6G is expected to run at around 7 GHz on the site grid that 5G already uses at around 3.5 GHz. That is the whole of the coverage problem in one sentence. The towers do not move, the frequency roughly doubles, and the link has to reach just as far.

RAN1 turned that into a measurement exercise rather than an argument. One formula, one reference channel, and a template that every company fills in. This note follows what the contributions below settled, what they measured, and where they still disagree.

Everything here comes from the RAN1 contributions listed in the Reference section at the foot of the page. Where the documents disagree with each other, both numbers are given.

• Executive Summary
• Acronym
• What does "same coverage" mean when 6G moves to 7 GHz?
• Why does the wording "comparable to same" matter?
• How large is the coverage gap between mid-band and 7 GHz?
• What do field measurements say about the 7 GHz penalty?
• Which channels are evaluated, and why is Msg3 the reference?
• Which signals and channels are on the list?
• What do the first Msg3 results look like?
• Why do company link budgets disagree by more than the gap they measure?
• What is in the collection of results?
• What does the Msg3 MIL row show?
• What data rate should the study treat as the coverage target?
• Why are the IMT-2020 rates not the answer?
• Which lower and upper bounds are proposed?
• What UE assumptions does the evaluation rest on?
• How small can the smallest maximum UE bandwidth be?
• How should SRS be designed for 6G coverage and capacity?
• Which sequence families are the starting point?
• Why is SRS capacity the limit that matters?
• Where does the separate NR coverage enhancement work item sit?
• How does the work item extend PRACH and modulation?
• How are repetition and beam indication handled?
• How large is the agreed gap, channel by channel?
• How does each channel stand against Msg3 in the same band?
• What separates the O2I and O2O columns?
• How much does receiver combining of SSB and PBCH recover?
• How much does four-shot combining recover indoors?
• What does it change outdoors, and what does it cost?
• Why does base station beamforming not fix the uplink?
• What do the antenna counts actually say?
• Why does the asymmetry leave the uplink short?
• Can a low PAPR uplink waveform buy back the link budget?
• What does the DFT-s-OFDM waveform contribute on its own?
• What is FDSS, and what separates the three options?
• How do CP-OFDM, DFT-s-OFDM and FDSS compare?
• Can a second carrier carry the uplink instead?
• Which use cases were agreed for study?
• Which carrier arrangements are under study?
• Can better coding of small uplink control payloads help?
• Why does the 5G Reed-Muller code run short?
• What is the counter-argument, and what is the prize?
• Can sparser SRS patterns improve uplink sounding coverage?
• Which comb values are under study?
• What limits and related agreements come with it?
• Can repetition close the gap, and at what price?
• What has the 6G study said about it so far?
• Where is repetition being specified, then?
• Reference

Executive Summary

The table below is a lookup rather than a narrative. Each row names an area, what the contributions say about it, and what follows from that.

Acronym

The list is short on purpose. Standard NR vocabulary is left out, because a reader of a 3GPP note already carries it. What remains is the vocabulary of the link budget, the deployment scenarios and the waveform proposals.

What does "same coverage" mean when 6G moves to 7 GHz?

A coverage target is only useful if everyone reads it the same way. RAN set the aim at re-using the existing 5G mid-band site grid, and RAN1 then had to turn that phrase into something a link budget can test. The gap between the two is what several contributions are about.

RAN#110 set two aims, and they point in different directions.

• Re-use the existing 5G mid-band site grid at around 3.5 GHz, for 6G deployments in at least around 7 GHz. The target there is the same coverage as 5G mid-band for initial access, and comparable to same coverage for data channels at the same data rate.
• Provide enhanced overall coverage compared to 5G in the same band. That aim is focused on cell-edge performance and on the uplink.

RAN1#123 turned the first aim into a method. Link budget template candidates 1 and 2 are the starting point for the metric. During initial access the coverage target refers to the bottleneck channel, which is Rel-15 NR Msg3 at 5G mid-band. Four deployment cases are considered: urban macro and sub-urban macro, each both outdoor-to-indoor and outdoor. The two carrier frequencies are 7 GHz for the 6G deployment and, for the existing 5G mid-band, a figure the agreement wrote in brackets as [4 GHz]. Those brackets mattered, and the section on the agreed tables below records where the number ended up.

An operator group then asked for the plainest possible reading of "same". Their proposal is that collocation is assumed, so same coverage for initial access means a UE that can access the 5G network can access the 6G network at the same location. For connected mode it means the same data rate per MHz at the same location.

The same contribution names what it expects to close the gap. It says the key technology is MIMO relying on a high number of antenna ports, giving 256 as an example, and more than 1000 antenna elements. It also asks that if repetitions turn out to be needed for initial access, the number of repetitions is written into the study outcome rather than left implicit.

Why does the wording "comparable to same" matter?

One word separates a target that can be missed from a target that cannot. The operator contribution asks for the RAN1#123 agreement to be reworded so that it matches what RAN actually decided.

The proposed change reads "targeting comparable to same coverage to 5G mid-band" in place of "targeting comparable coverage".

The same contribution asks for three more outputs: spectrum efficiency, latency and energy efficiency. An operator can then see the trade-off rather than a single coverage verdict.

• The target is relative, not absolute : coverage at 7 GHz is measured against an existing 3.5 GHz deployment, so the answer depends on what that deployment already achieves.
• Initial access and data channels have different targets : same coverage for the first, comparable to same coverage at the same data rate for the second.
• Collocation is the assumption that makes the test meaningful : the same UE at the same location is the comparison, not two separate network designs.
• Repetitions count as a cost, not a solution : the operator proposal asks for the required number to be stated explicitly if repetition is what closes the gap.

How large is the coverage gap between mid-band and 7 GHz?

Moving from 3.5 GHz to 7 GHz costs received power, and the study needs one number for how much. RAN1 agreed a formula that produces it. Field measurements then produced a larger number, and the difference between the two is the interesting part of this section.

The agreed calculation subtracts two maximum path losses and a frequency term. MPL1 is the maximum path loss of the evaluated signal or channel at around 7 GHz. MPL2 is the maximum path loss of the bottleneck channel at 5G mid-band, which is Rel-15 NR Msg3 for initial access. PL_diff is 20·log₁₀(f1 / f2), where f1 and f2 are the two carrier frequencies. An additional margin sits in the agreement in brackets and its value is still FFS.

Figure 1. The agreed coverage gap calculation. Three terms, one subtraction, and one value still marked FFS.

The frequency term is the part that can be computed without any assumptions. At 7 GHz against 3.5 GHz the ratio is exactly two, so PL_diff is 6.02 dB. Contributions that report the gap against a 3.5 GHz reference all carry that same 6.02 dB figure.

What do field measurements say about the 7 GHz penalty?

A formula gives the free-space part of the answer. Diffraction, scattering and building penetration are not in it, so an operator went and measured the difference instead.

Deutsche Telekom ran a propagation campaign at 3.6 and 6.8 GHz. The setup was deliberately plain.

• A continuous wave signal of very small bandwidth, with omni-directional antennas at both ends.
• Antenna patterns measured by an RF institute, with and without a ground plane.
• Sixteen transmit locations. A lift car gave several heights at some of them, so the analysis counts 29.
• The receive antenna 2 m above ground on a measurement van, which made most of the drive test non line of sight.

