCH message before a subscriber moves off the control channel.
BCH Forward only FCCH, SCH, BCCH Frequency sync, frame sync, cell info
broadcast
CCCH Both (PCH/AGCH fwd, PCH, RACH, AGCH Paging, random access, channel
RACH rev) assignment
For PSTN-originated calls, the process is similar but starts with the base station broadcasting a PCH
paging message. The mobile acknowledges on the RACH, receives an AGCH assignment, and the rest
of the procedure follows the same SDCCH → TCH path.
📌 The SDCCH is critical: it prevents the base station from assigning a precious TCH until the
subscriber has been fully authenticated and all call setup signaling is complete. This keeps TCH
resources available for active speech.
Time Slot (TS) 156.25 bits = 1 burst 576.92 µs Basic unit. 8 TSs make one
frame.
TDMA Frame 8 Time Slots 4.615 ms 8 users share one ARFCN per
frame.
The control multiframe is deliberately 51 frames (235 ms) while the speech multiframe is 26 frames
(120 ms). These are different lengths to guarantee that any GSM subscriber — regardless of which
time slot they are using for TCH data — will always receive the SCH and FCCH transmissions from the
BCH, since the two multiframe lengths are coprime. This ensures robust synchronization for all users.
The hyperframe is important for security: GSM encryption algorithms use the frame number (FN) as
part of the key stream generation. A hyperframe contains 2,715,648 TDMA frames, providing sufficient
key stream length to make the encryption secure without key repetition for any practical conversation
length.
Tail Bits (start) 3 Fixed zero bits to clear the convolutional decoder state
Guard Period 8.25 Empty period to prevent overlap with adjacent time
slots
📌 The 26-bit training sequence (midamble) is the key to GSM's robustness. By knowing exactly
what sequence was sent, the receiver's adaptive equalizer can measure the channel distortion
(multipath delay spread) and correct for it — crucial for reliable reception even in urban
environments with heavy reflections.
Speech Compresses voice from PCM Uses Residually Excited Linear Predictive
Coding(RELP-LTP) (64 kbps) down to 13 kbps. coder enhanced with a Long-Term Predictor.
The coder analyzes 20 ms Chosen after extensive subjective listening
blocks of speech and tests.
produces 260 bits.
Channel Adds redundancy to protect The 260 bits are classified by importance:
Coding(Error against channel errors. 260 Type Ia (50 most important) get 3 CRC parity
Protection) bits → 456 bits per 20 ms bits + convolutional coding. Type Ib (132 bits)
frame. get convolutional coding only. Type II (78
least important) pass uncoded. Rate 1/2 conv.
code, K=5.
Interleaving Spreads the 456 bits across 8 If a burst is lost due to a fade, only 1/8th of the
consecutive time slots (8 × 57- bits from each speech frame are lost — within
bit sub-blocks). Adjacent the error correction capability of the code.
bursts carry data from two Without interleaving, a single lost burst would
different speech frames. wipe out an entire speech frame.
Ciphering(A3 and Encrypts the interleaved data. Encryption key changes call-to-call. A5
A5) A3 authenticates the user. A5 algorithm is secret. Security enhanced by
generates the keystream that hyperframe-length key stream.
is XOR'd with the 114 data bits
in each burst.
Burst Formatting Adds tail bits, training The midamble is essential for equalization.
sequence (midamble), stealing Guard period prevents inter-slot interference.
flags, and guard period to form
the 156.25-bit normal burst.
Equalization At the receiver, the training Type of equalizer not specified in GSM
sequence is used to estimate standard — left to manufacturer. Typically a
the channel impulse response. Viterbi algorithm (MLSE) equalizer.
The equalizer compensates
for multipath distortion.
10.4 CDMA Digital Cellular Standard (IS-95)
IS-95 (Interim Standard 95), standardized by the Telecommunications Industry Association (TIA), is a
U.S. digital cellular system based on Code Division Multiple Access (CDMA). Unlike GSM's TDMA
approach — where users share a channel by taking turns in time — CDMA allows all users to transmit
simultaneously on the same frequency, distinguishing them by unique spreading codes. This is the
fundamental difference between the two systems.
IS-95 is designed to be backward-compatible with the existing U.S. AMPS analog system, so mobiles
and base stations can be produced economically for dual-mode operation. Each IS-95 CDMA channel
occupies 1.25 MHz of spectrum on each one-way link — compared to AMPS channels of 30 kHz. IS-95
is also fully compatible with the IS-41 networking standard used in North America.
