The shift from Non-Standalone (NSA) to Standalone (SA) 5G represents a major milestone for mobile network operators. Voice over New Radio (VoNR) promises better voice quality, lower latency, and more efficient spectrum use. But deploying VoNR means 4G becomes a fallback rather than your primary voice platform.

For operators planning this transition, the real question is: how do you maintain voice reliability if your 5G footprint is smaller than your legacy 4G coverage? Especially when running voice services on mid-band spectrum, which has less geographic reach than low-band LTE carriers.

The answer lies in careful planning, the right fallback mechanisms, and understanding where things can go wrong.

The reliability challenge: Building a better safety net

When a VoNR call starts on 5G, everything works beautifully - until the subscriber moves out of coverage. Without proper fallback mechanisms, you risk dropped calls. The solution involves multiple layers of protection, each one designed to catch calls that the previous layer couldn't handle.

Layer one: Low Band NR

Most NR deployments utilize mid-band carriers for 5G, since they offer the most capacity. Where possible, however, operators should convert at least one low-band carrier to 5G and use it for VoNR to improve overall voice performance. With low-band spectrum being used for VoNR, fewer transitions to VoLTE will be required, increasing overall VoNR usage & minimizing potential disruptions. And if the lowest available band (i.e., the one with the best coverage) is used for NR, transitions to VoLTE may not even be required. This is the most elegant option for ensuring voice-call continuity & reliability.

Take 600/700 MHz as an example. An operator that repurposes these bands as an NR carrier can carry VoNR traffic natively on the same spectrum that provides their deepest, widest coverage footprint. The strategic value is significant because 600/700 MHz propagates so far and penetrates buildings so well, a subscriber moving away from a mid-band 5G cell can be handed down to this low-band by the network – using fast & efficient inter-frequency handovers. These are preferable to inter-RAT transitions down to VoLTE, since those generally take longer and involves more steps. That said, the inter-RAT handovers, critical redirections, and VoLTE fallback machinery described later in this article are still worth having as a backstop; but, in a well-designed low-band NR deployment, they become last-resort mechanisms rather than everyday traffic handlers. The operational goal becomes straightforward: extend that low-band NR carrier as broadly as possible so that the VoNR layer is always the widest net available.

With this approach, one thing becomes critical – keeping low-band spectrum as free as possible for voice traffic in the first place. This vision is at least partially a load-balancing problem, and Samsung is addressing that with an AI-powered approach. Our Load Balancing Manager (LBM) solution is designed to move data users to higher bands to maximize throughput intelligently - leaving low-band capacity available for voice users. While LBM is not a voice-specific solution, its effect complements the fallback strategies described in this blog: the less low-band spectrum is congested with data traffic, the more reliably it can serve as the coverage layer, which VoNR and fallback mechanisms depend on.

Layer two: Inter-RAT handover

The second line of defense is graceful inter-RAT (Inter-Radio Access Technology) handover from 5G to 4G. As a subscriber moves out of 5G coverage, the network attempts to hand the call down to an LTE eNodeB. This handover sounds straightforward, but a lot is happening under the hood.

First, you need properly configured neighbor relations between your 5G gNodeBs and LTE eNodeBs. The Automatic Neighbor Relations (ANR) feature should handle this, but operators need to verify that these relations are built correctly before commercial launch. Voice calls are time-sensitive, and you can't afford to wait for Cell Global Identifier (CGI)-based ANR to build neighbor relations on the fly during an active call.

The handover procedure itself requires coordination between the Access and Mobility Management Function (AMF) in the 5G core and the Mobility Management Entity (MME) in the 4G core. When the device lands on LTE after handover, it must complete a Tracking Area Update (TAU) procedure to register with the 4G core. This transition point between 5G and 4G cores is actually one of the most common pain points in field testing. Get the core integration wrong, and even perfect radio-layer handovers won't save you.

One strategy that helps: requiring a handover to an LTE carrier with broader coverage than your mid-band 5G layer before the subscriber completely exits 5G coverage. By mandating a handover to that carrier while the subscriber is still within reach, you gain more margin for error and reduce the need for critical fallback procedures.

Layer three: Critical Redirection

But what happens when inter-RAT handover fails? Maybe the handover preparation didn't complete, or the device couldn't find any LTE cells to report. This rare condition is where critical redirection - sometimes called blind redirection - becomes your last line of defense.