Two evaluation methods were applied, and both produced almost the same answer.

• The first compares the 10th percentile of received power in 50 m distance intervals. Coverage is a level exceeded with a given probability, not an average, which is the reason for the percentile.
• The second compares average power per 10 m of travelled distance.

The campaign reports two numbers. The average difference in received power between 6.8 and 3.6 GHz is around 9 dB, which is what has to be recovered to close the gap with 50 per cent probability. The 10th percentile of that difference is around 14 dB, which is the figure for closing the gap with 90 per cent probability.

Figure 2. Three estimates of the same penalty. The formula term is the smallest of them, and it is the only one that assumes free space.

The two field numbers are not two measurements of the same thing, and the 5 dB between them is not experimental error. One is the average of a distribution. The other is a point in its lower tail. The campaign applied two evaluation methods, and each produced one of them.

Observation 3 averages the received power at each frequency and subtracts, which gives 9 dB. Observation 4 takes the 10th percentile of received power in 50 m distance bins at each frequency, and subtracts those two percentile curves instead, which gives 14 dB. A coverage limit is set by the weak locations rather than the typical one, so the second method is the one that matches the question.

The drawing below shows why the second number is the larger. The left panel puts the two received power distributions side by side, with the mean of each marked and the worst tenth of each shaded. The right panel is the same pair read cumulatively, so a horizontal line at any probability cuts both curves and the distance between the cuts is the gap at that probability.

Figure 3. What the two field numbers measure. The distributions are not simply shifted, so the distance between them at the 10 per cent line exceeds the distance between their means.

The point of the picture is that the two curves are not parallel. If the move to 6.8 GHz only shifted the distribution, every percentile would separate by the same 9 dB and the two observations would agree. They do not agree, so the spread must widen as well. Deeper diffraction and scattering losses at the higher frequency hurt the weak locations more than the typical one.

NOTE : the curves are an illustration rather than the measured histograms. The contribution publishes the two statistics and not the distributions behind them. The two normal curves drawn here are chosen to reproduce both numbers exactly, 9 dB between the means and 14 dB between the 10th percentiles, which needs the spread at 6.8 GHz to be about 3.9 dB wider. The shape is illustrative; the two separations are the measured ones.

Two caveats come with those numbers, and they pull opposite ways. Higher building penetration loss at around 7 GHz is not included, so an indoor target would be larger still. The receive antenna was also outdoors and mostly non line of sight. A use case with a high line of sight probability, such as fixed wireless access, would see less than the measured difference.

• The formula and the field disagree by roughly a factor of two : 6.02 dB from 20·log₁₀(f1/f2) against about 9 dB measured on average.
• The probability target changes the answer by 5 dB : 9 dB closes the gap half the time and about 14 dB closes it nine times out of ten.
• Two independent methods agreed : percentile-by-distance and average-per-travelled-interval gave 9 dB and 14 dB either way.
• Indoor is worse and fixed wireless is better : the measured numbers are for an outdoor receiver 2 m above ground, mostly without line of sight.

Which channels are evaluated, and why is Msg3 the reference?

A cell is only as large as its weakest channel, so the study picks that channel and measures everything against it. For initial access the choice is Rel-15 NR Msg3, and the agreements below list what else gets evaluated alongside it.

Which signals and channels are on the list?

Companies are asked to provide link budget results for a named set of signals and channels during initial access at around 7 GHz. On the downlink that is NR PSS/SSS, SSB, common PDCCH, SIB1 PDSCH and Msg2 and Msg4 PDSCH. On the uplink it is NR PRACH format C2, B4 and 0, Msg3 PUSCH, Msg5 PUSCH with the assumed payload reported, and Msg4 PUCCH format 1.

Two assumptions come with that list. The simulation assumptions in section A.1 of TR 38.830 are the starting point, and the payload for NR SIB1 PDSCH is 1200 bits. Other parameters are still FFS.

What do the first Msg3 results look like?

An early set of results shows what the comparison looks like in practice. The case is Msg3 PUSCH on 2 resource blocks, at code rate 120/1024, with a 56 bit transport block.

• Maximum coupling loss is 126.96 dB at both frequencies. The channel itself has not become harder to decode.
• Maximum isotropic loss rises from 136.73 dB to 139.74 dB.
• Range for urban macro outdoor-to-indoor falls from 67.16 m to 36.85 m.
• Range for urban macro outdoor falls from 522.01 m to 437.17 m.

A later contribution applies the agreed gap formula to two other channels. For urban macro outdoor-to-indoor it reports −6.77 dB for PRACH format 0 and −7.37 dB for PUCCH format 1, and for urban macro outdoor it reports −2.32 dB and −3.02 dB. In each case the 6.02 dB frequency term is subtracted from the difference of the two maximum path losses.

• Msg3 is the yardstick, not the only channel measured : nine other signals and channels are evaluated, and all of them are compared against the Msg3 reference.
• The comparison is a range, not a margin : the same coupling loss at both frequencies still produces a much shorter reach once the path loss model is applied.
• PRACH and PUCCH are evaluated too : the agreed formula is applied channel by channel rather than once for the cell.
• TR 38.830 supplies the starting assumptions : the 6G study did not start its link budget from scratch.

Why do company link budgets disagree by more than the gap they measure?

Three rounds of email discussion collected link budget results, and the collection itself turned into the finding. The numbers companies submitted for the same channel in the same scenario are further apart than the effect the study is trying to measure.

What is in the collection of results?

The post-RAN1#124bis collection carries columns from eighteen companies beside the template column. All eighteen submitted entries for the urban macro outdoor-to-indoor case: Huawei and HiSilicon, NTT DOCOMO, vivo, MTK, OPPO, CATT, LGE, IIT Madras and CEWiT, Spreadtrum, Xiaomi, ZTE, Tejas Networks, Nokia, Samsung, Qualcomm, Ericsson, Lenovo and CMCC.

What does the Msg3 MIL row show?

Reading the maximum isotropic loss row for the Msg3 columns shows the problem. At 5G mid-band the submitted values run from 140.04 dB to 157.70 dB, a spread of 17.7 dB. At around 7 GHz they run from 144.81 dB to 163.88 dB, a spread of 19.1 dB. The gap being measured is a few dB, so the disagreement between companies is several times larger than the quantity in question.

RAN1 responded by agreeing to align interpretations rather than by averaging the results. The alignment list names four things.

• Antenna gain correction factors at components 3 and 4, and separately at component 2, for both transmitter and receiver.
• Receiver interference density.
• Penetration margin, and whether indoor distance is counted in the path loss for outdoor-to-indoor scenarios.
• Shadow fading margin, including how the function is defined and whether a handoff gain is counted.

• The method is agreed and the inputs are not : one formula produced a 17 to 19 dB spread, which puts the disagreement entirely in the parameters.
• Antenna gain accounting is the largest single ambiguity : four of the seven parameters on the alignment list are antenna gain correction factors.
• Shadow fading and penetration margin are the other two : both are margins rather than measurements, and both were being applied differently.
• Three email rounds were needed before the meeting : the collection ran after RAN1#124 and again after RAN1#124bis.

