3GPP TSG RAN WG1 #120-bis    											              R1-2501749 Wuhan, China April 7th – 11th, 2025

Agenda Item:		9.8.1	
Source:				InterDigital, Inc.
Title:						Remaining Issues on Evaluation of FR3 Channel Modeling
Document for:		Discussion and Decision

Conclusions
In this contribution, we shared and discussed our views for Rel-19 FR3 NF channel modeling. Based on presented discussions, following observations and proposals are made,

Observation 1: For calibration of near field channel modeling at 15 GHz, support of a larger array of antennas at the BS is missing.

Proposal 1: Remove FFS in the following tables,
Table 7.8-7: Simulation assumptions for calibration for near field channel modeling

Table 7.8-8: Simulation assumptions for calibration for BS side spatial non-stationarity

Observation 2: A table containing the scale factors for a finite range of desired angular spreads would mean generating multiple tables which would introduce unnecessary complexity instead of applying a single linear equation which is common to all CDL profiles to generate the optimal scaling factor.     

Proposal 2: Do not introduce a table for selecting the optimal scaling factor for each desired angular spread.

Observation 3: Based on the conducted measurements in [3]-[4] and analysis [5], the level of cross-polarization interference can very according to the UE location and channel delay profile of a channel. 

Proposal 3: Support consideration of the proposed modeling of random power variability in each polarization.

3GPP TSG RAN WG1 #120bis                                                           R1-2501819
WuHan, China, April 7th – 11st, 2025

Source:	vivo, BUPT
Title:	Views on channel model validation of TR38.901 for 7-24GHz
Agenda Item:	9.8.1
Document for:	Discussion and Decision

Conclusions
In this contribution, we have expressed our views on the channel model validation of TR38.901 using measurements at least for 7-24 GHz. The observations and proposals are summarized as follows.
Observation 1: The average angle range corresponding to parameters  and  of each antenna on the smart user equipment are 115 and 85 degrees, respectively, based on the measurement results.
Observation 2:  The average values of each antenna element corresponding to  and  are 22dB and 26dB, respectively, based on the measurement result.
Observation 3: There are many misalignments and required updates for the table in the WA on the computation of intermediate angle values.
Observation 4: The calculation of the mean angle of the LOS channel model is not considered in the equation specified in Annex A.2 of TR38.901.

Proposal 1: RAN1 studies the impact of channel sparsity on the existing channel model based on the experiment result.
Proposal 2: The orientation angle value should be defined for each antenna placement in the new UE antenna model for calibration and further evaluation and Table 1 is introduced.
Proposal 3: 115 and 85 degrees can be used to define   and  in the new UE antenna model.
Proposal 4: The value of 25dB can be selected for both  and .
Proposal 5: Max transmission capability for UE is up to 2TX at least for FR1.
Proposal 6: Confirm the working assumption with the following update.

Proposal 7: Support the update of the power weighted mean angle for CDL-D/E channel model.

Proposal 8: RAN1 defines the upper bound of absolute delay in indoor office scenario as  where  is the largest dimension of the office hall while no upper bound is defined in UMa and UMi scenario.
Proposal 9: RAN1 should replace the existing parameters values for any new channel modelling parameter updating and the frequency continuity can be ensured by the new fitting mathematical model based on the new measurement results and the previous measurement results.

3GPP TSG RAN WG1 #120bis	R1-2501897
Wuhan, China, April 7th – 11th, 2025

Title 	:  Discussion on the channel model validation for 7-24 GHz
Source 	:  ZTE Corporation, Sanechips
Agenda item	:  9.8.1
Document for:  Discussion

Conclusion
In this contribution, we provide our analysis and proposals for Channel model validation of TR38.901 for 7-24 GHz。
Observation 1: The power imbalance of co-polarization and cross-polarization will be eliminated under  polarization slant assumption.
Observation 2: The polarization variability powers, if needed, should be generated in cluster level and needs to be normalized to avoid increasing cluster power.
Observation 3: Similar RANK can be achieved with and without adding polarization power variability for both H-V and  polarization with the updated approach.
Observation 4: For the concrete penetration loss model, compared to Option 1, when the frequency is in the millimeter-wave band and continues to increase, the penetration loss of Option 2 becomes significantly greater than that of electromagnetic calculation (EMC).
Observation 5: The penetration loss of IRR/Low-e glass is influenced not only by the thickness of the glass layers, but also more significantly by the thickness of the internal metal-coating.
Observation 6: For the IRR glass penetration loss model, Option 3 is not feasible because its penetration loss does not primarily depend on the thickness of the IRR glass. 
Observation 7: The fitted curve of delay spread based on mixed data set by combining new measurement results and existing data set across frequencies 0.5~100GHz can match the delay spread curve in TR 38.901.
Proposal 1: For LOS probability of SMa scenario, adopt the following Option 3:

Proposal 2: For the pathloss of SMa scenario, the working assumption can be revised as below to include the pathloss formula for dBP < d < 5000 m
Proposal 3:  Adopt the following absolute delay parameters for InH scenario

Proposal 4:  Introduce absolute delay parameters for RMa scenario with following parameters
Proposal 5: No need to change the polarization matrix in the channel realization formula in TR 38.901.
Proposal 6: For the concrete penetration loss model, either Option 1 or the updated Option 2 can be down-selected.
Proposal 7: For the wood penetration loss model, either Option 1 or the updated Option 2 can be down-selected.
Proposal 8: For the std glass penetration loss model, either Option 1, the updated Option 2 or the updated Option 3 can be down-selected.
Proposal 9: For the IRR glass penetration loss model, either Option 1 or Option 2 can be down-selected.
Proposal 10: To address the parameters validation and model updates, the mixed data set based approach should be considered, which is constructed by combining new measurement results and data set representing the existing model, which is obtained by one of following options:
Option-1: The original data source is collected;
Opiton-2: Regenerating the data from the existing model.
Proposal 11: For UE antenna radiation pattern modeling, the max directional gain of an antenna element should be well selected to ensure the antenna efficiency not exceeding 0dB.
Proposal 12: For angle scaling, confirm the working assumption made in RAN1#119.
Proposal 13: If the direction of the LOS ray in CDL-D and CDL-E needs to remain unchanged before and after the scaling of angles, the following formulas can be considered:
,

Proposal 14: To avoid misunderstanding on the granularity of angle scaling, the following text proposal can be adopted:

3GPP TSG-RAN WG1 Meeting #120-bis	R1- 2501937
Wuhan, China, April 7th – April 11th, 2025

Agenda Item:	9.8.1
Source:	NVIDIA
Title:	Channel model validation of TR 38.901 for 7-24 GHz
Document for:	Discussion
1	

Conclusion
To check and review the following results and measurement data provided in RAN1 #120 for further discussion in next RAN1 meeting. R1-2501426 contains the list of data sources for the results and measurements provided since RAN1 #116-bis until RAN1 #120.
Agreement
Adopt the following absolute delay parameters for InH scenarios.
Down-select between option 1, 2 and 3.
Table 7.6.9-1: Parameters for the absolute time of arrival model

Working Assumption
Intermediate angle values of the CDL channel is computed by 

where s is a scale factor chosen to change the distribution of the angles. Table X shows required scale factor for typical desired angular spread values for AOD, AOD, ZOA, and ZOD.

Table X: Scale factor values for each CDL model
Note: the values are computed by formula option 2A defined in RAN1#119.

Agreement
Further study updates to the mean angle calculation in Annex used for calculation of the mean angle values for CDL-D/E channel model
Potential update of the power weighted mean angle is

Agreement
Discuss further on the following suggested text changes:

Agreement
RAN1 has identified that standard multi-panel glass penetration loss model may require updates at least for 6-24 GHz frequency range. Further discuss changes.
For standard multi-panel glass penetration loss model, study the following model changes:
Option 1) thickness dependent model without resonation effect

where 
Default reference thickness for standard multi-panel glass is 3 cm
If model without resonation effect is selected, add note in TR that while glass material measurement may showcase resonation effect as a function of frequency and thickness, the model abstracts away the effect.
Option 2) thickness dependent model with resonation effect

For , 
where 
Default reference thickness for standard multi-panel glass is 3 cm
Option 3) thickness dependent model without resonation effect and with air interface loss
, 
Where X is the default thickness of the glass.
Parameters a, b, and X are FFS.
Note: for all options, consider the impact to overall building exterior penetration loss from the potential changes to glass penetration loss.


Agreement
RAN1 has identified that IRR glass penetration loss model may require updates at least for 6-24 GHz frequency range. Further discuss changes For IRR glass penetration loss model, study the following model changes:
Option 1) coating dependent model without resonation effect

For 
where  for single silver-coating;  for double silver-coating;  for triple silver-coating.
Default reference coating for standard multi-panel glass is single coating, i.e. 
Default reference thickness for IRR glass is 3 cm
If model without resonation effect is selected, add note in TR that while glass material measurement may showcase resonation effect as a function of frequency and thickness, the model abstracts away the effect.
Option 2) coating dependent model with resonation effect

For 
where  for single silver-coating;  for double silver-coating;  for triple silver-coating.
Default reference coating for IRR glass is single coating, i.e. 
Option 3) thickness dependent model without resonation effect and with air interface loss
, 
Where X is the default thickness of the glass.
Parameters a, b, and X are FFS.
Note: for all options, consider the impact to overall building exterior penetration loss from the potential changes to glass penetration loss.
FFS: whether and how to incorporate thickness dependent affect for the IRR model. In this context, incorporation of thickness impact is considered a minor refinement of option 1 or 2 to better match the measurement data.

Agreement
Discuss further on potential update of concrete penetration model based on measurement provided by companies.
For updates of the concrete penetration loss model, further study between the following model changes:
Option 1) model with no separate air to interface insertion loss


Default reference thickness for concrete is 23 cm
Option 2) model with separate air to interface insertion loss


Default reference thickness for concrete is 23 cm
Note: for all options, consider the impact to overall building exterior penetration loss from the potential changes to concrete penetration loss.

Agreement
Discuss further on potential introduction of new material mode “cinder block” penetration model based on measurement provided by companies for concrete based walls.

Agreement
Discuss further on potential update of wood penetration loss model based on measurement provided by companies. The following is an example of a potential update:
Option 1) thickness dependent model without air interface loss


Default reference thickness for wood is 6 cm
Option 2) thickness dependent model with air interface loss

 
Where X is the default thickness of the wood.
Parameters a, b, and X are FFS.
Agreement
Discuss further on whether to introduce additional model to handle LOS probability impact from vegetation for SMa and details of the additional modeling component.
For suburban scenario, the down-select among the following LOS probability
Option 1)


Option 2)










Option 3)


Working Assumption
For parameters with multiple options, choose from one of the options in meeting RAN1#120-bis.
For information: O2I are barrowed from TR38.901 UMa O2I, LOS and NLOS parameters are barrowed from ITU M.2135 SMa if values are available. For colored values listed, values are based on measurement data from companies.


Agreement
Further study on introduction of a loss outdoor-to-indoor (O2I) building penetration loss model for SMa:

Consider the impact to overall building exterior penetration loss

Working Assumption
Corresponding Working Assumption (made in RAN1#119) is dropped.
Agreement
For UE antenna modeling of handheld devices, at least support directional antenna radiation pattern for calibration purposes.
Further study the following directional radiation pattern parameters

Agreement
RAN1 has identified that angular spread for following scenarios require necessary updates at least for 6-24 GHz frequency range. Further discuss necessary changes. Note: For UMa ASD, weak dependency to frequency was observed and the potential necessary changes may include changes to values across wide frequency ranges.
UMi LOS ASA
UMi NLOS ASA
UMa LOS ASD
UMa LOS ASA
UMa LOS ASD
UMa NLOS ASD
UMa NLOS ASA
UMa NLOS ZSA

Agreement
Conclude in RAN1 #120bis among the following:
Alt 1)
RAN1 has identified that delay spread for UMi LOS/NLOS and UMa LOS/NLOS scenario is smaller compared to delay spread in the TR at least for 6-24 GHz frequency range. Further discuss, necessary changes for UMi LOS/NLOS and UMa LOS/NLOS
Proponent companies to provide detailed information on potentially necessary changes to DS for scenarios in question.
Data samples for discussion on necessary changes for UMi LOS
Alt 2)
For the following scenarios, there is no consensus to update delay spread models due to lack of consistent and significant observed difference between model and measurements.
UMi LOS/NLOS and UMa LOS/NLOS

Agreement
Conclude in RAN1 #120bis among the following:
Alt 1)
RAN1 has identified that number of clusters spread for [InH LOS/NLOS, UMi LOS/NLOS, and UMa LOS/NLOS] scenario is smaller compared to delay spread in the TR at least for 6-24 GHz frequency range. Further discuss, necessary changes for [InH LOS/NLOS, UMi LOS/NLOS, and UMa LOS/NLOS].
RAN1 to determine which scenarios among [InH LOS/NLOS, UMi LOS/NLOS, and UMa LOS/NLOS] requires necessary changes.
Data samples for discussion on necessary changes

Proponent companies to provide detailed information on potentially necessary changes to number of clusters for scenarios in question.
Alt 2)
For the following scenarios, while sources have observed smaller number of clusters for various deployment scenario from measurement taken for 6 – 24 GHz frequency range, due to impact on frequency continuity outside 6 – 24 GHz frequency range, RAN1 concludes that there is no census to update number of clusters for all existing scenarios.

Agreement
Conclude in RAN1 #120bis to either (1) introduce an optional modeling component for polarization variability for each cluster for NLOS component of the channel or (2) no consensus to introduce polarization variability for each cluster for NLOS component of the channel.

3GPP TSG RAN WG1 #120bis		                                                      R1-2502004
Wuhan, China, April 7th – 11st, 2025

Source:	CATT
Title:	On channel model validation for 7-24GHz
Agenda Item:	9.8.1
Document for:	Discussion and Decision

Conclusions
In this contribution, remaining aspects of channel model involved in the validation for 7-24GHz are discussed. Based on the discussion, the following proposals are made:
Proposal 1: For Suburban scenario, the following option is supported for ISD assumption for calibration purposes:
Option 1) ISD = 1299 m.
Proposal 2: Update the number of clusters for 7-24 GHz.
Proposal 3: Do not introduce intra-cluster power profile for 7-24 GHz.

