By David Thompson · Published March 28, 2026 · Updated June 8, 2026 · 11 min read
VoIP faxing has a reputation for being unreliable — and that reputation is completely earned. If you've ever watched a fax fail mid-transmission, arrive garbled, or disconnect for no apparent reason, the problem isn't a fluke. It's a fundamental technical mismatch that no amount of tweaking can fully resolve.
Fax was designed in the era of copper telephone lines. VoIP was designed for the internet. These two technologies have almost nothing in common — and when you try to run one through the other, the results are predictably bad.
This guide explains exactly why VoIP faxing fails, from the ground up.
The Bottom Line Up Front
VoIP fax failure rates run 15–20% under normal conditions — compared to ~5% on traditional analog lines. A single-page fax over unoptimized VoIP succeeds only about 80% of the time, according to Dialogic research. Cloud fax services like mFax bypass the problem entirely.
The Fundamental Mismatch: Two Incompatible Technologies
Fax was built for the Public Switched Telephone Network (PSTN) — a circuit-switched system. When you place a fax call on the PSTN, the network establishes a dedicated circuit between sender and receiver for the entire duration of the call. That circuit is continuous, exclusive, and lossless. The audio stream flows uninterrupted from one end to the other.
VoIP works completely differently. It breaks audio into small data packets, assigns each a sequence number, and sends them across a shared IP network. Routers forward packets independently — they may take different paths, arrive out of order, or be dropped entirely when a router becomes congested.
For human voice calls, this is fine. Your brain reconstructs speech even when a few milliseconds go missing. For fax, it is catastrophic.
A fax machine transmits a continuous stream of precisely timed analog tones. The receiving machine decodes those tones in real time to reconstruct the document. There is no buffering, no retry mechanism at the signal level, and zero tolerance for gaps or corruption in the audio stream. VoIP provides none of the guarantees this process requires.
Two Technologies, Two Eras
Fax protocol (T.30) was standardized in 1980. VoIP emerged in the mid-1990s. The two technologies were designed 15 years apart, for entirely different network architectures, with no thought given to compatibility.
How Fax Actually Works: The T.30 Protocol
To understand why VoIP breaks fax, you need to understand how fax works. Every Group 3 fax machine uses the T.30 protocol — an ITU standard that governs the entire session from dial-up to hang-up.
T.30 is not "send a file." It is an elaborate, half-duplex handshake with strict timing requirements at every step:
- Called Station Identification: The receiving machine sends a 2100 Hz CED tone
- Capability exchange: Both machines exchange DIS/DCS frames advertising their speeds and features
- Training: The sending machine transmits a Training Check Frame (TCF) to test line quality
- Confirmation: The receiver sends a CFR (Confirmation to Receive) signal
- Page transmission: Data is sent using modulated carriers — V.27ter at 2400/4800 bps, V.29 at 7200/9600 bps, or V.17 at up to 14,400 bps
- Post-page signaling: Acknowledgment, multi-page sequencing, final disconnect
Every step in this sequence has a hard timing requirement. The T.30 standard specifies that silence between signals must be 75 ±20ms. The T1 timeout (identifying the remote machine) allows only 35 ±5 seconds. The T2 timeout (changing transmit parameters) allows just 6 ±1 seconds.
On a POTS line, these timings are trivially met — the dedicated circuit provides consistent, sub-millisecond latency. On a VoIP network, they frequently are not.
T.30 also has no tolerance for data corruption. Unlike modern protocols such as TCP/IP, T.30 has no mechanism to retransmit individual signal frames. If a tone is corrupted mid-handshake, the session aborts entirely.
The Five Ways VoIP Destroys Fax Signals
1. Packet Loss
Packet loss is normal and expected behavior on IP networks. Routers drop packets when they become congested — it is how IP networks manage traffic. For voice calls, a dropped packet creates a tiny, barely perceptible gap. For fax, it is fatal.
