Massive MIMO Spatial Channel Model Dataset

On this Page

Introduction

Dataset – Massive MIMO Spatial Channel Model Dataset

Community

Get the Qualcomm^® newsletter straight to your inbox.

This dataset contains samples of the frequency domain channel matrix for the channel between user equipment (UEs) and their corresponding serving cell, synthetically generated using the 3GPP Spatial Channel Model defined in TR 38.901. The data can be used to better understand the statistical distribution of channel characteristics in a typical dense urban layout.

Product license agreement

ResourcesForum

No 3D asset configured

Dataset format details

The dataset is organized into several records, one per UE. Each record contains the following variables:

Name	Type	Description
Hf	Complex float array with size (20, 4, 50, 32) (# time samples, #UE ports, #RBs, #gNB ports)	20 time samples (20 ms apart) of frequency domain channel matrix per resource block
UeX_meters	Float	relative X location of UE (meters)
UeY_meters	Float	relative Y location of UE (meters)
UeZ_meters	Float	relative Z location of UE (meters)
ServingCellCouplingLoss_dB	Float	Coupling loss between serving cell and UE (dB)
Outdoor	Logical	1 if UE is outdoor, else 0
LineOfSight	Logical	1 if UE has line-of-sight condition to serving cell, else 0
ServingCellId	Integer	index of serving cell of UE

The following scripts provide examples for how the records can be retrieved and used.

The first script provides examples in Python:

import scipy.io as sio
import numpy as np 
#############################################
#                                           #
#   Example 1: Load all data from one file  #
#                                           #
#############################################
UeRecord = sio.loadmat('UE1.mat', squeeze_me=True)

# This loads the data in UE1.mat into a dictionary called UeRecord:
#  UeRecord['Hf'] : complex numpy array with shape (20, 4, 50, 32) - frequency domain channel matrix per resource block
#  UeRecord['UeX_meters'] : float - relative X location of UE (meters)
#  UeRecord['UeY_meters'] : float - relative Y location of UE (meters)
#  UeRecord['UeZ_meters'] : float - relative Z location of UE (meters)
#  UeRecord['ServingCellCouplingLoss_dB'] : float - coupling loss to serving cell (dB)
#  UeRecord['Outdoor'] : int - 1 if UE is outdoor, 0 otherwise
#  UeRecord['LineOfSight'] : int - 1 if UE has line-of-sight condition to serving cell, 0 otherwise
#  UeRecord['ServingCellId'] : int - index of serving cell of UE

#############################################
#                                           #
#   Example 2: Load Hf data from 10 files   #
#                                           #
#############################################
Hf_all = np.empty((10, 20, 4, 50, 32), dtype='complex')
for i in range(10):
   Hf = sio.loadmat(f'UE{i + 1}.mat', squeeze_me=True, variable_names='Hf')['Hf']
   Hf_all[i] = Hf

The next script provides examples in MATLAB:

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%                                           %
%   Example 1: Load all data from one file  %
%                                           %
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
UeRecord = load('UE1.mat');
% This loads the data in UE1.mat into a struct called UeRecord:
%  UeRecord.Hf : complex array with size (20, 4, 50, 32) - frequency domain channel matrix per resource block
%  UeRecord.UeX_meters : float - relative X location of UE (meters)
%  UeRecord.UeY_meters : float - relative Y location of UE (meters)
%  UeRecord.UeZ_meters : float - relative Z location of UE (meters)
%  UeRecord.ServingCellCouplingLoss_dB : float - coupling loss to serving cell (dB)
%  UeRecord.Outdoor : logical       - 1 if UE is outdoor, 0 otherwise
%  UeRecord.LineOfSight : logical   - 1 if UE has line-of-sight condition to serving cell, 0 otherwise
%  UeRecord.ServingCellId : int - index of serving cell of UE

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%                                           %
%   Example 2: Load Hf data from 10 files   %
%                                           %
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
Hf_all = zeros(10, 20, 4, 50, 32);
for i = 1:10
   load(sprintf('UE%d.mat', i), 'Hf');
   Hf_all(i, :, :, :, :) = Hf; 
end

