Manila Voice Network: Experience and Performance in a Controlled Network Test

The Philippine mobile telecommunications market is undergoing a transformation as operators expand 5G networks ahead of the planned 3G sunset in December 2026. This shift requires operators to migrate both data and voice traffic to modern 4G LTE and 5G architectures. The three operators follow different voice strategies that translate into measurable differences in call quality and call connection performance for consumers.

Ookla® conducted independent, controlled network testing across Manila from 17 to 22 February 2026 to assess real-world performance across the country’s three mobile operators, applying Ookla’s controlled testing methodology with the latest Samsung flagship handset. The evaluation spanned 1,200 kilometers and collected more than 23,000 controlled test samples across voice and data, capturing how each operator’s network architecture translates into measurable user experience. The drive route covered 14 cities and municipalities: Manila, Caloocan, Quezon City, Marikina, San Mateo, San Juan, Mandaluyong, Pasig, Makati, Pasay, Taguig, Parañaque, Las Piñas, and Muntinlupa.

Key Takeaways

VoLTE penetration in the Philippines remains well below global and regional levels. GSMA Intelligence data places VoLTE at 30.9% of total mobile connections in the Philippines in 2025, against a global average of 57.7% and regional peers such as Thailand at 65.2%, with Malaysia and Singapore both above 90%. VoLTE delivers HD-grade voice where deployed, so the gap reflects how many connections have yet to benefit from that experience, underscoring the migration required ahead of the December 2026 3G shutdown.

Operator voice architectures translate into measurable differences in call setup time, with all three operators delivering HD-grade voice quality once connected. Smart and Globe route over 99% of voice through mature VoLTE networks, delivering median call setup of 1.65 seconds on Globe and 2.30 seconds on Smart. DITO’s 5G Standalone (5G SA) core uses Evolved Packet System Fallback (EPS-FB) for voice, recording a median call setup of 4.59 seconds. All three operators delivered HD-grade voice quality, so the differences among them appear in call setup time rather than audio quality.

Consumers experience HD-grade audio quality on operator voice calls across all three networks, while Over-the-Top (OTT) voice apps such as WhatsApp deliver Fair audio quality. Mean Opinion Score (MOS), the standard 1-to-5 measure of perceived audio quality, rated HD-grade on mobile-to-mobile calls for every operator, while OTT calls rated Fair across operators, reflecting the typical exposure of VoIP packets to congestion in a shared internet pipeline.

Consumers in weaker coverage areas continue to experience HD-grade audio on operator-managed VoLTE calls, while OTT calls degrade further. In moderate-to-low Reference Signal Received Power (RSRP) conditions, VoLTE calls remained HD-grade across all three operators. OTT calls slipped lower within the Fair range at the cell edge. Charting MOS against RSRP shows VoLTE holding within a 4.0 to 4.5 band (Good to Excellent) as coverage weakens, while OTT slopes from around 3.7 toward 2.9 (Fair into Poor and Bad) in the weakest coverage bins. The pattern reflects how dedicated voice bearers protect call quality where best-effort data cannot.

The Philippine mobile market is transitioning to modern VoLTE infrastructure

Although 4G LTE has been commercially available in the Philippines since mid-2012, Voice over LTE (VoLTE)—the technology that enables voice calls over the high-speed 4G network—has taken longer to deploy. The shift toward modern voice standards accelerated significantly between 2020 and 2021. In November 2020, Smart Communications officially announced the completion of its nationwide VoLTE rollout, making the service accessible to both postpaid and prepaid subscribers. Globe Telecom followed a similar path, initially launching VoLTE in Manila in May 2020 and progressively expanding coverage nationwide throughout 2021. The 2021 commercial launch of DITO Telecommunity also relied on VoLTE for calls, as it does not operate legacy 2G/3G networks.

The Philippines is transitioning toward VoLTE and away from legacy networks, driven by wider 4G coverage and the rising adoption of VoLTE-compatible smartphones. GSMA Intelligence data places VoLTE connections at 30.9% of total mobile connections in the Philippines in 2025, below the global average of 57.7% and well behind regional peers such as Thailand (65.2%), Malaysia (96.1%), and Singapore (94.9%). The same source forecasts VoLTE will cross the majority threshold in 2027 and reach 70.4% of connections by 2030, pointing to the eventual phase-out of legacy 2G and 3G voice networks. The regulator, the National Telecommunications Commission (NTC), has mandated a phased shutdown, requiring the full nationwide decommissioning of 3G services by December 31, 2026.

