Short answer
Maximum coupling loss is how much signal a link can lose and still work. Qualcomm publishes NB-IoT at 164 dB against eMTC above 155.7 dB, and the extra budget is bought by repetition, sending the same message up to 2,048 times, which costs airtime and battery rather than being free.
Key takeaways
- Coupling loss is a budget in decibels, Qualcomm publishes NB-IoT at 164 dB and eMTC above 155.7 dB.
- Ericsson states both technologies can operate at 164 dB, matching the 5G coverage requirement.
- The mechanism behind the budget is repetition: up to 32 in LTE-M CE Mode A and up to 2,048 in CE Mode B.
- Every repetition spends cell airtime and device battery, so deep coverage has a running cost.
- A coverage claim without a stated coupling loss is a marketing sentence, not an engineering figure.
01What coupling loss is measuring
When a radio link is described as reaching further, the engineering statement underneath is about a budget. Maximum coupling loss is the total attenuation a signal can suffer between the transmitter and the receiver while still being recoverable, distance, walls, steel, ground, everything combined into one decibel figure. It is useful precisely because it does not care which of those things caused the loss. A hundred and sixty decibels lost to twelve kilometres of open desert and a hundred and sixty lost to three metres of concrete are the same problem to the modem. This is why coupling loss, rather than range in kilometres, is the honest way to compare radio classes: range depends on terrain that nobody can put in a datasheet, while the budget is a property of the technology and travels between deployments unchanged.
02The published numbers, and their small disagreement
Two credible sources put the figures slightly differently, and the gap between them is itself instructive. Qualcomm publishes NB-IoT at 164 dB and eMTC above 155.7 dB, which is a difference of roughly eight decibels in NB-IoT's favour, a real advantage, since eight decibels is a large fraction of a wall. An Ericsson analysis, written from inside the specification work, states that both technologies can operate at 164 dB, matching the 5G coverage requirement. These are not contradictory so much as differently scoped: one reports a typical published capability, the other what the standard permits when the coverage-enhancement features are fully applied. For a buyer the lesson is to ask which of the two a vendor means, because a claim of parity at 164 dB carries conditions that a claim of 155.7 dB does not.
03Repetition is the mechanism, and it is not free
The way these technologies buy extra decibels is unglamorous: they send the same information again and again until the receiver can accumulate enough of it to distinguish signal from noise. The GSMA LTE-M deployment guide caps this at 32 repetitions in coverage-enhancement Mode A and 2,048 in Mode B, and those two ceilings define the two ends of a spending range rather than two settings of equal cost. Reading 2,048 as a capability is only half the sentence. The other half is that a message which took one transmission at good signal now takes two thousand, occupying the cell for that whole duration and drawing transmit current from the device for the same period. Coverage bought this way therefore appears twice on your bill: once as spectrum efficiency the operator loses, and once as battery the device never recovers.
04Where the budget actually goes at your site
A budget is only useful if you can see what is drawing on it, and at a real installation the large expenditures are usually structural rather than distance-related. A metal cabinet, a concrete pit, a tank wall, an underground vault, a container stack, each of these subtracts from the same allowance the distance to the tower is already consuming. This reframes several familiar decisions. Mounting position stops being an aesthetic or protective choice and becomes an accounting one: moving an antenna outside a steel enclosure returns decibels directly to the budget. Likewise a site that works in winter and fails in summer is often not a heat fault at all but a budget already spent to its limit, with a small seasonal change tipping it over. Measuring the received level at the exact install point is the only way to know where you stand.
05What the budget costs you in latency
Spending airtime on repetition has a consequence that shows up as user experience rather than as a radio metric. The reported latency contrast between the two technologies is already wide before repetition enters: Cat-M1 at 10 to 15 milliseconds against NB-IoT at 1.6 to 10 seconds. Ericsson's statement of the 5G requirement both technologies meet puts the acceptable service latency ceiling at 10 seconds for a million devices per square kilometre, which tells you the standard itself expects seconds rather than milliseconds at scale. Layer heavy repetition on top and the delivery of a single small message becomes a measurable wait. That is fine for a scheduled reading and unacceptable for an actuation, so a device installed at the edge of its budget quietly changes what product experience it can support.
06The narrow channel is part of the same trade
Channel width and coupling loss are two faces of one design decision, which is why NB-IoT sits at the deeper end of the budget. Concentrating a transmitter's power into a narrower band raises the power spectral density, so the receiver sees a stronger signal per hertz and can pull it out of more noise. NB-IoT occupies 180 kHz, exactly one LTE physical resource block, against 1.4 MHz for a Cat-M1 device, and the throughput follows: Cat NB1 peaks at 26 kbit/s down and 66 kbit/s up in multi-tone half duplex, while Cat-M1 reaches 1 Mbps in each direction. So the extra reach is not a bonus feature sitting alongside the data rate. It is the same decision seen from the other side, and you cannot buy one without paying the other.
