3GPP TSG-RAN WG4 Meeting #98bis-e R4-2106868

Electronic Meeting, 12 April – 20 April 2021

Agenda Item: 8.14.2.2

Source: Ericsson

Title: Discussion on MMSE-IRC receiver for intra-cell inter-user interference

Document for: Discussion

1 Introduction
RAN#91-e approved the revision of WI Further enhancement on NR demodulation performance Error!
Reference source not found., where one of the objective is to UE demodulation requirements for MMSE-
IRC receiver for suppressing intra-cell inter-user interference as follows:

 MMSE-IRC receiver for suppressing intra-cell inter-user interference

o Phase I: Evaluate the performance under practical MU-MIMO interference profile for the
candidate reference receiver.

 Identify practical MU-MIMO interference modelling methodology

 Reference receiver: MMSE-IRC receiver. Use the DMRS-based interference

covariance estimation method as a starting point

 Prioritize slot-based transmission scenario

o Phase II: Define the requirements if needed based on the outcome of phase I

o Target frequency: FR1

o Rx antenna number: 2Rx and 4Rx for FR1

 Decide if it is feasible that the UE requirements are release independent from Rel-15.

According to WID, RAN4 will also decide the UE requirements are release independent from Rel-15, if it is
feasible.

In this contribution, we show our view on UE demodulation requirements with MMSE-IRC receiver
suppressing intra-cell inter-user interference.

2 Modeling of Intra-cell inter-user interference

Firstly, RAN4 need to decide how to model the intra-cell inter-user interference for MMSE-IRC receiver
performance requirements. LTE supports the multi-user transmission with TM7, TM8, TM9, and TM10, and
therefore RAN4 specifies UE demodulation requirements assuming inter-user interference with the
parameter ‘simultaneous transmission’. In LTE, the inter-user interference model is summarized as follows:

1
  Modulation symbols of an interference signal are random mapped onto one of the remaining
DMRS antenna port (e.g., 7, 8, 11, or 13, depending on the configured DMRS ports). The update
granularity for randomized mapping antenna port is 1 PRG in frequency domain and 1ms in time
domain.

RAN4 can reuse this model for intra-cell inter-user interference, but this model does not specify what
precoding is applied for the co-scheduled PDSCH. Since NR is designed to use MU-MIMO features explicitly,
we think RAN4 should specify more practical inter-user interference model.

In Rel-16 eMIMO WI and performance enhancement WI, RAN4 discussed the feasibility of PMI (e)Type-II
reporting requirements assuming MU-MIMO transmission setup. Although these WIs specified the PMI
reporting requirements based on SU-MIMO setup, we could reuse the methodology discussed in these WIs.

In MU-MIMO setup discussed in Rel-16, it was assumed the PDSCH of the co-scheduled (interfering) UE is
precoded with a precoder derived from the PMI reported by UE under test as follows:

𝑾 = 𝑯𝐻 (𝑯𝑯𝐻 + 𝜆𝑰)−1
𝑾 = [𝑾𝒂 𝑾𝒃 ]
𝑯𝒂
𝑯 = [ 𝒃]
𝑯
Where, 𝑾𝒂 is the precoder applied for UE under test, 𝑾𝒃 is the precoder applied for co-scheduled UE,
𝑯𝒂 is the channel estimates at UE under test, and 𝑯𝒃 is the channel at the co-scheduled UE. See
Appendix for the detailed definitions.
As we discussed in Rel-16, if we assume 𝜆 is large enough (i.e., high SNR test points), precoder is simply
given as follows:

𝑯𝒂 𝐻
𝑾 ≈ 𝑯𝑯 = [ ]
𝑯𝒃
That is,
𝐻
𝑾𝒂 ≈ (𝑯𝒂 )𝐻 , 𝑾𝒃 ≈ (𝑯𝒃)

Since 𝑾𝒂 = (𝑯𝒂 )𝑯 is equivalent to the PMI reported by UE under test, this is the same test setup as the
UE demodulation requirements, that is, TE selects 𝑾𝒂 randomly from the PMI codebook (e.g., Single
𝐻
panel type-I). Since 𝑾𝒃 = (𝑯𝒃) , TE also select 𝑾𝒃 randomly from the PMI codebook, but it is better to
ensure 𝑾𝒂 ≠ 𝑾𝒃 . This method is very simple and easily implemented in the TE.

Proposal 1: Specify intra-cell inter-user interference model as follows:

Precoded modulation symbols of an interference signal are randomly mapped onto one of the remaining
DMRS antenna port.
The update granularity for randomized mapping antenna port is PRB bundling size in frequency domain
and slot in time domain.

For each slot and each PRB bundling size, a precoding matrix for the number of layers, 𝑣, associated to
the selected rank shall be selected randomly from [Single Panel Type I]. The selected precoding matrix
shall not be identical to the precoding matrix applied for the UE under test.

Note we could also apply Type-II precoder if we assume 4Tx for UE demodulation.