Philippines VoLTE Connections as a Share of Total Mobile Connections (2021–2030)

Source: GSMA Intelligence | 2021 – 2030

Breaking down the Philippines market by operator, GSMA Intelligence data for 2025 shows VoLTE accounting for 41.9% of DITO’s connections, the highest of the three operators, and consistent with DITO’s 4G/5G-only network. Smart follows at 31.7% and Globe at 27.4%. With the regulatory push for a complete nationwide 3G shutdown looming, legacy operators are being forced to accelerate subscriber migration, directly shaping the country’s VoLTE adoption landscape.

VoLTE Share of Operator's Total Connections (%)

Source: GSMA Intelligence | 2025

As the Philippines’ third mobile operator, DITO operates on a 4G/5G-only network built from the ground up, making VoLTE mandatory for voice calls. Since DITO operates on 5G Standalone (5G SA) with limited native Voice over New Radio (VoNR) support, devices camped on DITO’s 5G network must temporarily step down to the 4G LTE network to establish a VoLTE voice call, a technical process called Evolved Packet System Fallback (EPS-FB).

In contrast, Smart and Globe manage massive subscriber bases that still include users with older devices. While both operators process nearly all of their active 4G voice traffic through mature VoLTE architecture, migrating the remaining legacy 3G devices ahead of the 2026 NTC deadline will depend heavily on regulatory support, since consumer device upgrades are constrained by affordability and cannot be driven by operators alone.

Core network architectures dictate how operators route voice calls

To capture the complete voice experience on a mobile network, controlled network testing evaluates both traditional mobile-to-mobile (M2M) connections and app-based digital calls using VoLTE-capable devices. Mobile-to-mobile voice testing measures standard phone calls that utilize dedicated, high-priority pathways established directly by the cellular network. These dedicated channels ensure reliable, immediate connections for standard voice traffic by separating it from standard internet usage.

Conversely, Over-the-Top (OTT) testing evaluates the performance of voice calls routed through third-party internet applications. For this controlled test, the methodology simulated OTT voice calls via WhatsApp, one of the most widely used messaging platforms. OTT applications function as Voice over IP (VoIP) services, compressing and transmitting voice data as digital packets over a standard internet connection.

An analysis of the mobile-to-mobile voice call distribution reveals a sharp contrast in how operators route standard voice calls. Smart and Globe rely almost exclusively on VoLTE for their voice traffic, with 99.94% and 99.85% of test samples on VoLTE, respectively. VoLTE allows these operators to connect calls using dedicated 4G voice channels without switching network types. Testing revealed that DITO routed 88.89% of its traditional voice calls using Evolved Packet System Fallback (EPS-FB), with only 11.11% routed directly over VoLTE.

Mobile-to-Mobile Voice Traffic Routing by Underlying Network Technology

Source: Speedtest Drive™ | February 2026

The distribution of call network technology shifts when examining OTT call results. The results show that Globe and Smart primarily used E-UTRAN New Radio-Dual Connectivity (EN-DC) for 90.12% and 84.15% of their application-based calls, respectively. EN-DC is a 5G Non-Standalone (5G NSA) technology that allows a mobile device to connect to both 4G and 5G networks simultaneously, leveraging existing 4G infrastructure to anchor the high-speed 5G data connection. In contrast, DITO completed 89.27% of its OTT calls exclusively over 5G SA, with the remaining 10.73% falling back to standard LTE.

OTT Voice Traffic Routing by Underlying Network Technology

Source: Speedtest Drive™ | February 2026

Operators delivered consistent high-definition voice clarity in Manila

The controlled test measured Mean Opinion Score (MOS) for both mobile-to-mobile and OTT calls using the Perceptual Objective Listening Quality Analysis (POLQA) algorithm to evaluate the overall consumer voice experience in Manila. MOS POLQA objectively analyzes audio signals to predict human perception of speech clarity, filtering out differences in device hardware, on a standardized 1-to-5 scale. Results are grouped into standard ITU-T MOS quality categories on the 1-to-5 scale: Excellent (4.3 to 5.0), Good (4.0 to 4.3), Fair (3.6 to 4.0), Poor (3.1 to 3.6), and Bad (1.0 to 3.1). Calls in the Good and Excellent range, with MOS at or above 4.0, are referred to as HD-grade and deliver clear, high-quality audio, while Fair reflects noticeable but tolerable quality loss, and Poor or Bad indicates degraded calls. Mobile-to-mobile testing measures the fidelity of dedicated cellular voice pathways, while OTT testing evaluates the audio quality of third-party platforms — in this case, WhatsApp — over data networks.

