The 6G waveform discussion has moved from choosing a starting waveform to evaluating specific uplink enhancements. This note follows that progression and separately examines the waveform and reference-signal work for integrated sensing and communication.

Document scope: This page uses the supplied 6G_TDoc_SearchResult.txt and selected original contributions, session notes, and moderator summaries from RAN1#122 through RAN1#126bis. The latest meeting covered is RAN1#126bis, held on October 12~16, 2026. Communication waveform discussion and ISAC waveform/frame-structure discussion are treated separately. Earlier communication milestones are reviewed directly in summary TDocs from #122, #122bis, #123, #124, #124bis, #125, and #126, with session notes used to cross-check #124–#126 outcomes. Earlier ISAC summaries cover #124bis, #125, and #126. The #126bis documents establish subsequent changes and the latest status covered here. This is a synthesis of selected passages, not an exhaustive review of every listed document or a completed 6G specification.

Status convention: An agreement or session conclusion is identified explicitly. A proposal remains a proposal, an observation retains its evaluation conditions, and a working assumption remains subject to further study. The communication conclusions use the endorsed R1-2607287 session notes; the ISAC agreement history uses R1-2607998. The later communication summary R1-2607944 was unavailable during this review. R1-2607287 R1-2607943 R1-2607998

• Executive Summary
  • Communication overview
  • Sensing overview
• Mind Map
  • Interactive topic map
  • Text outline and section links
• Agreed Items
  • Communication waveform agreements and conclusions
  • ISAC agreements and working assumptions
• What is the communication waveform baseline?
  • Baseline decisions and agenda boundaries
  • Enhancements within the agreed framework
• How does DFT-s-OFDM work, and what does it mean for 6G?
  • Waveform, modulation, and the LTE/NR relationship
  • Transmit procedure and the two transform sizes
  • Receive procedure and channel equalization
  • PAPR, single-carrier behavior, and practical constraints
  • What the 6G study changes around the baseline
• How is a low-PAPR benefit measured?
  • Net gain and equal-resource comparisons
  • PA behavior and the reference transmission
• What is FDSS, and how does it shape DFT-s-OFDM?
  • Basic operation: weight the DFT outputs
  • Why frequency-domain shaping can change time-domain peaks
  • Known processing, selectable filters, and receiver transparency
  • FDSS, spectrum adjustment, and CFR are separate choices
• What changes with FDSS, spectrum extension, and spectrum truncation?
  • Definitions and processing relationships
  • Examples with equal, extended, and truncated spectrum
• How does the receiver recover a truncated spectrum?
  • Symmetry and reconstruction of omitted outputs
  • Noise combining and implementation choices
• How do transform sizes affect scheduling and transport blocks?
  • DFT size and scheduling consequences
  • Transport-block sizing and MCS design
• When does lower PAPR become usable RF power?
  • Emission limits, maximum power, and power boosting
  • Spectral masks and mode-dependent power reporting
• Does the receiver need to know the transmit filter?
  • Transparent filtering and receiver knowledge
  • Data, DMRS, and specification responsibilities
• Which enhancement results remain convincing?
  • Allocation-dependent results and reference choices
  • Study progression and the 16QAM conclusion
• How do layers and waveform switching fit together?
  • Rank, precoding, and evaluation assumptions
  • Waveform switching and UE assistance
• What is being studied for sensing waveforms?
  • Baseline sequences and candidate scope
  • Sequence design, chirps, and pulse confinement
• How do sequence mapping and shared resources change sensing?
  • Root selection, comb mapping, and multiplexing
  • Receiver windowing and the composite transmitted signal
• How are sensing candidates compared?
  • Ambiguity function and sidelobe definitions
  • PAPR, oversampling, and realistic performance
• Does sensing require a different frame or longer CP?
  • Physical CP and effective observation interval
  • Transmit-receive separation and resource cost
• Update Note
  • Current revision
  • Earlier revisions
• Reference
  • Communication waveform sources
  • Sensing waveform and frame sources

Executive Summary

The main result is continuity with the NR waveform framework, together with targeted enhancement studies. Coverage, receiver complexity, scheduling flexibility, and sensing performance determine which changes can justify additional specification work.

Communication overview

The communication study combines an established waveform baseline with narrower uplink enhancement work. The main dependencies are receiver performance, transform and transport-block sizing, permitted RF power, and practical waveform selection.

Sensing overview

Sensing retains its own candidate and evaluation discussion. The communication waveform decision does not settle sequence design, pulse confinement, target separation, or the cost of sharing radio resources with communication.

The detailed status matters because incoming papers contain incompatible recommendations. Some request standardized low-PAPR filters; others prefer transparent filtering. The latest session conclusion assigns filter characterization to RAN4 and leaves receiver exploitation to implementation. Similarly, a candidate sensing sequence can remain under evaluation even after the communication baseline has been settled. R1-2607287 R1-2607943 R1-2607998

• Read the status before the performance number : a proposed method, a conditional gain, and an adopted baseline answer different questions.

Mind Map

The communication baseline is settled more firmly than the remaining uplink enhancements and sensing candidates. Use the branches to connect each proposal with its reference, receiver assumptions and evaluation limits. The sensing branches retain their separate study scope and open metric definitions.

Interactive topic map

Use + to expand a branch and select a topic for its description and section link. Drag the background or comment boxes with the left mouse button to pan. Hold the left mouse button and turn the wheel to zoom. Toolbar buttons and keyboard navigation are also available.

NOTE: Lines group related topics; they do not specify a mandatory processing chain. Final comment leaves explain each topic. An agreed baseline, a conditional performance result and a proposal have different status. The map follows this note's selected records through RAN1#126bis.

Text outline and section links

The outline provides the same topic descriptions and local section links without requiring interaction with the diagram. Use it to scan the expanded topics or navigate when JavaScript is unavailable.

Agreed Items

There are agreed items in this document set. They establish the communication baseline, narrow the enhancement work, and define parts of sensing evaluation without completing every waveform or reference-signal design.

Communication waveform agreements and conclusions

The communication discussion progressively narrowed its scope across meetings. The chronology below distinguishes the initial study basis from later support decisions and the specific conclusions reached at the latest meeting.

The chronology now uses the summaries and session notes of the relevant meetings, cited in each row. The #126bis conclusions come directly from the endorsed session notes. They do not establish universal support for every FDSS-SE or FDSS-ST variant proposed in the incoming papers. R1-2607943 R1-2607287

ISAC agreements and working assumptions

The sensing discussion uses its own baseline and evaluation decisions. Its candidate studies remain broader than the communication waveform agenda, while still aiming to reuse the communication waveform and frame wherever possible.

At #124bis, CP-OFDM as defined for 6GR became the sensing starting point and benchmark. The communication frame, subframe, slot, symbol, CP, and subcarrier spacing also became starting points. Enhancements and other sensing waveforms were not precluded. At #125 and #126, the study expanded its evaluation criteria and identified sequence and effective-CP enhancement families. R1-2602985 R1-2604765 R1-2606567 R1-2607998

At #126bis, circular ambiguity function and circular cross-ambiguity function option 2b were selected. Sidelobes exclude the first null-to-null mainlobe region; comb patterns use one period of the repeated regions. Cross-Ambiguity Peak and Integrated Cross Ambiguity were defined, with normalization still open. The ISLR definition was corrected to integrated sidelobe power divided by integrated mainlobe power. R1-2607998

Root-selection alternatives, ZC length changes, and quadratic-phase RS parameter choices remain working assumptions for further evaluation. Their presence in the meeting summary does not mean that a particular pulse, chirp, sequence family, or sensing mode has been finally selected. R1-2607998

• Separate agreed framework from candidate design : the baseline and evaluation rules are further advanced than several individual enhancement choices.

What is the communication waveform baseline?

The baseline question is already more settled than an early list of candidate waveforms suggests. The remaining discussion concentrates on improving uplink operation within the agreed CP-OFDM and DFT-s-OFDM framework.