2
3 Simulation setup for evaluation

3.1 DMRS configuration

For the MU-MIMO setup, in order to avoid the interference between two (or more) users, DMRS symbols of
UE under test and co-scheduled UE should be allocated in the different OFDM symbols. Therefore, when
DMRS ports 1000/1001 are assigned to the UE under test, DMRS ports 1002/1003 are assigned to the co-
scheduled UE.
Proposal 2: For inter-user interference scenario, configure 2 DMRS CDM groups without data, that is,
DMRS ports 1000/1001 are assigned for the UE under test and DMRS ports 1002/1003 are assigned for
the co-scheduled UE.

3.2 Simulation parameters

This is our initial view on the simulation parameters for the evaluation.

Antenna configuration

If we assume 2Tx for gNB, the maximum rank is limited to one. But we are also interesting to consider 4Tx
for gNB so that both the UE under test and co-scheduled UE are configured with rank 2.

MCS for UE under test

We would like to study the following two options for UE under test. Note we need discuss the modulation
used for co-scheduled UE.

 MCS4 (QPSK CR=0.3)

 MCS13 (16QAM CR=0.48)

 MCS19 (64QAM CR=0.5)

Proposal 3: For the MMSE-IRC receiver performance evaluation, RAN4 should assume the following
parameters as the starting point.

3
 Parameters FDD TDD
CBW 10MHz 40MHz
TDD configuration N/A 7DS2U,
S=6D+4G+4U
Schedule PDSCH in special slots
SCS 15kHz 30kHz
PDSCH configuration Type A, Start symbol 2, Duration 12 Type A, Start symbol 2, Duration 12
PRB bundling size 2 2
PDSCH DMRS configuration Type 1, 1+1 Type 1, 1+1
DMRS port assignment 1000 for UE under test for rank 1 1000 for UE under test for rank 1
1000/1001 for UE under test for rank 1000/1001 for UE under test for rank
2 2
1002 for co-scheduled UE for rank 1 1002 for co-scheduled UE for rank 1
1002/1003 for co-scheduled UR for 1002/1003 for co-scheduled UR for
rank 2 rank 2
MCS MCS 4 (QPSK 0.3) MCS 4 (QPSK 0.3)
MCS 13 (16QAM 0.48) MCS 13 (16QAM 0.48)
MCS 19 (64QAM 0.5) MCS 19 (64QAM 0.5)
Rank 1 and 2 1 and 2
PDSCH precoder Single Panel Type I, Random Single Panel Type I, Random
precoder selection updated per slot, precoder selection updated per slot,
with equal probability of each with equal probability of each
applicable i1, i2 combination, and applicable i1, i2 combination, and
with PRB bundling granularity with PRB bundling granularity
Antenna configuration 2x2, 2x4 for rank 1 2x2, 2x4 for rank 1
4x2, 4x4 for rank 2 4x2, 4x4 for rank 2
Propagation condition TDLA30-10 TDLA30-10
Metric SNR to achieve 70% of maximum SNR to achieve 70% of maximum
throughput throughput

Proposal 4: For evaluation RAN4 should compare the performance difference between the case with co-
scheduled UE and without co-scheduled UE.

4 Summary
Proposal 1: Specify intra-cell inter-user interference model as follows:

Precoded modulation symbols of an interference signal are randomly mapped onto one of the remaining
DMRS antenna port.

The update granularity for randomized mapping antenna port is PRB bundling size in frequency domain
and slot in time domain.

For each slot and each PRB bundling size, a precoding matrix for the number of layers, 𝑣, associated to
the selected rank shall be selected randomly from [Single Panel Type I]. The selected precoding matrix
shall not be identical to the precoding matrix applied for the UE under test.

Proposal 2: For inter-user interference scenario, configure 2 DMRS CDM groups without data, that is,
DMRS ports 1000/1001 are assigned for the UE under test and DMRS ports 1002/1003 are assigned for
the co-scheduled UE.

Proposal 3: For the MMSE-IRC receiver performance evaluation, RAN4 should assume the following
parameters as the starting point.

4
 Parameters FDD TDD
CBW 10MHz 40MHz
TDD configuration N/A 7DS2U,
S=6D+4G+4U
Schedule PDSCH in special slots
SCS 15kHz 30kHz
PDSCH configuration Type A, Start symbol 2, Duration 12 Type A, Start symbol 2, Duration 12
PRB bundling size 2 2
PDSCH DMRS configuration Type 1, 1+1 Type 1, 1+1
DMRS port assignment 1000 for UE under test for rank 1 1000 for UE under test for rank 1
1000/1001 for UE under test for rank 1000/1001 for UE under test for rank
2 2
1002 for co-scheduled UE for rank 1 1002 for co-scheduled UE for rank 1
1002/1003 for co-scheduled UR for 1002/1003 for co-scheduled UR for
rank 2 rank 2
MCS MCS 4 (QPSK 0.3) MCS 4 (QPSK 0.3)
MCS 13 (16QAM 0.48) MCS 13 (16QAM 0.48)
MCS 19 (64QAM 0.5) MCS 19 (64QAM 0.5)
Rank 1 and 2 1 and 2
PDSCH precoder Single Panel Type I, Random Single Panel Type I, Random
precoder selection updated per slot, precoder selection updated per slot,
with equal probability of each with equal probability of each
applicable i1, i2 combination, and applicable i1, i2 combination, and
with PRB bundling granularity with PRB bundling granularity
Antenna configuration 2x2, 2x4 for rank 1 2x2, 2x4 for rank 1
4x2, 4x4 for rank 2 4x2, 4x4 for rank 2
Propagation condition TDLA30-10 TDLA30-10
Metric SNR to achieve 70% of maximum SNR to achieve 70% of maximum
throughput throughput