What data rate should the study treat as the coverage target?

Coverage for a data channel only means something at a stated rate. The study needed that number, and the contribution below argues for two bounds rather than one value, because the obvious source turned out to be unrealistic.

Why are the IMT-2020 rates not the answer?

The ITU IMT-2020 figures were 100 Mbps downlink and 50 Mbps uplink. Those were not used when NR studied coverage enhancement in Rel-17. TR 38.830 instead used 10 Mbps downlink and 1 Mbps uplink for the urban scenario, and 1 Mbps downlink and 100 kbps uplink for rural. That is ten times lower downlink and fifty times lower uplink in urban, and one hundred and five hundred times lower in rural.

Which lower and upper bounds are proposed?

An operator group proposes reusing the TR 38.830 values as a lower bound, with the suburban scenario left FFS in both directions. The suburban case was added to the channel model in Rel-19, so it has no inherited target.

For the upper bound the same contribution uses measurements rather than a standard. The trials ran on an n78 carrier with 90 MHz bandwidth, using 32 and 64 TRx antennas from different vendors. Neighbouring n78 cells were deactivated to imitate a noise limited scenario in a dense urban environment.

At an RSRP below −115 dBm the trials saw around 100 Mbps downlink and around 10 Mbps uplink, for a single active UE. Those numbers become the proposed upper bound for a 100 MHz carrier.

• Two bounds, not one target : TR 38.830 supplies the floor and an operator trial supplies the ceiling.
• The ITU numbers were judged too high : the same figures were already passed over once, when NR studied coverage enhancement in Rel-17.
• The upper bound is a field observation : 100 Mbps down and 10 Mbps up came from a single UE in a cell with its neighbours switched off.
• Suburban has no target yet : the scenario entered the channel model in Rel-19 and both directions are still FFS.

What UE assumptions does the evaluation rest on?

A link budget is only as firm as the device at the far end of it. Three assumptions had to be pinned down before the numbers could be compared: how many antennas the UE has, how wide it can tune, and what the initial access messages contain.

Three antenna sets are under evaluation for study up to 7 GHz. None of them constrains a later discussion of UE capability.

• Up to 4 transmit and receive antenna elements.
• Up to 8 transmit and receive antenna elements.
• Up to 16 receive antenna elements, intended only for fixed wireless access.

An operator group argues for a floor under that. Every UE on n78 in 5G NR already carries at least four receive antenna ports. The proposal is to assume the same minimum of four at around 7 GHz.

The message contents are the third assumption, and they are not known yet. Little is settled about the format, payload and size of the 6G initial access messages. The proposal is to borrow the 5G NR values until RAN2 or SA2 says otherwise, which keeps the RAN1 evaluation moving.

How small can the smallest maximum UE bandwidth be?

A narrower UE is cheaper to build and harder to schedule around. The question has run across several RAN1 meetings, and the operator answer below splits it by direction and by device type.

RAN1#123 agreed three alternatives for study. Alternative 1 is 20 MHz RF and 20 MHz baseband. Alternative 2 is 5 MHz RF and 5 MHz baseband, with 10 or 20 MHz for the TDD combination. Alternative 3 is 20 MHz RF with less than 20 MHz baseband.

The operator position is that downlink and uplink behave differently.

• Cutting downlink RF bandwidth from 20 MHz to 5 MHz in FDD bands would hurt eMBB UEs and overall cell efficiency. A UE narrower than the carrier can only be scheduled where the SSB is, unless extra common signals are added, and the device saves little in return.
• In the uplink the same cut still restricts scheduling. It raises no problem for RACH or control signals, and the cost saving is real for half duplex FDD with a SAW-less implementation.

Two further proposals follow from the deployment side. One asks to study mechanisms that bar IoT devices from a particular carrier. The other asks to study mechanisms that move them onto a chosen carrier, during or before initial access. Both exist because operators run low bands as a coverage layer and wide high bands as a capacity layer.

• Four receive ports is the proposed floor : it matches what n78 already requires of every 5G UE.
• The message contents are borrowed until they are defined : 6G initial access is evaluated with NR formats, payloads and sizes as a placeholder.
• Downlink and uplink get different answers : 20 MHz both ways in general, and 20 MHz down with 5 MHz up only for half duplex FDD IoT devices.
• The saving only appears in one configuration : the contribution puts significant device cost reduction at half duplex FDD with a SAW-less design.

How should SRS be designed for 6G coverage and capacity?

Uplink sounding is a coverage problem twice over. The sounding signal has to reach the base station, and it has to stay separable from every other UE sounding at the same time. The second constraint is the one the contributions below attack.

RAN1#124 listed eight aspects to study for SRS in uplink and downlink CSI acquisition.

• Efficient support of larger channel bandwidth.
• Capacity enhancements.
• Coverage enhancements.
• Efficient resource utilisation.
• Dynamic and flexible adaptation of SRS parameters.
• Multi-TRP transmission and reception, fixed wireless access, high speed train and other high mobility scenarios.
• Interference mitigation.
• Energy efficiency.

Which sequence families are the starting point?

RAN1#124bis narrowed the starting point. ZC and CGS sequences in NR are the basis for the 6G SRS design. The 5G per-resource frequency domain pattern with combs 2, 4 and 8 is the basis for the pattern study.

Two items are listed for further study on top of that. One is multiplexing of multiple SRS ports. The other is sparse patterns, such as larger comb values and non-uniform patterns. A note asks that the effect on complexity, coverage, capacity and performance is considered.

Why is SRS capacity the limit that matters?

The capacity limit that the contributions target is concrete. SRS in NR uses at most 6, 8 or 12 cyclic shifts of a single root sequence, and that count caps how many users can sound at once. One contribution proposes non-orthogonal pilots built from multiple root indices, asking for more than 60 root indices to be supported. A later contribution from the same source proposes adjustable phase shift ZC, where the cyclic shifts are chosen from the statistical channel rather than spaced evenly.

The argument for the second proposal rests on sparsity. In a massive MIMO OFDM channel the dominant components in the angle-delay domain typically account for less than 5 per cent of all elements. Phase shifts can therefore be scheduled to keep different users almost non-overlapping there.

The supporting simulation is run at 3.5 GHz with 15 kHz subcarrier spacing. It uses a 16 by 16 dual-polarised planar array, 2048 subcarriers with 1200 of them used for training, and a cyclic prefix length of 144.

• The cyclic shift count is the capacity ceiling : at most 6, 8 or 12 shifts of one root sequence decides how many UEs can sound together.
• Both proposals trade orthogonality for capacity : more root indices or unevenly spaced shifts, in each case accepting some correlation.
• NR sequences remain the baseline : ZC and CGS are the agreed starting point, so these are extensions rather than replacements.
• Sparsity is what makes overloading work : fewer than 5 per cent of angle-delay elements carry the channel, which leaves room to separate users after the fact.

Where does the separate NR coverage enhancement work item sit?

One of the contributions behind this note is not part of the 6G study at all. It belongs to a NR coverage enhancement work item running under a different agenda item, and its agreements apply to NR. It is summarised here because the techniques are the same ones the 6G study will have to consider.