3GPP TSG RAN WG1 Meeting #120bis                                                 	R1-2502186
Wuhan, China, 7 April – 11 April, 2025
Source:           BUPT, Spark NZ Ltd, X-Net
Title:              Discussion on channel model validation of TR38.901 for 7-24 GHz
Agenda item:    	9.8.1
Document for:  	Discussion and Decision

Conclusion
In this contribution, we provide our views on the validation and the details for Rel-19 channel modeling enhancements for 7-24 GHz. Key consideration is the modification of the cluster structure. The observation and proposals are as follows:
Observation 1: The cluster structure described in Section 7.5 and 7.6 of 38901 and Example 1 underestimate channel sparsity. Among these approaches, the Section 7.6 methodology demonstrates better performance than Example 1, which in turn outperforms the Section 7.5 implementation. Only Example 3 yields result that closely align with the measurements.
Observation 2: Compared to the 3GPP model, the ICP model characterizes a few but strong eigenvalues, i.e., the ICP model characterizes sparsity more accurately.
Proposal 1: It is proposed to adopt the Example 3 method in Section 7.6 to modify the cluster structure, as it provides results that best match the measured data. This approach introduces the Intra-cluster power factor (ICP) to effectively model dominant rays. The ICP is introduced in section 7.6 as shown below:
To characterize the channel sparsity, the Intra-cluster power factor (ICP) is introduced. The ICP is applied to the generation of cluster powers in Step 6 in Subclause 7.5 to generate the ray power. The intra-cluster power factor () is modelled as:

where
-  represents the normalized cluster delay generated by Step 5.
-  represents the delay spread generated by Step 4.
-  and are the coefficients related to frequency and scenario.
The power of rays within a cluster is determined by the ICP, expressed as

3GPP TSG-RAN WG1 Meeting #120-bis                                                                                 	   R1-2502217
Wuhan, China, April 7-11, 2025

Agenda Item:	9.8.1
Source:	Huawei, HiSilicon
Title:	Considerations on the 7-24GHz channel model validation 
Document for:	Discussion and Decision

Conclusions
In this contribution, we introduce our parameters validation and update, calibration settings and initial results, several important scenarios, and other controversial aspects. Following observations and proposals are given:
Observation 1: Based on companies’ measurements, the measured delay spread under UMa LOS/NLOS and UMi LOS scenarios is smaller than that in TR 38.901.
Observation 2: Based on companies’ measurements, the measured number of clusters under UMa, UMi and InH scenarios is smaller than that in TR 38.901.
Observation 3: Updating the penetration losses of materials seems not that meaningful.
Observation 4: At least some fast fading parameters of low-altitude airspace scenario are altitude-dependent.
Observation 5: The fast fading parameters of low-altitude airspace scenario differ from that of UMa in TR 38.901/UMa-AV in TR 36.777.
Observation 6: The median intra-cluster power ratios are -1.1 dB and -2.5 dB for UMi LOS and NLOS scenarios, respectively, which means the power of the strongest ray in a cluster is not as dominant as the LOS path.

Proposal 1: Support Alt 1) to update the delay spread under UMa and UMi scenarios for 6-24 GHz.
Proposal 2: Support Alt 1) to update the number of clusters under UMa, UMi and InH scenarios for 6-24 GHz. 
Proposal 3: Update the to-be-updated fast fading parameters only for 6-24 GHz based on the data samples collected in Rel-19.
Proposal 4: Consider Equation (1) if pathloss and shadowing fading models are decided to be updated.
Proposal 5: Support Config C as a candidate UT antenna configuration for adequate calibration purpose.
Proposal 6: Capture 200-500 m ISD for UMa in section 7.2 of TR 38.901.
Proposal 7: Support to introduce low-altitude airspace scenario with:
Aerial UTs vertically ranging from 0 to 600 m.
Altitude-dependent fast fading parameters.
Proposal 8: For absolute delay, adopt Option 1 for InH scenario. 

3GPP TSG RAN WG1 #120bis	        	R1- 2502327
Wuhan, China, April 7th – 11th, 2025

Agenda Item:					    9.8.1
Source:	Sony
Title:	Further discussion of channel model validation of TR38.901 for 7–24GHz 
Document for:	Discussion and decision

Conclusions
We made the following observations and proposals:
 For UE antenna modeling of handheld devices, the values of the parameters , , ,  and  should be determined based on measurements.
 Measurements of the antenna ports of a mockup smartphone over the frequency range of 14.5–15.5 GHz show directivities in the range from 4.5 to 6.5 dBi.
 For UE antenna radiation pattern modeling for 7–24 GHz, for directional antenna radiation pattern, the parameter values SLA_V=A_m=20 dB are in good agreement with measurements of radiation patterns of a smartphone mockup over the 14.5–15.5 GHz frequency range.
 For UE antenna radiation pattern modeling for 7–24 GHz, for directional antenna radiation pattern, for 0-dBi radiation efficiency is zero dBi, i.e., the antenna element is lossless, the half-power beam width parameters  are uniquely determined from ,  and .
 For UE antenna radiation pattern modeling for 7–24 GHz, option 3 can also be selected, in addition to option 1, if needed.
 Measurements show power imbalances between antenna ports of a smartphone mockup in the 14.5–15.5 GHz in free space of up to 1.5 dB.
 It is essential to model antenna power imbalances, at least for handheld devices, as these are observed in actual device implementations.
 To accurately model the total power imbalance between antenna ports, one must also consider imbalances between the components of the respective RF chains, such as different amplification gains or insertion losses.
 Power imbalances may arise between the uplink and downlink directions of the same antenna port due to different RF components being used in uplink and downlink directions, as well as different routing due to, e.g., TX antenna switching.
 For an indoor scenario, ray-tracing simulations could predict the stronger scattering, which reflects the multiplexing richness of the environment, reasonably well. However, ray tracing did not predict lower-level scattering well.
 The 3GPP TR 38.901 delay profile mode is valid in the 7–24 GHz frequency for the InH office scenario.
 Ray-tracing simulations could predict the major features of the power delay profile of InH office channels well.
 For UE antenna radiation pattern modeling for 7–24 GHz, for directional antenna radiation pattern, select the parameter values  dBi,  dB, and .
 For UE antenna modeling of handheld devices, companies are encouraged to share antenna power imbalance data obtained from actual devices. The power imbalance can be between antenna ports or between the UL and DL directions of the same antenna port
 For UE antenna modeling of handheld devices, introduce power imbalance modeling to the UE antenna model as follows:
Randomized loss among {0dB, 1dB, 2dB, 3dB} in free space are applied per UE antenna port.
Randomized loss among {0dB, 1dB, 2dB, 3dB} in free space can be applied independently for the UL and DL directions.
Note: Further losses can be applied during the working item phase to capture imbalances between a device’s RF chains and other hardware.
 The delay spread model in TR 38.901 for the 7–24 GHz frequency range can be considered valid, and no updates are needed.

3GPP TSG RAN WG1 #120bis	R1-2502339
Wuhan, China, Apr. 7th – 11th, 2025

Source:	Lenovo
Title:	Channel model validation for 7-24 GHz
Agenda Item:	9.8.1
Document for:	Discussion and Decision
Discussion
In this document, we discuss further remaining aspects on channel validation for 7-24 GHz based on the agreements reached in RAN1#120  as well as prior RAN WG1 meetings, which are planned to be captured in TR 38.901 .
In RAN1#119 , it was proposed to introduce unequal powers per ray within a cluster. It was also proposed to maintain an unequal angle offset among rays to maintain the power angular spectrum per cluster. Furthermore, three alternatives were proposed in RAN1#120 regarding the support of unequal ray power ratios, in addition to updating the offset angles per ray, as follows: 
Alt 1)




Alt 2)

Alt 3)

In our opinion, introducing unequal powers per ray may complicates the design, since in practice the paths can be perceived as a continuum of cluttered components of the channel, wherein the simulation assumptions adopted in TR 38.901 are intended to capture the dominant paths that resonate around a cluster angle. Given that, unless it is observed via field tests by companies that the power variation across rays is richly distributed, i.e., the rays cannot be grouped to dominant rays with equal power and negligible rays with infinitesimal power, our first preference would be to maintain the current setup of equal powers across rays per cluster. Regarding the three alternatives above, we do not support Alt2, which introduces an additional ray with no phase offset, whose implications on the channel model are unclear to us. Alt3 would break the symmetry via assigning unequal phase offset values for the strongest two rays, leading to a shift in the average cluster angle. We are open to Alt1, which includes updated ray offset values that do not deviate from the legacy considerations for ray phase offset values, and whose ray power ratio is more straightforward compared with Alt2 and Alt3.  
For the intra-cluster ray power ratios and phase offset angles, support:
First preference: Legacy design; equal ray power within a cluster and no change to the current ray offset phase values.
Second preference: Alt1 (corresponding to Example 1) with 20 rays, symmetric phase offset values across rays and cluster-common ray power ratios.
Furthermore, in case any of the three alternatives above is supported, it was discussed whether the new table would replace the current table or be added as an optional modeling component. On one hand, having multiple alternatives for the same parameter set is inefficient, and the optional parameter set usually ends up being unused in most simulation evaluations. On the other hand, we believe some alternatives (specifically Alt2 and Alt3) deviate from legacy channel modeling principles and hence if supported, the legacy parameter set for ray powers and phase offset values should still be captured. Given that, our preference would depend on the selected alternative: if Alt1 is supported, we are fine to replace the legacy ray power and angle parameter set with that in Table 1, whereas if Alt2 or Alt3 are supported, we prefer them to be optional modeling components. Therefore, we propose the following:
Regarding whether the updated ray power/phase offset values within a cluster are supported to replace legacy model or be captured as additional modeling component:
If Alt1 (corresponding to Table 1) is supported, capture as a basic modeling component that replaces legacy approach in TR 38.901.
If Alt2 or Alt3 (corresponding to Example 2 or Example 3, respectively) are supported, capture as optional modeling component along with the legacy approach in TR 38.901.
The issue of device dimension and antenna placement applicability for CPE devices was revisited in RAN1#119. Although a CPE device may have a 3D cubical shape with comparable dimension values, the CPE can geometrically be modeled as a 2D planar device from the perspective of the gNB-to-CPE link under far-field channel assumptions. Hence, for channel modeling purposes, it suffices to model the CPE as a 2D device with equal length, width, with larger dimensions than a handheld UE, and ignoring the breadth, e.g., 30 cm x 30 cm x 0 cm. 
For channel modeling purposes, the CPE is modeled as a 2D device with equal length, width, with larger dimensions than a handheld UE, and ignored breadth, e.g., 30 cm x 30 cm x 0 cm.
It was also discussed in RAN1#120 whether to introduce an optional modeling component for polarization variability for each cluster for NLOS component of the channel. It is not clear how the polarization variability distribution would be modeled, and whether values across different clusters would be correlated. In general, this component can be helpful for modeling impairments in UL transmission causing power imbalance across two cross-polarized ports, however the variability would then be correlated across the clusters. To summarize, we are open to consider this polarization variability as an additional modeling component, however we believe more discussion is needed on the underlying distribution, as well as the correlation of the values across the clusters.
Polarization variability is considered as an optional modeling component per NLoS cluster for 7-24 GHz channel model, with further discussion on the underlying distribution and the correlation of the values across clusters.
An email discussion was initiated following RAN1#120 to discuss the simulation calibration assumptions for 7-24 GHz channel model validation, including computing the coupling loss and geometry. Since both metrics are computed for LoS components of the channel, the frequency response is expected to be flat. Hence, the BW specification for LoS channel metrics is unnecessary. 
For simulation calibration assumptions for 7-24 GHz channel model, bandwidth specification for LoS channel metrics, e.g., LoS-based coupling loss and geometry, can be dropped.
On the other hand, for metrics including NLoS channels, the BW specification in the calibration is important. However, there needs to be clarity on the window size for each measurement. For example, when computing the CDF of the coupling loss over a 200 MHz BW, the number of samples taken, e.g., whether fixed for a given BW or fixed to be every K RBs, etc., needs to be specified. The sampling of the points over the BW needs to be done in a unified way. One suggestion below is as follows:
For 6/7 GHz, assume SCS 30 kHz, and one sample /4 RBs for 20 MHz (13 samples) and one sample/40 RBs for 200 MHz (13 samples). 
The metric, e.g., coupling loss is the average of the coupling loss over the 13 samples within the BW.
We do not have strong opinion on the numerology or simulation assumptions, however the sampling assumptions over the BW need to be unified across simulations. We also do not have strong opinion towards adding this to the TR; it is OK to include it as an agreement or note without being added explicitly to the calibration table. 
For simulation calibration assumptions for 7-24 GHz channel model, for BW specification for channel metrics comprising NLoS channels, the sampling method of the points constituting the CDF needs to be specified, e.g., whether each point in the CDF plot is based on an average over multiple pre-defined subband size, per RB, or otherwise.
Conclusion
This contribution studied aspects on channel model validation for 7-24 GHz. We have the following proposals:
For the intra-cluster ray power ratios and phase offset angles, support:
First preference: Legacy design; equal ray power within a cluster and no change to the current ray offset phase values.
Second preference: Alt1 (corresponding to Example 1) with 20 rays, symmetric phase offset values across rays and cluster-common ray power ratios.
Regarding whether the updated ray power/phase offset values within a cluster are supported to replace legacy model or be captured as additional modeling component:
If Alt1 (corresponding to Table 1) is supported, capture as a basic modeling component that replaces legacy approach in TR 38.901.
If Alt2 or Alt3 (corresponding to Example 2 or Example 3, respectively) are supported, capture as optional modeling component along with the legacy approach in TR 38.901.
For channel modeling purposes, the CPE is modeled as a 2D device with equal length, width, with larger dimensions than a handheld UE, and ignored breadth, e.g., 30 cm x 30 cm x 0 cm.
Polarization variability is considered as an optional modeling component per NLoS cluster for 7-24 GHz channel model, with further discussion on the underlying distribution and the correlation of the values across clusters.
For simulation calibration assumptions for 7-24 GHz channel model, bandwidth specification for LoS channel metrics, e.g., LoS-based coupling loss and geometry, can be dropped.
For simulation calibration assumptions for 7-24 GHz channel model, for BW specification for channel metrics comprising NLoS channels, the sampling method of the points constituting the CDF needs to be specified, e.g., whether each point in the CDF plot is based on an average over multiple pre-defined subband size, per RB, or otherwise.
RAN1#120 Agreements/Conclusions
The following has been agreed in RAN1#120 for the 7-24 GHz frequency range channel model validation study:

TDoc file conclusion not found

3GPP TSG RAN WG1 Meeting #120bis	R1-2502341
Wuhan, China, April 7th – 11st, 2025

Source:	Intel Corporation
Title:	Discussion on channel modeling verification for 7-24 GHz
Agenda item:	9.8.1
Document for:	Discussion

Conclusions
In this contribution, we discussed the potential aspects that may require channel modeling validation effort for frequencies from 7 to 24 GHz. The following is a summary of proposals and observations made in this contribution.
Proposal 1:
Support updates to glass, concrete, and wood penetration model.
Use models without penetration loss resonance effect for glass penetration models.
RAN1 to conclude on observing glass penetration loss resonance effect but conclude not to model the effect.
Proposal 2:
Support introduction of cinder block penetration loss model based on concrete measurements from companies.

Proposal 3:
Update the SMa LOS Pathloss as
For 10m ≤ d < ,


For   ≤ d < 5000m,
 

  is the center frequency in Hz, m/s

Proposal 4:
After cross-checking, we suggest updating the working assumption for CDL angle tables as follows:


Proposal 5:
Update the angle spread and mean angle calculation for CDL-D and CDL-E cases as follows:


Proposal 6:
Do not update “CDL ray” to “CDL cluster” in Section 7.7.5.1 of TR38.901.