Even 1% packet loss can kill an entire fax transmission. Here's why:
The receiving fax machine is continuously decoding the modulated audio stream into binary data. A dropped packet creates a gap in that stream. The decoder cannot reconstruct the binary data from an incomplete signal. With ECM (Error Correction Mode) enabled, the machine detects the error and aborts. With ECM disabled, the fax technically "completes" — but pages arrive with missing lines, garbled text, or blank sections where the dropped packets were.
The longer the fax, the more chances for a disruptive packet drop. Single-page faxes fail at ~20% over unoptimized VoIP. Multi-page faxes (5+ pages) fail at substantially higher rates.
The 1% Rule
On a traditional POTS line, packet loss is zero — it is a circuit-switched connection. On a shared IP network, even "good" network conditions involve occasional packet loss. And for fax, "occasional" is enough to cause failure.
2. Jitter
Jitter is the variability in packet arrival timing. Even when no packets are dropped, packets that arrive unevenly — some early, some late — produce an audio stream with timing irregularities that fax modems cannot decode.
Fax modems lock onto the carrier frequency and timing of the incoming signal. Jitter disrupts that lock. When the jitter is severe enough, the modem loses synchronization and the handshake fails.
A few important points about jitter:
- Jitter buffers partially compensate by holding incoming packets briefly and releasing them at a steady rate — but jitter buffers add latency, and excessive latency violates T.30 timeouts
- Increasing bandwidth does not reduce jitter — jitter is caused by variable queuing delays in routers, not by insufficient capacity
- On the public internet, controlling end-to-end jitter is essentially impossible — you have no control over how third-party routers queue and forward your packets
3. Codec Compression
This is arguably the single most immediately destructive problem. By default, VoIP uses compressed audio codecs designed to reduce bandwidth consumption:
| Codec | Bitrate | Suitable for Fax? |
|---|---|---|
| G.729 | 8 kbps | No — fax signals at 9,600 bps exceed this |
| G.723.1 | 5.3–6.3 kbps | No — same problem |
| G.711 | 64 kbps (uncompressed) | Yes — the only viable option |
The math for G.729 is blunt: a fax signal running at 9,600 bps cannot survive being compressed through an 8 kbps codec. It is arithmetically impossible. G.729 was engineered to compress human speech by exploiting characteristics unique to the voice — it has no understanding of modem tones and treats them as noise to be discarded.
Even when a VoIP provider technically uses G.711, the switch matters. If a call starts on G.729 (common for voice calls) and the system attempts to renegotiate to G.711 after detecting the CED fax tone, the brief negotiation gap can cause the receiving machine to lose synchronization and abort.
4. Voice Processing Features That Break Fax
VoIP systems include several audio processing features that improve voice call quality — and actively destroy fax transmissions. All of them must be disabled for fax to have any chance:
Voice Activity Detection (VAD) / Silence Suppression: VAD detects when there is no speech and stops transmitting packets to save bandwidth. During a T.30 fax session, there are precise timing pauses between signal exchanges. VAD interprets these as silence and suppresses them — destroying the very timing that T.30 depends on.
Echo Cancellation: Designed to remove line echo from voice calls. Echo cancellers work by identifying signals that resemble the outgoing audio and suppressing them in the incoming stream. Fax tones are steady-frequency signals that can appear similar to echo. The canceller may actively suppress or distort them — or, in some systems, terminate the call when it detects what it believes to be an echo condition.
Comfort Noise Generation (CNG): When VAD suppresses silence, CNG inserts artificial background noise to prevent callers from experiencing unsettling dead air. This injected noise corrupts the fax signal during the gaps the receiving machine is relying on.
Settings You Cannot Always Change
VAD, echo cancellation, and CNG are often enabled by default at the carrier level — not in your equipment. Even if you configure your ATA correctly, your VoIP provider may apply these processes in their infrastructure without telling you.
