Overview — Evolving in Software and Antennas

The challenge facing operators in 2007 was straightforward: smartphone adoption was accelerating rapidly, LTE standardisation was underway in Rel-8 but commercial deployment was 3–4 years away, and existing HSPA networks were already running close to their spectral efficiency limits. Upgrading every cell site to a new radio standard was not practical on that timescale.

Rel-7 answered this with HSPA+ (HSPA Evolution) — a set of enhancements that could be deployed largely through software upgrades on existing UMTS base stations combined with new antenna hardware. The two headline radio features are 2×2 MIMO (two transmit, two receive antenna chains) and 64-QAM downlink modulation. Either alone roughly doubles peak throughput; combined, they produce a nearly fourfold increase over the baseline HSPA defined in Rel-5/6.

Beyond raw speed, Rel-7 also addressed battery life — an increasingly critical issue as smartphones sat connected to the network continuously — with CPC (Continuous Packet Connectivity), which introduced a way for devices to stay registered while letting their radios sleep between data bursts.

2×2 MIMO on the Downlink

MIMO (Multiple-Input Multiple-Output) is a technique that uses multiple antenna chains at both the transmitter and receiver simultaneously — not as diversity backup (which was already used in UMTS), but to transmit genuinely independent data streams at the same time on the same frequency.

In Rel-7 HSPA+ downlink MIMO, the Node B uses two transmit antennasand the UE uses two receive antennas. The two streams are transmitted simultaneously on the same 5 MHz W-CDMA carrier:

• The Node B precodes the two streams using a precoding matrix, selected to exploit the spatial channel between the antenna pairs
• The UE feeds back a Precoding Control Indicator (PCI) and a CQI for each stream, telling the Node B which precoding matrix to use and what data rate each stream can support
• The two received signals are a mixture of both transmitted streams, distorted by the channel. The UE applies MIMO detection (e.g. MMSE or successive interference cancellation) to mathematically separate the streams
• In good channel conditions — high SNR and low spatial correlation between the antenna paths — both streams decode cleanly, effectively doubling the data rate with no additional spectrum
• In poor conditions (low SNR or highly correlated paths), the Node B can fall back to transmit diversity mode, sending one stream with redundancy across both antennas for robustness

The spatial separation that makes MIMO work depends on the antenna elements being sufficiently separated and the propagation environment having enough scattering to create distinct channel paths. In practice, MIMO gains are highest in rich scattering urban environments and lower in line-of-sight or low-scatter conditions.

64-QAM Downlink Modulation

QAM (Quadrature Amplitude Modulation) encodes data by varying both the amplitude and phase of the carrier signal, producing a constellation of distinct symbols. Each symbol carries a fixed number of bits, determined by the size of the constellation:

• QPSK (Rel-5 baseline) — 4 symbols in a 2×2 pattern; 2 bits per symbol. The symbols are well-separated and easy to distinguish even with significant noise — ideal for poor radio conditions
• 16-QAM (introduced with HSDPA in Rel-5) — 16 symbols in a 4×4 grid; 4 bits per symbol. Requires a better signal than QPSK but delivers twice the spectral efficiency per symbol
• 64-QAM (new in Rel-7) — 64 symbols in an 8×8 grid; 6 bits per symbol. Each symbol carries 50% more bits than 16-QAM, but the symbols are much more closely spaced — any noise or interference can cause the receiver to misidentify a symbol, so 64-QAM is only used when the received SNR is high

The Node B selects the modulation order for each 2 ms HSDPA slot based on the CQI reported by the UE. For a UE close to the base station with an excellent channel, 64-QAM is selected automatically; for a UE at the cell edge, the Node B drops back to 16-QAM or QPSK to maintain error-free delivery.

The practical consequence is that the inner portion of each cell — where most of the users are during peak hours — benefits substantially from 64-QAM, while coverage range is unchanged because the fallback to lower-order modulation is automatic.

Peak Rates with MIMO and 64-QAM Combined

The two Rel-7 radio techniques compound each other:

• 64-QAM alone (no MIMO): approximately 21.6 Mbps DL peak with five HS-DSCH codes (UE Category 13)
• 2×2 MIMO alone (16-QAM on each stream): approximately 28.8 Mbps DL peak — two streams each carrying the equivalent of HSDPA Cat 12 (UE Category 14)
• 64-QAM + 2×2 MIMO together: theoretical peak of 28.8 Mbps DL (Category 14) — since in Rel-7 the two techniques are not combined simultaneously on the same device; the device supports one or the other depending on its category

Rel-8 later added dual-carrier HSPA+ (two adjacent 5 MHz W-CDMA carriers aggregated), which combined with MIMO reaches 42 Mbps DL and represents the practical ceiling for HSPA+ before it becomes directly comparable to early LTE deployments.

On the uplink, Rel-7 introduced 16-QAM modulationfor HSUPA, raising the peak from 5.76 Mbps (QPSK) to 11.5 Mbps — a significant improvement for video calling, cloud backup, and large file uploads.

CPC — Continuous Packet Connectivity

The smartphone era created a new radio resource problem that had not existed in the feature-phone world: devices maintain a packet connection continuously — for push email, messaging, background sync, keep-alive signals — but the actual data exchange happens in short, infrequent bursts. Under basic HSPA, maintaining a connection means the UE keeps its radio active between bursts, consuming battery power even while no data flows.

CPC (Continuous Packet Connectivity) introduced two complementary mechanisms to let the UE's radio sleep between bursts while remaining registered and reachable:

The UE remains in the Cell_DCH state — fully registered with a dedicated channel, reachable within milliseconds — but its radio hardware is powered down for the majority of the time between data bursts. Measurements from commercial deployments showed battery life improvements of 30–50% for devices that are connected but data-idle (the dominant state for a smartphone between active uses).

The DTX/DRX pattern that CPC introduced became the standard mechanism for radio power saving in every subsequent generation. LTE defined its own DRXcycles for the PDCCH in Rel-8, using the same fundamental principle. 5G NRextends this further with connected-mode DRX configurable at very fine granularity — all tracing back to the concept proven in Rel-7.

EDGE Evolution

Rel-7 also enhanced the 2G data path — EGPRS (EDGE) — for operators who still ran large GSM-only coverage areas and needed to improve the data experience there without deploying 3G infrastructure.

EDGE Evolution (also called EGPRS2 in some contexts) introduced:

• Higher-order modulation — adding 16-QAM and 32-QAM modulation schemes to EDGE's existing GMSK and 8-PSK, increasing the theoretical peak from 473 kbps to over 1 Mbps on the downlink in good conditions
• Latency reduction — reducing the round-trip time for EDGE connections from roughly 200–300 ms to around 100 ms, making interactive applications significantly more responsive
• Receiver diversity — allowing dual-antenna receive configurations in EDGE base stations to improve sensitivity and reduce interference

While not as dramatic as the HSPA+ improvements, EDGE Evolution meaningfully extended the commercial life of GSM networks and improved the experience for hundreds of millions of users who remained on 2G-only coverage for years after 3G was available in urban areas.