That work item sits at agenda item 10.4.1 and covers coverage enhancements for NR phase 3. Its RAN1#122bis agreements are about multiple PRACH transmissions and about modulation.

How does the work item extend PRACH and modulation?

Separate ROs, or separate preambles on a shared RO, distinguish multiple PRACH transmissions with the same and with different transmit beams. The working assumption is that one common solution covers three cases: contention-based random access, the reconfiguration-with-sync case of contention-free random access, and system information request.

The modulation agreements extend pi/2-BPSK to more MCS entries. The extension covers entries with spectral efficiency no larger than N, where N is at most 0.8770 and the exact value is FFS. The code rate is doubled so that the spectral efficiency stays the same.

RRC signalling is introduced to enable the extension. Link level simulation decides which entries qualify, and it weighs two gains: the link level gain, and the power boosting gain that the lower peak to average ratio allows.

A further agreement reuses NR machinery rather than inventing new. Five things carry across from the same-beam case to the different-beam case.

• The definition and determination of the RO group.
• The definition of the time period.
• The SSB-to-RO mapping rule.
• The time offset between RO groups.
• The PRACH power control rule, with power ramping left FFS.

How are repetition and beam indication handled?

The Msg3 repetition rules are reused in the same spirit. Redundancy version determination and available slot determination carry over to PUSCH repetition scheduled by DCI 0_0 with C-RNTI, for repetition type A only.

The open question in that contribution is how the network tells the UE which uplink beam to use. Five options were on the table.

• Implicit indication by RA-RNTI.
• Explicit indication in DCI format 1_0 scrambled by RA-RNTI.
• Repurposing fields in the UL grant in the RAR.
• A new field in the MAC RAR.
• Repurposing bits in the MAC RAR.

The contribution prefers the third. The first option adds UE complexity, and the second and fourth reduce the network’s freedom to pack several RARs into one MAC PDU.

The same work item is covered in full elsewhere on this site. Coverage Enhancement - NR Phase 3 follows all three of its objectives through the RAN1 agreements from RAN1#122bis to RAN1#125. This section summarises only the one contribution that reached the 6G study.

• This work item is NR, not 6G : it runs at agenda item 10.4.1 and its agreements do not carry into the 6G study automatically.
• The techniques overlap with the 6G coverage problem : PRACH repetition across beams and a lower peak to average waveform both address uplink reach.
• pi/2-BPSK is extended by doubling the code rate : spectral efficiency is held constant, so the gain comes from the waveform rather than from the coding.
• Reuse is the default : RO groups, time periods, SSB-to-RO mapping and power control all carry over from the same-beam case.

How large is the agreed gap, channel by channel?

Everything above measures the problem. RAN1#126 finished that job and wrote the answer into four tables, one per scenario. The numbers below are the target that every technique in the rest of this page has to hit.

Every table carries the same title: the coverage gap between 7.0 GHz and 5G mid-band at 3.5 GHz. That 3.5 GHz is the reference, not the [4 GHz] the RAN1#123 agreement had carried in brackets. Reading these numbers as a 4 GHz comparison would overstate the frequency penalty, because 20·log₁₀(7/3.5) is 6.02 dB while 20·log₁₀(7/4) is 4.86 dB.

Each entry is a trimmed mean across the companies that submitted, with the sample count and standard deviation recorded beside it. A negative number is a shortfall against that reference, and a positive number means the channel already has margin to spare.

The tables also say how to move the reference to another mid-band carrier. Comparing 3.5 GHz with 2.6 GHz gives a 4.5 dB difference for Msg3 in urban macro outdoor-to-indoor, and 2.58 dB in urban macro outdoor. In each case the difference is assumed to apply to all signals and channels, so it shifts a whole column rather than one row.

The agreed coverage gap in dB for each signal and channel, across the four scenarios. Negative is a shortfall.

Reading the urban macro outdoor-to-indoor column top to bottom shows where the difficulty concentrates. The drawing below plots it.

Figure 4. The urban macro outdoor-to-indoor column drawn to scale. The uplink data channels sit at the bottom, and they are the ones that decide the cell.

Two structural facts sit behind the shape of that chart. The first is that uplink channels dominate the deficit: Msg5 PUSCH at −13.00 dB and Msg3 PUSCH at −8.68 dB are the two worst entries in the whole table. The second is that the outdoor scenarios are far easier than the indoor ones, because building penetration loss at 7 GHz is what makes outdoor-to-indoor hard.

How does each channel stand against Msg3 in the same band?

The chart above answers a question that spans two frequencies. A different question sits underneath the whole study, and it lives at one frequency: which channel actually limits the cell? Msg3 is named the reference channel throughout, and the collection can be read to show how far the others sit from it.

The measurement is a subtraction inside each company column. Every company that submitted a full channel set at around 7 GHz also submitted its own Msg3 PUSCH. Subtracting that company’s Msg3 maximum isotropic loss from each of its other channels gives a relative figure, and the company’s own assumptions cancel out. Thirteen companies can be read this way, and the chart shows the median of those differences.

Maximum isotropic loss is the right row for this comparison. It stops before the shadow fading and penetration margins, so the ranking below does not change between the outdoor-to-indoor and outdoor scenarios. Only the range each figure buys changes.

Figure 5. Every channel measured against Msg3 in its own band, rather than against itself at another frequency. Positive means the channel reaches further than Msg3.

The downlink sits well clear of the uplink. Msg4 PDSCH, PBCH, Msg2 PDSCH and the two SIB1 channels all land between 7 and 9 dB above Msg3. That margin is the antenna asymmetry of the previous section, read from the other side: the base station spends 768 or more elements on the downlink and the UE answers with one.

The uplink channels spread much further than the downlink ones. PUCCH format 1 sits 6.80 dB clear, because two bits over one resource block is the smallest payload on the list. The three PRACH formats fall in order of how much energy they carry, from 6.02 dB for format 0 down to 0.10 dB for format C2.

Two rows carry the conclusion. PRACH format C2 lands on the line, so it is the only non-data channel that matches Msg3 rather than beating it. Msg5 PUSCH sits 6.04 dB below, which makes it the weakest channel in the set and not Msg3.

That last point deserves care, because it looks like it contradicts the study. Msg3 is the reference for initial access, not the weakest channel overall. Msg5 carries a larger payload and arrives after the connection is up, so the study measures it but does not size the cell with it.

• The downlink has 7 to 9 dB in hand : every broadcast and downlink data channel reaches further than Msg3 in the same band.
• Msg3 is second from the bottom : only Msg5 PUSCH is worse, and it is worse by 6.04 dB.
• PRACH format C2 is the borderline case : at +0.10 dB it neither helps nor hurts, while format 0 has 6.02 dB of margin.
• The ranking survives the scenario : maximum isotropic loss excludes the penetration margin, so indoors and outdoors give the same order.

NOTE : these medians are computed from the collection rather than agreed by RAN1. Each median rests on between six and thirteen company submissions, because companies filled in different channel sets. The per-company spread is wide too, because they assume different bandwidths and beam counts. PBCH runs from −4.61 to +16.92 dB across the thirteen entries. The ordering is the finding here, not any single figure.

What separates the O2I and O2O columns?