Proposal 4: For evaluation RAN4 should compare the performance difference between the case with co-
scheduled UE and without co-scheduled UE.

References
[1] RP-210920, “Revised WID: Further enhancement on NR demodulation performance”, China Telecom.

Appendix
Consider gNB schedule PDSCH to two UEs (one for UE under test (user a) and another for co-schedule UE
(user b).

5
 AWGN

gNB h11 y11

UE under test
(user a)
s1 y12
Precoder
W
s2 y21
Co-scheduled
UE
H (user b)
y22

gNB has 4 Tx antenna ports and each UE has 2 Rx antenna ports. The channel is modeled as follows:
(𝑎) (𝑎) (𝑎) (𝑎)
ℎ11 ℎ12 ℎ13 ℎ14 𝒉𝒂𝟏
(𝑎) (𝑎) (𝑎) (𝑎)
ℎ21 ℎ22 ℎ23 ℎ24 𝒉𝒂𝟐 𝑯𝒂
𝑯= (𝑏) (𝑏) (𝑏) (𝑏) = 𝒃 = [ 𝒃]
ℎ11 ℎ12 ℎ13 ℎ14 𝒉𝟏 𝑯
𝒃
(𝑏) (𝑏) (𝑏) (𝑏) [𝒉 𝟐 ]
[ℎ21 ℎ22 ℎ23 ℎ24 ]

That is, the channel vectors to j-th Rx antenna port for user a and b are given as follows:
(𝑎) (𝑎) (𝑎) (𝑎)
𝒉𝒂𝒋 = [ℎ𝑗1 ℎ𝑗2 ℎ𝑗3 ℎ𝑗4 ]

(𝑏) (𝑏) (𝑏) (𝑏)

𝒉𝒃𝒋 = [ℎ𝑗1 ℎ𝑗2 ℎ𝑗3 ℎ𝑗4 ]

If gNB schedules PDSCH with 2 layers for each UE, the transmit signal is given by
𝑠1𝑎
𝑠2𝑎 𝒔𝒂
𝑺 = 𝑏 = [ 𝒃]
𝑠1 𝒔
𝑏
[𝑠2 ]

Where 𝑠𝑖𝑎 means the PDSCH with layer i scheduled to user a, and 𝑠𝑖𝑏 means the PDSCH with layer i
scheduled to user b.

If gNB applies the precoder 𝑾 before transmission, the received signals given as follows:
𝑦1𝑎
𝑦2𝑎 𝒚𝒂 𝑯𝒂 𝒔𝒂
𝒀 = 𝑏 = [ 𝒃 ] = 𝑯𝑾𝒔 + 𝒖 = [ 𝒃 ] 𝑾 [ 𝒃 ] + 𝒖
𝑦1 𝒚 𝑯 𝒔
𝑏
[𝑦2 ]
Where 𝑦𝑖𝑎 means the received signals at i-th Rx antenna for user a, and 𝒖 is AWGN given by
𝑢1𝑎
𝑢2𝑎 𝒖𝒂
𝒖 = 𝑏 = [ 𝒃]
𝑢1 𝒖
𝑏
[𝑢 2 ]
The precoder is given as follows:
6
 (𝑎) (𝑎) (𝑏) (𝑏)
𝑤11 𝑤21 𝑤11 𝑤21
(𝑎) (𝑎) (𝑏) (𝑏)
𝑤12 𝑤22 𝑤12 𝑤22
𝑾= (𝑎) (𝑎) (𝑏) (𝑏) = [𝒘𝒂𝟏 𝒘𝒂𝟐 𝒘𝒃𝟏 𝒘𝒃𝟐] = [𝑾𝒂 𝑾𝒃 ]
𝑤13 𝑤23 𝑤13 𝑤23
(𝑎) (𝑎) (𝑏) (𝑏)
[𝑤14 𝑤24 𝑤14 𝑤24 ]

That is, the precoding vectors of i-th layer for user a and b is given as follows.
(𝑎) (𝑎) (𝑎) (𝑎) 𝑇
𝒘𝒂𝒊 = [𝑤𝑖1 𝑤𝑖2 𝑤𝑖3 𝑤𝑖4 ]

(𝑏) (𝑏) (𝑏) (𝑏) 𝑇

𝒘𝒃𝒊 = [𝑤𝑖1 𝑤𝑖2 𝑤𝑖3 𝑤𝑖4 ]