3GPP TSG-RAN WG1 Meeting #120bis		R1-2502380
Wuhan, China, April 7th – 11th, 2025

Agenda item:	9.8.1 
Title:	Discussion on channel model validation of TR38.901 for 7-24 GHz 
Source:	Samsung
Document for:	Discussion and Decision

Conclusion
The marginal difference compared to pathloss for UMi scenario in existing channel model is confirmed
RAN1 discuss whether the updates of pathloss in UMi scenario is needed
As shown in below table, in the UMi scenario, the mean delay spread in the LOS environment was found to be lower compared to the existing channel model parameter
Even in a visually LOS environment between Tx and Rx, the delay spread can vary depending on the surrounding environment. In particular, if an update of delay spread in the LOS environment is required, it will be necessary to collect and analyze sufficient measurement data from various LOS environment to reflect actual conditions
 RAN1 discuss whether the updates of delay spread in UMi LOS scenario is needed
As shown in below table, slight difference compared to mean and standard deviation of ASD for UMi LOS scenario in the existing channel model were confirmed


As shown in below table, significant difference compared to mean and standard deviation of ASA for UMi scenario in the existing channel model were confirmed

RAN1 discuss whether the updates of azimuth spread of departure/arrival angles in UMi LOS scenario is needed
Currently available channel parameters reflect stochastic properties of measurement results focused on both Below 6 GHz band around 3.5 GHz and mmWave bands around 28 GHz. New measurements across 7 – 24 GHz band may exhibit nonlinear traits inconsistent with prior values
RAN1 considers replacement method for channel parameter values/model equations by aggregating old measurement results and new measurement results, extracting new statistical outcomes
Upon plotting the pathloss across distance per frequency bands utilizing these equations, it was observed that beyond a particular distance (i.e., breakpoint), the values tend to converge 
In the case of a UE located at ground level, (effective environment height) is set to 1 m. As a result, the breakpoint distance becomes a dominant parameter for frequency rather than the height of the UE. In other words, as the frequency increases,  also increase, and at the same time, the correction term for  in UMa increases, so the absolute value change of  decreases
Pathloss comparison with long distance will mainly be used for link budget calculation or coverage analysis over frequency band rather than system-level simulation.
One alternative is to set the  (effective environment height) to 0 m. The meaning of  being 0 m is to assume reflection at a completely flat ground level and another alternative is to consider using only , which does not take into account the breakpoint
If needed, RAN1 discuss the methods with two alternatives: one is setting  to 0 m and another one is applying only PL1 in the case of link budget and/or coverage analysis.
Since coupling loss and geometry are not implying the bandwidth relevant parameters, the exact values will not be required
RAN1 consider the 20 MHz bandwidth from the perspective for alignment with the existing assumption
SMa has a BS located above the surrounding environment to allow wide area coverage, and has a high BS height. In addition, the environment constituting the scenario is dominated by residential buildings
RAN1 consider a lower BS antenna down-tilting value than the 102 degrees for SMa scenario
RAN1 consider a low-loss O2I penetration model for SMa scenario
RAN1 prioritize the basic assumption and discuss the necessity of optional and FFS parts before deciding the assumption for full calibration
Some readers of channel model might want to know the difference in performance between having the FR1 antenna at the edge versus the corner, which could also apply to the FR2 antenna array. Additionally, whether antenna for new frequency band become antenna elements or arrays remains unknown.
RAN1 do not consider the detailed guidelines for how antenna elements and antenna modules are mapped to antenna location candidates
With varying the value of HPBW, it tends that the larger the HPBW becomes, the weaker the directional properties get in both horizontal and vertical planes. 
With varying  and _while maintaining HPBW, the lower bound boundary increases in both horizontal and vertical planes.
RAN1 consider Option1 as a baseline for UE antenna modelling with the values of parameter values as below:
 90 degrees
 25 dB
 25 dB
 5 dBi
Combining method for 3D antenna element pattern: same with TR38.901
UE antenna modelling in existing channel model sets a bearing angle for the rotation of the UE. It is necessary to confirm that each antenna elements for the UE applies the same bearing angle
RAN1 confirm that maximal gain directivity of the antenna radiation pattern to be aligned with direction of the antenna candidate location
RAN1 consider that same bearing angle is applied to all antenna element candidates
Power imbalance arises due to device hardware characteristics and is reflected in the air interface. To incorporate this characteristic accurately, specific values would be necessary. Such studies should be conducted by another working group rather than RAN1
RAN1 deprioritize to introduce power imbalance modelling

3GPP TSG RAN WG1 #120bis			R1-2502625
Wuhan, China, April 7th – 11st, 2025
 
Agenda Item:	9.8.1
Source:	Apple
Title:	Validation of Channel Model
Document for:	Discussion/Decision

Conclusion
In this contribution, we provided our simulation results for channel model validation of TR 38.901 for 7-24 GHz and our views on suburban scenario definition. Our observations and proposals are as follows:

Proposal 1: Include “Vegetation is modest” in the description of suburban scenario.

Proposal 2: For suburban scenario calibration purpose, support ISD of 1299 m. 

Proposal 3: For SMa, the LOS probability is given by . 

Proposal 4: For SMa, the O2I penetration loss for residential buildings is given by . 

Proposal 5: Confirm the working assumption on SMa pathloss model, without the potential changes.

Proposal 6: For SMa NLOS, the delay spread (in logarithm) mean value is -6.98 (Option 3) and the delay spread (in logarithm) standard deviation is 0.74 (Option 3). 

Proposal 7: For SMa NLOS, 
AoA spread (in logarithm) mean value is 1.27 (Option 3) and AoA spread (in logarithm) standard deviation is 0.86 (Option 3).
AoD spread (in logarithm) mean value is 0.86 and AoD spread (in logarithm) standard deviation is 1.
ZoA spread (in logarithm) mean value is -0.388 (Option 2) and ZoA spread (in logarithm) standard deviation is 1.17 (Option 2). 

Observation 1: The measured penetration loss of glass does not linearly increase with frequency. 

Proposal 8: The glass penetration loss model is given by 
,
where , with (vacuum permeability), (vacuum permittivity),  (angular frequency in Hz),  is material thickness in unit of meter and . 

Observation 2: The measured penetration loss of wood with thickness of 1.75 cm is up to 4 dB below TR 38.901 at frequencies of 7.5 GHz to 30 GHz. 

Proposal 9: The wood penetration loss model is given by 
,
where .

Observation 3: The measured penetration loss of concrete with thickness of 4.7 cm is between 30 dB and 100 dB below TR 38.901 at frequencies of 7.5 GH to 30 GHz.

Proposal 10: The concrete penetration loss model is given by 
,
where .

Observation 4: The measured penetration loss of IRR glass of one layer with thickness of 0.55 cm is between 4 dB and 9 dB above TR 38.901 at frequencies of 7.5 GH to 30 GHz. 

Proposal 11: The IRR glass penetration loss model should take Option 1: , where  for single silver-coating,  for double silver-coating,  for triple silver-coating. 

Observation 5: For indoor office scenario at 13 GHz, the measured delay spread is approximately aligned with TR 38.901 with 25 dB dynamic range.

Proposal 12: The update on delay spread model for indoor office LOS/NLOS scenarios is not needed. 

Observation 6: The simulated delay spread for UMa NLOS scenario is smaller than TR 38.901 at frequency 8 GHz. 

Proposal 13: The delay spread for UMa NLOS scenario is to be reduced.

Observation 7: The measured delay spread for UMi LOS/NLOS scenarios is smaller than TR 38.901 at frequency 13 GHz. 

Proposal 14: The delay spread for UMi LOS/NLOS scenario is to be reduced.

Observation 8: The simulated AoA/AoD/ZoA spread for UMa NLOS scenario is smaller than TR 38.901 at frequency 8 GHz. 

Observation 9: The simulated AoA/AoD/ZoA spread for UMi NLOS scenario is smaller than TR 38.901 at frequency 8 GHz. 

Observation 10: The measured AoA/AoD/ZoA spread for UMi LOS scenario is smaller than TR 38.901 at frequency 13 GHz. 

Observation 11: For UMi NLOS scenario at 8 GHz, the number of clusters is 18 at 95% percentile. 

Proposal 16: For UE antenna radiation pattern modeling, support , , , .

3GPP TSG RAN WG1 #120	bis	             			            		R1-2502737
Wuhan, China, April 7th –11th, 2025						
Agenda Item:	9.8.1
Source:	AT&T
Title:	Discussion on Validation of the Channel Model in 38901
Document for:	Discussion/Decision

Conclusion
In this contribution, we discussed the validation of the channel model for 7-24GHz in TR 38.901. We made the following proposals and observations.


Observation 1: In TR38.901, the O2I parameters distributions in UMa and UMi channel models are the same. 

Observation 2: The SMa O2I channel parameters distributions are expected to be considerably different from UMa and UMi O2I parameters distributions.

Observation 3: In suburban macro deployments, vegetation varies across cities and states.  


Observation 4: For suburban scenario, the Pathloss model in ITU M.2135 model is a better match to the Pathloss model at BS height of 35m for NLoS environments than the pathloss model in 3GPP UMa in TR 38.901.

Observation 5: As can be seen from Table 3, the mean of the large scale DS for NLoS for 7, 8 and 15GHz at BS height of 35m more closely match those of the ITU model than the 3GPP UMa model in TR 38.901. 

Observation 6: The pathloss for LoS observed from the SMa measurements largely matches the pathloss model predicted by the formula expression in ITU M.2135-1 for hBS = 25m.

Observation 7: As can be observed from Table 4, the mean of the large scale DS for NLoS for 7, 8 and 15GHz at BS height of 25m is a better match to the mean of the DS in the ITU M.2135-1 model than the 3GPP UMa model in TR 38.901.


Observation 8: For the SMa channel model, the ITU model is a better fit for the large-scale path loss and delay spread parameters than the 3GPP UMa model in TR 38.901.

Observation 9: The ITU model with hBS=35 m is a better fit for the pooled pathloss model derived from the suburban macro scenario measurements

Observation 10: The ITU model is a better fit for the pathloss model for suburban model because it captures the height dependency.


Observation 11: The ITU model does not model ZSA. The ZSA model in the 3GPP UMa model can be considered to model the ZSA in the SMa channel model.  

Observation 12: The 3GPP UMa model underestimates the path loss in urban macrocell deployments.


Observation 13: The 3GPP UMi SCM model underestimates the NLoS path loss in urban microcell  deployments.

Observation 14: Measurements conducted at 7, 8, 11, and 15 GHz over 650 RX locations on multiple floors of an office building show that the PLE for LoS and NLoS environments agree with the previously proposed 3GPP SCM InH channel model.

Observation 15: Measurements conducted at 7, 8, 11, 15 GHz over 650 RX locations on floors of an office building show that shadow fading distribution for LoS and NLoS environments agree with the previously proposed the 3GPP SCM InH channel model. 

Observation 16: Measurements conducted at 7, 8, 11 and 15 GHz over 650 RX locations on floors of an office building show that the mean delay spread for LoS and NLoS environments agree with the 3GPP SCM InH channel model. 

Observation 17: Measurements conducted at 8GHz and 15 GHz 650 RX locations on floors of an office building show that the mean angular spread (ASA) for LoS and NLoS environments agree with the 3GPP SCM InH channel model. 

Observation 18: Measurements conducted at 8GHz and 15 GHz 650 RX locations on floors of an office building show that the mean angular spread (ZSA) for LoS and NLoS environments do not agree with the 3GPP SCM InH channel model. 
Observation 19: Measurements conducted at 7, 8, 11 and 15 GHz over 650 RX locations on floors of an office building show that the correlation between the ASA and the DS for NLoS at 8GHz and 15GHz is considerable, and not 0, as the current 3GPP InH model suggests

Observation 20: Measurements conducted at 7, 8, 11 and 15 GHz over 650 RX locations on floors of an office building show that the correlation between the ZSA and the SF for NLoS at 8GHz and 15GHz is considerable, and not 0, as the current 3GPP InH model suggests



Proposal 1: For the SMa channel model, reuse the NLoS parameters distributions shown in Table 1 as the O2I parameters distribution. 

 
Proposal 2: Adopt the following values for the channel model parameters for SMa 
For information: O2I is borrowed from NLoS parameter values, LOS and NLOS parameters are borrowed from ITU M.2135 SMa when measurement values are not available.
Note: O2I parameters are subject to change based on new measurement results 






Proposal 3: For suburban scenario, the following LOS probability is assumed



Proposal 4: for suburban scenario, additional model to handle LoS probability impact from vegetation for SMa is optionally included and used depending on the vegetation distribution.


Proposal 5: Consider low outdoor-to-indoor (O2I) building penetration loss model for SMa. 

Proposal 6: For suburban macro scenarios, adopt the pathloss model in ITU M.2135 as the pathloss model for 3GPP SMa channel model

Proposal 7: If needed, consider changing the breakpoint frequency for  the  frequency dependency for the NLOS pathloss model for SMa to 7GHz. 


Proposal 8: Use the following table for evaluation parameters in SMa.
Table: Evaluation parameters for SMa 

Note 1: The ISD values can be used for channel model calibration. The choice of ISD in the evaluation metrics beyond calibration, is made in the corresponding evaluation study/work items, e.g. Study item on IMT-2030 evaluations.  

Proposal 9: For SMa channel model calibration, consider the following base station electric antenna downtilt values:





Proposal 10: For suburban macro scenarios, adopt the parameters for the delay spread distribution in ITU M.2135 as the delay spread distribution parameters for the 3GPP SMa channel model.


Proposal 11: For suburban macro scenarios, adopt the parameters for the ASA and ZSA distributions in 3GPP UMa channel model in TR38.901 for the ASA and ZSA distributions in 3GPP SMa channel model.

Proposal 12: Consider modifying the PLE and the intercept for the 3GPP UMa pathloss model. 

Proposal 13: Do not update delay spread models between model and measurements for UMa LOS/NLOS

Proposal 14: Companies are encouraged to check the cross correlation parameters for ZSA and ASA with SF and DS in 3GPP InH channel model  

3GPP TSG RAN WG1 #120bis                                         R1-2502751 

Wuhan, China, April 7th – 11th, 2025

Source:	Sharp
Title:	Views on Channel model validation of TR38.901 for 7-24 GHz
Agenda Item:	9.8.1
Document for:	Discussion and Decision
Umi
Curve fittings for UMi DS, ASA LOS/NLOS Mean and Standard deviation (std) are present in R1-2502415_UMi_curve_fitting.pptx (Agenda Item 9.8).
UMa
Curve fittings for UMa DS, ASA, ASD LOS/NLOS Mean and Standard deviation (std) are present in R1-2502415_UMa_curve_fitting.pptx (Agenda Item 9.8)..
Measurement Data
Raw measurement data for UMi DS, ASA LOS/NLOS Mean/Std and UMa DS, ASA, ASD LOS/NLOS Mean/Std are present in R1-2502415_consolidated_meas_data_rel14_rel19.xlsx (Agenda Item 9.8).
O2I Building Penetration Loss Model for SMa
R1-2406139 conducted penetration loss measurements at 28 GHz in New Jersey Suburb. The details of the measurement and measured building penetration loss can be found in R1-2406139. 
As can be clearly seen from Fig 1. the O2I BPL measurements conducted at 28 GHz in SMa scenario is much lower than that estimated by current low loss BPL model in TR 38.901. The mean BPL estimated by the current TR 38.901 low loss model at 28 GHz is ~18 dB whereas the mean BPL estimated by the proposed SMa BPL is ~13 dB. The proposed O2I BPL for SMa scenario is shown in Table 1. The proposed O2I BPL model for SMa scenario in Table 1 (Blue dashed line in Fig 1 is the average of green, magenta and black lines in Fig. 1) accurately captures the average O2I BPL observed in SMa scenario from measurements conducted at 28 GHz. 