The two scenarios sit side by side in every table above, and O2I is always the worse of the pair. It is worth knowing exactly where that difference enters, because it is not where most readers would guess.

Neither scenario is measured. Both are computed from the same link budget template, and the template is filled in twice. Reading the two sheets of the collection side by side shows how little actually changes between them.

The same Msg3 link budget under the two scenarios. The last column counts the 96 columns that carry the same company, carrier frequency and channel in both sheets.

The radio is the same equipment in both sheets, and most rows say so. Transmit power, antenna counts and noise figure never move. A quarter of the columns do re-run the noise and sensitivity chain, so the MIL differs in 27 of the 96. Row 27 differs in all 96.

The O2I cells are not typed constants. Thirty-three of them still carry the formula that produced them, and it reads as follows.

PL_tw = 5 − 10 log₁₀ ( 0.7 · 10^{−(25.4 + 0.11 f )/10} + 0.3 · 10^{−(5 + 4 f )/10} )

That is the high-loss model of TR 38.901 Table 7.4.3-2, with f the carrier frequency in GHz. The first term is infrared reflecting glass at 30 per cent weight, using the Release 19 value of 25.4 + 0.11 f. The second is concrete at 70 per cent weight, at 5 + 4 f. Evaluating it gives 27.50 dB at 3.5 GHz and 32.35 dB at 7 GHz.

The penetration margin, split into the two terms TR 38.901 defines for it.

Subtracting the wall formula from every entry leaves a residue. That residue is constant for each company and does not move with frequency, which identifies it as the indoor depth term. Six companies add nothing at all. Five add 6.25 dB, two add 4.17 dB, one adds 5.6 dB and one adds 10.18 dB. Halving those figures recovers the depth each company assumed, from 0 m to 20.4 m inside the building.

So the wall is not in dispute. Every company uses the same formula for it, and the apparent 18 dB spread is two separate effects read as one. Part is carrier frequency, since the collection spans 2.5 GHz to 8.4 GHz. The rest is how far inside the building the user is assumed to stand. At a single frequency the spread is 10.18 dB, and all of it is depth.

The outdoor value is not zero either. RAN1 agreed 9 dB of car penetration loss for the outdoor scenarios, from TR 38.901 clause 7.4.3.2. O2O therefore models a user inside a vehicle rather than standing in the open.

One consequence reaches the coverage gap directly. The wall costs 4.85 dB more at 7 GHz than at 3.5 GHz, and nothing in the outdoor column moves with frequency. The indoor gap is therefore larger than the outdoor gap by roughly that amount. IIT Madras and CEWiT show it in their own tables: PRACH format 0 sits at −6.77 dB indoors against −2.32 dB outdoors, and PUCCH format 1 at −7.37 dB against −3.02 dB.

• O2I and O2O are computed, not measured : both come from the same link budget template, filled in twice.
• One row carries the difference : the penetration margin, and it differs in all 96 matched columns.
• The wall is agreed, the depth is not : every company uses the same wall formula, and differs only on how far inside the user stands.
• It shows up as range : a median of 196 m for Msg3 indoors against 642 m outdoors, from the same hardware.
• Outdoor is not open air : the 9 dB is car penetration loss, so the outdoor user sits in a vehicle.

NOTE : the Deutsche Telekom propagation campaign quoted earlier is a real measurement rather than a template, and its receiver sat outdoors 2 m above ground. Its 9 dB and 14 dB figures therefore describe the outdoor case. The difference becomes larger once the higher building penetration loss at around 7 GHz is counted, which that contribution states explicitly.

Two supporting assumptions were fixed at the same meeting. The receiver noise figure is 9 dB for downlink and 5 dB for uplink, and RAN1 sent a liaison statement to RAN4 to say so without implying RAN4 should adopt it. Cell area reliability stays at 95 per cent for control channels, SSB and SIB1, and 90 per cent for channels capable of HARQ retransmission, with retransmission not assumed.

• The uplink data channels set the cell size : Msg5 PUSCH and Msg3 PUSCH carry the two largest deficits in every scenario.
• Outdoor-to-indoor is the hard case : the same channel can sit at −8.68 dB indoors and −3.75 dB outdoors in the same urban macro scenario.
• Some channels already have margin : PRACH format 0 reaches +2.54 dB in urban macro outdoor, so it is not the channel to spend effort on.
• The spread is now small enough to act on : standard deviations mostly sit between 0.3 and 2.6 dB, against the 17 to 19 dB spread that the raw submissions had before alignment.

How much does receiver combining of SSB and PBCH recover?

The first technique in the list is the one the agreed tables already measure, because two rows appear twice. PSS/SSS and PBCH are each reported with and without combining, so the gain is a subtraction between two rows of the same table rather than a claim.

How much does four-shot combining recover indoors?

Combining across four SSB transmissions moves the numbers a long way. In urban macro outdoor-to-indoor, PSS/SSS goes from −7.24 dB to −4.08 dB, and PBCH goes from −8.32 dB to −3.09 dB. The gains are 3.16 dB and 5.23 dB.

What does it change outdoors, and what does it cost?

In the outdoor scenarios the same step changes the verdict rather than only the number. Urban macro outdoor PSS/SSS moves from −2.81 dB to +1.13 dB and PBCH from −1.51 dB to +1.69 dB. Both cross zero, so with combining those two channels no longer have a gap at all.

The cost is time rather than spectrum or power. Four SSB transmissions have to be received and combined before the UE can decode, which lengthens initial access. That trade is the reason the study records both rows instead of assuming combining everywhere.

• The gain is measured, not projected : both rows come from the same agreed table, so the difference between them is the technique.
• Three to five dB from four combinations : 3.16 dB for PSS/SSS and 5.23 dB for PBCH in urban macro outdoor-to-indoor.
• It closes the outdoor gap entirely : both channels move from negative to positive in urban macro outdoor.
• It does nothing for the uplink : combining applies to downlink broadcast, and the worst entries in the table are uplink data channels.

Why does base station beamforming not fix the uplink?

Antenna gain is the obvious answer to a link budget deficit, and the agreed templates already assume a great deal of it. The reason the gap survives anyway is visible in a single row of the link budget spreadsheet.

What do the antenna counts actually say?

Reading the transmit antenna element count across every company submission gives two values and no others. Downlink entries use 768 or 1024 elements at the base station. Uplink entries use 1 element at the UE, in every single column.

Why does the asymmetry leave the uplink short?

That asymmetry is the whole explanation. The network can add array gain to the downlink almost without limit, and it cannot add any to the UE transmitter. A base station array helps the uplink only on reception, and that receive gain is already counted in the same templates.

An operator contribution names the same technology as the intended fix for the site grid target. The key technology for matching 3.5 GHz coverage at 7 GHz is MIMO with a high number of antenna ports. The contribution gives 256 ports as an example, with more than 1000 antenna elements.

• The templates already assume 768 to 1024 elements downlink : the deficit in the agreed tables is what remains after that gain is counted.
• The UE transmits from one element : every uplink column in the collection uses 1, which caps what beamforming can do for the uplink.
• Base station arrays help the uplink only on receive : the gain is real but it is already inside the numbers.
• This is why the remedies below are uplink remedies : waveform, coding, carrier choice and sounding all act where array gain cannot.