📌 The key insight of CDMA: interference from other users looks like broadband noise. By
spreading data over a wide band (1.25 MHz), each user's signal appears as low-level noise to
others. The receiver uses the unique code to "despread" and recover the desired signal, while all
other users' signals remain as noise — the near-far problem must be carefully managed via
power control.
• User data rate varies: Unlike other standards where the channel chip rate is fixed and data
rate varies, in IS-95 the chip rate is always 1.2288 Mchip/s. User data rates (9600, 4800, 2400,
1200 bps) vary based on voice activity, and the spreading factor adjusts accordingly.
• No frequency planning: Because every cell uses the same 1.25 MHz channel, operators do
not need to plan frequency reuse patterns (unlike GSM where adjacent cells must use different
ARFCNs to avoid co-channel interference).
• Soft handoff: A CDMA mobile can maintain simultaneous links with two or more base stations
during a cell boundary transition. RAKE receivers at both the mobile and base stations combine
the signals, providing diversity gain rather than the abrupt "make-before-break" of GSM hard
handoffs.
• Variable data rate: The speech coder (QCELP) reduces its output rate to 1200 bps during
silence. The mobile transmits at lower power during these periods, directly reducing interference
to all other users in the system.
Parameter Value
Maximum User Data Rate 9.6 kbps (14.4 kbps with QCELP13)
Total Spreading Factor (fwd) 128 (from 9600 bps → 1.2288 Mcps)
Guard Band (AMPS boundary) 270 kHz (9 AMPS channels each side)
Paging Channels 1 – 7 (up to 7) 9600, 4800, or Send paging and control messages from
2400 bps base station to mobiles. Also used for
system parameter messages.
Forward Traffic Remaining 9600 / 4800 / Carry actual user voice/data. Variable rate
Channels codes 2400 / 1200 bps depending on voice activity.
• Convolutional Encoding (Rate 1/2, K=9): User data at 9600/4800/2400/1200 bps is encoded
using a rate 1/2 convolutional code with constraint length 9. Generator vectors are G0=753
(octal) and G1=561 (octal). This doubles the bit rate but adds error correction capability.
• Symbol Repetition: To maintain a constant baseband symbol rate of 19.2 kbps, symbols are
repeated for lower data rates. At 4800 bps, each symbol is repeated once (×2). At 2400 bps,
each is repeated 3 times (×4). At 1200 bps, each is repeated 7 times (×8).
• Block Interleaving: Symbols are written into a 24×16 array (20 ms block) by columns and read
out by rows. This separates originally adjacent symbols so that a burst error affects spread-out
positions, within the correction capability of the convolutional code.
• Data Scrambling (Long PN): The interleaved data is XOR'd with a decimated version of the
2^42-1 chip long PN code. The long code is unique per user (based on their ESN/MIN),
providing privacy. The 1.2288 Mcps long code is decimated by 64 to produce 19.2 ksps for
scrambling.
• Power Control Subchannel Insertion: One power control bit is inserted every 1.25 ms (800
b/s total) by puncturing (replacing) a data symbol. This bit instructs the mobile to raise (0) or
lower (1) its transmit power by 1 dB.
• Orthogonal Covering (Walsh Function): Each traffic channel is spread by a 64-chip Walsh
function at 1.2288 Mcps. Each data symbol (52.083 µs) is replaced by 64 Walsh chips. Different
users in the same cell use different Walsh functions — providing perfect code-division
separation on the forward link.
• Quadrature Spreading (Pilot PN): The Walsh-spread signal is further spread by the short I-
channel and Q-channel pilot PN sequences (period 2^15 - 1 chips). All base stations in the
system use the same pilot PN sequences but with different time offsets, allowing mobiles to
distinguish base stations.
• Public Mask: Based on the mobile's Electronic Serial Number (ESN). Used for all initial CDMA
calls. The upper 9 bits are set to 1100011000 and the remaining 32 bits are a permuted version
of the 32-bit ESN.
• Private Mask: Based on the mobile's MIN (Mobile Identification Number). Transition to private
mask occurs after authentication. Provides stronger privacy than public mask.
• Open-Loop Power Control: The mobile estimates path loss by measuring the pilot channel
received power and sets its initial transmit power inversely proportional to the received signal.
This handles large-scale path loss and shadowing.
• Closed-Loop Power Control: The base station measures each mobile's received signal-to-
interference ratio every 1.25 ms and sends a 1-bit update (via the forward traffic channel power
control subchannel). The mobile adjusts in 1 dB steps. With 800 updates/second, this can track
rapid fading variations.