Fig. 1. CDFs of measured BPL for SMa scenario from R1-2406139 and comparison with proposed SMa BPL and existing TR 38.901 low loss BPL model.

Table 1. Proposed O2I building penetration loss model for SMa scenario

Observation 1: The proposed O2I BPL for SMa scenario in Table 1. accurately captures the average O2I BPL observed in SMa scenario from measurements conducted at 28 GHz. 

Proposal 1: RAN1 to introduce a low loss O2I building penetration loss model for SMa scenario as in Table 1.

TDoc file conclusion not found

3GPP TSG RAN WG1 #120-bis	R1-2502777
Wuhan, China, Apr 7th – 11th, 2025

Title 	:  Discussion on channel model validation for 7-24 GHz
Source 	:  NTT DOCOMO, INC.
Agenda item	:  9.8.1
Document for:  Discussion

Conclusion
In this contribution, we provided our views and discussions on channel model validation of TR38.901 for 7-24GHz related to suburban use cases (SMa) and UMa. The following proposals are made:
Proposal 1: For calibration purposes, the following modeling parameters of SMa scenario are supported:
Option 1) ISD = 1299 m

Proposal 2: For LOS probability of SMa scenario the following option should be used.
Option 3)


Proposal 3: The values in following table should be used for SMa scenario:




Proposal 4: The pathloss model of UMa in TR38.901 is validated and does not need to be updated.

Agenda item:	9.8.1
Source: 	Qualcomm Incorporated
Title: 	Channel model validation of TR 38.901 for 7-24GHz
Document for:	Discussion/Decision

Conclusion 
We make the following observations and proposals in this document:	
Observation 1: For SMa scenario, the proposed changes to the pathloss model in ITU M.2135-1 lead to more optimistic pathloss estimate (i.e., smaller pathloss values) for sub-10GHz frequency range compared to ITU M.2135-1, and a more conservative pathloss estimate (i.e., larger pathloss values) for the > 10 GHz frequency range. 

Proposal 1: Avoid ad hoc changes to the suburban pathloss model in ITU M.2135-1. Reuse the pathloss model in ITU M.2135-1 without any changes. 
Proposal 2: For SMa scenario, suggest using the LOS probability model in Option 1:


Proposal 3: For SMa scenario, an ISD of 1299m is preferred. Higher ISDs can be considered if admission control policies for out of coverage UEs are allowed.
Proposal 4: The radiation pattern of UE antennas can be modelled using the templated pattern provided in Table 7.3-1. The parameters for the antenna radiation pattern provided in Table 7.3-1 can be set as follows:
 = 25 dB
 = 25 dB
 = 180 to 270 
 = 90
Max gain = 6.53 dBi 
Maximum directional gain is chosen to normalize the antenna pattern to unit gain (0 dBi).
Antenna-specific efficiency is included as a separate parameter. 
Proposal 5: For UEs with 2, 3, and 6 antennas, the default antenna placements are given as follows:
 
Proposal 6: Model CPEs as having a cubical form factor with 15-20 cm sides. The antennas are placed along the central horizontal plane and are oriented outwards from the centre of the cube. The relative antenna locations are similar to that of a handheld device.
Proposal 7: To model potential differences in antenna gain/efficiency and variations in insertion loss across tx/rx ports, introduce an antenna imbalance factor (in dB) when enhancing the UE antenna modelling framework. 
Proposal 8: Support updating the penetration losses due to wood and concrete to be a function of thickness of the material. The losses (in dB) are modelled to scale linearly with thickness. Exact model is chosen based on the fit to measurement data.
Proposal 9: For penetration losses incurred due to IRR glass and standard glass, adopt a model that captures the average loss across different incident angles.
Observation 2: Ground reflection model in 38.901 offers a mode to realize polarization power imbalance in the channel realizations. 
Proposal 10: Due to lack of clarity on whether the observed imbalance across polarization is due to ground reflection or due to other effects, it is suggested that changes to introduce polarization imbalance be deprioritized. 
Proposal 11: Due to the large impact of any changes to the ray-cluster framework and impact to frequency continuity of the channel model, RAN1 strives to retain the existing ray-cluster framework.
Proposal 12: When updating parameters based on observations from Release 19 measurements, also take the Release 14 measurements into consideration. When determining the frequency dependence of the parameters, consider a weighted least squares fit that gives equal weightage to FR1, FR2, and 6-24 GHz frequency range. 
Observation 3: Current 38.901 specification provides the values for number of clusters to consider for three different UE categories: LOS outdoor UE, NLOS outdoor UE, and O2I UE (indoor). When considering any update to the number of clusters for existing deployment scenarios, it is important to consider all three UE categories with specific emphasis on O2I UEs which constitute 80% of UEs in typical system-level evaluations. It is noted that currently there are no measurement data for O2I UEs. 
Observation 4: From a system-level evaluation perspective, reducing the number of clusters to a lower number either by reducing the nominal number of clusters or by increasing the threshold to drop weaker clusters does not seem to significantly impact the communication metrics. 
Observation 5: Changing the number of clusters for a specific frequency range will make it difficult for cross-frequency comparisons in future evaluations.
Proposal 13: Given the negligible system-level impact due to a reduction in the number of clusters and the need to ensure frequency continuity, it is suggested that the number of clusters remain unchanged in TR 38.901.

3GPP TSG-RAN WG1 Meeting #120bis	Tdoc R1- 2502878
Wuhan, China, April 7th – 11th, 2025
Agenda Item:	9.8.1
Source:	Ericsson
Title:	Discussion on validation of channel model
Document for:	Discussion

Conclusion
In the previous sections we made the following observations: 
Observation 1	In contrast to urban scenarios, vegetation and foliage makes up a signicant amount of the ground clutter in Suburban Macro scenarios.
Observation 2	The UMa and UMi scenarios in TR 38.901 clause 7.2 do not specify any particular ratio of low-loss and high-loss buildings, leaving such considerations for future decisions depending on study needs.
Observation 3	For calibration, a particular 50/50 ratio of low-loss and high-loss buildings was used for the UMa and UMi scenarios in TR 38.901 (clause 7.8).
Observation 4	In a suburban residential scenario, measurements show that the excess path loss has a  frequency dependence in the sub 6 frequency range and a rather flat frequency dependence at higher frequencies.
Observation 5	The model in option 1 does not capture the dependence on UE height, and the model in option 3 does not capture the dependence on either UE or BS height.
Observation 6	The SMa LOS probability is strongly dependent on the tree density in the scenario.
Observation 7	The LOS probability in an SMa scenario may create unique interference conditions, hence addition of an SMa scenario can support better resiliency to these in future 3GPP releases.
Observation 8	In a large country like the US there are very diverse levels of foliage in different suburban areas.
Observation 9	Typical foliage densities, assuming a 30 m bin size, are in the 10—20% range.
Observation 10	The LSPs for SMa O2I are unreasonably different compared to the LSPs for SMa NLOS.
Observation 11	The ZSD values in the Working assumption are not correct due to an unfortunate copy-paste error.
Observation 12	The correlation distances in the two existing microcell scenarios UMa and RMa are very similar.
Observation 13	The measured angular ASD at 3.4 GHz, 3.5 GHz, 13 GHz, and 28 GHz is lower than expected from TR 38.901.
Observation 14	The ASD and ZSD for 13 and 28 GHz are very similar, which is in line with TR 38.901.
Observation 15	In the TR 38.901 model, the two co-polar components in the channel always have exactly equal power, and the two cross-polar components are equally attenuated according to a stochastic XPR.
Observation 16	Measurements and ray tracing experiments show a slow variability around the mean co-polar and cross-polar power that is independent between different components.
Observation 17	The potential rank reduction of the channel due to polarization power variability is independent on whether the Tx or Rx uses V/H-polarized or ±45° polarized antennas. It manifests as a power imbalance in the former case and a correlation between fading in the latter case.
Observation 18	More paths can be detected in measurements than in comparable channels generated by the UMa model.
Observation 19	In the indoor-office scenario, raytracing experiments show that the excess delay is zero for 40% of the channels. For the other 60% of the channels, a lognormal distribution with the parameters ,  provides a good fit to the experimental data.
Observation 20	Neither of Option 1 nor Option 2 can reproduce the excess delay distribution from the experiments.

Based on the discussion in the previous sections we propose the following:
Proposal 1	The Suburban Macro scenario description is updated to reflect the higher importance of the vegetation compared to urban scenarios.
Proposal 2	For the Suburban Macro scenario definition, align with the principle of UMa and UMi by not specifying any particular low-loss/high-loss ratio of buildings.
Proposal 3	For calibrations involving the Suburban Macro scenario, use 100% low-loss for UEs in residential buildings, and 100% high-loss for UEs in commercial buildings. Note: This should not be seen as restricting the use of other ratios or models for future evaluations.
Proposal 4	Adopt both options for Suburban Macro ISD, i.e. use ISD = 1299 m or ISD = 1732 m.
Proposal 5	Add additional frequency dependence,  to SMa NLOS pathloss.
Proposal 6	Set the anchor frequency in the SMa NLOS pathloss to 7 GHz.
Proposal 7	Update the working assumption on SMa path loss with fixed m and fixed  m, limit BS height to 25—35 m. and increase UT height from 1—10 m to 1—14 m.
Proposal 8	Use the LOS probability in Table 3 as the LOS probability for the SMa scenario.
Proposal 9	Use the parameters in Table 4 as typical parameters.
Proposal 10	Keep the amount of vegetation as a parameter and use 0%, 10%, and 20% as standard options illustrating no, sparse, and rich vegetation scenario, respectively.
Proposal 11	Update the working assumption by changing the SMa O2I LSPs to be copies of the SMa NLOS LSPs.
Proposal 12	Update the working assumption with lgZSD = 0.14 and lgZSD = 0.16 for LOS, NLOS, and O2I in the Suburban Macro scenario.
Proposal 13	The NLOS ZOD offset in the Suburban Macro scenario is determined by:  ,  where  is the clutter height and is one of .
Proposal 14	For the Suburban Macro scenario, use the same values as for UMa for cross-correlations between zenith angles and other large scale parameters that are currently FFS and where no new measurements are provided.
Proposal 15	For the SMa scenario, use the same correlation distances for spatial consistency as in the RMa scenario.
Proposal 16	For the SMa scenario, use the following model parameters for the absolute time of arrival.
Proposal 17	Consider the delay spread model in the TR 38.901 UMa scenario to be validated at 3.5 GHz, and that Alt 2) in the RAN1#120 agreement can be agreed.
Proposal 18	The ASD parameters for the UMa model are adjusted according to Table 6 to better represent measurements at 3.4 and 3.5 GHz in two different cities, and at 13 GHz and 28 GHz in a third city.
Proposal 19	The NLOS path loss in the UMa model is considered to be validated by new measurements at 0.8 GHz, 2 GHz, 5 GHz, 10 GHz, 22 GHz, and 37 GHz, therefore no changes are needed.
Proposal 20	Introduce a random variability of the co- and cross polar powers in the TR 38.901 model, such as an i.i.d zero-mean Gaussian with 2-3 dB standard deviation, via the following changes to step 9 and eqs (7.5-22) and (7.5-28) in clause 7.5 in TR 38.901.
Proposal 21	Encourage companies to perform measurements to further study whether the existing mechanisms for generating clusters and rays are inaccurate when simulating large antenna arrays.
Proposal 22	While measurements support increasing the number of clusters in the UMa channel, for complexity reasons it may be better to keep this number unchanged, i.e. support Alt 2) in the RAN1#120 agreement.
Proposal 23	Adopt Option 3 for the absolute delay modeling in the Indoor office scenario.
Proposal 24	Add to TR 38.901 a stochastic model of antenna imbalance, with random antenna imbalance according to some distribution per antenna or per set of antennas.
Proposal 25	The stochastic model of antenna imbalance may be applied to one link direction (e.g., UL) but not the other (e.g., DL), or both.

3GPP TSG RAN WG1 #120bis	R1-2502899
Wuhan, China, 7 – 11 April 2025

Agenda item:		9.8.1
Source:	Nokia
Title:	Discussion on Channel Model Validation of TR38.901 for 7-24GHz
WI code:	FS_NR_7_24GHz_CHmod
Release:	Rel-19
Document for:		Discussion and Decision

Conclusion
In this paper we further elaborate about the open issues in the discussion of channel modeling for 7-24GHz frequencies.
The following proposals and observations are made:
Proposal 1: For any updates of the parameters the frequency continuity should be maintained so that the new parameters are applicable for the whole 0.5-100 GHz range.
Proposal 2: RAN1 needs to consider the standard deviation of the data and the scale of potential changes when deciding about the need for the updates of the parameters.
Proposal 3: RAN1 to keep the reference Material penetration losses fixed and close to the values in Table 7.4.3-1 (i.e., regardless of whether the material thickness is introduced or not) because they should match low and high loss O2I penetration losses derived based on measurements of buildings.

Observation 1: Material penetration loss of IRR glass weakly depends both on frequency and thickness. The loss in IRR glass is mostly caused by the metalized layer and not by the thickness of the glass. Consideration of more than two layers for IRR glass is not practical due to very high losses.
Proposal 4: RAN1 does not need to introduce resonance effects in glass because the proposed models cannot capture different angles of incidence.
Table 2: Summary of proposed thickness-dependent Material penetration loss models. 

Proposal 5: If thickness dependency for material penetration losses are agreed to be introduced, RAN1 to use models from Table 2 above.

Observation 2: The breakpoint distance phenomena is observed in the UMi scenario, but the breakpoint distances are generally larger than 200 m.
Observation 3: The breakpoint distance phenomena is observed in the UMa scenario, but the breakpoint distances are generally larger than 500 m.
Observation 4: The breakpoint distance phenomena are not observed in the SMa or RMa scenarios that can be used instead if larger ISDs are considred.
Proposal 6: RAN1 does not need to handle the breakpoint distances issue in the current TR 38.901 models for UMi and UMa scenarios, since the breakpoint distances are usually larger than the corresponding ISDs.

Observation 5: LOS probability model Option 3 aligns well with the dense vegetation case of Option 2, which captures the impact of vegetation. Whereas LOS probability model Option 1 aligns well with the no vegetation case of Option 2, which does not capture the impact of vegetation.
Proposal 7: RAN 1 to specify ITU-based model(Option 3) due to it’s simplicity and capability to captures well enough the impact of vegetation and other blockages on LoS probability.