Can a low PAPR uplink waveform buy back the link budget?

A UE transmits at a power its amplifier can sustain, not at its nominal maximum. The distance between the two is the backoff that the waveform forces, so a waveform with a lower peak to average ratio is a direct uplink gain that costs no spectrum.

The size of that distance is recorded in the evaluation results. A vivo entry reports a maximum power reduction of 5 dB for DFT-s-OFDM against 8 dB for CP-OFDM in the same configuration. That setup is 4 GHz, with a 64 TxRU base station and a 2-transmit UE at 31 dBm. The 3 dB between them is available to the link budget before anything else is changed.

What does the DFT-s-OFDM waveform contribute on its own?

The shaping options below are applied to a waveform that already has a lower peak to average ratio than the alternative. It is worth separating the two gains, because only one of them survives the move to 7 GHz.

DFT-s-OFDM spreads the modulation symbols through a transform before they reach the subcarriers, which is why the agreement can speak of a DFT size at all. The contributions treat it as the alternative to CP-OFDM rather than as an addition to it, and they evaluate the two side by side.

One vivo entry runs that comparison twice, at two carriers, with everything else held as close as the deployment allows.

• At 4 GHz the setup is 20 MHz with 30 kHz spacing, a 64 TxRU base station and a two-transmit UE at 31 dBm, at maximum rank 2 with MU-MIMO.
• At 7 GHz the setup is 100 MHz with the same subcarrier spacing, a 256 TxRU base station and the same UE at the same power and rank.

The maximum power reduction is the same in both: 5 dB for DFT-s-OFDM against 8 dB for CP-OFDM. At 256QAM the pair reads 4.5 dB against 6.5 dB, with the entry recording a delta of 0.5 dB and 1.5 dB. The waveform advantage itself does not change with the carrier.

The same entry carries throughput figures, and they measure something else. They are the gain from enabling UL 1024QAM, against a 256QAM baseline, computed separately under each waveform. They are not a comparison of the two waveforms.

At 4 GHz that 1024QAM gain is positive under both. It runs from +5.38 to +11.92 per cent under DFT-s-OFDM and from +3.48 to +11.49 per cent under CP-OFDM. The entry attributes the difference to the lower maximum power reduction of DFT-s-OFDM, which is the waveform advantage showing through indirectly.

At 7 GHz that gain turns. It is recorded as negative or near zero, from −3.1 to +1.75 per cent under DFT-s-OFDM and from −5.08 to −1.14 per cent under CP-OFDM. The stated reason is that the uplink power is the same while the bandwidth is not.

That reason is worth holding onto. The UE still transmits 31 dBm, but at 7 GHz it spreads that power across 100 MHz instead of 20 MHz. Power per subcarrier falls, and a waveform advantage measured in dB of backoff buys less once the link is short of power to begin with.

• Two gains, not one : the waveform contributes 3 dB of backoff before any shaping is applied, and the shaping options below are additional to it.
• The backoff advantage is carrier independent : 5 dB against 8 dB at both 4 GHz and 7 GHz, in the same entry.
• The 1024QAM gain is not carrier independent : positive at 4 GHz and negative to near zero at 7 GHz, under both waveforms.
• The gain is smallest where this page needs it most : 7 GHz is the carrier with the coverage gap, and it is the one where the measured waveform benefit nearly disappears.

RAN1#126 agreed the list of uplink PAPR reduction techniques to carry forward for down-selection. All three are forms of frequency domain spectral shaping, and each is to be studied in both a transparent and a non-transparent form.

• Option 1 is FDSS with no spectrum extension or truncation, for pi/2-BPSK and QPSK. The applicable spectral efficiency range is FFS.
• Option 2 is FDSS with spectrum extension, for pi/2-BPSK, QPSK and 16QAM. The applicable range is FFS.
• Option 3 is FDSS with spectrum truncation, for pi/2-BPSK only. The applicable range is FFS.

What is FDSS, and what separates the three options?

All three surviving options carry the same four letters, so the choice between them is not a choice of technique. It is a choice about what happens to the occupied bandwidth, and the agreement encodes that in the option names.

One thing to settle first, because the two terms sit side by side in this section. FDSS is not a waveform, and it is not an alternative to DFT-s-OFDM. DFT-s-OFDM is the waveform, and the contributions evaluate it against CP-OFDM as its alternative. FDSS is a shaping step applied on top of that waveform.

Three details in the record say so. The agreement fixes a DFT size before the shaping is applied, so a DFT has already run by then. The evaluation tables compare CP-OFDM against DFT-s-OFDM as two waveforms, not as two shaping schemes. And the contribution devoted to the subject is titled PAPR reduction for DFT-s-OFDM.

Frequency domain spectral shaping acts on the uplink signal after the transform and before it reaches the subcarriers. The agreement says as much when it fixes the DFT size before spectrum extension or truncation is applied, which places the shaping between the two. Changing the shape of the spectrum changes the shape of the time domain signal, and a flatter time domain envelope is a lower peak to average ratio.

What separates the options is the width of the result.

• Option 1 keeps the same occupied bandwidth. The shaping redistributes energy inside the allocation and nothing spills outside it.
• Option 2 extends the spectrum. The signal occupies more subcarriers than the data alone would need, and the extra width is what buys the reduction.
• Option 3 truncates the spectrum. The signal occupies fewer subcarriers, so some of the shaped spectrum is discarded before transmission.

The modulations attached to each option follow that width directly. Truncation is listed for pi/2-BPSK only, which is the most robust modulation and the one that can best survive losing part of its spectrum. Extension reaches as far as 16QAM, because the added width gives the receiver something to work with. Option 1 sits between them at pi/2-BPSK and QPSK.

That ceiling is worth comparing with the waveform underneath. DFT-s-OFDM itself is evaluated at high order modulation, and the same entry that reports its backoff also reports a figure at 256QAM and a gain from enabling 1024QAM. The shaped variants under study stop at 16QAM.

So the limit belongs to the shaping, not to the waveform. It is also not settled. The agreement records a target to down-select the modulation for these options at RAN1#126bis, and it singles out 16QAM as the case to decide, which marks the top of the range as the contested part.

The transparent and non-transparent split is about the receiver rather than the transmitter. Every option is listed in both forms, and the agreement targets a down-selection between them at RAN1#126bis together with the modulation question.

NOTE : the agreement uses the letters SE in two different senses on the same page. In FDSS with spectrum extension and in DFT size before SE/ST it means spectrum extension. In Applicable SE range: FFS, which appears under all three options, it means spectral efficiency. Only the surrounding words separate them.

Two further decisions frame the work. The condition for applying the finally selected scheme is FFS, and the target is to down-select the transparent or non-transparent form and the modulation, at least for 16QAM, at RAN1#126bis. For the DFT size before spectrum extension or truncation, 12 × 2^x3^y5^z subcarriers is supported, and whether a second option with X = 4 is also supported will be decided by RAN1#127.

How do CP-OFDM, DFT-s-OFDM and FDSS compare?

Three names have appeared in this section and only two of them are the same kind of thing. The table sets them side by side so the mismatch is visible rather than implied, and so the empty cells can be accounted for.