The index of the serving cell (ServingCellId) refers to an integer from 1 to 57 that indicates the serving sector (or cell) in a two-tier hexagonal layout with 3 sectors per site. The cell locations are according to the following table:

ServingCellId	X_meters	Y_meters
1	0	0
2	0	0
3	0	0
4	0	200
5	0	200
6	0	200
7	173.205	100
8	173.205	100
9	173.205	100
10	173.205	-100
11	173.205	-100
12	173.205	-100
13	0	-200
14	0	-200
15	0	-200
16	-173.205	-100
17	-173.205	-100
18	-173.205	-100
19	-173.205	100
20	-173.205	100
21	-173.205	100
22	0	400
23	0	400
24	0	400
25	173.205	300
26	173.205	300
27	173.205	300
28	346.41	200
29	346.41	200
30	346.41	200
31	346.41	0
32	346.41	0
33	346.41	0
34	346.41	-200
35	346.41	-200
36	346.41	-200
37	173.205	-300
38	173.205	-300
39	173.205	-300
40	0	-400
41	0	-400
42	0	-400
43	-173.205	-300
44	-173.205	-300
45	-173.205	-300
46	-346.41	-200
47	-346.41	-200
48	-346.41	-200
49	-346.41	0
50	-346.41	0
51	-346.41	0
52	-346.41	200
53	-346.41	200
54	-346.41	200
55	-173.205	300
56	-173.205	300
57	-173.205	300

Dataset organization details

Each UE’s record is stored as a separate .MAT file. The .MAT files are grouped into .ZIP files each containing the files for 550 UEs. The total number of UEs is 42180.

Dataset generation details

The data is based on the dense urban layout (macrocell layer only). A two-tier hexagonal layout with 3 sectors per site is assumed, and wrap-around modeling is incorporated. The inter-site distance is 200 meters, the base station antenna height is 25 meters, and the minimum BS-UE distance is set to 35 meters. Spatial consistency is modeled, as described in TR 38.901. Mobility aspects are not considered.

In the spatial domain, massive MIMO is assumed wherein each base station is modeled with 128 antenna elements mapped to 32 ports, and each UE is modeled with 4 antenna elements mapped to 4 ports.

In the frequency domain, a 4 GHz carrier frequency is assumed. The channel data is computed for 20 MHz bandwidth, mapped to 50 resource blocks (RBs).

In the time domain, for each UE, 20 time-samples of the channel are recorded, with successive time samples being 20 milliseconds apart.

The following table captures the above assumptions:

Parameter	Value
Scenario	Dense Urban (Macro only)
Frequency Range	FR1, 4 GHz
Inter-BS distance	200m
Minimum BS-UE 2D distance	35 m
Channel model	UMa, according to TR 38.901
Antenna setup and port layouts at gNB	32 ports: (8,8,2,1,1,2,8), (dH,dV) = (0.5, 0.8)λ
Antenna setup and port layouts at UE	4 ports: (1,2,2,1,1,1,2), (dH,dV) = (0.5, 0.5)λ
BS antenna height	25m
UE antenna height & gain	Follow TR 36.873
Bandwidth	20 MHz (50 RBs)
UE distribution	80% indoor (3km/h), 20% outdoor (30km/h)

Dataset license

Massive MIMO Spatial Channel Model Dataset is available for research purposes.

Massive MIMO Spatial Channel Model Dataset License Agreement

Qualcomm AI Research

AI is shifting from simply seeing what is happening in front of the camera to understanding it. Data is the effective force behind these deep learning breakthroughs and is integral to the human-level performance of neural networks. Our crowd-acting approach to data collection overcomes the typical limitations of crowdsourcing, resulting in high-quality video data that is densely captioned, human-centric and diverse.

Qualcomm AI Research continues to invest in and support deep-learning research in computer vision. The publication of the Jester dataset for use by the AI research community is one of our many initiatives.

Find out more about Qualcomm AI Research.
For any questions or technical support, please contact us at [email protected]