Coverage Enhancement
How Repetition Buys Range in Cellular IoT
The fundamental problem Coverage Enhancement solves is that low-power, low-cost IoT endpoints cannot afford big antennas, high transmit power, or expensive receivers, yet they are frequently installed exactly where signals are weakest: smart meters in concrete basements, agricultural sensors miles from a tower, asset trackers inside shipping containers. To reach these locations, 3GPP Release 13 introduced narrowband technologies (LTE-M, also called Cat-M1, and NB-IoT) with an explicit design target of 164 dB maximum coupling loss, about 20 dB beyond the roughly 144 dB that conventional LTE supports.
That 20 dB cannot come from raw power because the device is constrained to 20 or 23 dBm. Instead it comes from time. Each physical channel is transmitted repeatedly, and the receiver coherently combines the copies. Because thermal noise is uncorrelated between repetitions while the signal adds in phase, every doubling of the repetition count yields about 3 dB of processing gain. NB-IoT supports up to 2048 repetitions on the downlink shared channel (NPDSCH) and 128 on the random-access channel (NPRACH); LTE-M CE mode B supports up to 2048 repetitions on key channels. A second contribution is power spectral density boosting: squeezing the transmit power into a single 180 kHz NB-IoT carrier or a 1.08 MHz LTE-M narrowband raises the energy per resource element well above what a full 20 MHz LTE carrier would deliver.
The cost of this margin is throughput and latency. Hundreds of repetitions stretch a single transport block over seconds and drop the effective data rate to a few hundred bits per second. For this reason the standard defines graduated coverage classes so that a device in good signal does not waste airtime: only endpoints that genuinely need the deepest coverage pay the full repetition penalty.
Maximum Coupling Loss and Processing Gain
MCL = PTX − (S + NF + 10·log10(B) − 174) dB
Repetition Processing Gain:
Grep ≈ 10 × log10(Nrep) dB (≈ 3 dB per doubling)
Effective SNR after combining:
SNReff = SNR1 + Grep = SNR1 + 10·log10(Nrep)
Where PTX = transmit power (dBm), S = required SNR, NF = receiver noise figure (dB), B = effective noise bandwidth (Hz), −174 = thermal noise floor (dBm/Hz at 290 K), Nrep = number of repetitions. Example: going from 1 to 64 repetitions adds ≈ 18 dB, lifting a device near the 164 dB MCL into a decodable SNR.
Coverage Classes and Repetition Budgets
| Coverage Class | Technology | Target MCL | Typical Repetitions | Approx. Gain | Throughput / Latency |
|---|---|---|---|---|---|
| Normal (CE level 0) | NB-IoT / LTE-M mode A | ~144 dB | 1 to 8 | 0 to 9 dB | Highest / lowest |
| CE level 1 | NB-IoT | ~154 dB | 16 to 64 | ~12 to 18 dB | Moderate |
| CE level 2 | NB-IoT | ~164 dB | 128 to 2048 | up to ~20 dB | Lowest / seconds |
| CE mode A | LTE-M (Cat-M1) | ~155.7 dB | 1 to 32 | up to ~15 dB | Higher rate |
| CE mode B | LTE-M (Cat-M1) | ~164 dB | up to 2048 | up to ~20 dB | Deep indoor |
Frequently Asked Questions
How much extra link budget does Coverage Enhancement provide versus standard LTE?
It raises maximum coupling loss from about 144 dB for legacy LTE to 164 dB for NB-IoT, roughly a 20 dB gain. That margin comes from time-domain repetition (up to 2048 repetitions on NB-IoT NPDSCH), power spectral density boosting into a 180 kHz or 1.08 MHz narrowband, and robust coding. Each doubling of repetitions adds about 3 dB, so 16 repetitions yields roughly 12 dB while the full 20 dB needs hundreds of repetitions at the cost of throughput and latency.
What is the difference between CE mode A and CE mode B in LTE-M?
CE mode A is the lighter mode, supporting 1 to 32 repetitions for a moderate 5 to 15 dB extension with low latency and higher data rates. CE mode B is the deep-coverage mode, supporting up to 2048 repetitions for the full 15 to 20 dB extension needed in the most challenging deep-indoor sites, at the expense of latency measured in seconds and very low throughput. NB-IoT uses an analogous scheme of CE levels 0, 1, and 2.
How does the device choose its Coverage Enhancement level during random access?
Before sending a preamble the device measures RSRP and compares it against two thresholds broadcast in SIB2. Above the upper threshold it picks CE level 0; between the thresholds, CE level 1; below the lower threshold, CE level 2. Each level maps to a configured PRACH or NPRACH repetition count and a dedicated preamble resource set. If access fails after the configured attempts, the device escalates to the next level, trading battery and latency for more repetitions.