Baseline decisions and agenda boundaries

The #122 starting point left room for additional OFDM-based proposals. By #125, RAN1 supported CP-OFDM for downlink and both CP-OFDM and DFT-s-OFDM for uplink, as defined for NR. It also closed further discussion of other waveforms in this agenda unless tasked by RAN. Earlier alternatives therefore belong to the study history, rather than a current list of equally open communication choices. R1-2506595 R1-2508043 R1-2605036

The transition occurred in stages. At #124, the study gave high priority to uplink coverage improvements and multi-rank DFT-s-OFDM. Downlink DFT-s-OFDM discussion ended in the waveform agenda, with a separate allowance for signal-specific agendas such as synchronization, WUS, or sensing. This was an agenda boundary, not a blanket decision on every downlink signal. R1-2601696 R1-2601512

At #124bis, the prioritized low-PAPR families became plain FDSS, FDSS with spectrum extension, and π/2-BPSK FDSS with spectrum truncation. Transparent and non-transparent filters were both still study options. A separate conclusion deferred the listed other proposals unless RAN requested further work. The #125 report then consolidated the communication baseline and the work that would continue after June. R1-2603315 R1-2603279 R1-2603729 R1-2605036

Enhancements within the agreed framework

This narrowing does not eliminate meaningful waveform work. A DFT-s-OFDM transmission can still change its frequency-domain shaping, occupied spectrum, transform-size options, layer support, or waveform-selection procedure. Each change must demonstrate a benefit against the established reference and identify its effect on the receiver and scheduler. R1-2608624 R1-2608337

Different names can also describe related transmitted signals. Proposed π/2-BPSK with spectrum truncation and I/Q-offset QPSK with spectrum extension can produce equivalent waveforms under corresponding constructions. Their transform sizes and implementation steps may differ. That equivalence helps organize proposals, but it does not prove equal complexity or equal performance for arbitrary filters and receivers. R1-2607460 R1-2608743

The sensing discussion needs a separate reading. CP-OFDM is its starting point, but sensing-specific enhancements and additional waveforms were not ruled out by that agreement. Applying the communication agenda's closure to every sensing proposal would therefore misrepresent the document scope. R1-2607998

• Baseline first, enhancement second : assess each proposed change against the adopted communication framework and the correct reference transmission.

How does DFT-s-OFDM work, and what does it mean for 6G?

DFT-s-OFDM describes how the transmitter processes modulation symbols. It does not restrict the input to π/2-BPSK. Understanding that distinction helps separate ordinary QAM transmission from the additional low-PAPR techniques discussed elsewhere on this page.

This section adds established signal-processing background to the TDoc discussion. The LTE and NR specifications supply historical processing examples. The 6G conclusions establish the study baseline and enhancement scope; they do not yet specify every future implementation detail.

Waveform, modulation, and the LTE/NR relationship

The modulation mapper converts bits into complex symbols. DFT spreading then processes a block of those symbols before OFDM generation. Changing the constellation therefore does not remove the DFT-s-OFDM processing structure.

LTE PUSCH already combines QPSK, 16QAM, or 64QAM with transform precoding. Its SC-FDMA uplink is a DFT-spread OFDM realization. These modulation choices appear before the DFT step in the LTE specification. Actual use also depends on UE capability and configuration. TS 36.211 V8.9.0, sections 5.3.2 and 5.3.3

For π/2-BPSK, the symbol position determines the orientation, while the bit selects one of two opposite points. A fixed phase rotation can place those points on diagonal axes. The NR modulation definition provides that convention. TS 38.211 V16.5.0, section 5.1.1

Transmit procedure and the two transform sizes

Start with one data block on one layer and omit optional shaping for the moment. Let M be the spreading-DFT size and N the OFDM-IFFT size. They perform different jobs and usually have different sizes.

1. Prepare modulation symbols. Channel coding, rate matching, scrambling, and modulation produce the data symbols. The scheduled modulation determines how many coded bits each input symbol represents.

2. Apply an M-point DFT. Transform M input symbols into M frequency-domain values. Each output generally combines all M input symbols. This is transform precoding, also called DFT spreading.

3. Map the DFT outputs. Place them on the allocated subcarriers within an N-bin OFDM grid. The simple localized example uses M contiguous bins. Reference signals follow their own defined generation and mapping rules.

4. Apply an N-point IFFT. Convert the mapped frequency-domain grid into N time-domain samples. The IFFT still generates an OFDM symbol; the preceding DFT changes the relationship among its subcarrier values.

5. Add the cyclic prefix and transmit. Copy the end of the useful symbol to its beginning, then perform the required RF processing. The CP supports block equalization under the appropriate channel and synchronization conditions.

The NR PUSCH specification places transform precoding after layer mapping and before antenna precoding and resource mapping. The diagram isolates a single-layer data path, so it omits antenna precoding, reference-signal generation, and optional processing. TS 38.211 V16.5.0, sections 6.3.1 and 5.3.1

Figure 1. DFT-s-OFDM processing and an illustrative receiver. The receiver algorithm is not a mandated 6G implementation.

In compact notation, D = F_Md and x = F_N^HPD. Here d contains M input symbols, F denotes a unitary DFT, and P maps the M outputs into N bins. The superscript H denotes the inverse transform through the conjugate transpose. This expression describes the useful samples before CP insertion.

For example, take M = 24 and N = 128. The transmitter spreads 24 QPSK symbols into 24 DFT outputs and maps them into 24 of the 128 IFFT bins. That represents 48 coded data bits before accounting for pilot placement and coding overhead. With 16QAM, the same 24 input positions represent 96 coded bits. Neither change requires removing the DFT. These sizes illustrate the mathematics, not a prescribed 6G allocation.

Receive procedure and channel equalization

The receiver must undo the radio channel before recovering the input symbols. Simply applying an inverse DFT to the received bins would also spread their unequal channel gains and noise across the symbol estimates.

An illustrative receiver first corrects timing and frequency offset, removes the CP, and computes the N-point FFT. It then selects the scheduled bins and estimates their channel responses using reference signals. Frequency-domain equalization compensates those responses. An M-point inverse DFT recovers modulation-symbol estimates, followed by soft demapping, descrambling, and channel decoding.

Zero-forcing equalization inverts the channel but can strongly amplify noise near a spectral null. MMSE equalization balances channel inversion against noise amplification. Residual interference and noise therefore depend on the equalizer and channel. DFT spreading alone does not guarantee a particular BLER improvement or diversity order. DFT-precoded-OFDM receiver analysis

This simple per-bin description assumes that the CP and synchronization are adequate and that the channel changes little within the symbol. Excess delay, Doppler, or residual frequency error can break that approximation. Multiple layers also require spatial separation; the scalar single-layer illustration does not specify the complete MIMO receiver.

PAPR, single-carrier behavior, and practical constraints

The DFT creates a structured relationship among the subcarrier values. This produces single-carrier-like behavior, while retaining an FFT-based resource grid and frequency-domain equalization. It does not turn the transmitted signal into one occupied subcarrier.

The main uplink motivation for DFT-s-OFDM is its generally lower PAPR compared with CP-OFDM. The UE power amplifier must reproduce occasional peaks without excessive distortion. A signal with larger peaks needs more margin between its average operating power and the amplifier's limiting region. This margin is commonly discussed as PA back-off.

Lower peaks can reduce the required back-off. Depending on the operating constraint, the UE may then transmit more average power or operate with better PA efficiency at the same output power. Those outcomes can improve uplink coverage or reduce power consumption. They still depend on the actual PA, emission limits, permitted UE power, and receiver performance. R1-2608211 Net-gain evaluation

PAPR = max(|x(t)|²) / average(|x(t)|²), or 10 log₁₀(PAPR) in dB. It measures the highest instantaneous power relative to the average over the observation interval. Scaling an entire waveform changes both powers by the same factor, so scaling alone does not change its PAPR.

Figure 2 plots instantaneous signal power rather than a processing block diagram. The upper pair uses the same QPSK data and equal average power. The lower pair scales each waveform to the same illustrative peak ceiling. The black dashed averages then show how smaller peaks can permit a higher average operating level.

Figure 2. Generated QPSK power traces illustrate the PAPR and back-off relationship. The lower plots use a common illustrative peak ceiling; they do not model an actual PA.

In this generated symbol, CP-OFDM has a PAPR of 8.27 dB, while DFT-s-OFDM has 5.19 dB. With the peak ceiling normalized to one, their average powers become 0.149 and 0.303. This follows directly from P_average = P_peak / PAPR, using the linear PAPR ratio. The resulting 3.08 dB difference illustrates available average-power margin in this example; it is not a measured coverage gain.