OTT voice quality was measured using a simulated WhatsApp client rather than the live application, because end-to-end encryption, proprietary codecs, and frequent app updates make the production app impractical to benchmark consistently. The simulated client replicates WhatsApp’s codec and audio behavior. Because the test placed two co-located devices on shared radio conditions and routed calls through a hosted SIP server, the OTT MOS values represent a controlled, best-case path for OTT voice — real-world WhatsApp calls between parties in different locations on live internet infrastructure may perform below these figures.

Mean Opinion Score (MOS) Rating by Operator, Manila

Source: Speedtest Drive® | February 2026

Across the full test sample, all three operators delivered HD-grade audio on mobile-to-mobile voice calls, while OTT calls rated Fair across operators. Mobile-to-mobile calls rated Excellent on DITO and Smart and Good on Globe. OTT applications like WhatsApp use the OPUS codec to compress voice into data packets that travel alongside standard web traffic, and OTT calls rated Fair on all three operators, reflecting the expected exposure of VoIP packets to congestion in a shared internet pipeline.

Codec note: During the drive, Globe’s mobile-to-mobile calls predominantly negotiated the AMR-WB codec, while EVS was the dominant codec on the other two networks in the same drive. EVS is a mature, widely deployed codec in the region, and Globe’s network is understood to be EVS-ready. The codec used on any given call is determined by the negotiation between the device and the network. On the Samsung test devices used in this drive, the IMS/VoLTE profile applied for Globe, as provisioned through the device vendor’s certification process, did not enable EVS, so the negotiation defaulted to AMR-WB. This is consistent with a device-side profile outcome, specific to the Samsung test devices, not a limitation of the codecs supported by Globe’s network. The MOS result above reflects perceived quality under AMR-WB and is not directly comparable to an EVS-based measurement.

For Manila subscribers, mobile-to-mobile calls deliver consistently HD-grade clarity across all three operators, while OTT calls rate Fair across the full Manila drive test.

Operator voice holds up in weak coverage where OTT calls degrade

Voice network performance varies most where coverage weakens. We filtered the test sample to measurements collected in moderate-to-low RF coverage, defined as Reference Signal Received Power (RSRP) at or below -100 dBm and Signal-to-Interference-plus-Noise Ratio (SINR) at or below 5 dB. These conditions occur on outdoor routes that sit at distance from the nearest cell site, along corridors where dense high-rise construction blocks line-of-sight at street level, and on peripheral stretches of the drive route where cell density is lower. Approximately 9% of VoLTE samples and 10% of OTT samples fell within this band.

The share of measurements falling into the moderate-to-low band varied by operator. Globe recorded the smallest cell-edge sample share on both call types, with 5.8% of its VoLTE measurements and 7.3% of its OTT measurements in this coverage band, indicating more uniform RF coverage across the tested Manila routes. Smart recorded approximately 10% on VoLTE and 12% on OTT in the moderate-to-low band, while DITO recorded approximately 11% on both call types.

Cell-Edge Mean MOS Rating by Operator and Call Type, Manila

Source: Speedtest Drive™ | February 2026 | RSRP ≤ -100 dBm and SINR ≤ 5 dB

Within the cell-edge sample, VoLTE held HD-grade across all three operators, sustaining the clarity recorded on the full mobile-to-mobile sample. OTT calls slipped lower within the Fair range across operators, below the level they recorded on the full sample. The pattern was consistent across operators — operator-managed VoLTE preserved audio quality at the cell edge, while shared-bandwidth OTT voice degraded further.

Mean MOS by RSRP and SINR, Cell-Edge Subset, Manila

Source: Speedtest Drive™ | February 2026

Mobile-to-mobile VoLTE mean MOS held within a stable 4.0 to 4.5 band (Good to Excellent) across most of the RSRP distribution, sustaining HD-grade audio across the cell-edge band. The curve is largely flat, which means VoLTE quality stayed near its full-sample baseline regardless of how the radio link moved.

OTT mean MOS tracked a different shape. From around 3.7 at the cell-edge band entry, OTT MOS declined as RSRP weakened, falling below 3.0 (into Poor and Bad) in the weakest, well-populated bins. The same divergence holds when MOS is plotted against SINR — mobile-to-mobile VoLTE remains largely in the 3.8 to 4.5 band (Good to Excellent), while OTT spans a wider range that dips below 3.0 in the deepest interference bins. The shape of the curves matters as much as the cell-edge averages. VoLTE’s flat profile reflects active protection of the voice bearer. OTT’s downward slope reflects a connection that responds directly to the underlying radio link.