Proposal 8: RAN1 to use new parametrization for low loss outdoor-to-indoor (O2I) building penetration loss model for the Sma scenario which consists of 70% wood and 30% plain glass, and the penetration loss through wall is:


Proposal 9: RAN1 to consider BS antenna down tilt matching the ISD: 95 for the ISD of 1299m and 93 for ISD of 1732m.

Table 4: Radiation power pattern of a single UE antenna element

Proposal 10: RAN1 to consider radiation power pattern  in Table 4 for handheld UE antenna modelling in 7-24 GHz band.

Proposal 11: Include a random implementation loss between 0 and 3 dB for each configured antenna on each UE.

Proposal 12: If CPE needs to be described in the TR, then consider typical dimensions (HxWxD) or (X,Y,Z) of 18x18x6 cm.
Proposal 13: If there is a need to illustrate antenna locations for CPE devices in the TR, then an example of possible individual antennae locations is shown in the figure below for the CPE device.


Figure 10: Potential CPE dimensions and candidate antenna locations

3GPP TSG-RAN WG1 Meeting #120-bis	R1-2502909
Wuhan, China, April 7th – 11th, 2025
Agenda Item:	9.8.1
Source:	Vodafone, Ericsson
Title:	Measurements of the angular spreads in urban and suburban macrocells
Document for:	Discussion

Conclusion
In the previous sections we made the following observations: 
Observation 1	The measured elevation angular spreads (ZSD) at 3.4 GHz for a very large number of communication links in an operational urban macro 5G NR network match the 38.901 UMa model.
Observation 2	The measured azimuth angular spreads (ASD) at 3.4 GHz for a very large number of communication links in an operational urban macro 5G NR network are several times lower than predicted by the 38.901 UMa model.
Observation 3	The measured UMa ZSD vs ASD correlation is similar to the model.
Observation 4	The measured ZSDs at 3.4 GHz in a suburban 5G NR macrocell are very similar to the ZSDs in urban macrocells, however at the upper percentiles there are less high outlier values.
Observation 5	The measured suburban ZSD can be well represented by a lognormal distribution with lgZSD = 0.14 and lgZSD = 0.16.
Observation 6	The measured ASDs at 3.4 GHz in a suburban 5G NR macrocell are very similar to the ASDs in urban macrocells, however at the upper percentiles there are less high outlier values.
Observation 7	The WINNER II Suburban Macro channel model overestimates the ASD by 2-3 times.
Observation 8	The measured suburban ASD can be well represented by a lognormal distribution with lgASD = 0.55 and lgASD = 0.25.
Based on the discussion in the previous sections we propose the following:
Proposal 1	The ASD parameters for the UMa model are adjusted according to Table 2 to better represent measurements.
Proposal 2	Use the ASD and ZSD parameters according to Table 4, Table 5, and Table 6 as a starting point for the Suburban Macro scenario.
Proposal 3	Further measurements of ASD and ZSD in a Suburban Macro scenario, including measurements that distinguish LOS vs NLOS vs O2I, can later be used to refine the suggested parameter values.

3GPP TSG RAN WG1 Meeting #120bis                                                 	R1-2502951
Wuhan, China, 7 April – 11 April, 2025
Source:           BUPT, X-Net
Title:              Discussion on channel model validation of TR38.901 for 7-24 GHz
Agenda item:    	9.8.1
Document for:  	Discussion and Decision

Conclusion
In this contribution, we provide our views on the validation and the details for Rel-19 channel modeling enhancements for 7-24 GHz. Key consideration is the modification of the cluster structure. The observation and proposals are as follows:
Observation 1: The cluster structure described in Section 7.5 and 7.6 of 38901 and Example 1 underestimate channel sparsity. Among these approaches, the Section 7.6 methodology demonstrates better performance than Example 1, which in turn outperforms the Section 7.5 implementation. Only Example 3 yields result that closely align with the measurements.
Observation 2: Compared to the 3GPP model, the ICP model characterizes a few but strong eigenvalues, i.e., the ICP model characterizes sparsity more accurately.
Proposal 1: It is proposed to adopt the Example 3 method in Section 7.6 to modify the cluster structure, as it provides results that best match the measured data. This approach introduces the Intra-cluster power factor (ICP) to effectively model dominant rays. The ICP is introduced in section 7.6 as shown below:
To characterize the channel sparsity, the Intra-cluster power factor (ICP) is introduced. The ICP is applied to the generation of cluster powers in Step 6 in Subclause 7.5 to generate the ray power. The intra-cluster power factor () is modelled as:

where
-  represents the normalized cluster delay generated by Step 5.
-  represents the delay spread generated by Step 4.
-  and are the coefficients related to frequency and scenario.
The power of rays within a cluster is determined by the ICP, expressed as

3GPP TSG RAN WG1 Meeting #120bis	R1- 2502957
Wuhan, China, April 7th – 11st, 2025															(revision of R1-2502341)

Source:	Intel Corporation
Title:	Discussion on channel modeling verification for 7-24 GHz
Agenda item:	9.8.1
Document for:	Discussion

Conclusions
In this contribution, we discussed the potential aspects that may require channel modeling validation effort for frequencies from 7 to 24 GHz. The following is a summary of proposals and observations made in this contribution.
Proposal 1:
Support updates to glass, concrete, and wood penetration model.
Use models without penetration loss resonance effect for glass penetration models.
RAN1 to conclude on observing glass penetration loss resonance effect but conclude not to model the effect.
Proposal 2:
Support introduction of cinder block penetration loss model based on concrete measurements from companies.

Proposal 3:
Update the SMa LOS Pathloss as
For 10m ≤ d < ,


For   ≤ d < 5000m,
 

  is the center frequency in Hz, m/s

Proposal 4:
Discuss the following 4 options for mean angle and angular spread definitions for CDL angle calculations.
Based on selection option for mean angle and angular spread definition, update the CDL angle table appropriately.
Option 1) Scaling applies to both NLOS and LOS angles for all subpaths, Mean Angle and AS include both NLOS and LOS subpaths







Option 2) Scaling applies to only NLOS for all subpaths, Mean Angle and AS include both NLOS and LOS subpaths
Note: for CDL-A/B/C, this is same as Option 1







Option 3) Scaling applies to only NLOS for all cluster path (not subpath), Mean Angle and AS include both NLOS and LOS subpaths







Option 4) Scaling applies to only NLOS for all cluster path (not subpath), Mean Angle and AS include only NLOS subpaths
Note: for CDL-A/B/C, this is same as Option 3








Proposal 5:
Use the following formula for computing the scaling factor.

3GPP TSG RAN WG1 #120bis                                         R1-2502986 

Wuhan, China, April 7th – 11th, 2025

Source:	Sharp
Title:	Views on Channel model validation of TR38.901 for 7-24 GHz
Agenda Item:	9.8.1
Document for:	Discussion and Decision
Umi
Curve fittings for UMi DS, ASA LOS/NLOS Mean and Standard deviation (std) are present in R1-2502415_UMi_curve_fitting.pptx (Agenda Item 9.8).
UMa
Curve fittings for UMa DS, ASA, ASD LOS/NLOS Mean and Standard deviation (std) are present in R1-2502415_UMa_curve_fitting.pptx (Agenda Item 9.8).
Measurement Data
Raw measurement data for UMi DS, ASA LOS/NLOS Mean/Std and UMa DS, ASA, ASD LOS/NLOS Mean/Std are present in R1-2502415_consolidated_meas_data_rel14_rel19.xlsx (Agenda Item 9.8).
O2I Building Penetration Loss Model for SMa
O2I SMa penetration loss measurements were conducted at 28 GHz in a New Jersey suburb [1]. As can be clearly seen from Fig. 1, the O2I BPL measurements conducted at 28 GHz in the SMa scenario are much lower than that estimated by the current low loss BPL model in TR 38.901. The mean BPL estimated by the current TR 38.901 low loss model at 28 GHz is ~18 dB, whereas the mean BPL estimated by the proposed SMa BPL is ~13 dB. The proposed O2I BPL for the SMa scenario is shown in Table 1. Table 1 accurately captures the average O2I BPL observed in the SMa scenario from measurements conducted at 28 GHz. The proposed SMa BPL loss model (blue dashed lines) is the average of the 3 solid lines indicated by green, magenta, and black colours.


Fig. 1 CDFs of measured BPL for SMa scenario from [1] and comparison with proposed SMa BPL and existing TR 38.901 low-loss BPL model.

Table 1. Proposed O2I building penetration loss model for SMa scenario

Observation 1: The proposed O2I BPL for SMa scenario in Table 1 accurately captures the average O2I BPL observed in SMa scenario from measurements conducted at 28 GHz as shown in Fig. 1. 

Proposal 1: RAN1 to introduce a low loss O2I building penetration loss model for SMa scenario as shown in Table 1.
Penetration Loss Modelling
Overview on creation of existing TR 38.901 O2I BPL Model:
The building penetration loss model according to TR 38.901 [2] consists of the following parts:

where PLb is the basic outdoor path loss given by the UMa or UMi path loss models, PLtw is the building penetration loss through the external wall, PLin is the inside loss dependent on the depth into the building, and  is the standard deviation.


Table 2. O2I building penetration loss model from TR 38.901 [2]

Observation 2: 
In Table 7.4.3-2 in TR 38.901 an additional loss of 5 dB has been added to the external wall loss to account for non-perpendicular incidence [3]. 
In Table 7.4.3-2 in TR 38.901 the indoor loss has been selected at 0.5 dB/m to maintain consistency with the 3D SCM TR 36.873 [4]. 
In Table 7.4.3-2 in TR 38.901 the standard deviation has been tentatively selected based on reported measurements [3].
The standard deviation in Table 7.4.3-2 in TR 38.901 (shown in Table 2 here) is intended to capture two types of variations: the variation of the building penetration loss within a single building and the variation between different but similar buildings. To properly characterize this variation, it is important to consider how well different measurements and compilations of measurements capture the two types of variations. Fig. 2 (b) contains a comparison of the model with various measurements. The error bars represent the variations of measurements within a single building while the similarly colored markers represent measurements from buildings of the same or similar compositions. The heterogeneity of the measurement procedures makes it quite difficult to assess the standard deviation. Nevertheless, an attempt based on the mean value from each measurement has been performed. In this assessment, the measurements from buildings with IRR glass were compared to the high loss model while the measurements from buildings with standard glass were compared to the low loss model. The resulting standard deviations are summarized were adopted in Table 7.4.3-2 in TR 38.901 [5].


Fig. 2 (a) Comparison between the material loss model and measurements (left) and (b) the composite penetration loss model for normal incidence and measurements (right). The bars indicate variability for a given building.

Views on modelling resonant behaviour for standard multi-pane glass and IRR glass: 
From Fig. 2 (a) we can clearly observe the resonance behavior for standard multi-pane and IRR glass. However, the loss trends with frequency can still be approximated to a first order. It was noted in Rel-14 [3] that variations around the linear trend for materials can be understood from multiple reflections within the material or between different layers which cause constructive or destructive interference depending on the frequency and incidence angle. Similar observation can be made based on the data provided by companies in Rel-19 as well.

Observation 3: Resonance behavior for standard multi-pane glass and IRR glass was observed in Rel-14 [3] material penetration loss data and a simple 1st order linear approximation was used to reflect the overall BPL. 

Proposal 2: RAN#1 to avoid modelling resonance behavior for standard multi pane glass and IRR glass.

Proposal 3: A note can be made in the TR 38.901 for Table 7.4.3-1 stating that “Variations around the linear trend for standard multi-pane glass and IRR glass can be understood from multiple reflections within the material or between different layers which cause constructive or destructive interference depending on the frequency and incidence angle”.

 

Views on modelling interface loss for concrete, wood, standard multi-pane glass:
For modelling material penetration loss based on interface loss modelling the parameters a, b was FFS for wood, glass, IRR glass based on agreements from RAN#120 [6]. For concrete the values of a, b proposed in RAN1#120 were 1.4 and 4.6, respectively. [7] and [8] further provided values for a, b in RAN1#120b which are listed in Table 3.

Table 3. Values of a, b for modelling penetration loss of materials using interface loss modelling.

Observation 4: There is discrepancy in values reported for a, b for wood, glass and concrete in [7] and [8]. 


Fig 3. Comparison of penetration loss for different materials based on values reported for a, b with TR 38.901 from two different sources [7, 8].

Due to the discrepancies in reported values of a and b for modelling penetration loss using interface loss modelling, we analyse the proposed values of a, b with the current TR 38.901 material penetration loss model. Fig. 3 demonstrate that for concrete and glass, the value proposed for a and b in [7] aligns more closely with the existing TR 38.901 penetration loss model for concrete and glass over the frequency range of 0.5-100 GHz. Conversely, for wood, the value proposed in [8] aligns more closely with the existing TR 38.901 penetration loss model for wood over the frequency range of 0.5-100 GHz.

Proposal 4: For modelling penetration loss based on interface loss for concrete, wood or glass adopt Table 4.

Table 4. Values of a, b for material penetration loss modelling using the interface loss option from [7, 8].

Views on impact of incorporating thickness in material penetration loss:
The main goal of introducing material thickness is to align the measurement results reported by different companies in Rel-19. However, it is imperative to study the impact of introducing material thickness on the O2I building penetration loss models. Table 5 lists the potential options for down selection for concrete, glass and IRR glass based on [6] and [7, 8]. Fig. 4 and Fig. 5 shows the impact of potential down selections on the low loss and high loss O2I building penetration loss models, respectively. 

Observation 5: It can be clearly seen from Fig. 4 that there is no impact of introducing material thickness for concrete, glass and resonant effect of glass on low loss O2I building penetration loss model if we assume the default thickness of the material in simulations. Any variations around the mean low loss O2I building penetration loss are within 1 standard deviation as defined in Table 7.4.3-2 in TR 38.901.

Observation 6: It can be clearly seen from Fig. 5 that there is no impact of introducing material thickness for concrete and resonant effect of IRR glass on high loss O2I building penetration loss model if we assume the default thickness of the material in simulations. Any variations around the mean high loss O2I building penetration loss are within 2 standard deviations as defined in Table 7.4.3-2 in TR 38.901.

Table 5. Impact on low-loss and high-loss O2I BPL based on various options available for down selection for concrete, IRR glass and Standard multi-pane glass.















Fig 4. Impact of various options listed in Table 5 on low loss O2I building penetration loss model.













Fig 5. Impact of various options listed in Table 5 on high loss O2I building penetration loss model.

Proposal 5: RAN#1 to avoid introducing material thickness and resonant effects to model the material penetration loss as there is no impact on low and high loss O2I building penetration loss models on introducing material thickness or resonant effects. The standard deviations defined in the existing low and high loss BPL model in Table 7.4.3-2 in TR 38.901 can account for any variations observed due to material thickness, resonant effects and other phenomenon around the mean frequency dependent penetration loss. The only advantage of introducing thickness or resonant effects is to align measurements results submitted in Rel-19.