NOTE: These traces use one reproducible random block of 64 QPSK symbols, contiguous subcarrier mapping, and four-times oversampling of a 256-point base IFFT. The useful symbol excludes the CP. Neither waveform uses FDSS, spectrum extension/truncation, clipping, a channel, or a PA model. A practical comparison needs many symbols, PAPR statistics, and the applicable RF constraints. Actual PA back-off is not automatically equal to the maximum PAPR of one symbol.

The DFT-s-OFDM trace still fluctuates, so reduced PAPR does not mean constant envelope. A favorable PAPR result also does not guarantee a positive link-budget improvement. Any usable transmit-power increase must be considered together with the SNR required to reach the target BLER. This is why the 6G enhancement study evaluates net gain.

A useful limiting case is M = N with identity mapping and no shaping. The DFT and IFFT cancel, leaving x = d at the sample instants. With M smaller than N, mapping and interpolation change the time-domain samples, so that exact cancellation no longer applies. This explains why the two transforms are not generally redundant.

DFT-s-OFDM commonly has lower PAPR than comparable CP-OFDM, but it is not inherently constant-envelope. QAM amplitude variation, interpolation peaks, mapping, filtering, and antenna processing affect the result. Lower PAPR can reduce PA back-off, yet the usable power gain still depends on RF limits. DFT-precoded-OFDM background R1-2608211

In the plain construction, an M-point DFT preserves M complex dimensions. Spreading therefore does not add coding redundancy or reduce the modulation order. It also does not provide extra information capacity by itself. The scheduler and receiver must still agree on allocation, modulation, coding, and reference signals.

Localized allocation is the example used here. Other mathematical mappings exist, but their existence does not imply support in a particular release. Allowed allocations, DFT sizes, reference-signal arrangements, and precoders remain system design constraints. Check those constraints before treating DFT-s-OFDM as interchangeable with CP-OFDM in every scheduling situation.

What the 6G study changes around the baseline

The reviewed 6G records retain DFT-s-OFDM as an uplink waveform and study improvements around it. The added work includes rank-two operation, waveform switching, shaping, and spectrum adjustment. Each addition has its own applicability conditions.

Plain FDSS changes the spectrum weights without changing the occupied-bin count. Spectrum extension and truncation change the relationship between the input-transform size A and occupied size B. In the unmodified diagram above, A = B = M. With these enhancements, the generic transmit and receive chains need corresponding additional processing. R1-2607943

The proposed π/2-BPSK truncation receiver uses conjugate-related transform outputs to recover selected omitted components. That property is specific to the construction. It cannot be assumed for arbitrary QPSK or QAM inputs. The truncation recovery section explains this distinction. R1-2608743

NOTE: The #126bis conclusion stops pursuing additional low-PAPR enhancement for 16QAM. It does not establish that 16QAM DFT-s-OFDM is impossible or remove ordinary higher-order modulation from the waveform concept. Likewise, selecting DFT-s-OFDM does not automatically select π/2-BPSK. R1-2607943

The rank-two agreement also needs separate treatment. Transform precoding and antenna precoding serve different functions. The reviewed agreement limits each transmit port to at most one layer for a given precoder. Detailed codebooks and switching procedures require their own decisions. See layers and waveform switching. R1-2605036

• Keep the three choices separate : modulation selects the input alphabet; DFT spreading defines the waveform processing; FDSS and SE/ST add conditional enhancements.

How is a low-PAPR benefit measured?

A waveform can reduce its peaks and still lose link performance. The evaluation therefore combines the usable transmit-power improvement with the SNR change needed to achieve the same target error rate.

Net gain and equal-resource comparisons

Net gain [dB] = transmit-power gain relative to the reference − SNR degradation relative to the reference at 10% BLER. Comparisons use similar spectral efficiency and the same occupied bandwidth. A positive power term can be cancelled by a receiver penalty or by the higher coding rate needed to carry the same information. R1-2508043 R1-2509531

The comparison uses the following definitions. Identify the reference and candidate before interpreting a gain, because the waveform name alone does not specify the modulation, filtering, receiver, or operating conditions.

Actual SNR versus required SNR: Actual received SNR comes from the received signal power and noise power. Required SNR is the threshold at which a particular transmission and receiver achieve the target BLER. The candidate can increase actual SNR through higher transmit power while also needing a higher required SNR. Net gain accounts for both changes.

Figure 3 illustrates a proposed DFT-s-OFDM enhancement compared with its applicable DFT-s-OFDM baseline. It is not a CP-OFDM-versus-DFT-s-OFDM result. For π/2-BPSK, the evaluation reference uses Release-15 transparent FDSS. For QPSK or the earlier 16QAM evaluations, it uses the corresponding modulation without FDSS. The candidate applies the proposed shaping, spectrum extension, or spectrum truncation where applicable. R1-2605036 R1-2607943

The three cases below are hypothetical enhancement results. Each assumes that the candidate can use 23 dBm average transmit power, compared with 20 dBm for the reference. The reference requires 2 dB received SNR to achieve 10% BLER. Each candidate requires a different SNR at that same target. These numbers explain the calculation; they do not specify an inherent penalty for DFT-s-OFDM or a measured gain for any particular enhancement.

Figure 3. Net gain is the usable transmit-power increase minus the increase in required SNR. These values illustrate the calculation; they are not reported TDoc results.

Left example: positive net gain. The enhanced DFT-s-OFDM candidate provides 3 dB more usable transmit power than the baseline. It requires 3 dB received SNR at 10% BLER, compared with the baseline's 2 dB requirement. Its required-SNR penalty is therefore 1 dB, leaving 3 − 1 = +2 dB net gain. At a location where the baseline just meets its 2 dB target, the extra transmit power raises the candidate's received SNR to 5 dB. The candidate needs only 3 dB, so it has 2 dB of additional margin.

Middle example: break-even. The enhanced candidate again provides 3 dB more usable transmit power. However, it requires 5 dB received SNR at 10% BLER, which is 3 dB above the baseline's requirement. At the same baseline-limit location, its received SNR rises from 2 to 5 dB. It just meets its own target and has no additional margin: 3 − 3 = 0 dB net gain.

Right example: net loss. The enhanced candidate still provides 3 dB more usable transmit power, but now requires 6 dB received SNR at 10% BLER. That requirement is 4 dB above the baseline's 2 dB target. At the same location, the candidate receives 5 dB SNR and falls 1 dB short of its target. The result is 3 − 4 = −1 dB net gain. It needs 1 dB less path loss, or an equivalent improvement elsewhere in the link budget, to meet the target.

NOTE: The candidate does not physically lose received SNR because of the stated penalty. In all three examples, 3 dB more transmit power produces 3 dB more received SNR at the same path loss and noise power. The penalty describes how much higher the candidate's required SNR is at 10% BLER. Its value depends on the enhancement, modulation, coding, channel, and receiver; it is not a fixed cost of using DFT-s-OFDM.

Use the additional transmit power actually available under the applicable PA and RF constraints. A 3 dB PAPR reduction does not automatically provide 3 dB more permitted transmit power. Compare the same occupied bandwidth and similar spectral efficiency. With the other link-budget terms fixed, positive net gain represents additional tolerable path loss.

PA behavior and the reference transmission

The reference transmission must be stated before a gain can be interpreted. Modulation, filtering, power-amplifier behavior, clipping, allocation position, and receiver processing all affect the comparison and can change which candidate performs best.

The agreed reference treatment includes π/2-BPSK with Release-15 transparent FDSS, and QPSK or 16QAM without FDSS for the corresponding comparison. Evaluation reports also identify the PA model and whether crest factor reduction, or CFR, is used. Comparing a new filter against an unnecessarily weak baseline can exaggerate the apparent benefit. R1-2605036 R1-2608743

The #123 assumptions include a 4 GHz carrier, at least 100 MHz channel bandwidth, 30 kHz SCS, TDL-C with 300 ns delay spread, and 3 km/h mobility. They also include one transmit antenna, one or four receive antennas, and two DMRS plus twelve data symbols. HARQ and frequency hopping are excluded from that evaluation setup. These are comparison assumptions, not universal 6G deployment limits. R1-2509531

Inner and outer RB allocations require separate attention because emissions constrain the available transmit power differently. Contributions also ask for results across RB sizes and positions. A favorable allocation does not establish the same benefit for every scheduler choice. R1-2608624 R1-2607460

Power boosting introduces another condition. One proposal evaluates tighter ACLR at a boosted operating point and asks RAN4 to address the corresponding power-class and transmission requirements. That proposal does not authorize every UE to transmit above its existing maximum power. The useful comparison must identify the permitted output-power limit as well as the modeled PA response. R1-2608211

• Keep the complete comparison : report occupied bandwidth, spectral efficiency, BLER target, PA/RF limits, allocation, CFR, and receiver assumptions together.

