High BLER in UL with 4x2 100Mhz and OFH

[IbrahimBabiker87]

with OFH and 100Mhz / RAN650 benetel UL UDP/TCP is hsowing very High UL BLER ~16% while LA settings were set to target 1% BLER in UL
The maximum uplink throughput with 7:2 TDD pattern should be around ~72Mbps
Below is the logs:
gnb.zip

           -----------------DL-----------------------|------------------UL--------------------
 pci rnti  cqi  ri  mcs  brate   ok  nok  (%)  dl_bs | pusch  mcs  brate   ok  nok  (%)    bsr
   1 4601   15   4   27    14k    4    0   0%      0 |  24.5   28    60M  310   40  11%   300k
   1 4601   15   4   27    11k    3    0   0%      0 |  24.4   28    57M  297   53  15%   300k
   1 4601   15   4   27    14k    4    0   0%      0 |  24.3   28    57M  294   56  16%   300k
   1 4601   15   4   27   9.6k    3    0   0%      0 |  24.3   28    58M  300   50  14%   300k
   1 4601   15   4   27    11k    3    0   0%      0 |  24.1   28    58M  302   48  13%   300k
   1 4601   15   4   27    14k    4    0   0%      0 |  24.4   28    54M  280   70  20%   300k
   1 4601   15   4   27   9.6k    3    0   0%      0 |  24.6   28    58M  301   49  14%   300k
   1 4601   15   4   27   9.6k    3    0   0%      0 |  24.6   28    56M  291   59  16%   300k
   1 4601   15   4   27    12k    4    0   0%      0 |  24.5   28    56M  291   59  16%   300k
   1 4601   15   4   27   9.6k    3    0   0%      0 |  24.5   28    55M  285   65  18%   300k
   1 4601   15   4   27    11k    3    0   0%      0 |  24.5   28    54M  282   68  19%   300k

           -----------------DL-----------------------|------------------UL--------------------
 pci rnti  cqi  ri  mcs  brate   ok  nok  (%)  dl_bs | pusch  mcs  brate   ok  nok  (%)    bsr
   1 4601   15   4   27   9.6k    3    0   0%      0 |  24.4   28    54M  280   70  20%   300k
   1 4601   15   4   27    12k    4    0   0%      0 |  24.3   28    58M  301   49  14%   300k
   1 4601   15   4   27   9.6k    3    0   0%      0 |  24.5   28    58M  299   51  14%   300k
   1 4601   15   4   27    14k    4    0   0%      0 |  24.6   28    56M  293   57  16%   300k
   1 4601   15   4   27   9.6k    3    0   0%      0 |  24.5   28    58M  300   50  14%   300k
   1 4601   15   4   27   8.6k    3    0   0%      0 |  24.2   28    55M  288   62  17%   300k
   1 4601   15   4   27    14k    4    0   0%      0 |  24.3   28    58M  301   49  14%   300k
   1 4601   15   4   27    12k    3    0   0%      0 |  24.6   28    56M  293   57  16%   300k
   1 4601   15   4   27    15k    4    0   0%      0 |  24.5   28    58M  299   51  14%   300k
   1 4601   15   4   27   9.6k    3    0   0%      0 |  24.4   28    57M  297   53  15%   300k
   1 4601   15   4   27    12k    3    0   0%      0 |  24.3   28    57M  298   52  14%   300k

           -----------------DL-----------------------|------------------UL--------------------
 pci rnti  cqi  ri  mcs  brate   ok  nok  (%)  dl_bs | pusch  mcs  brate   ok  nok  (%)    bsr
   1 4601   15   4   27    12k    4    0   0%      0 |  24.2   28    61M  318   32   9%   300k
   1 4601   15   4   27   9.6k    3    0   0%      0 |  24.3   28    56M  291   59  16%   300k
   1 4601   15   4   27    11k    4    0   0%      0 |  24.2   28    59M  305   45  12%   300k
   1 4601   15   4   27   8.6k    3    0   0%      0 |  24.3   28    57M  298   52  14%   300k
   1 4601   15   4   27    14k    4    0   0%      0 |  24.6   28    56M  292   58  16%   300k
   1 4601   15   4   27    12k    3    0   0%      0 |  24.6   28    57M  297   53  15%   300k
   1 4601   15   4   27   9.6k    3    0   0%      0 |  24.5   28    57M  294   56  16%   300k
   1 4601   15   4   27    12k    4    0   0%      0 |  24.5   28    56M  291   59  16%   300k
   1 4601   15   4   27   9.6k    3    0   0%      0 |  24.5   28    57M  298   52  14%   300k
   1 4601   15   4   27   9.6k    3    0   0%      0 |  24.1   28    57M  296   54  15%   300k
   1 4601   15   4   27    15k    4    0   0%      0 |  24.5   28    58M  299   51  14%   300k

[IbrahimBabiker87]

Is RX diversity implemented in Srsran, If so is there ant specific settings other than nof_antennas_ul:
for instance for below settings Im using two antennas for UL ( MIMO is not supported yet ), this will be treated as rx diversity with 2 antennas or 4 antennas, The more the receiving antennas the more the gain, is it recommended to use such settings or its better to have only 1 tx port and antenna.

dl_port_id: [0,1,2,3]                
ul_port_id: [0,1,2,3]      
nof_antennas_dl: 4                                              
nof_antennas_ul: 2                                            

[IbrahimBabiker87]

HI @pgawlowicz ,
Still there is BLER in Uplink but when Using benetel " worse with RX diversity i.e. two port in RX", while for N310 its works fine with RX diversity " 2 RX" but with BLER when using one port " 1 RX"

Benetel with one RX >>> BLER but better than with 2RX
Benetel with two RX >>> High BLER
N310 with one RX >>> Low BLER.
N310 with two RX >>> No BLER

[pgawlowicz]

Interesting, thanks for your feedback!
Could you try Benetel with the new firmware (v1.0.4)? You can find more details here #487

[IbrahimBabiker87]

@pgawlowicz ,

with Benetel 1.0.4 for both 20 & 40 MHz and with RX diversity the BLER is almost eliminated " didn't test yet with 80 & 100 MHz",

20 MHz:

40 MHz:

[pgawlowicz]

@andrepuschmann please check nice results here related to the new Benetel firmware

[IbrahimBabiker87]

Dear @pgawlowicz ,

Its good to know that for above results I was using additional DMRS as part of my config:
dmrs_additional_position: 2

I think this is why the BLER went down at a cost of less throughput,

I tested again with Additional DMRS = 0

20Mhz with additional DMRS=0:

20Mhz with additional DMRS=3:

40Mhz with additional DMRS=0:

40Mhz with additional DMRS=2:

[andrepuschmann]

Hey @IbrahimBabiker87,
thanks for testing with the new firmware and configs. Those results look very good, for 40Mhz it seems DMRS=0 vs DMRS=2 is almost no difference.

On a separate note @IbrahimBabiker87 . I know you've shared some details about your test setup before but would you mind explaining the cabling, components, attenuation, etc. in full detail and maybe share a drawing? Feel free to reach out to me directly if that suits you better.

[aibtw]

Hey @IbrahimBabiker87, we're trying to achieve this kind of system performance with the RAN550, but the BLER is still high, +30% (as in my issue #513). Were there other modifications you made, other than DMRS parameter, upgrading firmware and uplink diversity?
Would really appreciate sharing working configs if you don't mind.