Thus, a note can be added in the TR in Table 7.4.3-1 that could list all the potential observations and provided some reference for material thickness to align the measurement results provided in Rel-19.

Proposal 6:  The following can be adopted as a potential change to capture the observations regarding thickness and resonant behavior without explicitly modelling them in the material penetration loss model could be:

Table 7.4.3-1: Material penetration loss

Views on Updating Number of Clusters
Background on number of clusters for UMa scenario:
In Rel-14 [9], an investigation into the clustering of the rays using ray-tracing study was performed. To determine clusters, the agglomerative algorithm was employed.  The results showed that the average number of clusters and the average number of rays per cluster were both fairly consistent across the different carrier frequencies. However, the cluster delay spread tended to decrease with increasing frequency in both LOS and NLOS, and hence it was determined to model that frequency dependency. In interpreting these results, especially the average number of rays per cluster, it should be noted that the number of modelled rays was limited to 20 in the simulations. Hence, it was decided to still retain the 3GPP TR 36.873 modelling of the number of clusters and the number of rays per cluster until more study with additional measurements or more detailed ray tracing is available.

Background on number of clusters for UMi scenario:
In Rel-14 [10], the cluster parameters were preliminarily derived from some measurement and ray-tracing data. The ray-tracing data predicts fewer clusters and weak-frequency-dependent characteristics. However, the clustering results based on raytracing are limited due to some simulation conditions, such as a small number of observed paths in simulation and the deterministic characteristic of propagation simulation. Further measurement campaigns will be needed to verify these observations. Meanwhile, as a starting point, the number of clusters and number of rays in 3D-SCM TR 36.973 parameters can be used. Furthermore, in some cases, it is also required to modify the clustering model to support large BWs and large-sized arrays. It is recommended to further study this through measurements with large bandwidths and/or large antenna arrays. 


Observation 7: Based on observations made in Rel-14 for updating the number of clusters for the UMi and UMa scenario, it is evident that there was very limited real-world measurement data available in Rel-14 to adopt the reduction in the number of clusters. Most of the data in Rel-14 to study the reduction in the number of clusters came from ray tracing, which showed that the average number of clusters and the average number of rays per cluster were both consistent across the different carrier frequencies. Additionally, the clustering results based on ray tracing were limited due to some simulation conditions, such as a small number of observed paths in the simulation and the deterministic characteristic of propagation simulation. Thus, it was recommended that further measurement campaigns would be needed to verify these observations and, as a starting point, the number of clusters and the number of rays in 3D-SCM TR 36.873 parameters were adopted, which became the basis of the current TR 38.901.

On the other hand, in Rel-19, various sources have conducted measurements in real-world scenarios to determine the number of clusters. However, these measurements are limited to the frequency range of 6-24 GHz only. Whereas the number of clusters in TR 38.901 is captured for all scenarios and under each scenario specifically for LOS, NLOS, and O2I UEs. Additionally, it is important to note that the number of clusters in the current TR 38.901 is applicable to the entire frequency range of 0.5-100 GHz. Therefore, any changes made solely based on measurements conducted for 6-24 GHz are inappropriate, as the applicability of those cluster numbers for frequencies less than 6 GHz and above 28 GHz cannot be determined. 

Observation 8: A limitation in Rel-14 was not to change the number of clusters, due to the limited data available. All the data for reducing the number of clusters primarily came from ray tracing, which has limitations, and thus the reduction in the number of clusters was not adopted. However, in Rel-19, we have real-world measurement data to study the number of clusters observed in a real-world scenario, but they are limited to the 6-24 GHz range only.

Proposal 7: RAN1 should consider the number of clusters based on different scenarios, channel conditions, and frequencies in the 6-24 GHz range as well before deciding whether to reduce the number of clusters in the existing TR 38.901. Additionally, whether to consider the number of clusters based on ray tracing data from Rel-14 and Rel-19 in conjunction with measurements should also be discussed before any updates to the number of clusters are made.

Observation 9: Companies are encouraged to check the values reported in Table 6 for the number of clusters reported by different sources. 

Table 6. Number of Clusters for different scenarios.
[10] 5G workshop paper (http://www.5gworkshops.com/2016/5GCMSIG_White%20Paper_r2dot3.pdf)

Based on Table 6 above, it can be clearly observed that for UMi NLOS scenario based on measurement data provided in Rel-19 for 6-24 GHz only the mean number of clusters is 10. However, the number of clusters observed at 28 GHz and 73 GHz based on measurement data provided in Rel-19 in UMi NLOS scenario is 5. Thus, the mean number of clusters based on measurement data only for 6-24 GHz UMi NLOS is still 2 times higher than the mean number of clusters observed at 28 and 73 GHz for UMi NLOS scenario. Thus, simply taking the mean of the reported number of clusters across various sources may not be the best approach as it can still lead to overestimation and underestimation of the number of clusters. More measurements result across different scenarios and frequencies are required to accurately capture the number of clusters across the entire frequency range of 0.5-100 GHz. 

Further investigating the data at 28 GHz and 73 GHz for UMi NLOS scenarios in [11], we found that the 95% value from the CDF for the number of clusters at 28 GHz and 73 GHz is 14 and 16, respectively. Whereas in TR 38.901, the number of clusters for UMi NLOS is 19, as shown in Table 7. Thus, from Table 7, we observe that if we compare the mean values of the number of clusters with TR 38.901, they are significantly lower. However, if we compare the 95% value from the CDF, they are much higher than the mean values and somewhat closer to existing TR 38.901.

Table 7. Number of Clusters for UMi NLOS scenario based on real-world measurement.

Proposal 8: Companies to provide more details on the mean, max, 90% or beyond values for the number of clusters to make fair comparisons with TR 38.901. It is unclear whether the mean values were adopted in TR 38.901 for the number of clusters as shown in Table 7. 

Proposal 9: RAN#1 currently lacks comprehensive measurement data for the entire frequency range of 0.5-100 GHz to adopt the reduction in the number of clusters. The updates solely based on data from 6-24 GHz may not be applicable to frequencies above 24 GHz or below 6 GHz. Consequently, RAN#1 is advised to refrain from making any modifications to the number of clusters. 
Views on Changing Power Angular Spectrum
In Section 7.5 in TR 38.901, clusters are characterized by a joint delay-angle probability density function, such that a group of traveling multipath components must depart and arrive from a unique Angle of Departure (AOD) - Angle of Arrival (AOA) combination centred around a mean propagation delay. The power of the cluster is equally divided among 20 multipath components/rays. On the other hand, in Section 7.6.2.2 in TR 38.901 which is stated as an optional modelling component the clusters are characterized like section 7.5 in TR 38.901, but the power of the cluster is unequally distributed among the rays as defined in equation (7.6-6) in TR 38.901. Section 7.6.2.2 in TR 38.901 clearly states that “with large antenna arrays or large bandwidths, the angle and/or delay resolution can be larger than what the fast-fading model in Section 7.5 in TR 38.901 is designed to support”. This implies that a larger antenna array will generally provide a larger spatial resolution, and a large bandwidth will provide a larger time resolution and the combination of these would results in more rays being detected. Based on simulation parameters provided in [13], the number of rays (M) per cluster in LOS and NLOS for UMa scenario using equation (7.6-8) in TR 38.901 is always 20. Based on this value of M here we generate a single channel realization of a single UE and consider there are only two clusters, each cluster employing M = 20, i.e., 20 rays per cluster. The first cluster comprises 75% of the total power, while the second cluster holds 25% of the total power. Furthermore, the power distribution among rays is computed using the equation (7.6-6) in TR 38.901. Consequently from Fig. 6, we observe that the rays within both clusters possess unequal powers, and the overall power between the clusters decay exponentially. Although Fig. 6 presents an example with two clusters, each containing unequal power/ray, the underlying concept can be generalized to any number of clusters.



Fig. 6. Unequal distribution of power/ray in 2 different clusters for UMa NLOS scenario as per equation (7.6-6) in TR 38.901. R1, C1 and R1, C2 indicate Ray 1 in Cluster 1 and Ray 2 in Cluster 2, respectively.

From Fig. 6, it is evident that each cluster contains some rays with higher power compared to others. However, in Fig. 6 there is no dominant ray that possesses most of the cluster power. This observation has also been corroborated by various measurement campaigns conducted in mmWave frequencies, employing higher bandwidth and spatial resolution in UMi scenario in NYC [14, 15]. Measurements [14, 15] have demonstrated that the clusters exhibit varying power among their rays due to the reflection, diffraction, and scattering phenomena, and this is accurately modelled in Section 7.6.2.2 of TR 38.901. While it is true that some clusters may possess a dominant ray, there is no evidence from measurements whether all clusters possess a dominant ray. Even if such a case exists where all clusters possess a dominant ray it could be environment specific, limitation of the measurement system, or due to other factors. Further measurements in diverse scenarios and environments are necessary to ascertain whether all clusters observe a dominant ray. In this regard, [13] posits that most the cluster’s power is concentrated within the dominant ray, while the other rays within the cluster possess negligible power and proposes the introduction of the ICP to model this phenomenon observed through measurements conducted in the UMa scenario. 

In Fig. 6, for instance, when R1 in C1 and R1 in C2 are allocated 95% of the total power of the cluster, while the other rays in C1 and C2 are assigned only 5% of the total power of the cluster, the scenario becomes equivalent to the proposal in [13]. In practice, during simulations when multiple channel realizations are generated using the procedures outlined in Section 7.6.2.2 of TR 38.901, certain clusters will possess a dominant ray or at least very few rays consisting of 95% of the total power of the cluster. Consequently, as demonstrated in [13], the Gini index values obtained from simulations using the modelling framework outlined in Section 7.6.2.2 exhibit better accuracy compared to Section 7.5 of TR 38.901 and approximate actual measurements. It is impractical to precisely construct a model that perfectly aligns with measurements, as the measurements are constrained by system design and may also be environment specific.

Observation 10: 
Dominant Ray Existence: Uncertain whether all clusters have a dominant ray; further measurements in different scenarios and environments are needed.
Power Distribution among Rays in Clusters: The current implementation of unequal power distribution among rays in a cluster in Section 7.6.2.2. in TR 38.901 is more generic as the power among rays varies in a cluster due to reflection, diffraction, and scattering phenomena. Thus, ICP [13] is a special case of the existing implementation in Section 7.6.2.2 in TR 38.901 and few of the clusters generated using Section 7.6.2.2. in TR 38.901 will possess the dominant ray in the cluster.
Proposal 10: Section 7.6.2.2 of TR 38.901 modestly captures the unequal distribution of power among rays in the cluster. Therefore, altering the unequal power distribution of rays in a cluster to model dominant rays for all clusters is premature and not needed. 

Conclusion
Observation 1: The proposed O2I BPL for SMa scenario in Table 1 accurately captures the average O2I BPL observed in SMa scenario from measurements conducted at 28 GHz as shown in Fig. 1. 

Observation 2: 
In Table 7.4.3-2 in TR 38.901 an additional loss of 5 dB has been added to the external wall loss to account for non-perpendicular incidence [3]. 
In Table 7.4.3-2 in TR 38.901 the indoor loss has been selected at 0.5 dB/m to maintain consistency with the 3D SCM TR 36.873 [4]. 
In Table 7.4.3-2 in TR 38.901 the standard deviation has been tentatively selected based on reported measurements [3].
Observation 3: Resonance behavior for standard multi-pane glass and IRR glass was observed in Rel-14 [3] material penetration loss data and a simple 1st order linear approximation was used to reflect the overall BPL. 

Observation 4: There is discrepancy in values reported for a, b for wood, glass and concrete by R1-2501897 [8] and R1-2502899 [9]. 

Observation 5: It can be clearly seen from Fig. 4 that there is no impact of introducing material thickness for concrete, glass and resonant effect of glass on low loss O2I building penetration loss model if we assume the default thickness of the material in simulations. Any variations around the mean low loss O2I building penetration loss are within 1 standard deviation as defined in Table 7.4.3-2 in TR 38.901.

Observation 6: It can be clearly seen from Fig. 5 that there is no impact of introducing material thickness for concrete and resonant effect of IRR glass on high loss O2I building penetration loss model if we assume the default thickness of the material in simulations. Any variations around the mean high loss O2I building penetration loss are within 2 standard deviations as defined in Table 7.4.3-2 in TR 38.901.

Observation 7: Based on observations made in Rel-14 for updating the number of clusters for the UMi and UMa scenario, it is evident that there was very limited real-world measurement data available in Rel-14 to adopt the reduction in the number of clusters. Most of the data in Rel-14 to study the reduction in the number of clusters came from ray tracing, which showed that the average number of clusters and the average number of rays per cluster were both consistent across the different carrier frequencies. Additionally, the clustering results based on ray tracing were limited due to some simulation conditions, such as a small number of observed paths in the simulation and the deterministic characteristic of propagation simulation. Thus, it was recommended that further measurement campaigns would be needed to verify these observations and, as a starting point, the number of clusters and the number of rays in 3D-SCM TR 36.873 parameters were adopted, which became the basis of the current TR 38.901.

Observation 8: A limitation in Rel-14 was not to change the number of clusters, due to the limited data available. All the data for reducing the number of clusters primarily came from ray tracing, which has limitations, and thus the reduction in the number of clusters was not adopted. However, in Rel-19, we have real-world measurement data to study the number of clusters observed in a real-world scenario, but they are limited to the 6-24 GHz range only.

Observation 9: Companies are encouraged to check the values reported in Table 6 for the number of clusters reported by different sources. 

Table 6. Number of Clusters for different scenarios.
[10] 5G workshop paper (http://www.5gworkshops.com/2016/5GCMSIG_White%20Paper_r2dot3.pdf)

Observation 10: 
Dominant Ray Existence: Uncertain whether all clusters have a dominant ray; further measurements in different scenarios and environments are needed.
Power Distribution among Rays in Clusters: The current implementation of unequal power distribution among rays in a cluster in Section 7.6.2.2. in TR 38.901 is more generic as the power among rays varies in a cluster due to reflection, diffraction, and scattering phenomena. Thus, ICP [13] is a special case of the existing implementation in Section 7.6.2.2 in TR 38.901 and few of the clusters generated using Section 7.6.2.2. in TR 38.901 will possess the dominant ray in the cluster.

Proposal 1: RAN1 to introduce a low loss O2I building penetration loss model for SMa scenario as shown in Table 1.
Table 1. Proposed O2I building penetration loss model for SMa scenario

Proposal 2: RAN#1 to avoid modelling resonance behavior for standard multi pane glass and IRR glass.

Proposal 3: A note can be made in the TR 38.901 for Table 7.4.3-1 stating that “Variations around the linear trend for standard multi-pane glass and IRR glass can be understood from multiple reflections within the material or between different layers which cause constructive or destructive interference depending on the frequency and incidence angle”.

Proposal 4: For modelling penetration loss based on interface loss for concrete, wood or glass adopt Table 4.

Table 4. Values of a, b for material penetration loss modelling using the interface loss option from [7, 8].