What is FDSS, and how does it shape DFT-s-OFDM?

FDSS means frequency-domain spectrum shaping. It changes the weights of the frequency-domain signal before OFDM generation. The aim in this uplink study is to improve power behavior without losing too much receiver performance.

The basic operation is simple, but selecting a useful filter is not. A filter can reduce peaks and still deliver a poor net gain. This overview separates the multiplication operation from its filter design, receiver assumptions, and optional spectrum adjustment.

Basic operation: weight the DFT outputs

Start with the ordinary DFT-s-OFDM data path. The modulation mapper produces symbols, and the spreading DFT produces frequency-domain values X[k]. Plain FDSS multiplies those values by filter coefficients before subcarrier mapping and IFFT processing.

Y[k] = F[k] × X[k], where X[k] is the unshaped DFT output, F[k] is the shaping coefficient, and Y[k] is the shaped output. The coefficient can describe amplitude and phase. If F[k] = 1 everywhere, this operation leaves the signal unchanged.

Figure 4 uses eight equal input magnitudes to isolate the multiplication. The toy filter gives smaller weights to the edge bins. All eight bins remain present, so this is shaping without spectrum extension or truncation. Actual data generally produces unequal DFT-output magnitudes.

Figure 4. An illustrative FDSS multiplication. The plotted weights explain the operation; they are not a proposed 6G filter or evidence of a particular PAPR gain.

For example, X[k] = 1 + j and F[k] = 0.5 produce Y[k] = 0.5 + 0.5j. The amplitude halves and the power becomes one quarter. A fair waveform comparison then uses an explicit normalization convention. Merely reducing every sample amplitude would reduce average power and peak power together, leaving PAPR unchanged.

Why frequency-domain shaping can change time-domain peaks

The IFFT combines the weighted bins into a time-domain signal. Changing their relative weights changes how they combine at each instant. FDSS therefore changes the pulse shape and envelope even though its multiplication occurs in the frequency domain.

The interactive illustration below makes this relationship visible. Move the shaping-strength slider and compare the three plots. At zero strength, the filter weights are all one and the two waveforms coincide. Increasing the strength changes their relative frequency weights and therefore changes the time-domain sum.

Figure 5. Interactive frequency-weight and time-waveform comparison. The upper plot shows raw filter weights; the lower plots include equal-average-power normalization.

Each occupied subcarrier contributes a rotating complex component. In simplified notation, x(t) is proportional to the sum of X[k] exp(j2πf_kt). After shaping, the sum uses F[k]X[k] instead. The components reinforce or cancel each other differently as time changes. Reducing an edge weight therefore need not reduce every time-domain sample: it can also weaken a component that previously helped cancel another one.

The middle plot shows only the real component, so its largest visible excursion does not necessarily locate the largest signal-envelope peak. The bottom plot uses |x(t)|², including both real and imaginary components. Both signals have average power one, so their maximum plotted powers directly determine their linear PAPR. Normalization prevents simple attenuation from being mistaken for a PAPR improvement.

NOTE: This teaching example uses 32 π/2-BPSK symbols, a spreading DFT, contiguous mapping, and 512 time samples over one useful symbol. This represents four-times oversampling of a 128-point base IFFT. The real, non-negative toy filter tapers the edges as strength increases; it is not a specified 6G filter. CP, reference signals, SE/ST, CFR, channel effects, and PA distortion are omitted. Select New data block to see that peak locations and PAPR depend on the data. A smooth spectrum does not guarantee a PAPR reduction, and this illustration does not predict net gain.

For a fixed mapped grid, frequency-domain multiplication corresponds to circular convolution of the useful time-domain block with the inverse transform of the filter response. This signal-processing relationship explains why frequency weights can change time-domain peaks. It does not mean every smooth spectral shape produces lower PAPR.

A useful filter must suit the modulation and waveform construction. The reviewed proposals include GMSK-approximation shaping and other steep-roll-off filters. Their peak behavior and required-SNR results depend on more than the visible spectrum shape. Some constructions add processing beyond the single multiplication shown here. R1-2608211 R1-2607943

Strong attenuation can also weaken useful observations. A receiver that inverts a small effective response may amplify noise. The evaluation must therefore measure the power advantage and the required-SNR change together. A favorable spectrum or a lower peak alone does not establish better coverage. See net-gain evaluation.

Known processing, selectable filters, and receiver transparency

FDSS names a processing family rather than one universal coefficient table. To reproduce a candidate, we need its filter response, normalization, transform size, mapping, and any extension or truncation rules. A name such as FDSS alone is insufficient.

The reviewed #126bis conclusion leaves filter characterization to RAN4, for example through a spectrum mask. It states that RAN1 will not define transmit-filter coefficients. Individual evaluations can still disclose exact filter designs; their disclosure does not turn one design into a common adopted 6G filter. R1-2607287

Transparent FDSS still applies a filter. The receiver does not need separate, explicit knowledge of its coefficients. When data and DMRS share the filter, channel estimation can observe the combined radio-channel and filter response. Non-transparent processing explicitly exploits knowledge of the transmit filter. Neither label alone defines a signaling procedure. See the filter-transparency definition table. R1-2608337 R1-2607287

FDSS, spectrum adjustment, and CFR are separate choices

Several techniques can affect PAPR, but they change different parts of the transmission. Keep their roles separate when reading a candidate description or choosing a reference for a performance comparison.

FDSS does not automatically imply extra occupied bandwidth. It also does not automatically invoke a peak-clipping operation. A design can combine shaping and CFR, so the reference and candidate descriptions must state both. The reviewed evaluations explicitly distinguish comparisons with and without CFR. R1-2607943

• Understand the operation before the gain claim : identify the weights, spectrum mapping, normalization, and receiver treatment, then compare usable power and required SNR.

What changes with FDSS, spectrum extension, and spectrum truncation?

The proposals manipulate the frequency-domain signal around DFT spreading. Their main difference is whether shaping keeps the same spectrum size, expands it, or removes part of it before transmission.

Definitions and processing relationships

FDSS means frequency-domain spectrum shaping. In this section, SE means spectrum extension and ST means spectrum truncation; spectral efficiency is written out to avoid confusing the two uses of SE. Let A be the number of subcarriers before extension or truncation, and B the occupied number afterward. The agreed extension factor is (B − A)/B. The truncation factor is (A − B)/A. The different denominators matter when comparing reported percentages. R1-2509531

The diagram below groups the three constructions by their spectrum-size relationship. It summarizes the comparison variables rather than prescribing the detailed ordering of every transmitter implementation. Equivalent signal constructions can use different processing arrangements. R1-2607460 R1-2608743

Figure 6. Spectrum-size relationships for the FDSS enhancement families. A lower PAPR must still translate into positive net gain.

Examples with equal, extended, and truncated spectrum

The following examples keep the input size fixed so that the occupied-bin count is easy to compare. The accompanying notes explain why a truncation illustration alone cannot define a valid information-preserving mapping.

Figure 7 makes the three size relationships visible using individual frequency bins. Each example starts with A = 12. Blue boxes show the unchanged-size region or retained bins, green boxes show added bins, and crossed red outlines show removed bins. The counts are illustrative, not prescribed 6GR allocations.

Figure 7. Plain shaping keeps B = A; extension makes B > A; truncation makes B < A. Only occupied output bins count toward B.

NOTE: These examples hold A fixed to explain the size change. A net-gain evaluation instead compares candidates at the same occupied bandwidth and similar spectral efficiency, as described in Figure 6. The extension and truncation percentages use different denominators.

