The Inefficiency of Circuit-Switched Data

Standard GSM (2G) is fundamentally architected as a circuit-switched network, engineered exclusively to guarantee the real-time delivery of human voice. When a user initiates a call, the Mobile Switching Center (MSC) allocates a dedicated, physical copper or fiber circuit through the core network, and simultaneously locks a specific TDMA time slot on the radio tower exclusively for that user.

This circuit remains permanently locked for the entire duration of the call. This architecture is perfect for voice, as it guarantees zero latency jitter. However, it is catastrophically inefficient for internet data. Internet traffic is intensely bursty. A user downloading a webpage receives a massive burst of data for two seconds, and then spends two minutes reading it. Under a circuit-switched paradigm (like early HSCSD data services), those physical circuits and radio time slots remain rigidly locked and entirely unusable by anyone else during those two minutes of silent reading, wasting massive amounts of incredibly expensive radio bandwidth.

GPRS: The Packet-Switched Pivot (2.5G)

General Packet Radio Service (GPRS), often referred to as 2.5G, represents a monumental architectural pivot. It marked the transition of cellular networks from legacy circuit-switched voice systems into the modern era of packet-switched internet routing.

GPRS did not replace GSM; it overlaid a completely separate, parallel packet-switched routing infrastructure directly onto the existing GSM radio towers. It introduced two massive, intelligent new routing nodes into the core network to handle IP traffic, completely bypassing the MSC.

Serving GPRS Support Node (SGSN)

The SGSN is the packet-switched equivalent of the MSC/VLR. It handles all complex mobility management, packet routing, logical link management, and authentication specifically for data traffic within its assigned geographical area. As the mobile user moves, the SGSN tracks their location at the routing area level, ensuring that IP packets arriving from the internet are funneled down to the correct cell tower.

Gateway GPRS Support Node (GGSN)

The GGSN is the crucial boundary node. It acts as the colossal IP router that physically connects the walled-garden mobile network to the external public Internet (or corporate intranets). The GGSN is responsible for assigning dynamic IP addresses to mobile phones, executing deep packet inspection, managing firewall rules, and acting as the Home Agent for Mobile IP tunnels.

The Power of Statistical Multiplexing

The true superiority of GPRS lies in its adoption of Statistical Multiplexing on the radio interface. Time slots are no longer permanently assigned to a user. Instead, they are placed in a shared pool.

When a user clicks a link, the network dynamically assigns time slots for the few milliseconds required to burst the IP packets. The exact millisecond the download finishes, those time slots are instantly released back into the pool. While User A is reading their downloaded webpage, User B can instantly utilize those exact same time slots to download their email. This allows the network to serve vastly more users on the exact same limited spectrum, and allows operators to charge users based on data volume (megabytes downloaded) rather than connection time (minutes connected).

UMTS and Wideband CDMA (3G)

While GPRS introduced packet-switched routing, it was still fundamentally bottlenecked by the narrow 200 kHz radio channels of GSM, limiting maximum speeds to roughly 114 kbps. The transition to 3G required a massive leap in radio access technology to support high-speed internet and video. The resulting standard was the Universal Mobile Telecommunications System (UMTS).

The UTRAN Architecture

To support the massive bandwidth demands, UMTS required the complete replacement of the legacy GSM Base Station Subsystem (BSS) with a radically new architecture called the UMTS Terrestrial Radio Access Network (UTRAN).

  • Node B: The Node B serves as the UMTS equivalent of the GSM BTS. However, it is significantly more computationally intensive. It handles the brutal mathematics of the physical layer, specifically the massive spreading and despreading operations of the W-CDMA orthogonal codes, executing complex Forward Error Correction (FEC), and managing ultra-fast inner-loop power control.
  • Radio Network Controller (RNC): Above the Node Bs sits the RNC, the highly intelligent equivalent of the GSM BSC. The RNC acts as the absolute dictator of the radio spectrum within its domain. It manages sophisticated radio resource allocation, executes admission control algorithms, handles cryptographic encryption, and orchestrates complex Soft Handoffs.

Wideband CDMA (W-CDMA)

UMTS entirely abandoned the narrow-band time-slicing (TDMA) of GSM in favor of massive, mathematics-driven spectral spreading known as Wideband CDMA (W-CDMA).

W-CDMA allocates massive, continuous 5 MHz frequency blocks. Every single user within the cell transmits on the exact same 5 MHz frequency, at the exact same time. They do not interfere with each other because their digital signals are multiplied by mathematically unique, orthogonal pseudo-noise codes, allowing the receiver to filter out everyone else as background static.

Combating Multipath Fading with Chipping Rates

This transition to massive bandwidth specifically engineered immense resilience against multipath fading. In W-CDMA, the data is spread using a chipping rate of 3.84 Mega-chips per second (Mcps). This extreme speed means the physical duration of a single chip (T_c) is incredibly short—approximately 0.26 microseconds.

This short chip duration is a powerful weapon against multipath reflections (echoes bouncing off buildings). If a delayed reflection arrives separated by more than 0.26 μs, the receiver’s mathematical correlator can easily distinguish it from the primary signal. The correlator treats the delayed reflection as uncorrelated, orthogonal noise and effectively rejects it.

Furthermore, UMTS utilizes a RAKE receiver. The RAKE receiver actively hunts for these delayed multipath reflections (called “fingers”). Instead of rejecting them, it captures them, delays the primary signal to perfectly align the phases, and sums their energy constructively. The massive bandwidth of W-CDMA allows the system to turn multipath fading from a catastrophic cause of Inter-Symbol Interference into a measurable signal gain.

UMTS Handoff Strategies: Make-Before-Break

The shift to W-CDMA also fundamentally revolutionized how handoffs are executed at the cell boundary.

The GSM Hard Handoff (Break-Before-Make)

GSM utilizes a Hard Handoff. Because adjacent GSM cells operate on entirely different FDMA frequency channels, a phone must completely sever its connection to the old tower, physically retune its radio to the new frequency, and connect to the new tower. This causes a mandatory ~50ms blackout, which drops TCP packets and can drop calls if the new cell is congested.

The UMTS Soft Handoff (Make-Before-Break)

UMTS utilizes a Soft Handoff. Because W-CDMA spreads all users across the exact same 5 MHz frequency universally, adjacent cells share the same spectrum. When a user approaches a cell boundary, the mobile phone uses its correlators to connect to the new Node B while simultaneously maintaining its active, locked connection to the old Node B.

This simultaneous connection enables Macrodiversity. During the transition zone, the mobile phone actively transmits its data stream to both Node Bs at the same time. Both Node Bs receive the signal, decode it, and forward the packets up the fiber backhaul to the RNC.

The RNC acts as a master synthesizer. It aligns the two incoming data streams in time, evaluates their checksums, and dynamically selects the highest-quality frames from either stream on a millisecond-by-millisecond basis. This “Make-Before-Break” approach practically eliminates dropped calls at cell edges and provides massive resilience against deep fades. Only when the mobile station has moved firmly into the new cell is the old link finally dropped.