Hyper Open Edge Cloud

Enhancing MIMO Precoding in Amarisoft vRAN for Real-World Network Performance

Field and lab tests reveal that Amarisoft’s strict 3GPP-compliant MIMO precoding, while optimal in lab conditions, can underperform in real-world networks with non-compliant UEs. Adapting mode selection logic with additional quality metrics could significantly improve downlink throughput.
  • Last Update:2025-09-02
  • Version:001
  • Language:en

Introduction

MIMO precoding strategies are critical to achieving optimal throughput in modern 4G and 5G networks. While Amarisoft’s vRAN solution adheres closely to 3GPP specifications and performs well in test-and-measurement laboratory conditions, recent field observations indicate that strict compliance may not always translate to optimal performance in live deployments.

Through extensive testing in both dense-urban pilot sites and controlled laboratory environments, we have identified an opportunity to enhance Amarisoft’s MIMO mode selection logic to better adapt to real-world conditions, particularly when UEs do not comply strictly with standards.

Background

Amarisoft’s vRAN and eNodeB/gNodeB implementations are designed primarily for lab testing and research, with a strong emphasis on 3GPP compliance. In such controlled environments, all network elements and UEs behave predictably, making standard-compliant algorithms optimal.

However, real-world deployments often introduce deviations:

  • Third-party core network, such as certain Huawei CN deployments, have been observed to deviate from 3GPP specifications, with concrete evidence from field measurements. (The topic is not covered in this blog)
  • User equipment variability, different UE models and firmware versions may report CQI or RI (Rank Indicator) in ways that do not fully reflect current radio conditions.

Under these circumstances, a purely standard-compliant implementation may miss opportunities to adapt transmission modes for higher efficiency. This is not a bug in Amarisoft’s software, but rather a deployment adaptation opportunity.

Current Implementation and Limitations

Amarisoft’s downlink transmission mode selection relies strictly on UE-reported RI values:

  • RI = 1: Transmit Diversity (Tx-Div)
  • RI = 2: Tx-CDD (open-loop MIMO)
  • RI = 3: MIMO 3 layers
  • Ri = 4: MIMO 4 layers
  • etc.

While correct per specification, this approach assumes that the UE’s RI reporting is both timely and accurate. In practice, some UEs may continue reporting RI=2 even when link quality has deteriorated, resulting in suboptimal throughput.

By contrast, some vendors of ZHEN (ZTE, HUAWEI, Ericsson, Nokia)  network incorporate additional real-time metrics into their decision-making, allowing them to revert to more robust transmission modes when needed.

Test Methodology

We conducted a series of comparative tests between Amarisoft and ZHEN in the lab, field test report as supplementary material.

Laboratory setup included:

  • 2×2 MIMO (TM=3) configuration with identical radio hardware
  • Simbox UE to create controlled RI reporting scenarios
  • Keysight Propsim for precise AWGN noise addition
  • ZHEN B3 RRU 2212 vs Amarisoft Callbox, tested well above hardware noise floor
  • 64QAM table at low SINR (ZHEN adjusts table automatically; Amarisoft changes manually)
  • Note: Tests conducted well above the RRU and SDR radios’ noise floor confirmed that the performance gap was unrelated to radio hardware quality.

Key Findings

  • Forced RI=1 scenario:
    ZHEN dynamically switched between Tx-Div and spatial multiplexing ~90% of the time, maintaining BLER <10% and achieving roughly twice the throughput of Amarisoft.
    Amarisoft remained in Tx-Div for the entire duration, resulting in significantly lower throughput.
  • With CQI reporting enabled:
    Amarisoft experienced a 20–40% throughput loss at SINR <13 dB compared to ZHEN.
  • With CQI reporting disabled:
    Amarisoft matched theoretical benchmark performance, confirming that hardware was not a limiting factor.


orange line = ZHEN eNB; grey line = Amarisoft Callbox.

Performance matches theoretical benchmark when CQI is disabled.

Analysis

The root cause appears to be Amarisoft’s reliance solely on UE CQI/RI reporting for both MCS selection and MIMO mode switching. In cases where this reporting is inaccurate or delayed, throughput suffers.

Evidence suggests that ZHEN, for example, uses dynamic thresholds when selecting the best MCS and deciding whether to switch transmission modes. This allows their scheduler to adapt to imperfect UE behavior in real time.

Potential Improvement Approaches

  • Hybrid MIMO mode selection logic: Incorporate additional downlink quality metrics (CQI trends, HARQ performance, BLER) alongside RI to trigger fallback when conditions degrade.
  • Adaptive thresholds: Use dynamically adjusted thresholds rather than fixed values for MCS and transmission mode switching.
  • Field validation during development: Simulate proposed algorithms in the lab, then validate them on pilot sites where non-3GPP-compliant UE behavior can be observed.

Amarisoft's Next Steps

Amarisoft has acknowledged the potential for improvement in cases where UEs do not behave as expected. They plan to:

  • Simulate possible enhancements to scheduler strategy
  • Implement and test changes to ensure no unintended side effects
  • Provide an update in September

We have offered Amarisoft access to one of our pilot sites should they wish to conduct real-world validation during the research stage.

Conclusion

Amarisoft’s strict 3GPP-compliant approach ensures reliability in lab environments, but real-world deployments demand adaptive strategies to handle non-ideal UE (and CN) behavior. By incorporating additional real-time performance indicators into MIMO mode selection, Amarisoft could significantly close the throughput gap observed in the deployment. Such an enhancement would not be a bug fix, but a deployment adaptation improvement, bringing lab-grade software closer to carrier-grade performance in the field.