NOTE: Does B < A mean losing data? It would if the removed bins carried independent information. In the proposed π/2-BPSK DFT-s-OFDM construction, however, the DFT outputs have conjugate-symmetry relationships. Selected outputs therefore contain redundant information. The transmitter can omit those outputs, and the receiver can reconstruct them through repetition and conjugation after equalization, before the inverse DFT. This is not permission to delete arbitrary subcarriers from a general OFDM signal. The crossed-out bins in Figure 7 illustrate a count reduction; the actual retained bins must follow the waveform's symmetry and mapping rules. R1-2608743 R1-2607460

Preserving recoverable information does not mean preserving the same reception performance. The omitted bins no longer provide additional noisy observations for combining, and truncation changes PAPR. The receiver can therefore incur an SNR penalty even when the underlying information remains recoverable. Any usable transmit-power advantage must outweigh that penalty in the net-gain comparison. R1-2608743

The bin relationships describe signal construction. The receiver must also recover the information correctly, while the scheduler must choose a supported transform and a consistent transport-block size. These constraints are developed in the following sections.

• Separate spectrum size from recoverability : A and B count bins; valid truncation also requires the defined symmetry and receiver mapping.

How does the receiver recover a truncated spectrum?

Spectrum truncation needs a defined relationship between the retained and omitted transform outputs. The proposed receiver exploits that relationship; it does not recover arbitrary independent data after the transmitter has discarded it.

Symmetry and reconstruction of omitted outputs

The proposed π/2-BPSK construction starts from real-valued information symbols with a prescribed phase rotation. Its DFT outputs consequently contain conjugate-related components, so every output does not represent an independent complex degree of freedom. R1-2608743

The transmitter retains a permitted subset of those outputs. After FFT processing and channel estimation, the illustrated receiver equalizes the received components, extends the vector using repetition and conjugation, and then performs the inverse DFT. Demapping and channel decoding follow. The omitted values are reconstructed from the signal relationship, not guessed by the channel decoder. R1-2608743

The selected indices and phase conventions are essential. A count such as A = 12 and B = 8 does not specify them. Figure 7 should therefore be read as an illustration of sizes; it is not an implementation recipe for deleting both edges of an arbitrary DFT output. R1-2608743 R1-2607460

Noise combining and implementation choices

Information recoverability describes the noiseless mapping. In a radio link, each transmitted component also supplies a channel-dependent noisy observation. Removing redundant components can reduce the observations available to the equalizer and change error performance. R1-2608743

The proposed MMSE processing can exploit the same symmetry when combining observations. Its result depends on channel estimation, the retained spectrum, the filter, and the truncation factor. A stronger truncation ratio therefore cannot be justified by symbol counting alone: the required SNR and the available transmit-power improvement must be evaluated together. R1-2608743

Equivalent I/Q-offset QPSK constructions can implement the same transmitted signal with a smaller complex transform and additional mapping operations. This matters when comparing hardware cost. A direct large-DFT block diagram and an equivalent realization can have different processing requirements, even when their over-the-air waveform agrees. R1-2607460 R1-2608743

• Recoverable does not mean penalty-free : verify the symmetry mapping, noisy receiver behavior, and implementation before interpreting a truncation gain.

How do transform sizes affect scheduling and transport blocks?

The network allocates occupied spectrum, but the transmitter processes an information-bearing transform that may have a different size. Spectrum adjustment therefore connects waveform design with allocation granularity, coding rate, and hardware support.

DFT size and scheduling consequences

Changing the spectrum size also changes the relationship between the information-bearing transform and the scheduled allocation. The resulting design question includes scheduler flexibility, transform hardware, transport-block sizing, and achievable extension or truncation ratios.

The #126 agreement supports A values of 12 × 2^x × 3^y × 5^z, with non-negative integer exponents. The additional factor-four family remains for a #127 decision, while the upper bound also needs consideration. The latest documents therefore should not be read as already adopting unrestricted transform sizes. R1-2606806 R1-2606760

Proposals favoring the factor-four family seek finer matching between the desired and realizable spectrum-adjustment ratios. Other proposals prefer the factor-twelve baseline because of implementation cost. A separate approach proposes DFT segmentation and selected small transforms. These are competing ways to obtain scheduling flexibility, rather than one agreed implementation. R1-2608337 R1-2607460 R1-2607555 R1-2608624

Transport-block sizing and MCS design

Transport-block sizing and MCS design remain connected to this choice. Incoming proposals use the pre-adjustment number A in resource calculations and relate MCS entries to the extension or truncation factor. Large ST allocations also raise transform-size questions. An equivalent offset-modulation implementation may change the processing burden, so equal transmitted signals do not automatically imply equal hardware cost. R1-2608337 R1-2607460 R1-2608743

For a fixed occupied allocation B, extension leaves fewer input symbols A, while truncation permits more input symbols within the proposed recoverable construction. A comparison carrying the same information must therefore account for the changed coding rate and modulation order. Copying an MCS entry without checking this relationship can make two nominally equal-resource simulations carry different information amounts. R1-2608337 R1-2607460

The factor-four proposal allows finer adjustment of A, but the input transform no longer always corresponds to an integer number of twelve-subcarrier groups. This affects how a resource-element count is derived for transport-block sizing. The proposed use of A and the associated MCS adjustments remain design proposals; they are not a finalized 6G transport-block calculation. R1-2608337

• Spectrum adjustment is more than filtering : its useful gain must survive coding-rate, transform-size, allocation, and implementation constraints.

When does lower PAPR become usable RF power?

Lower peaks can give the power amplifier more operating margin, but the UE still has to satisfy its output-power and emission requirements. Waveform comparisons must identify which constraint actually limits the transmitted power.

Emission limits, maximum power, and power boosting

A PA evaluation may find a higher output level that still meets ACLR or in-band emission requirements. That result establishes a modeled operating point. It does not by itself increase the maximum output power permitted for a UE power class. R1-2608211

The power-boosting proposal therefore asks for RF requirements to be considered together with the waveform. If the maximum permitted UE power is already binding, additional PA margin cannot simply be added to the link budget. If emission performance is binding instead, the waveform and allocation can affect how much of that margin is usable. R1-2608211 R1-2608624

Power consumption is a separate outcome. At a specified radiated output, a more efficient PA operating point may reduce DC consumption. At a specified DC budget, it may support a different output. The comparison must say which quantity is held fixed; neither result automatically establishes an increase in permitted radiated power. R1-2608211

Spectral masks and mode-dependent power reporting

Intentional spectrum shaping changes the amplitude across the allocation. A flatness requirement that assumes an unshaped signal can therefore constrain the proposed filter, even when the PA itself introduces little additional distortion. R1-2608211

The proposed characterization separates the intended shaping response from RF impairment. The session conclusion leaves filter characterization to RAN4 and does not require RAN1 filter coefficients. This distinction allows receiver implementation and RF conformance to be discussed without treating a proposed filter as an adopted coefficient table. R1-2607287 R1-2608211

The same dependency reaches scheduling. Proposed power-headroom information distinguishes waveform, layer count, and filter options because maximum-power reductions and the applicable maximum power can change with the mode. A scheduler selecting a new mode needs an appropriate power estimate for that mode. The proposed reporting tables and UE preferences remain under discussion. R1-2608211

• Identify the binding power limit : PA margin, permitted output power, DC consumption, and reported headroom answer different questions.

Does the receiver need to know the transmit filter?

The filter debate concerns both performance and specification responsibility. A receiver may obtain a better result by exploiting the transmit filter, but that does not automatically require RAN1 to standardize filter coefficients.

Transparent filtering and receiver knowledge

Transparent FDSS does not mean no FDSS. The transmitter still shapes the frequency-domain signal. Transparency describes whether the receiver needs explicit knowledge of that shaping response, rather than whether shaping occurs.

