CQI Table
How the CQI Table Drives Adaptive Transmission
The CQI table is the shared codebook that lets a base station and a mobile device agree on how aggressively to load the radio channel. In LTE the canonical version is Table 7.2.3-1 of 3GPP TS 36.213; 5G NR carries the same idea forward in TS 38.214. Each of the 16 possible CQI values is bound to a triplet: a modulation scheme, an approximate code rate expressed as a fraction times 1024, and the spectral efficiency that combination yields. CQI 0 is reserved to signal that the channel is so poor that no entry is usable (out of range), while CQI 1 through 15 step monotonically upward in efficiency from a heavily coded QPSK link to lightly coded 64-QAM or 256-QAM.
The reporting loop runs continuously. A user equipment measures the cell-specific or CSI reference signals, estimates the effective signal-to-interference-plus-noise ratio, and selects the largest CQI index whose required SINR is still met given a 10% first-transmission block error target. That single number is fed back on the uplink control channel, periodically or aperiodically, and the scheduler converts it into a modulation and coding scheme (MCS) index for the actual grant. Because the CQI is a quantized, delayed snapshot, schedulers add an outer-loop offset driven by HARQ ACK/NACK statistics, nudging the operating point up or down to hold the true block error rate near target.
The practical payoff is throughput that scales with channel conditions. A device near the cell edge under deep fading might report CQI 2 or 3 and run QPSK at a low code rate for robustness, while a stationary device with strong line-of-sight reports CQI 14 or 15 and runs 64-QAM or 256-QAM near rate 0.93. The same table, evaluated independently by millions of devices, lets a network squeeze close to the Shannon limit of each link without any per-user manual tuning.
Spectral Efficiency and Required SINR
Each table entry implies a minimum SINR needed to meet the 10% block error rate target. The mapping is not the bare Shannon bound; it includes the coding gain of the turbo or LDPC code and an implementation margin, so a practical CQI-to-SINR curve sits a few decibels above the theoretical capacity line. The governing relationships below show how efficiency, code rate, and modulation order combine.
η = log2(M) × R (bits per symbol)
Code Rate from the 1024-Scaled Field:
R = R1024 / 1024
Shannon Reference Bound:
ηmax = log2(1 + SINR)
Where M = modulation order (4 for QPSK, 16 for 16-QAM, 64 for 64-QAM, 256 for 256-QAM), R = code rate, R1024 = the integer code-rate field the table lists. Example: 64-QAM at R = 666/1024 ≈ 0.65 gives η = 6 × 0.65 ≈ 3.90 bits per symbol, matching CQI 12. Practical SINR to reach this entry is ≈ 13 to 14 dB, a few dB above the ηmax bound.
LTE CQI Table (64-QAM, TS 36.213 Table 7.2.3-1)
| CQI Index | Modulation | Code Rate × 1024 | Efficiency (bits/sym) | Typical SINR |
|---|---|---|---|---|
| 0 | Out of range | n/a | n/a | < -6 dB |
| 1 | QPSK | 78 | 0.1523 | ≈ -6 dB |
| 4 | QPSK | 308 | 0.6016 | ≈ 1 dB |
| 7 | 16-QAM | 378 | 1.4766 | ≈ 6 dB |
| 10 | 64-QAM | 466 | 2.7305 | ≈ 11 dB |
| 12 | 64-QAM | 666 | 3.9023 | ≈ 14 dB |
| 15 | 64-QAM | 948 | 5.5547 | ≈ 20 dB |
Frequently Asked Questions
What is the difference between a CQI value and an MCS index?
CQI is a 4-bit report (0 to 15) the UE sends indicating the highest modulation and code rate that keeps block error rate at or below 10%; the CQI table defines what each value means. The MCS index (0 to 28 in LTE, up to 27 or 31 in 5G NR) is the actual setting the scheduler grants, taken from a separate MCS table. The scheduler maps reported CQI to MCS, usually applying an outer-loop offset from HARQ ACK/NACK feedback, so the chosen MCS sits a few steps below the raw CQI to hold the target BLER.
Why does the CQI report target a 10 percent block error rate?
Per 3GPP TS 36.213 and TS 38.214, the UE picks the highest-efficiency CQI whose single initial transmission would not exceed 10% transport block error rate. That target pushes the link aggressively while leaving HARQ retransmissions to recover the roughly one-in-ten failed blocks at low average overhead. With chase combining or incremental redundancy, residual error after one or two retransmissions falls well below 1%, so throughput stays near maximum without sacrificing reliability.
How does 256-QAM change the LTE CQI table?
The original LTE table tops out at 64-QAM, with CQI 15 reaching about 5.55 bits per symbol. LTE-Advanced added an alternative CQI table where high indices map to 256-QAM at code rates up to about 0.93, raising peak efficiency to roughly 7.41 bits per symbol; the UE signals which table it uses via RRC. 5G NR extends this to three tables: a QPSK to 64-QAM table, a 256-QAM table, and a low-spectral-efficiency table with extra low-rate QPSK entries for URLLC links at very low SINR.