The two waveforms against the shaping step applied to one of them. The first four rows are relationship and capability, the next four are measurements, and the last is status. The two 1024QAM rows are measured against a 256QAM baseline, not against each other.

The FDSS column is empty in the middle because no figure exists yet. RAN1#126 agreed the template for collecting uplink PAPR reduction results, with companies asked to fill it by 27 August and a further email round running from 28 to 30 September. The numbers were still being gathered when the meeting closed.

The two percentage rows need their baseline stated, because a gain in per cent is meaningless without one. Both are the gain from enabling UL 1024QAM where 256QAM was used before. They come from a results table whose other rows compare 1024QAM against the 256QAM MCS table. The waveform is the condition under which that gain was measured, not the thing being compared.

That also explains why the two maximum power reduction rows carry the coverage argument on their own. Backoff is a property of the waveform and applies at any modulation. A 1024QAM gain is a peak rate result that a cell edge UE will never see.

NOTE : the four measurement rows come from a single company entry, so they are one evaluation rather than an agreed value. That is the opposite of the coverage gap tables earlier on this page, which are trimmed means across as many as twenty submissions. Treat the waveform figures as indicative and the gap figures as settled.

• Three dB is available from the waveform alone : 5 dB maximum power reduction for DFT-s-OFDM against 8 dB for CP-OFDM in one reported configuration.
• All three surviving options are FDSS : the choice is between no shaping change, spectrum extension and spectrum truncation.
• The modulation order is the open question : pi/2-BPSK is in all three options, and 16QAM only in the extension option.
• Transparent and non-transparent are both still live : a transparent scheme needs no receiver change, which matters for early deployment.

Can a second carrier carry the uplink instead?

If the uplink cannot reach at 7 GHz, one answer is to stop asking it to. A UE can transmit on a lower carrier whose propagation is kinder and receive on the 7 GHz carrier, and RAN1 has opened that as a study item for idle mode.

Which use cases were agreed for study?

Three use cases were agreed for further study of initial access with idle mode multi-carrier operation. The first is uplink coverage enhancement and downlink to uplink coverage imbalance. The second is RACH capacity, offloading and load balancing. The third is network energy saving and common signalling reduction. Other use cases are not precluded.

The agreement also sets a test that each use case has to pass. For each one, the study asks whether and under which conditions multiple carrier operation in idle mode provides additional benefit over existing mechanisms. That wording keeps the burden of proof on the new feature.

Which carrier arrangements are under study?

Three carrier arrangements are under study for the case where a UE uses resources on a carrier other than the one that carried the system information. S1 is one downlink physical carrier with more than one uplink carrier. S2 is more than one downlink carrier with a single uplink carrier. S3 has more than one of each.

• The imbalance is named as a use case in its own right : UC-1 covers uplink coverage enhancement and downlink to uplink coverage imbalance together.
• The feature has to beat what already exists : each use case is studied for additional benefit over existing mechanisms, not in isolation.
• S1 is the shape that matches the coverage problem : one downlink carrier with several uplink carriers lets the uplink move down in frequency on its own.
• This is idle mode work : the agreement is about initial access, which is where the Msg3 bottleneck sits.

Can better coding of small uplink control payloads help?

Uplink control carries a handful of bits and pays a disproportionate price for them. The block code that NR uses for small payloads runs out of room at low code rates, and the fix it falls back on is the one that costs coverage.

Why does the 5G Reed-Muller code run short?

Several sources record the same chain of causes. The mother code length of the 5G Reed-Muller code is limited to 32, which was observed by five sources. Four sources then observed that repetition is needed at low code rates, and that performance and coverage degrade as a result. Small uplink control payloads of 3 to 11 bits are generally scheduled at exactly those low code rates.

Other weaknesses were recorded alongside it. Four sources observed BLER loss from DMRS overhead, at least for PUCCH. Two observed loss from the high peak to average ratio, again at least for PUCCH. Three observed that the code cannot support DMRS-less transmission, because of the all-one vector in its first column.

What is the counter-argument, and what is the prize?

The counter-argument is also on the record. Four sources observed that the Reed-Muller code was used in 5G NR, and that 6G has no new requirement in this region to justify a redesign. One source added that the code is optimised for good minimum distance, and that fast Hadamard transform based maximum likelihood decoding handles it at low complexity.

One source puts a number on the prize. R1-2605288 reports that an enhanced parity-check Polar code improves BLER by 0.6 to 0.9 dB, and increases coverage by more than 10 per cent over the 5G Reed-Muller code at low code rates. A second source, R1-2605909, reports about 0.7 dB from the same code family at K = 8 bits and E = 256 bits. Enhanced Reed-Muller with optimised basis sequence selection and rate matching is the other family under discussion.

• The limit is the mother code length of 32 : below the rate that supports, the only tool left is repetition.
• Repetition is what costs the coverage : four sources tie the degradation directly to it rather than to the code itself.
• More than 10 per cent coverage is claimed for enhanced Polar : one source, at low code rates, alongside a 0.6 to 0.9 dB BLER gain.
• Redesign is contested : four sources argue 6G has no new requirement here that would justify replacing a working 5G code.

Can sparser SRS patterns improve uplink sounding coverage?

Sounding has the same power problem as any other uplink transmission, and one more besides. Spreading the same transmit power over fewer subcarriers raises the power per subcarrier, so a sparser comb is a coverage technique as much as a capacity one.

Which comb values are under study?

RAN1#126 agreed to study sparse SRS patterns with comb values larger than 8, naming comb 12 and comb 16, for the combined purposes of increasing SRS capacity, reducing overhead and improving coverage. Six aspects are listed for consideration.

• SRS sequences.
• MPR and PAPR or cubic metric impact.
• Maximum number of cyclic shifts for the corresponding comb value.
• Minimum available bandwidth and its corresponding sequence.
• Channel estimation performance.
• Applicable RB sizes, for example an RB size of 8 or more.

What limits and related agreements come with it?

Two notes come with the agreement. Other comb values are still FFS, and RAN1 records that input from RAN4 on maximum power boosting is needed. That second note is the coverage part stated plainly, because a sparser comb only converts into reach if the amplifier is allowed to raise the power on the subcarriers that remain.

Two related agreements sit beside it. Cross-slot SRS transmission is supported at least for consecutive S and U slots, which lengthens the time a sounding transmission can occupy. The study has not yet decided whether it extends to any two consecutive slots, such as U and U.

• A sparser comb concentrates power : the same transmit power over fewer subcarriers is more power per subcarrier, which is reach.
• The gain depends on RAN4 : the agreement records that input on maximum power boosting is needed before the benefit can be claimed.
• Comb 12 and comb 16 are the named candidates : both are larger than anything NR supports today.
• Cross-slot sounding adds time as well : consecutive S and U slots are supported, with wider spans still open.

Can repetition close the gap, and at what price?

Every technique above tries to buy dB without spending time. Repetition spends time instead, and it always works, because the receiver accumulates energy from every copy it receives. So the interesting question is not whether repetition closes the gap. It is what the network sacrifices when it does.

Start with the size of the deficit. Msg3 PUSCH sits at −8.68 dB in urban macro outdoor-to-indoor, and Msg5 PUSCH sits at −13.00 dB in the same scenario. Those are the two worst entries in the agreed tables, and both of them are uplink data channels.

