Failures in first transmission of pdsch when using 256qam

Hello 5G Experts.
In 5G NR SA.
I have serious issue, I’m getting at first transmission pdsch almost more than half attempts of first transmitions "failures " (failure decoding).
But on second/third re-transmissions almost succeeded in all failed previous attempts.

Why it’s happenning? Any idea or suggestion how do I fix that issue?

Note- this issue is happen just when I use 256qam , with 64 qam all fine!!!
My (ss-PBCH-BlockPower) is:

Range : MIN_SSB_POWER - MAX_SSB_POWER --> 
<!-- ssb_power for Cell 0 --> 
<SsbPower>-25</SsbPower>

Try to change SLIV.
You can also check, when network used 256QAM what is EVM, check whether EVM is high or low.

To what value?
I mean, how SLIV could affect here?

Bler is high so EVM is high of course.
What I see currently is, with MCS 20, I see ~26% BLER and MCS 22, its ~36% BLER
Again issue happen (decoding pdsch failure at first transmition) just when using 256qam.
When I use 64qam all is fine and no issue seen.

If UE reports a specific CQI, for 64 QAM, why do you expect to use 256 QAM?

Issue seen when I use 256qam, which UE also reports specific CQI for 256qam not for 64qam.
At 64 qam UE report specific CQI for 64 qam and all works fine.
Now, when used 256qam, which UE support and reports specific CQI for 256qam issue seen. :frowning:

Raise this issue to network L1.
Are you testing in lab area?

How did you force the UE to report CQI for 64qam and for 256qam?

Could it be due to EVM related to 256QAM.
In 4G eNb used to reduce power for all UEs during high load and there was a switch in Huawei and feature in Nokia to reduce power only in the TTIs of high load and for the UEs in need and not all Users.
Is your cell high loaded?

I checked EVM for 256qam, it’s fine.

Yes.

I haven’t forced UE to report CQI. It was only reporting HARQ ACK/NACK, not CQI.
Attach was not going through when I completely removed the attenuation.
So I kept it at 5 dB.
RSRP was around -75 dBm.
In all cases irrespective of SSB power configured (issue isn’t with SSB power - all fine with it).
There is no impact to DL throughput with SSB powers of -21, -15, -10, -5. In fact, it has slightly degraded with higher SSB power.
In UE logs, in PDSCH stats, we see that BLER is better with increased SSB power (BLER of 23% with -5 or -10 and BLER of 30% with -21).

Checked EVM, it’s fine.

It’s one single UE.