CPRI
How CPRI Moves Radio Samples Over Fiber
CPRI was defined in 2003 by a cooperation of base-station vendors (Ericsson, Huawei, NEC, Nokia, and others) to standardize the internal interface between the baseband processing of a base station and its radio frontend. Before CPRI, that connection was proprietary and forced operators to source the baseband and the radio from a single vendor. By digitizing the antenna signal into IQ samples and shipping them over a generic serial fiber link, CPRI enabled the radio head to be mounted at the antenna (cutting the lossy coaxial feeder run) and the baseband to sit in a sheltered cabinet, the architecture that underpins centralized RAN deployments.
The frame structure is hierarchical and clocked to the LTE/NR 10 ms radio frame. A basic frame spans one UMTS chip period (260.42 ns) and contains 16 words; one control word per basic frame carries management, Layer 1 in-band signaling, and the vendor-specific channel, while the remaining words carry user-plane IQ. Sixty-four basic frames form a hyperframe, and 150 hyperframes form a 10 ms radio frame, giving the receiving end a deterministic reference for frame and phase alignment. Because the link is constant-bit-rate, there is no statistical multiplexing gain: a radio that needs 9.8 Gbps of payload occupies a full CPRI option 7 wavelength even when idle.
Line coding adds the only meaningful overhead. CPRI options 1 through 7 use 8B/10B coding (80 percent efficiency), while the higher options 8, 9, and 10 use 64B/66B coding (about 97 percent efficiency) to keep the optics practical at multi-gigabit rates. The protocol distributes a recovered clock to the radio, so the radio unit needs no local high-stability oscillator; the cable delay is measured and calibrated (the T14 round-trip measurement) so the radio can compensate for the fixed fiber propagation when timestamping.
Line-Rate and Latency Equations
Rpayload ≈ fs × bIQ × 2 × AxC (bps)
CPRI line rate (with coding overhead):
Rline = Rpayload / ηcode, η8B/10B = 0.8, η64B/66B ≈ 0.97
One-way fronthaul fiber latency:
tfiber ≈ 5 μs/km × Lkm
Where fs = sampling rate, bIQ = bits per I or Q sample (typically 15), the factor 2 accounts for I and Q, and AxC = number of antenna-carrier streams. Example: 30.72 Msps × 15 × 2 × 1 ≈ 0.92 Gbps payload → option 3 (2.4576 Gbps) after overhead and additional paths. The ±250 μs HARQ budget limits reach to roughly 15 to 20 km.
CPRI Line-Rate Options
| Option | Line Rate | Line Coding | Coding Efficiency | Typical Use |
|---|---|---|---|---|
| Option 1 | 614.4 Mbps | 8B/10B | 80% | Single narrowband carrier |
| Option 3 | 2.4576 Gbps | 8B/10B | 80% | 20 MHz LTE, 1 to 2 antennas |
| Option 5 | 4.9152 Gbps | 8B/10B | 80% | 4T4R LTE radio |
| Option 7 | 9.8304 Gbps | 8B/10B | 80% | Multi-band macro radio |
| Option 8 | 10.1376 Gbps | 64B/66B | ~97% | High-capacity radio |
| Option 9 | 12.16512 Gbps | 64B/66B | ~97% | Wideband multi-carrier |
| Option 10 | 24.33024 Gbps | 64B/66B | ~97% | Peak CPRI; beyond this use eCPRI |
Frequently Asked Questions
How do you calculate the CPRI line rate for a given antenna configuration?
Compute the payload as R ≈ fs × bIQ × 2 × AxC, then divide by the coding efficiency (0.8 for 8B/10B) and round up to the nearest standard option. A single 20 MHz LTE carrier at 30.72 Msps with 15-bit IQ needs ~0.92 Gbps and maps to option 3 (2.4576 Gbps). A 4T4R 20 MHz radio typically lands at option 5 (4.9152 Gbps), while a 100 MHz 64T64R massive-MIMO radio would need ~157 Gbps of raw CPRI, which is why such radios moved to eCPRI.
What is the latency and synchronization budget on a CPRI fronthaul link?
CPRI is constant-bit-rate, so its dominant delay is fiber propagation at ~5 μs/km one way. The 3GPP HARQ round-trip budget of ~250 μs limits practical reach to about 15 to 20 km. The link also carries timing: it must hold ±2 ppb frequency accuracy and ~±8 ns round-trip delay accuracy so the radio can recover phase for TDD frame alignment and MIMO. The T14 cable-delay measurement calibrates out the fixed fiber propagation.
How does CPRI differ from eCPRI, and why did 5G adopt eCPRI?
Classic CPRI ships raw time-domain IQ over dedicated fiber at a constant rate, so bandwidth grows linearly with antennas and channel bandwidth; a 100 MHz, 64-antenna radio would need well over 100 Gbps. eCPRI raises the functional split (typically split 7-2x) and sends partially processed data, cutting the bitrate 5 to 10 times, and it is packetized over standard Ethernet or IP for statistical multiplexing and shared transport. CPRI persists on legacy 4G radios; eCPRI dominates 5G massive MIMO.