07How to read a coverage claim
Three questions turn a coverage sentence into something checkable. First, what coupling loss is the claim made at, a figure in decibels can be verified against Qualcomm's or Ericsson's published numbers, while a claim of deep indoor coverage cannot be verified against anything. Second, how many repetitions are assumed to reach it, since a budget met at 2,048 repetitions in CE Mode B is a different product from the same budget met at 32 in Mode A. Third, what the resulting message latency and per-report energy are at that setting, because those are the two costs the claim does not mention. A vendor who can answer all three is describing engineering. A vendor who answers none is describing an aspiration, and the difference will surface at the worst-sited asset on the site.
| Lever | Published figure | What it costs |
|---|---|---|
| NB-IoT coupling loss | 164 dB | Throughput capped at Cat NB1 rates |
| eMTC coupling loss | Above 155.7 dB | About 8 dB less reach than NB-IoT |
| Standard-permitted parity | 164 dB for both | Requires the enhancement features applied |
| CE Mode A repetition | Up to 32 | Modest airtime and energy |
| CE Mode B repetition | Up to 2,048 | Heavy airtime and energy per message |
| Narrow channel | 180 kHz | Cat NB1: 26 kbit/s DL, 66 kbit/s UL |
| Wider channel | 1.4 MHz | 1 Mbps, but less reach |
Figures and sources
- Maximum coupling loss of each technology, the roughly 8 dB link-budget gap: NB-IoT 164 dB; eMTC above 155.7 dB (Source: Qualcomm, Leading the LTE IoT evolution, as of 2026-08-31)
- Coupling loss both technologies can operate at, per a 3GPP specification co-author: 164 dB, matching the 5G coverage requirement (Source: Ericsson, LTE-M and NB-IoT meet the 5G performance requirements, as of 2026-08-31)
- LTE-M coverage-enhancement repetition ceilings by mode, the mechanism that buys the budget: CE Mode A up to 32 repetitions; CE Mode B up to 2,048 (Source: GSMA LTE-M Deployment Guide v3 (June 2019), as of 2026-08-31)
- 5G connection-density requirement both technologies meet, and its latency ceiling: 1,000,000 devices per km² with service latency of at most 10 seconds (Source: Ericsson, LTE-M and NB-IoT meet the 5G performance requirements, as of 2026-08-31)
- Reported latency contrast between the two technologies before repetition is applied: Cat-M1 10–15 ms; NB-IoT 1.6–10 s (Source: IoT For All, What Is the Difference Between Cat-M1 and NB-IoT?, as of 2026-08-31)
- NB-IoT occupied bandwidth, the narrow channel that raises power spectral density: 180 kHz (one LTE physical resource block) (Source: GSMA NB-IoT Deployment Guide v3 (June 2019), as of 2026-08-31)
- Cat-M1 device bandwidth and peak rate, the other side of the same trade: 1.4 MHz bandwidth; 1 Mbps peak in DL and UL (Source: GSMA LTE-M Deployment Guide v3 (June 2019), as of 2026-08-31)
- NB-IoT peak data rates by 3GPP device category, and its duplex mode: Cat NB1: DL 26 kbit/s, UL 66 kbit/s multi-tone, half duplex; Cat NB2: DL 127, UL 159 kbit/s (Source: 3GPP device category specification table (Narrowband IoT), as of 2026-08-31)
Questions
What does 164 dB of coupling loss actually mean?
It means the link still works after the signal has lost that much strength between transmitter and receiver, whatever caused the loss, distance, a wall, a steel cabinet or all three. Qualcomm publishes NB-IoT at 164 dB and eMTC above 155.7 dB, so the figure is a budget rather than a distance.
If both technologies reach 164 dB, why choose between them?
Because the parity comes with conditions. Ericsson states both can operate at 164 dB when the coverage-enhancement features are applied, while Qualcomm's published figures show an eight-decibel gap. Reaching the deeper budget takes repetition, and the repetition costs airtime, energy and latency that the headline figure does not show.
Does more repetition mean better coverage for free?
No. The GSMA LTE-M guide caps repetition at 32 in CE Mode A and 2,048 in CE Mode B, and each repetition occupies the cell and draws transmit current from the device. A message that took one transmission at good signal can take two thousand at the edge, so deep coverage has a running cost in spectrum and battery.
How do I check whether my site has budget left?
Measure the received level at the exact install point rather than at the site entrance, and treat structures as expenditure: a metal cabinet, a concrete pit or a container stack all draw on the same allowance the distance to the tower is already using. Moving an antenna outside an enclosure returns decibels to the budget directly.