For the simple per-subcarrier model, H_effective(k) = H_channel(k) × F_FDSS(k). Applying the same filter to data and DMRS lets channel estimation observe this combined response. The receiver can compensate it without separately identifying the shaping filter. This does not guarantee identical performance to a receiver that explicitly exploits filter knowledge. R1-2608337 R1-2607287

NOTE: The π/2-BPSK reference with Release-15 transparent FDSS is already shaped. It must not be treated as a no-FDSS reference. Transparency also does not by itself specify whether spectrum extension or truncation is present; those are separate construction choices. R1-2605036 R1-2607943

Incoming positions differ substantially. One group favors transparent shaping and considers the additional benefit of filter knowledge too limited. Another requests specified or nominal filters so that receiver processing, MCS design, and related procedures can rely on a known response. These requests appear in the papers as proposals, not as a common conclusion. R1-2608624 R1-2607252 R1-2607555 R1-2607895 R1-2608211

The #126bis observations isolate the benefit of knowledge by comparing the same steep-roll-off transmit filter with and without its use at the receiver. Without SE or ST, the additional gain is typically 0–0.5 dB. Two exceptional results use 30 ns delay spread and four PRBs, reaching 1.05 dB and 2.15 dB. These exceptions cannot be generalized to the 300 ns comparison baseline. R1-2607287

With spectrum extension, the observed additional gain ranges from 0 to 1.3 dB. For spectrum truncation, one source reports 0.3–1.62 dB. These are gains from receiver knowledge relative to the same filter and spectrum adjustment. They are not the total gain over an NR reference waveform. Receiver processing and its additional complexity remain part of the interpretation. R1-2607287

Data, DMRS, and specification responsibilities

The session conclusion requires the same transmit filter on data and DMRS. It treats the receiver's exploitation of filter knowledge as an implementation choice. RAN4 decides how to characterize filters, for example through a spectrum mask, while RAN1 will not define coefficients. Capability indication and RRC activation are mentioned as a possible alignment mechanism, not a finalized signaling design. R1-2607287

• Do not equate filter knowledge with specified coefficients : the latest conclusion separates receiver implementation from RAN4 filter characterization.

Which enhancement results remain convincing?

The incoming results do not identify one enhancement that wins across all allocations and spectral efficiencies. They instead show why the study must retain its baseline, receiver assumptions, and operating range.

Allocation-dependent results and reference choices

Some evaluations favor π/2-BPSK with truncation over an intermediate spectral-efficiency range. Others find that spectrum extension or truncation provides only limited benefit once transparent FDSS and optimized legacy operation are included. A further position emphasizes plain FDSS with a GMSK-approximation filter and extending π/2-BPSK coding rates. These remain different technical recommendations. R1-2608743 R1-2607877 R1-2608211

Allocation size can reverse the preferred candidate. One evaluation favors QPSK with extension for a smaller inner allocation but π/2-BPSK with truncation for a larger one at low spectral efficiency. Its outer-allocation results have a different ordering. Another evaluation stresses that no single filter is best across PA models, MCS values, and bandwidths. A single maximum-gain number would hide these conditions. R1-2608337 R1-2607460

Study progression and the 16QAM conclusion

The #125 milestone had already recognized transparent FDSS as beneficial for π/2-BPSK from the RAN1 perspective and liaised RAN4. However, its observations on other PAPR enhancements were explicitly interim. Spectrum extension, truncation, and filter transparency still needed further work before definite conclusions. R1-2603729 R1-2605036

At #126, the list for further selection retained plain FDSS for π/2-BPSK and QPSK, FDSS-SE also for 16QAM, and FDSS-ST for π/2-BPSK. Applicable spectral-efficiency ranges and selection conditions remained open. This explains why 16QAM appears in the August summary but is no longer pursued after the October conclusion. R1-2606806 R1-2606760

The #126bis record is more decisive for 16QAM. For plain FDSS, two sources show net losses against unshaped 16QAM without CFR over the evaluated spectral-efficiency range. Other cases include a small isolated gain and mixed results when CFR is included. Comparisons against QPSK also fail to establish a compelling benefit. The session therefore concludes that UL low-PAPR enhancement for 16QAM is not pursued. R1-2607287

Practical takeaway: The evaluated 16QAM enhancements offered too little benefit to justify the additional implementation effort and complexity. This is a practical interpretation of the results, rather than a quantified effort-to-gain criterion adopted by RAN1. Many cases produced a net loss, not merely a small gain: the usable transmit-power improvement did not outweigh the increase in required SNR. The decision concerns additional low-PAPR enhancements; it does not exclude ordinary 16QAM DFT-s-OFDM. R1-2607943

That conclusion should replace an older description of 16QAM FDSS-SE as an equally active candidate in this study. It does not remove ordinary 16QAM operation from 6G. For π/2-BPSK and QPSK, the remaining discussion still needs to connect net gain with scheduling, RF requirements, and implementation cost. R1-2607943 R1-2607287

• Preserve conditional results : identify where a gain applies, and use the latest session conclusion when the study has narrowed its scope.

How do layers and waveform switching fit together?

Waveform support also depends on how the UE transmits multiple layers and how the network selects the operating mode. These procedures affect the practical value of a low-PAPR uplink design.

Rank, precoding, and evaluation assumptions

The earlier evaluation work explains the later rank decision. At #124, link comparisons used equal rank, allocation, and transmit power, with HARQ retransmissions disabled. System-level metrics included user-perceived throughput and cell-edge percentiles. By #124bis, reported gains still depended on traffic, scheduling, receiver antennas, and the reference configuration. A power-limited link gain therefore did not establish a universal system-throughput gain. R1-2601512 R1-2603279

The #125 agreement supports a maximum rank of two for UL DFT-s-OFDM. Each transmit antenna port carries at most one layer for a given precoder. Final codebook decisions belong to the uplink-transmission agenda. A UE supporting a DFT-s-OFDM maximum rank of one or two must support at least that rank for CP-OFDM. R1-2605036

Waveform switching and UE assistance

The agreed waveform-selection scope includes semi-static determination through RRC for all RRC states. Dynamic switching is supported for RRC_CONNECTED, with detailed operation and capability aspects left for further study. These statements establish the intended behavior without selecting one DCI encoding or every switching restriction. R1-2605036

RRC configuration and dynamic selection can serve different roles in the same procedure. The table separates the agreed scope from proposed ways to select a waveform, and identifies where the referenced documents leave the mechanism open.

NOTE: RRC and DCI are not mutually exclusive alternatives. RRC can configure the permitted operation or candidate modes, while scheduling DCI selects or implies the waveform dynamically. The agreement supports dynamic PUSCH waveform determination in RRC_CONNECTED, but does not settle the DCI encoding, activation timing, or a MAC CE switching procedure. R1-2605036

One proposal uses an implicit association with DCI format or frequency-domain resource allocation to avoid an extra explicit waveform bit. Another asks that switching account for coherent and non-coherent UE precoding capabilities and multilayer operation. Neither proposal can be presented as the agreed switching mechanism. R1-2607252 R1-2607877

Related proposals ask the UE to report waveform, layer-count, and FDSS preferences. They also suggest power-headroom information for different operating combinations instead of one undifferentiated value. These proposals highlight the scheduling dependency: the network needs an appropriate power and capability picture for the selected mode. They do not establish new finalized information elements. R1-2608211

• Rank support does not finish the procedure : precoders, waveform selection, capabilities, and power reporting require their own detailed treatment.

What is being studied for sensing waveforms?

Sensing adds requirements that a communication BLER comparison does not capture. The waveform must support target detection and estimation while controlling interference, transmit power, resource consumption, and disruption to communication.

Baseline sequences and candidate scope

The agreed starting point is CP-OFDM and the communication frame structure. The study considers enhancements to Gold and Zadoff-Chu sequences, quadratic-phase sequences, transform-precoded signals, processing or shaping changes, and signals spanning multiple symbols. Their applicability differs between TRP and UE transmission. The comparison benchmarks are NR Gold for TRP transmission and NR ZC for UE transmission. R1-2607998

One design direction improves existing RS reuse through root selection, sequence mapping, or shaping. Another seeks chirp-like and pulse-like signals within an OFDM generation framework. A joint proposal asks for the quadratic-phase option to support both pulse and chirp behavior, while allowing competing constructions. That is a joint proposal, not an adopted universal sensing waveform. R1-2607483 R1-2607590 R1-2607989

Sequence design, chirps, and pulse confinement

The key question is what the transmitted signal actually does after mapping and waveform generation. A sequence name alone does not establish its power behavior, delay-Doppler response, or ability to confine energy within a pulse.