When inter-RAT handover fails, the network can perform a critical redirection to an LTE frequency from a predefined list. The network essentially tells the device: "We're releasing you. Go find this LTE frequency. If you can't find it, try the others on your list." From the network's perspective, this is treated as a release because the Radio Access Network (RAN) no longer controls what happens next.

The term "blind" means the network doesn't know whether a frequency is available at the subscriber's location. It's making an educated guess based on your deployment strategy & configuration. In most cases - field testing shows it works extremely well - the device finds an LTE carrier, and the call continues. But unlike a graceful handover, you lose visibility into the success rate from RAN KPIs. You'll need to rely on IMS (IP Multimedia Subsystem) metrics to understand how well critical redirections are working.

The other downside: these redirections cause audio interruptions. A brief period might occur where audio packets are lost but it’s a far better alternative to a complete dropped call.

The optimization challenge here is setting the right threshold (Received Signal Reference Power (RSRP)/Received Signal Reference Quality (RSRQ)/Signal to Noise Ratio (SINR) for triggering critical redirection. Set it too high, and you'll redirect unnecessarily when a graceful handover might have worked. Set it too low, and you risk dropping calls before the redirection kicks in. Finding that sweet spot requires careful drive testing and tuning.

Quality matters: Maintaining performance on VoNR

Reliability is only half the equation. Once subscribers are on VoNR, you need to ensure voice quality matches or exceeds what they experienced on VoLTE. This functionality is especially important at the cell edge, where signal quality degrades.

VoLTE solved this problem by bundling Transmission Time Interval (TTI) and improving uplink coverage in extreme cell-edge conditions. Physical Uplink Shared Channel (PUSCH) repetition (or slot aggregation) performs a similar operation in VoNR.

⦁ TTI bundling in LTE allows multiple consecutive TTIs to be bundled into a single transport block, improving the uplink link budget for VoLTE by enabling higher transmission power over a longer duration.

⦁ PUSCH repetition in 5G NR achieves similar coverage enhancement by repeating the same PUSCH transmission across multiple slots, providing the network with multiple opportunities to decode voice packets successfully.

NR's approach is more flexible than LTE's TTI bundling because 5G already has variable slot durations and numerologies. The repetition can be configured dynamically based on coverage needs using DCI, eliminating the need for an additional RRC Reconfiguration Message that was previously necessary with TTI-B. NR also supports configured grant transmissions for VoNR that can include repetition parameters.

Samsung's approach: Extensive commercial experience

Samsung has been working with global operators to refine VoNR deployment. What started as lab testing evolved into field trials and eventually commercial deployment across multiple markets and countries. The experience has been valuable in identifying real-world pain points and developing solutions that work across different deployment scenarios.

All of the features described here: inter-RAT handover optimization, critical redirection, PUSCH repetition, are supported in Samsung's 5G RAN portfolio on both virtualized RAN (vRAN) and traditional RAN deployments. Samsung also offers VoNR specific features that intelligently move users between bands, based on metrics such as congestion, UL voice quality, and more. This helps the network adapt to changing traffic patterns automatically, minimizing the need for user intervention.

Currently, this software operates on a deterministic basis rather than relying on AI or ML models, allowing operators to configure and tune it to fit their specific network architecture. That said, the industry’s evolutionary path points firmly toward AI/ML augmentation. 3GPP Release 17 established the study framework for AI-enabled RAN intelligence, and Release 18 (5G-Advanced) moved that work into specifications - targeting mobility optimization, load balancing, and handover failure reduction as priority use cases. For VoNR specifically, AI/ML-driven mobility prediction promises to replace today’s static RSRP/RSRQ/SINR thresholds with dynamic, context-aware handover triggers that can anticipate when a subscriber is about to leave VoNR coverage and initiate fallback earlier and more accurately. Samsung’s roadmap tracks this trajectory, and operators deploying VoNR today on deterministic foundations will be well-positioned to layer in AI/ML optimization as those capabilities mature.

Planning your VoNR transition

The move to VoNR touches your core network, IMS platform, and operational procedures. Operators who succeed will verify neighbor relations before launch, thoroughly test core integration, carefully tune RSRP/RSRQ/SINR thresholds, monitor both RAN and IMS metrics, and leverage LTE bands as a bridge for smaller 5G footprints. Looking ahead, operators should also plan for AI/ML-driven optimization of handover and fallback decisions - a capability being normalized in 3GPP Release 18 and beyond that will bring dynamic threshold management and proactive mobility prediction to VoNR deployments.

With the right preparation and technology partner, like Samsung, operators can make this transition without compromising reliability.