The contributions behind this page never state a repetition factor, so the counts here are arithmetic rather than an agreed result. Ideal combining of N copies recovers 10·log₁₀(N) dB, which makes each doubling worth about 3 dB. Closing 8.68 dB then takes roughly eight transmissions, and closing 13.00 dB takes roughly twenty. Real combining falls short of the ideal, so read both counts as a floor.

Repetition counts set against the urban macro outdoor-to-indoor column of the agreed gap table. The gain column is 10·log₁₀(N) with ideal combining, so every count is a floor rather than a prediction.

That is the price, and an operator group has already asked for it to be visible. Their proposal is simple. If repetitions turn out to be needed for initial access, the number of repetitions is written into the study outcome rather than left implicit. A coverage target met by twenty transmissions is not the same target as one met by a single transmission.

What has the 6G study said about it so far?

Repetition is absent from the list of techniques the study carries forward, and that absence is worth reading as a position rather than as an oversight. It appears twice in the contributions behind this page, and both times as something to be counted rather than something to be adopted.

The first is the operator proposal just described. The second sits in channel coding, where repetition is already the fallback and already costs. The mother code length of the 5G Reed-Muller code is limited to 32. Repetition is the only tool left below the rate that length supports. Four sources tie the performance and coverage degradation directly to it, rather than to the code itself.

The study has already accepted accumulation over time in one place. Four-shot combining of SSB recovers 3.16 dB for PSS/SSS and 5.23 dB for PBCH in urban macro outdoor-to-indoor, and that gain costs a longer initial access. The difference is direction rather than principle. SSB is broadcast periodically, so those copies are transmitted whether or not a given UE combines them. An uplink repetition instead occupies resources the scheduler could have given to another UE.

Where is repetition being specified, then?

The mechanisms themselves are not waiting for 6G. NR coverage enhancement, or NR CE, is a work item for NR phase 3 at agenda item 10.4.1, and it runs alongside the 6G study. That is where the repetition rules are actually being written.

Two of its RAN1#122bis agreements matter here. Msg3 repetition rules are reused for PUSCH repetition scheduled by DCI 0_0 with C-RNTI, for repetition type A only. Redundancy version determination and available slot determination carry over with them. PRACH repetition is distinguished either by separate ROs or by separate preambles on a shared RO, for transmissions with the same and with different transmit beams.

Read those as precedent rather than as inheritance. Agreements from that work item are NR, and they do not carry into 6G automatically. What they do show is that repetition raises design questions which already have answers. One of them is how the network indicates the uplink beam for a repeated Msg3.

• Repetition always works, and that is not the question : energy accumulates with every copy, so the only issue is how much time and capacity the network will spend.
• The deficit it would have to cover is −8.68 dB : that is Msg3 PUSCH in urban macro outdoor-to-indoor, with Msg5 PUSCH further behind at −13.00 dB.
• Nobody has stated a repetition count : an operator group asks for the number to be written into the study outcome, which is exactly why it is missing today.
• The 6G study is buying dB another way : waveform shaping, a second uplink carrier, better short block coding and a sparser SRS comb all act on the link, not on the schedule.
• The specification work is happening in NR : agenda item 10.4.1 holds the Msg3 and PRACH repetition agreements, and they do not carry into 6G automatically.

Reference

Thirteen of the twenty-one documents listed below were downloaded and read: ten of the thirteen supplied for this page, both sets of draft minutes, and TR 38.901. Every RAN1#126 agreement quoted on this page, including the coverage gap tables, comes from the RAN1#126 minutes.

Three documents are listed but could not be retrieved, because they are absent from the 3GPP server: R1-2604037, R1-2604041 and R1-2606727. The titles of the first two come from the RAN1#125 minutes, and no title is recorded anywhere for the third. Five further documents are named inside the RAN1#126 minutes and were not opened separately: R1-2606693, R1-2606804, R1-2606880, R1-2605288 and R1-2605909.

References are listed in meeting order, earliest first, with moderator documents first within each meeting, and the specification last. The agenda item number changed during the study. RAN1#123 to RAN1#125 used 10.5.0 for general aspects and 10.5.3.2 for uplink-based CSI acquisition. RAN1#126 renumbered the study, so 10.5 is now Multi-antenna system and no separate coverage item exists. R1-2509379 sits outside the study altogether, at 10.4.1, which is the NR coverage enhancement work item.

• R1-2509379 (RAN1#123) : Discussion on coverage enhancement (DENSO CORPORATION)
• R1-2601360 (RAN1#124) : On 6GR coverage target (Vodafone, Deutsche Telekom, Orange, Bouygues Telecom, Telecom Italia, British Telecom)
• R1-2603197 (RAN1#124-bis) : Summary of [Post-124-R20-General Aspects-Coverage] (Moderator, Huawei)
• R1-2603242 (RAN1#124-bis) : Summary of [Post-124-R20-General Aspects-Coverage] (Moderator, Huawei)
• R1-2601979 (RAN1#124-bis) : Discussion on 6G SRS coverage, capacity and overhead aspect (PML)
• R1-2603085 (RAN1#124-bis) : Initial Evaluation Results on 6G Coverage (IITM, CEWiT)
• R1-2605061 (RAN1#125) : FL summary of [Post-124b-R20-General Aspects-Coverage] (Moderator, Huawei)
• R1-2604037 (RAN1#125) : Discussion on 6G SRS coverage, capacity and overhead aspect (PML)
• R1-2604041 (RAN1#125) : Discussion on 6G SRS coverage, capacity and overhead aspect (PML)
• R1-2604053 (RAN1#125) : Discussion on 6G SRS coverage, capacity and overhead aspects (PML)
• R1-2604783 (RAN1#125) : 6GR Coverage Gap Evaluations (IITM, CEWiT, Tejas Networks)
• R1-2604799 (RAN1#125) : u6GHz coverage gap and smallest max. UE bandwidth (Deutsche Telekom, Vodafone)
• Draft minutes report RAN1#125 v0.8.0 (RAN1#125) : Chairman’s notes for the meeting (3GPP RAN WG1). Source of the agreed simulation assumptions, including car penetration loss and building penetration loss
• R1-2606693 (RAN1#126) : On PAPR reduction for DFT-s-OFDM (Sony)
• R1-2606804 (RAN1#126) : Feature Lead summary #3 on 6G waveform (Moderator, Nokia)
• R1-2606880 (RAN1#126) : Moderator summary on improvement of SRS capacity and coverage: Round3 (Moderator, CATT)
• R1-2605288 (RAN1#126) : Channel coding for 6GR air interface (Huawei, HiSilicon)
• R1-2605909 (RAN1#126) : Discussion on 6G channel coding (OPPO)
• R1-2606727 (RAN1#126) : Title not recorded
• Draft minutes report RAN1#126 v0.1.0 (RAN1#126) : Chairman’s notes for the meeting (3GPP RAN WG1)
• TR 38.901 v19.4.0 (Release 19) : Study on channel model for frequencies from 0.5 to 100 GHz (3GPP). Clause 7.4.3 defines both penetration loss models the link budget uses