Proposal 5: RAN#1 to avoid introducing material thickness and resonant effects to model the material penetration loss as there is no impact on low and high loss O2I building penetration loss models on introducing material thickness or resonant effects. The standard deviations defined in the existing low and high loss BPL model in Table 7.4.3-2 in TR 38.901 can account for any variations observed due to material thickness, resonant effects and other phenomenon around the mean frequency dependent penetration loss. The only advantage of introducing thickness or resonant effects is to align measurements results submitted in Rel-19.

Proposal 6:  The following can be adopted as a potential change to capture the observations regarding thickness and resonant behavior without explicitly modelling them in the material penetration loss model could be:

Table 7.4.3-1: Material penetration loss

Proposal 7: RAN1 should consider the number of clusters based on different scenarios, channel conditions, and frequencies in the 6-24 GHz range as well before deciding whether to reduce the number of clusters in the existing TR 38.901. Additionally, whether to consider the number of clusters based on ray tracing data from Rel-14 and Rel-19 in conjunction with measurements should also be discussed before any updates to the number of clusters are made.

Proposal 8: Companies to provide more details on the mean, max, 90% or beyond values for the number of clusters to make fair comparisons with TR 38.901. It is unclear whether the mean values were adopted in TR 38.901 for the number of clusters as shown in Table 7. 

Table 7. Number of Clusters for UMi NLOS scenario based on real-world measurement.

Proposal 9: RAN#1 currently lacks comprehensive measurement data for the entire frequency range of 0.5-100 GHz to adopt the reduction in the number of clusters. The updates solely based on data from 6-24 GHz may not be applicable to frequencies above 24 GHz or below 6 GHz. Consequently, RAN#1 is advised to refrain from making any modifications to the number of clusters. 

Proposal 10: Section 7.6.2.2 of TR 38.901 modestly captures the unequal distribution of power among rays in the cluster. Therefore, altering the unequal power distribution of rays in a cluster to model dominant rays for all clusters is premature and not needed. 

References
[1] R1-2406139, “Discussion on Channel model validation of TR38.901 for 7-24GHz”, Nokia.
[2] TR 38.901 TR 38.901 “Study on channel model for frequencies from 0.5 to 100 GHz”.
[3] R1-161687 “Building penetration loss modelling”, Ericsson, Nokia Networks, AT&T, CMCC, ETRI, Huawei, HiSilicon, Intel, KT Corporation, NTT DOCOMO, Qualcomm, Samsung.
[4] 3GPP TR 36.873 “Study on 3D channel model for LTE”, v12.0.0, June 2015.
[5] R1-163256 “Removing square brackets in the O2I penetration loss modeling, part 2”, Ericsson, NTT DOCOMO, Samsung.
[6] R1-2501638 “Summary #5 of discussions for Rel-19 7-24 GHz Channel Modeling Validation”, Moderator (Intel Corporation).
[7] R1-2501897 “Discussion on the channel model validation for 7-24 GHz”, ZTE Corporation, Sanechips.
[8] R1-2502899 “Discussion on Channel Model Validation of TR38.901 for 7-24GHz”, Nokia.
[9] R1-161661, “Joint proposal on large-scale parameters for UMa scenario”, Nokia Networks, Samsung, AT&T, CMCC, Ericsson, Huawei, HiSilicon, Intel, KT Corporation, NTT DOCOMO, Qualcomm, ETRI.
[10] 5G workshop paper (http://www.5gworkshops.com/2016/5GCMSIG_White%20Paper_r2dot3.pdf
[11] R1-2410319 “Channel model validation of TR38.901 for 7-24 GHz”, Sharp, Nokia.
[12] Czink, N., Cera, P., Salo, J., Bonek, E., Nuutinen, J. P., & Ylitalo, J. (2006, September). A framework for automatic clustering of parametric MIMO channel data including path powers. In IEEE Vehicular Technology Conference (pp. 1-5). IEEE.
[13] R1-2502951, “Discussion on channel model validation of TR38.901 for 7-24 GHz”, BUPT, X-Net.
[14] Samimi, M. K., & Rappaport, T. S. (2014). Characterization of the 28 GHz millimeter-wave dense urban channel for future 5G mobile cellular. NYU Wireless TR, 1, 1-322.
[15] Shakya, D., Ju, S., Kanhere, O., Poddar, H., Xing, Y., & Rappaport, T. S. (2024). Radio propagation measurements and statistical channel models for outdoor urban microcells in open squares and streets at 142, 73, and 28 GHz. IEEE Transactions on Antennas and Propagation.

TDoc file conclusion not found

3GPP TSG RAN WG1 Meeting #120bis	R1- 2503060
Wuhan, China, April 7th – 11st, 2025

Source:	Moderator (Intel Corporation)
Title:	Data source descriptions for 7 – 24 GHz SI
Agenda item:	9.8
Document for:	Information
Coversheet
In RAN Plenary #102, study for channel modeling verification for 7 – 24 GHz was approved [1]. The Study includes two objectives as described below.
Validate using measurements the channel model of TR38.901 at least for 7-24 GHz
Note: Only stochastic channel model is considered for the validation.
Note: The validation may consider all existing scenarios: UMi-street canyon, UMa, Indoor-Office, RMa and Indoor-Factory.

Adapt/extend as necessary the channel model of TR38.901 at least for 7-24 GHz, including at least the following aspects for applicable scenarios: 
Near-field propagation (with consideration being given to consistency between near-field and far-field)
Spatial non-stationarity

Note 1: Continuity of the channel model in the frequency domain below 7 GHz and above 24 GHz shall be ensured.

Note 2: Mathematical and/or theoretical aspects (if any) may be studied before results of measurement campaigns are available. While measurement results may be available and submitted at any time, the study of measurement results may start later (e.g., Q3 2024).

This document is the coversheet for the excel spreadsheet for collecting measurement descriptions for 7 – 24 GHz channel modeling efforts. The attached excel sheet provides measurement/simulation descriptions for the SI.

References
RP-234018, “New SID: Study on channel modelling enhancements for 7-24GHz for NR”, Nokia, Nokia Shanghai Bell, RAN#102, December 2023.

TDoc file conclusion not found

3GPP TSG RAN WG1 Meeting #120bis	R1-2503063
Wuhan, China, April 7th – 11st, 2025

Source: 	Moderator (Intel Corporation)
Title:	
Agenda item:	9.8.1
Document for:	Discussion

Conclusions from RAN1 #118bis
[To be filled]

3GPP TSG RAN WG1 Meeting #120bis	R1-2503064
Wuhan, China, April 7th – 11st, 2025

Source: 	Moderator (Intel Corporation)
Title:	
Agenda item:	9.8.1
Document for:	Discussion

Conclusions from RAN1 #118bis
[To be filled]

3GPP TSG RAN WG1 Meeting #120bis	R1-2503065
Wuhan, China, April 7th – 11st, 2025

Source: 	Moderator (Intel Corporation)
Title:	
Agenda item:	9.8.1
Document for:	Discussion

Conclusions from RAN1 #118bis
Conclusion
Based on measurement data provided, RAN1 concludes that following delay spread parameters for 6 – 24 GHz frequency range are validated and no updates to TR are needed.
InH LOS and NLOS

Agreement
Update the UMi LOS and NLOS delay spread as follows:
Note: the update parameter was generated using all measurement and ray tracing data set from Rel-14 SI and (current) Rel-19 SI and dividing the data points into 3 groups, below 6 GHz, 6 to 24 GHz, and above 24 GHz, and weighting the data sets for each group to perform weighted least square curve fit. If a group has fewer data points, higher weight per data point is calculated. All points within a group have same weight. Sum of weights for all groups is equal to 1.
Note: Each group is given equal weightage. 

Agreement
Update the UMa LOS and NLOS delay spread as follows:
Note: the update parameter was generated using all measurement and ray tracing data set from Rel-14 SI and (current) Rel-19 SI and dividing the data points into 3 groups, below 6 GHz, 6 to 24 GHz, and above 24 GHz, and weighting the data sets for each group to perform weighted least square curve fit. If a group has fewer data points, higher weight per data point is calculated. All points within a group have same weight. Sum of weights for all groups is equal to 1.
Note: Each group is given equal weightage.

Agreement
Update the UMi LOS and NLOS AOA spread as follows:
Note: the update parameter was generated using all measurement and ray tracing data set from Rel-14 SI and (current) Rel-19 SI and dividing the data points into 3 groups, below 6 GHz, 6 to 24 GHz, and above 24 GHz, and weighting the data sets for each group to perform weighted least square curve fit. If a group has fewer data points, higher weight per data point is calculated. All points within a group have same weight. Sum of weights for all groups is equal to 1.
Note: Each group is given equal weightage.

Agreement
Update the UMa LOS and NLOS AOA spread as follows:
Note: the update parameter was generated using all measurement and ray tracing data set from Rel-14 SI and (current) Rel-19 SI and dividing the data points into 3 groups, below 6 GHz, 6 to 24 GHz, and above 24 GHz, and weighting the data sets for each group to perform weighted least square curve fit. If a group has fewer data points, higher weight per data point is calculated. All points within a group have same weight. Sum of weights for all groups is equal to 1.
Note: Each group is given equal weightage.

Working Assumption
Adopt the following absolute delay parameters for RMa scenarios.

Agreement
Defines upper bound of absolute delay in indoor office scenario as 2L/c where L is the largest dimension of the office hall.
No upper bound of absolute delay in UMi, UMa, and RMa scenario.


Conclusion
For the following scenarios, there is no consensus to update pathloss models due to lack of consistent and significant observed difference between model and measurements.
UMa LOS/NLOS

3GPP TSG RAN WG1 Meeting #120bis	R1-2503066
Wuhan, China, April 7th – 11st, 2025

Source: 	Moderator (Intel Corporation)
Title:	
Agenda item:	9.8.1
Document for:	Discussion

Conclusions from RAN1 #118bis
Conclusion
Based on measurement data provided, RAN1 concludes that following delay spread parameters for 6 – 24 GHz frequency range are validated and no updates to TR are needed.
InH LOS and NLOS

Agreement
Update the UMi LOS and NLOS delay spread as follows:
Note: the update parameter was generated using all measurement and ray tracing data set from Rel-14 SI and (current) Rel-19 SI and dividing the data points into 3 groups, below 6 GHz, 6 to 24 GHz, and above 24 GHz, and weighting the data sets for each group to perform weighted least square curve fit. If a group has fewer data points, higher weight per data point is calculated. All points within a group have same weight. Sum of weights for all groups is equal to 1.
Note: Each group is given equal weightage. 

Agreement
Update the UMa LOS and NLOS delay spread as follows:
Note: the update parameter was generated using all measurement and ray tracing data set from Rel-14 SI and (current) Rel-19 SI and dividing the data points into 3 groups, below 6 GHz, 6 to 24 GHz, and above 24 GHz, and weighting the data sets for each group to perform weighted least square curve fit. If a group has fewer data points, higher weight per data point is calculated. All points within a group have same weight. Sum of weights for all groups is equal to 1.
Note: Each group is given equal weightage.

Agreement
Update the UMi LOS and NLOS AOA spread as follows:
Note: the update parameter was generated using all measurement and ray tracing data set from Rel-14 SI and (current) Rel-19 SI and dividing the data points into 3 groups, below 6 GHz, 6 to 24 GHz, and above 24 GHz, and weighting the data sets for each group to perform weighted least square curve fit. If a group has fewer data points, higher weight per data point is calculated. All points within a group have same weight. Sum of weights for all groups is equal to 1.
Note: Each group is given equal weightage.

Agreement
Update the UMa LOS and NLOS AOA spread as follows:
Note: the update parameter was generated using all measurement and ray tracing data set from Rel-14 SI and (current) Rel-19 SI and dividing the data points into 3 groups, below 6 GHz, 6 to 24 GHz, and above 24 GHz, and weighting the data sets for each group to perform weighted least square curve fit. If a group has fewer data points, higher weight per data point is calculated. All points within a group have same weight. Sum of weights for all groups is equal to 1.
Note: Each group is given equal weightage.

Working Assumption
Adopt the following absolute delay parameters for RMa scenarios.

Agreement
Defines upper bound of absolute delay in indoor office scenario as 2L/c where L is the largest dimension of the office hall.
No upper bound of absolute delay in UMi, UMa, and RMa scenario.


Conclusion
For the following scenarios, there is no consensus to update pathloss models due to lack of consistent and significant observed difference between model and measurements.
UMa LOS/NLOS

Agreement
Update the UMa LOS and NLOS AOD spread as follows:
Note: the update parameter was generated using all measurement and ray tracing data set from Rel-14 SI and (current) Rel-19 SI and dividing the data points into 3 groups, below 6 GHz, 6 to 24 GHz, and above 24 GHz, and weighting the data sets for each group to perform weighted least square curve fit without frequency dependency. If a group has fewer data points, higher weight per data point is calculated. All points within a group have same weight. Sum of weights for all groups is equal to 1.
Note: Each group is given equal weightage.


Agreement
For UE antenna modeling, support the following direction antenna radiation pattern for calibration purposes.

Agreement
Reference UE orientation vector of the handheld UE is perpendicular to the plane of the flat UE handheld device, and reference point for near field phase calculation of the UE is assumed to be the center of the plane of the UE handheld.

Agreement
For cases when a candidate antenna placement location is used for two distinct antenna polarization field pattern:
Reference radiation pattern of the UE antenna model is
For first antenna field pattern: and .
For second antenna field pattern:  and .

Agreement
Update and agree to the following the SMa LOS pathloss working assumption:

Agreement
Update and agree to the following the SMa NLOS pathloss:

Agreement
For suburban scenario, adopt the LOS probability

3GPP TSG RAN WG1 Meeting #120bis	R1-2503067
Wuhan, China, April 7th – 11st, 2025

Source: 	Moderator (Intel Corporation)
Title:	
Agenda item:	9.8.1
Document for:	Discussion

Conclusions from RAN1 #118bis
Conclusion
Based on measurement data provided, RAN1 concludes that following delay spread parameters for 6 – 24 GHz frequency range are validated and no updates to TR are needed.
InH LOS and NLOS

Agreement
Update the UMi LOS and NLOS delay spread as follows:
Note: the update parameter was generated using all measurement and ray tracing data set from Rel-14 SI and (current) Rel-19 SI and dividing the data points into 3 groups, below 6 GHz, 6 to 24 GHz, and above 24 GHz, and weighting the data sets for each group to perform weighted least square curve fit. If a group has fewer data points, higher weight per data point is calculated. All points within a group have same weight. Sum of weights for all groups is equal to 1.
Note: Each group is given equal weightage. 

Agreement
Update the UMa LOS and NLOS delay spread as follows:
Note: the update parameter was generated using all measurement and ray tracing data set from Rel-14 SI and (current) Rel-19 SI and dividing the data points into 3 groups, below 6 GHz, 6 to 24 GHz, and above 24 GHz, and weighting the data sets for each group to perform weighted least square curve fit. If a group has fewer data points, higher weight per data point is calculated. All points within a group have same weight. Sum of weights for all groups is equal to 1.
Note: Each group is given equal weightage.