📌 IS-95 power control is far more aggressive than GSM's SACCH-based power control (which
updates every ~480 ms). The 800 updates/second closed-loop rate is fast enough to track
Rayleigh fading at typical mobile speeds, which GSM cannot do.
[Link] Orthogonal Covering
The 64 Walsh functions used for forward-link orthogonal covering form a 64×64 Hadamard matrix.
Each row is a binary sequence of length 64 that is completely orthogonal (zero cross-correlation) with
every other row. Walsh function 0 (all zeros) is assigned to the pilot channel. Walsh function 32 goes to
the synchronization channel. Paging channels use the lowest-numbered remaining Walsh functions,
and traffic channels use the rest. Each 52.083 µs data symbol is spread to 64 chips at 1.2288 Mcps,
providing a processing gain of 64.
Pilot Channel Yes (shared by cell) No pilot — base uses special correlator
Power Control Pilot-based open loop + forward Open loop + closed loop (800 bits/s from
subchannel base)
• Rate 1/3 Convolutional Code (K=9): Stronger coding rate than forward link (1/3 vs 1/2)
because the reverse link has no pilot-aided coherent detection — noncoherent detection is
used, which requires more coding gain to achieve the same BER. Fixed output = 28,800 sps.
• Block Interleaving (32×18 array): 20 ms block interleaver. Symbols written by columns, read
by rows.
• 64-ary Orthogonal Modulation: Every group of 6 coded bits selects one of 64 Walsh functions
to transmit. The Walsh function number = c0 + 2c1 + 4c2 + 8c3 + 16c4 + 32c5. Instead of
transmitting 6 bits directly, the system transmits 64 Walsh chips — providing significant coding
gain (6 bits → 64 chips = ~10.7 dB gain). Walsh chips are transmitted at 307.2 kcps.
• Variable Rate / Data Burst Randomizer: At lower data rates, only some power control groups
(1.25 ms slots) are transmitted. A 14-bit block taken from the long PN code determines
randomly which slots are active — the data burst randomizer ensures each repeated code
symbol is transmitted exactly once.
• Direct Sequence Spreading: Walsh chips are spread by the user's long PN code running at
1.2288 Mchip/s. Each Walsh chip is spread by 4 long code chips (1.2288/0.3072 = 4).
• Quadrature Modulation (OQPSK): The spread signal is modulated using Offset QPSK. The Q
channel is delayed by half a chip (406.9 ns) relative to the I channel. This offset prevents the
signal envelope from passing through zero during transitions, reducing the peak-to-average
power ratio and making the power amplifier more efficient.
• Forward Link: Convolutional code rate changed from 1/2 to 3/4 by puncturing (deleting) 2 out
of every 6 symbols from the original rate 1/2 encoded stream. Effective information rates
become 14,400 / 7200 / 3600 / 1800 bps.
• Reverse Link: Convolutional code rate changed from 1/3 to 1/2. Effective information rates
similarly increase.
The 14.4 kbps channel is used with the QCELP13 variable-rate speech coder:
• QCELP13 (Qualcomm Code Excited Linear Predictive Coder 13): An improved version of
the original QCELP. It operates at variable rates — highest rate (13.4 kbps equivalent) when
speech is active, lowest when idle — just like the original QCELP. Improvements include better
spectral quantization, improved voice activity detection, better pitch prediction, and pitch post-
filtering. The coder has multiple operating modes — Mode 0 behaves identically to the original
QCELP for backward compatibility.
• Intermediate rates: Used for stationary voiced frames, unvoiced frames, and transition frames,
reducing the average data rate and thus average interference in the system — directly
increasing system capacity.
📌 QCELP13 in Mode 0 is identical to the original 9600 bps QCELP — this backward
compatibility means networks can use QCELP13 without breaking connections to older handsets.
Channel Bandwidth 200 kHz per ARFCN 1.25 MHz per CDMA channel
Speech Coder RELP-LTP @ 13 kbps (22.8 QCELP @ 9.6 kbps variable (QCELP13
kbps with coding) @ 14.4 kbps)
Channel Coding Rate 1/2 conv., K=5 (speech) Rate 1/2 K=9 (fwd), Rate 1/3 K=9 (rev)
Frequency Planning Required — adjacent cells use Not needed — frequency reuse factor = 1
different ARFCNs
Power Control Slow (~2 updates/s via Fast closed-loop (800 updates/s)
SACCH)