The quadratic-phase study uses a frequency-domain form x(m) = A exp(jBm² + jCm + jD). The #126bis working assumptions associate B with chirp rate and duration, C with time placement, and D with phase relationships. A can be constant or depend on m, including for pulse confinement. Candidate selection, multiplexing parameters, sequence counts, and comb details remain open. R1-2607998

Proposals use enhanced ZC constructions to create continuous or pulsed chirp-like signals. They emphasize energy inside and outside the intended pulse, edge decay, and PAPR during the on-duration. Another contribution points out that a shorter nominal duration alone does not guarantee adequate energy confinement. These distinctions matter when transmit and receive periods must be separated. R1-2608440 R1-2608313

Other approaches study Golay sequences, constant-amplitude sequences, or different receiver processing. One proposal applies FMCW-style processing to an OFDM-based ZC signal. Another studies chirp-like sequences and π/2-BPSK reference signals while favoring receiver-side windowing over transmit-side shaping. These examples illustrate alternative implementations and trade-offs rather than a single agreed receiver. R1-2607084 R1-2607279 R1-2608243 R1-2608369

Shared sensing and communication resources require special care. When both components occupy the transmitted signal, a sensing-only sequence's PAPR or ambiguity response may not describe the composite transmission. A proposal therefore asks evaluations to include the communication component when assessing shared-resource operation. R1-2607910

• Evaluate the transmitted signal and sensing mode : sequence properties must survive mapping, multiplexing, PA behavior, and the intended receiver.

How do sequence mapping and shared resources change sensing?

A sensing sequence acquires its practical properties after subcarrier mapping, symbol placement, pulse construction, and receiver processing. Those choices can change the transmitted peaks and the delay-Doppler response even when the sequence name stays the same.

Root selection, comb mapping, and multiplexing

Selecting a Zadoff-Chu root for low PAPR alone can leave strong Doppler-induced range sidelobes. Proposed sensing-oriented root selection therefore considers both power behavior and ambiguity response, while retaining enough useful sequences for multiplexing. R1-2608313 R1-2607279

Sequence length and extension also matter. One proposal studies even-length sequences without the cyclic extension used in a legacy construction. Another emphasizes retaining useful cross-correlation behavior when several TRPs, beams, or UEs transmit. A favorable single-sequence autocorrelation result does not establish reliable separation between all simultaneous transmitters. R1-2607279 R1-2608313

A comb in frequency produces repeated portions in the time-domain signal. A pulse proposal transmits one portion and leaves a receive interval after it. Choosing the comb and pulse duration changes the timing opportunity, but simply mapping a sequence onto a comb does not itself silence the remaining repetitions. The pulse construction and its resulting spectrum need their own evaluation. R1-2607279

Receiver windowing and the composite transmitted signal

Windowing can suppress delay sidelobes at the cost of a broader mainlobe. The proposed location of that operation matters: transmit-side shaping changes the radiated waveform and its PAPR, whereas receiver-side weighting changes the processing of an already received signal. R1-2608369 R1-2607279

One contribution favors receiver-side windowing as an implementation option to avoid the transmit-PAPR penalty seen in its comparison. Other studies compare one-sided and matched transmit/receive shaping. These are conditional design results; the window, normalization, receiver, and ambiguity definition must be stated before comparing their sidelobe numbers. R1-2608369 R1-2607279

Shared communication and sensing resources add another dependency. A sensing-only signal can have low peaks while the sum with communication data has different peaks and interference. Evaluations must include the transmitted composite signal where the resources are shared. Dedicated sensing occasions instead need their resource cost and transmit-receive separation to be reported. R1-2607910 R1-2608243 R1-2607084

• Evaluate the mapped signal : include sequence diversity, Doppler, receiver weighting, and communication content where they apply.

How are sensing candidates compared?

Sensing evaluation combines waveform characteristics with detection and estimation performance. The latest decisions improve comparability by specifying how ambiguity functions, sidelobes, cross-ambiguity, and PAPR are measured and which assumptions must be reported.

The sensing evaluation framework also developed over several meetings. At #124bis, CP-OFDM became the benchmark and the communication frame became the starting point. The #125 work identified enhancement objectives and their costs, including emissions, interference, processing complexity, and communication impact. The #126 summary still contains discussion of common scenarios and competing metric definitions. The later #126bis choices must therefore be distinguished from those earlier proposals and open questions. R1-2602985 R1-2604765 R1-2606567 R1-2607998

The study considers transmit-power gain, PAPR, peak and integrated sidelobes, mainlobe width, cross-correlation, interference resilience, and resource efficiency. It also considers RF emissions, ISI/ICI, receiver complexity, and communication impact. Link-level or simplified system-level evaluation complements theoretical analysis; an attractive ambiguity plot alone does not establish a deployment benefit. R1-2607998 R1-2607483 R1-2607590

Ambiguity function and sidelobe definitions

The ambiguity function describes the response to delay and Doppler changes. Its definition must match the assumed signal processing, otherwise two evaluations can report different performance for reasons unrelated to the candidate sequence.

The #126bis agreement selects circular AF and CAF option 2b over one or more OFDM symbols. Delay shifts are cyclic within the useful symbol interval, while Doppler is evaluated across the observation. Reports identify receiver processing, observation window, and oversampling. The same oversampling must be used for the NR reference and candidate sequences. Operation beyond the CP-associated delay remains for further study. R1-2607998

The mainlobe region uses the first null-to-null interval. Mainlobe width is separately measured at the half-power points. The sidelobe region excludes the mainlobe; for a comb pattern, the relevant full region is one period among the periodic replicas. This distinction prevents the region definition from changing silently between candidates. R1-2607998

PSLR uses 10 log₁₀(P_ML/P_SL), where the powers are the maximum mainlobe and sidelobe values. The corrected ISLR uses 10 log₁₀(I_SL/I_ML), where the powers are integrated over the respective regions. The ratios have opposite orientations. Older ISLR values using the reciprocal definition must therefore not be compared without checking their convention. R1-2607998

Cross-Ambiguity Peak takes the maximum cross-ambiguity power over the declared region. Integrated Cross Ambiguity integrates that power. Reports specify the delay-Doppler region or Doppler cut used. Normalization remains open, so the agreed names alone do not make all submitted numerical values directly interchangeable. R1-2607998

PAPR, oversampling, and realistic performance

PAPR also needs a consistent measurement interval and sequence population. This becomes particularly important for a pulse-like signal, whose power statistics differ depending on whether the interval includes its silent portion.

The #126bis text defines instantaneous PAPR per OFDM symbol, retaining a bracketed [1%] operating point. It requires an IFFT oversampling factor of at least four to be reported and a CCDF over the RS sequence family. Reports also include PAPR variation and the worst sensing-symbol PAPR within the coherent processing interval. R1-2607998

Two details remain open: handling sequence families too small for the stated probability point, and measuring pulse-like PAPR over the whole symbol or only the pulse duration. The brackets and open items should remain visible in a summary. They prevent a provisional metric from appearing to be a fully closed requirement. R1-2607998

Candidate proposals also examine detection and false-alarm behavior under PA nonlinearity, Doppler robustness, and sequence-family cross-correlation. A root choice that reduces PAPR may still leave strong Doppler-induced sidelobes. Improving one metric therefore does not eliminate the need to test the others under common assumptions. R1-2607590 R1-2608313

• Match definitions before comparing results : correlation model, sampling, region, normalization, and pulse interval can materially change the reported outcome.

Does sensing require a different frame or longer CP?

The sensing papers examine cases where echoes do not fit comfortably within communication timing assumptions. They do not establish that every sensing mode needs a new frame or a physically longer cyclic prefix.

Physical CP and effective observation interval

The agreed starting point remains the communication frame, slot, symbol, CP, and SCS. Further study considers effective-CP enhancement if necessity and benefits are identified. The candidate families include successive OFDM symbols, frequency-domain comb structures, and time gaps within an unchanged OFDM-symbol duration. Detailed parameters and applicable sensing modes remain open. R1-2607998

Contributions disagree about how often the problem requires specification changes. One emphasizes absolute round-trip delay in monostatic sensing and relative path delay in bistatic sensing. Another argues that timing alignment or receiver timing adjustment can address much of the concern. A third stresses that reception beyond the communication CP does not itself prove a need for a standardized longer CP. R1-2607279 R1-2607483 R1-2608313

Transmit-receive separation and resource cost

Pulse-like signals also introduce a transmit/receive scheduling question. Time separation can help manage self-interference, but the pulse must fit the intended resources and leave sufficient usable reception time. Proposals for dedicated sensing gaps and flexible sensing occasions address that scheduling problem. They remain proposals rather than an agreed replacement frame structure. R1-2607989 R1-2608243 R1-2607084

The same distinction applies to repeated symbols. Repetition or joint design across symbols can extend useful sensing operation, but affects resources, coherent processing, and sequence design. The evaluation must show the benefit and communication cost for the relevant mode. The open excess-CP treatment in the latest AF/CAF agreement is another reason to avoid claiming that this issue is already resolved. R1-2608369 R1-2607590 R1-2607998

• Establish the timing problem before changing the frame : distinguish a receiver solution, effective-CP enhancement, and a new standardized signal structure.