Agreement
Update the UMi LOS and NLOS AOA spread as follows:
Note: the update parameter was generated using all measurement and ray tracing data set from Rel-14 SI and (current) Rel-19 SI and dividing the data points into 3 groups, below 6 GHz, 6 to 24 GHz, and above 24 GHz, and weighting the data sets for each group to perform weighted least square curve fit. If a group has fewer data points, higher weight per data point is calculated. All points within a group have same weight. Sum of weights for all groups is equal to 1.
Note: Each group is given equal weightage.

Agreement
Update the UMa LOS and NLOS AOA spread as follows:
Note: the update parameter was generated using all measurement and ray tracing data set from Rel-14 SI and (current) Rel-19 SI and dividing the data points into 3 groups, below 6 GHz, 6 to 24 GHz, and above 24 GHz, and weighting the data sets for each group to perform weighted least square curve fit. If a group has fewer data points, higher weight per data point is calculated. All points within a group have same weight. Sum of weights for all groups is equal to 1.
Note: Each group is given equal weightage.

Working Assumption
Adopt the following absolute delay parameters for RMa scenarios.

Agreement
Defines upper bound of absolute delay in indoor office scenario as 2L/c where L is the largest dimension of the office hall.
No upper bound of absolute delay in UMi, UMa, and RMa scenario.


Conclusion
For the following scenarios, there is no consensus to update pathloss models due to lack of consistent and significant observed difference between model and measurements.
UMa LOS/NLOS

Agreement
Update the UMa LOS and NLOS AOD spread as follows:
Note: the update parameter was generated using all measurement and ray tracing data set from Rel-14 SI and (current) Rel-19 SI and dividing the data points into 3 groups, below 6 GHz, 6 to 24 GHz, and above 24 GHz, and weighting the data sets for each group to perform weighted least square curve fit without frequency dependency. If a group has fewer data points, higher weight per data point is calculated. All points within a group have same weight. Sum of weights for all groups is equal to 1.
Note: Each group is given equal weightage.


Agreement
For UE antenna modeling, support the following direction antenna radiation pattern for calibration purposes.

Agreement
Reference UE orientation vector of the handheld UE is perpendicular to the plane of the flat UE handheld device, and reference point for near field phase calculation of the UE is assumed to be the center of the plane of the UE handheld.

Agreement
For cases when a candidate antenna placement location is used for two distinct antenna polarization field pattern:
Reference radiation pattern of the UE antenna model is
For first antenna field pattern: and .
For second antenna field pattern:  and .

Agreement
Update and agree to the following the SMa LOS pathloss working assumption:

Agreement
Update and agree to the following the SMa NLOS pathloss:

Agreement
For suburban scenario, adopt the LOS probability

3GPP TSG RAN WG1 Meeting #120bis	R1-2503130
Wuhan, China, April 7th – 11st, 2025

Source: 	Moderator (Intel Corporation)
Title:	
Agenda item:	9.8.1
Document for:	Discussion

Conclusions from RAN1 #118bis
Conclusion
Based on measurement data provided, RAN1 concludes that following delay spread parameters for 6 – 24 GHz frequency range are validated and no updates to TR are needed.
InH LOS and NLOS

Agreement
Update the UMi LOS and NLOS delay spread as follows:
Note: the update parameter was generated using all measurement and ray tracing data set from Rel-14 SI and (current) Rel-19 SI and dividing the data points into 3 groups, below 6 GHz, 6 to 24 GHz, and above 24 GHz, and weighting the data sets for each group to perform weighted least square curve fit. If a group has fewer data points, higher weight per data point is calculated. All points within a group have same weight. Sum of weights for all groups is equal to 1.
Note: Each group is given equal weightage. 

Agreement
Update the UMa LOS and NLOS delay spread as follows:
Note: the update parameter was generated using all measurement and ray tracing data set from Rel-14 SI and (current) Rel-19 SI and dividing the data points into 3 groups, below 6 GHz, 6 to 24 GHz, and above 24 GHz, and weighting the data sets for each group to perform weighted least square curve fit. If a group has fewer data points, higher weight per data point is calculated. All points within a group have same weight. Sum of weights for all groups is equal to 1.
Note: Each group is given equal weightage.

Agreement
Update the UMi LOS and NLOS AOA spread as follows:
Note: the update parameter was generated using all measurement and ray tracing data set from Rel-14 SI and (current) Rel-19 SI and dividing the data points into 3 groups, below 6 GHz, 6 to 24 GHz, and above 24 GHz, and weighting the data sets for each group to perform weighted least square curve fit. If a group has fewer data points, higher weight per data point is calculated. All points within a group have same weight. Sum of weights for all groups is equal to 1.
Note: Each group is given equal weightage.

Agreement
Update the UMa LOS and NLOS AOA spread as follows:
Note: the update parameter was generated using all measurement and ray tracing data set from Rel-14 SI and (current) Rel-19 SI and dividing the data points into 3 groups, below 6 GHz, 6 to 24 GHz, and above 24 GHz, and weighting the data sets for each group to perform weighted least square curve fit. If a group has fewer data points, higher weight per data point is calculated. All points within a group have same weight. Sum of weights for all groups is equal to 1.
Note: Each group is given equal weightage.

Working Assumption
Adopt the following absolute delay parameters for RMa scenarios.

Agreement
Defines upper bound of absolute delay in indoor office scenario as 2L/c where L is the largest dimension of the office hall.
No upper bound of absolute delay in UMi, UMa, and RMa scenario.


Conclusion
For the following scenarios, there is no consensus to update pathloss models due to lack of consistent and significant observed difference between model and measurements.
UMa LOS/NLOS

Agreement
Update the UMa LOS and NLOS AOD spread as follows:
Note: the update parameter was generated using all measurement and ray tracing data set from Rel-14 SI and (current) Rel-19 SI and dividing the data points into 3 groups, below 6 GHz, 6 to 24 GHz, and above 24 GHz, and weighting the data sets for each group to perform weighted least square curve fit without frequency dependency. If a group has fewer data points, higher weight per data point is calculated. All points within a group have same weight. Sum of weights for all groups is equal to 1.
Note: Each group is given equal weightage.


Agreement
For UE antenna modeling, support the following direction antenna radiation pattern for calibration purposes.

Agreement
Reference UE orientation vector of the handheld UE is perpendicular to the plane of the flat UE handheld device, and reference point for near field phase calculation of the UE is assumed to be the center of the plane of the UE handheld.

Agreement
For cases when a candidate antenna placement location is used for two distinct antenna polarization field pattern:
Reference radiation pattern of the UE antenna model is
For first antenna field pattern: and .
For second antenna field pattern:  and .

Agreement
Update and agree to the following the SMa LOS pathloss working assumption:

Agreement
Update and agree to the following the SMa NLOS pathloss:



Agreement
For suburban scenario, adopt the LOS probability










Agreement
Update the UMi LOS and NLOS AOA spread as follows:
Note: the update parameter was generated using all measurement and ray tracing data set from Rel-14 SI and (current) Rel-19 SI and dividing the data points into 3 groups, below 6 GHz, 6 to 24 GHz, and above 24 GHz, and weighting the data sets for each group to perform weighted least square curve fit. If a group has fewer data points, higher weight per data point is calculated. All points within a group have same weight. Sum of weights for all groups is equal to 1.
Note: Each group is given equal weightage.

Agreement
Update the UMa LOS and NLOS AOA spread as follows:
Note: the update parameter was generated using all measurement and ray tracing data set from Rel-14 SI and (current) Rel-19 SI and dividing the data points into 3 groups, below 6 GHz, 6 to 24 GHz, and above 24 GHz, and weighting the data sets for each group to perform weighted least square curve fit. If a group has fewer data points, higher weight per data point is calculated. All points within a group have same weight. Sum of weights for all groups is equal to 1.
Note: Each group is given equal weightage.

Agreement
Adopt the following absolute delay parameters for InH scenarios.
Table 7.6.9-1: Parameters for the absolute time of arrival model

Agreement
Adopt the following correlation distance for SMa spatial consistency
Table 7.6.3.1-2 Correlation distance for spatial consistency
Agreement
Update SMa description as follows:
In suburban macro-cells base stations are located above the rooftops and clutter (e.g. vegetation) to allow wide area coverage, and mobile stations are outdoors at street level. Buildings are typically low residential detached houses with one or two floors, or blocks of apartments/condos or commercial buildings with a few floors. Occasional open areas such as parks or playgrounds between the houses make the environment rather open. Streets do not form urban-like regular strict grid structure. In suburban areas, vegetation is more prevalent than in urban areas with a high variability of foliage density across different suburban areas.

Agreement
For suburban scenario, adopt the following assumptions for calibration purposes:
ISD = 1299m and 1732m
Note: For evaluation in study items/work items, if needed, admission control policies may be considered in conjunction with the ISD choice to address out of coverage UEs.

Agreement
Endorse R1-2503060 to be captured as reference for TR38.901.
R1-2503060 contains the latest update of measurement data sources for channel model as part of the Rel-19 channel modeling enhancements for 7-24 GHz SI.

Agreement
Draft CR R1-2503129 to TR38.901 is endorsed in principle.
Agreement:
For standard multi-panel glass penetration loss model, clarify that reference thickness is 3 cm. no changes to the model.
For IRR glass penetration loss model, update the model equation as follows:

Single coating for IRR glass is assumed.
For concrete penetration loss model, clarify that reference thickness is 23 cm. no changes to the model.
For wood penetration loss model, clarify that reference thickness is 3.3cm. no changes to the model.
For plywood penetration loss model:
1.03 + 0.17f
Reference thickness is 1.27 cm
Note: plywood penetration loss model is intended to be used for SMa penetration loss modeling.

Agreement
To potentially resolve the UMi/UMa pathloss convergence beyond breakpoint distance, RAN1 to further discuss following options:
Option 1) Set  to 0 m
Option 2) Applying only PL1 when performing link analysis
Option 3) add note that provide information on pathloss convergence phenomena beyond the breakpoint distance across different frequencies.
Other options are not precluded.


Working assumption
Proposed Working assumption
Update the UMa LOS, NLOS, and O2I Cluster AOD spread as follows:
Note: the update parameter for LOS and NLOS was generated using all measurement and ray tracing data set from Rel-14 SI and (current) Rel-19 SI and dividing the data points into 3 groups, below 6 GHz, 6 to 24 GHz, and above 24 GHz, and weighting the data sets for each group to perform weighted least square curve fit. If a group has fewer data points, higher weight per data point is calculated. All points within a group have same weight. Sum of weights for all groups is equal to 1.  Each group is given equal weightage.
Note: the update parameters were generated from scaling of updated NLOS UMa O2I value with the ratio of NLOS UMa NLOS and O2I measurement data fit from (current) Rel-19 SI at 3.7 GHz.



Working assumption
Proposed Working assumption
Update the UMa O2I AOD spread as follows:
Note: the update parameters were generated from scaling of updated NLOS UMa O2I value with the ratio of NLOS UMa NLOS and O2I measurement data fit from (current) Rel-19 SI at 3.7 GHz.
Note: UMa O2I ASD values of Rel-14 TR38.901 used values from UMa O2I ASD values of TR36.873. UMa O2I ASD values of TR36.873 used UMi O2I ASD values from ITU-RM.2135-1. IMT-2020 modeling used the same values from TR36.873. TR25.996 does not contain UMa O2I values. The UMi O2I ASD parameters were derived from Winner II report.

Conclusion
RAN1 to continue study the number of clusters for InH, UMi, and UMa scenarios, and intra-cluster power angular profile modeling for all scenarios.
Study to further consider distribution of the number of clusters, resolvable clusters and subpaths of the measurements.

Conclusion
Further study on introduction of polarization variability for each rays of cluster for NLOS component of the channel as an optional modeling component.
Polarization variability is introduced by amplitude scaling parameter  for XPR matrix as follows:


The study should include distribution of the amplitude scaling parameters, power normalization of the amplitude scaling.

Agreement
Adopt the following equations for scaling of angles for CDL channel model. (This updates existing agreements and working assumption).
Option 0) Scaling applies to both NLOS and LOS angles for all cluster paths, Mean Angle and AS include both NLOS and LOS cluster paths but not subpaths








Table X: Example Scale factor values for each CDL model
Note: the values are computed by
		(A-3)
		(A-4)


Working Assumption
SMa O2I will use the same values as SMa NLOS


Agreement
For suburban scenario, adopt the following ZSD parameters as updated working assumption:
Values in [ ] are working assumption



Working Assumption
Introduce new penetration loss outdoor-to-indoor (O2I) building penetration loss model applicable for SMa:
Values in [ ] are working assumption

Agreement
Adopt the following absolute time of arrival parameters for SMa
Values in [ ] are working assumption

Agreement
Adopt the following changes to Clause 7.2 of TR38.901.
Adopt the following changes to Clause 6.2 of TR38.901.

Working Assumption
Reference orientation of the handheld UE as follows:
FFS how to capture the reference orientation of the handheld UE


For calibration of handheld UE, use the following UE rotation based on reference UE orientation
For calibration with blockage:
For one-hand blockage, ΩUT,α = 0 – 360 deg, ΩUT,β = 45 deg, ΩUT,γ = 0 deg, 
For dual-hand blockage, ΩUT,α = 0 – 360 deg, ΩUT,β = 45 deg, ΩUT,γ = 90 deg, 
For hand and head blockage, ΩUT,α = 0 – 360 deg, ΩUT,β = 90 deg, ΩUT,γ = 0 deg, 
For all other calibration cases:
ΩUT,α = 0 – 360 deg, ΩUT,β = 45 deg, ΩUT,γ = 0 deg, 
Note: UT array orientation is defined by three angles ΩUT,α (UT bearing angle), ΩUT,β (UT downtilt angle) and ΩUT,γ (UT slant angle).
Example of ΩUT,α = 0 - 360 deg, ΩUT,β = 90 deg, ΩUT,γ = 0 deg

Example of ΩUT,α = 0 - 360 deg, ΩUT,β = 90 deg, ΩUT,γ = 90 deg



Agreement
For CPE devices adopt the following device dimensions for UE antenna modeling:
(x cm × y cm × z cm) = (0 cm × 20 cm × 20 cm)
Total 9 candidate antenna locations in 4 corners of the vertical square plane, 4 center of the edges of the vertical square plane, and center of the vertical square plane. The figure below captures the side view on the device.
Note: a candidate antenna location (e.g., 9) can be used for multiple antennas


Agreement
For UE antenna modeling of handheld devices, introduce optional antenna imbalance modeling as part of antenna field pattern as follows:
No imbalance is modelled by default.
If modelled, Randomized loss is applied per UE antenna port. Details of the randomized value and it’s distribution will not be specified in TR 38.901 as part of Rel-19 SI.
If modelled, Randomized loss can be applied independently for the UL and DL directions.


Agreement
Confirm the downtilt value for SMa with ISD of 1299m, and introduce downtilt value [93] for SMa with ISD of 1732m
Downtilt value for SMa with ISD of 1732m is working assumption