The divergence reflects how each voice path handles network resources. VoLTE calls are carried on a dedicated quality-of-service bearer that the network prioritizes over best-effort traffic, and the IMS core can refuse a call setup if it cannot guarantee quality. OTT voice rides over standard data bearers, sharing capacity with web browsing, video streaming, and other internet traffic. As RF degrades, OTT packets compete for limited radio resources, which translates into higher packet loss and lower mean MOS. The flat VoLTE curve under weakening RSRP is the visible signature of admission control and dedicated bearer prioritization at work.

OTT voice calls connect faster than traditional cellular calls, but shared internet bandwidth raises call block rates

Beyond the cell-edge view, the test captured three call connection metrics across the full sample to characterize how successfully calls connect and persist under representative network conditions. Call setup time tracks how quickly a network establishes a connection, the block rate measures initial connection failures, and the drop rate identifies active calls that terminate unexpectedly.

VoLTE Call Connections Metrics, Per Operator, Manila

Source: Speedtest Drive® | February 2026

VoLTE voice performance reflects the operator’s underlying network infrastructure. Globe led in initial connectivity, with a median VoLTE call setup time of 1.65 seconds, followed by Smart at 2.30 seconds. Both operators benefit from mature VoLTE networks that provide dedicated voice channels. VoLTE has benefited from a decade of global tuning, including optimized SIP signaling (the protocol that initiates and manages voice calls), dedicated QoS bearers (reserved data channels that guarantee consistent call quality), and robust SRVCC handovers (Single Radio Voice Call Continuity, which seamlessly transfers calls to older networks when needed).

DITO recorded a median VoLTE call setup time of 4.59 seconds. As the only provider operating on a 5G SA network in the market, DITO relies heavily on EPS-FB. In controlled testing, 88.89% of mobile-to-mobile call samples transitioned back to 4G upon connection, acting as a bridge when the SA footprint or operational maturity is incomplete.

Call block rates varied across operators on the VoLTE path. Globe recorded the lowest at 0.47%, with Smart at 0.59% and DITO at 0.73%. Once connected, call drop rates were uniformly low across the VoLTE sample, ranging from 0.04% on DITO to 0.20% on Globe, with Smart at 0.09%.

OTT Voice Call Connection Metrics, Per Operator, Manila

Source: Speedtest Drive™ | February 2026

OTT voice calls achieved faster median call setup times than VoLTE calls on all three operators. Globe recorded a median OTT call setup time of 1.62 seconds, narrowly faster than its 1.65-second VoLTE median. DITO bypassed EPS-FB for OTT calls and recorded a median OTT setup time of 1.72 seconds, compared with its 4.59-second VoLTE median. Smart recorded a median OTT setup time of 1.91 seconds compared with 2.30 seconds on VoLTE.

The size of the VoLTE-versus-OTT setup time gap reflects each network’s architectural maturity. Globe and Smart, with mature VoLTE deployments, recorded VoLTE setup times within 0.03 to 0.39 seconds of their OTT setup times — the dedicated voice bearer adds little overhead on top of the underlying data path. DITO’s 2.87-second gap reflects the structural cost of the EPS-FB step-down required to establish a VoLTE call on its 5G SA core.

OTT voice packets share bandwidth with regular internet traffic such as web browsing and video streaming, leaving them more vulnerable to network congestion. Heavy data traffic can disrupt data packets, impacting initial connections and causing interruptions to active calls. OTT block rates were materially higher than VoLTE block rates across all operators — Globe at 1.64%, DITO at 2.33%, and Smart at 2.76% — consistent with the higher congestion exposure of best-effort data bearers. OTT call drop rates ran in a similar low range to VoLTE, between 0.12% on Globe, 0.16% on Smart, and 0.19% on DITO.

OTT’s marginal setup-time advantage at the median is best read alongside the cell-edge MOS finding established earlier. The OTT path achieves a small speed savings on the median by routing voice through best-effort data bearers, but the same path lacks the dedicated bearer protection that holds VoLTE call quality at the cell edge.

Each operator’s architecture shapes how consumers experience voice in Manila

Manila’s controlled benchmarking captures how each operator’s voice architecture translates into measurable consumer experience. Globe and Smart route voice over mature VoLTE, delivering fast call setup. DITO’s 5G SA core uses EPS Fallback to route mobile-to-mobile voice through its LTE layer, recording a longer median setup time. Across the full sample, operator voice calls on every network deliver HD-grade audio to consumers, while OTT calls rate Fair. In weaker coverage areas, operator-managed VoLTE held HD-grade across all three operators, while OTT fell further. These voice architecture differences will shape how each operator handles the December 2026 NTC 3G shutdown.

To learn more about Ookla’s controlled drive and walk testing with Speedtest Drive and first-party crowdsourced data from Speedtest®, please contact us.