Update Note

The entries below record substantive expansions and their source scope. They distinguish the present topic-directed review from earlier additions so that later readers can follow changes in coverage and status.

Current revision

This revision revisits selected technical passages in the referenced contributions and compares them with the session conclusions. The expansion adds receiver, scheduling, RF, and sensing-mapping detail while retaining the existing agreement history.

• 2026-10-25 - Added DFT-s-OFDM fundamentals with transmit and receive processing, a worked transform-size example, a flow diagram, characteristics, and the distinction between baseline modulation support and 6G enhancements. Added selected historical specification and receiver references.
• 2026-10-25 - Added dedicated sections on truncation recovery, transform and scheduler design, RF power constraints, and sensing resource mapping. Revisited relevant passages from the listed communication and sensing contributions, including Ericsson, Nokia, Qualcomm, Huawei, and Samsung. Organized every h1 under at least two h3 subsections; updated the contents and mind map. This is a topic-directed review, not a page-by-page audit of every referenced document.

Earlier revisions

The earlier revisions established the baseline and agreement history and introduced the original comparison sections. Their entries are retained below to show which material predates the more detailed implementation and resource discussions.

• 2026-10-20 - Reviewed earlier summary TDocs and session notes from RAN1#122–#126. Expanded Agreed Items, baseline history, enhancement progression, multilayer evaluation, and sensing evaluation history; added direct historical citations and references.
• 2026-10-20 - Added Executive Summary, Agreed Items, the communication baseline, net-gain evaluation, FDSS/SE/ST, filter knowledge, enhancement results, layers and switching, and ISAC waveform, metrics, and frame/CP discussions from selected RAN1#122–#126bis source passages and agreement histories.

Reference

These sources include historical summaries, session records, and individual technical contributions. Read an individual proposal together with the applicable later conclusion; listing a document does not imply that its recommendations were adopted.

Communication waveform sources

The communication sources cover the baseline, low-PAPR enhancement, receiver knowledge, transform sizing, and multilayer operation. Session records and feature-lead summaries supply the status against which individual recommendations are interpreted.

• R1-2506595 (RAN1#122) : Feature Lead summary #2 on 6GR waveform (Moderator (Nokia))
• R1-2508043 (RAN1#122bis) : Feature Lead summary #3 on 6G waveform (Moderator (Nokia))
• R1-2509531 (RAN1#123) : Feature Lead summary #5 on 6G waveform (Moderator (Nokia))
• R1-2601512 (RAN1#124) : Session Notes of AI 10.2.1(Waveform for 6GR air interface) (Ad-Hoc Chair (NTT DOCOMO, INC.))
• R1-2601696 (RAN1#124) : Feature Lead summary #5 on 6G waveform (Moderator (Nokia))
• R1-2603279 (RAN1#124bis) : Session Notes of AI 10.2.1(Waveform for 6GR air interface) (Ad-Hoc Chair (NTT DOCOMO, INC.))
• R1-2603315 (RAN1#124bis) : Feature Lead summary #6 on 6G waveform (Moderator (Nokia))
• R1-2603729 (RAN1#125) : Feature Lead summary #5 on 6G waveform (Moderator (Nokia))
• R1-2605036 (RAN1#125) : Session Notes of AI 10.2.1(Waveform for 6GR air interface) (Ad-Hoc Chair (NTT DOCOMO, INC.))
• R1-2606760 (RAN1#126) : Session Notes of AI 10.2.1 (Ad-Hoc Chair (NTT DOCOMO, INC.)
• R1-2606806 (RAN1#126) : Feature Lead summary #5 on 6G waveform (Moderator (Nokia))
• R1-2607252 (RAN1#126bis) : On waveform enhancements/additions for 6G Radio (OPPO)
• R1-2607287 (RAN1#126bis) : Session Notes of AI 10.2.1 (Ad-Hoc Chair (NTT DOCOMO, INC.))
• R1-2607460 (RAN1#126bis) : Waveform for 6GR air interface (Huawei, HiSilicon)
• R1-2607555 (RAN1#126bis) : Discussion on waveform for 6GR (Samsung)
• R1-2607877 (RAN1#126bis) : On waveform for 6GR (Ericsson)
• R1-2607895 (RAN1#126bis) : Waveform for 6GR air interface (MediaTek Inc.)
• R1-2607943 (RAN1#126bis) : Feature Lead summary #5 on 6G waveform (Moderator (Nokia))
• R1-2608211 (RAN1#126bis) : On Waveforms for 6GR air interface (Apple)
• R1-2608337 (RAN1#126bis) : Discussion on Waveform (NTT DOCOMO INC.)
• R1-2608624 (RAN1#126bis) : On remaining aspects of waveform for 6GR (Nokia)
• R1-2608743 (RAN1#126bis) : Waveforms for 6GR (Qualcomm Incorporated)
• 3GPP TS 36.211 V8.9.0 : LTE Physical channels and modulation. Sections 5.3.2 and 5.3.3 reviewed for historical modulation and transform-precoding background.
• 3GPP TS 38.211 V16.5.0 : NR Physical channels and modulation. Selected portions of sections 5.1.1, 5.3.1, and 6.3.1 used for established processing background, not as a finalized 6G procedure.
• Performance limits of conventional and widely linear DFT-precoded-OFDM receivers in wideband frequency-selective channels : Kiran Kuchi, EURASIP Journal on Wireless Communications and Networking, 2014, article 159. Introduction and receiver discussion used for equalization background; no numerical result generalized to 6G.

Sensing waveform and frame sources

The sensing sources cover sequence design, waveform generation, ambiguity evaluation, and frame timing. Their proposals include different transmission modes and receiver assumptions, so performance results should be read with their stated conditions.

• R1-2602985 (RAN1#124bis) : Summary #5 on waveform and frame structure for 6G ISAC (LG Electronics)
• R1-2604765 (RAN1#125) : Summary #3 on waveform and frame structure for 6G ISAC (Moderator (LGE))
• R1-2606567 (RAN1#126) : Summary #5 on waveform and frame structure for 6G ISAC (Moderator (LGE))
• R1-2607084 (RAN1#126bis) : On waveform for sensing in 6GR (Nokia)
• R1-2607279 (RAN1#126bis) : Discussion on waveform and frame structure for sensing (OPPO)
• R1-2607483 (RAN1#126bis) : Waveform and signal for 6GR ISAC (Huawei, HiSilicon)
• R1-2607590 (RAN1#126bis) : Discussion on waveform and frame structure (Samsung)
• R1-2607910 (RAN1#126bis) : Discussion on sensing waveform and frame structure (MediaTek Inc.)
• R1-2607989 (RAN1#126bis) : Support pulse and chirp signal for ISAC (NERCDTV, CMCC, China Telecom, China Unicom, Turk Telekom, ABS, ABP, Pengcheng Laboratory, VIAVI, Tongji University, Shanghai Jiao Tong University, CATT, Qualcomm, ZTE)
• R1-2607998 (RAN1#126bis) : Summary #5 on waveform and frame structure for 6G ISAC (Moderator (LGE))
• R1-2608243 (RAN1#126bis) : On ISAC Waveforms and Frame Structure (Apple)
• R1-2608313 (RAN1#126bis) : Discussion on ISAC waveform and frame structure (Ericsson)
• R1-2608369 (RAN1#126bis) : Discussion on Waveform and frame structure for ISAC (NTT DOCOMO INC.)
• R1-2608440 (RAN1#126bis) : Waveforms for sensing in 6G (Qualcomm Incorporated)