5. Latency and Clock Drift
T.30's strict timing requirements mean that latency alone can kill a fax session. If acknowledgment signals take too long to arrive — due to network congestion, multi-hop routing, or gateway transcoding delays — the sending machine hits a T.30 timeout and hangs up.
A subtler issue is PCM clock drift: VoIP gateways and ATAs contain independent clocks for their PCM audio processing. These clocks are never perfectly synchronized. Over time — particularly on long multi-page faxes — they drift relative to each other. Jitter buffers can only absorb so much drift before they overflow or underflow, causing audio glitches that corrupt the fax signal. This is why long faxes often fail on VoIP lines even when short faxes succeed.
T.38: The Intended Fix and Its Limitations
The ITU recognized the fax-over-VoIP problem early. In 1998, they standardized T.38 — a protocol designed specifically to carry fax over IP networks without relying on raw audio transport.
Instead of sending the raw fax tones as audio packets, T.38 converts the fax data stream at the gateway into IFP (Internet Fax Protocol) packets. These are sent via UDP or TCP and reconstructed into audio tones at the far-end gateway. This approach:
- Eliminates codec compression entirely — the signal is data, not audio
- Tolerates up to ~1,000ms of network delay (vs. a few hundred ms for G.711 passthrough)
- Uses redundant packet transmission to recover from packet loss without retransmission delays
- Is unaffected by VAD, echo cancellation, and comfort noise
So why doesn't T.38 solve the problem? Several reasons:
Both ends must support it. T.38 requires the gateway at each end of the call to implement the protocol. If the call traverses multiple carrier networks — which it almost always does — and any carrier in the path does not support T.38, the call silently falls back to G.711 passthrough with all its problems.
Many providers never implemented it. Despite being standardized in 1998, T.38 support remains inconsistent across carriers more than 25 years later. Many ITSPs (Internet Telephony Service Providers) simply do not offer it.
Interoperability problems are common. The T.38 standard leaves room for implementation variation. Different vendors' gateways frequently fail to negotiate successfully. Technology forums are full of T.38 interoperability troubleshooting threads involving specific Cisco, AudioCodes, and Patton gateway combinations.
T.38 activation can fail. For T.38 to activate on a SIP call, the originating gateway must send a SIP re-INVITE mid-call when it detects the fax CED tone. This re-INVITE must successfully propagate back to the originating SIP endpoint. In complex multi-carrier architectures, this renegotiation frequently fails — leaving the call on G.711 with no indication that T.38 was never activated.
Sustained packet loss still causes failures. T.38 uses redundancy to compensate for occasional lost packets, but it has limits. If packet loss exceeds what the redundancy settings account for, the transmission fails. And the standard provides no graceful degradation — it either works or it doesn't.
T.38 at a Glance
T.38 was standardized in 1998. It significantly improves VoIP fax reliability — but only when both ends of every carrier hop in the call path support it. That end-to-end guarantee is rarely achievable on public VoIP networks.
VoIP Fax vs. Traditional POTS: The Real Numbers
The reliability gap between traditional and VoIP faxing is not theoretical. It has been measured:
| Factor | POTS (Analog) | VoIP Fax (Unoptimized) |
|---|---|---|
| Circuit type | Dedicated, continuous | Packet-switched, shared |
| Packet loss | Zero | 1% can kill transmission |
| Codec | Native analog (no compression) | G.711 required; G.729 fails |
| Jitter | None | Highly variable |
| Latency | Consistent, predictable | Variable, can exceed T.30 timeouts |
| VAD/Echo cancellation | Not applicable | Must be disabled |
| Failure rate | ~5% | 15–20% |
| Single-page success rate | ~95% | ~80% |
| Multi-page reliability | High | Degrades with page count |
| Max fax speed | 33.6 kbps (V.34) | Effectively 9,600 bps recommended |
| Power dependency | Works without power/internet | Requires both |
The numbers tell a consistent story. Infotel Systems, citing research from Reed University, found POTS fax fails roughly 5% of the time. VoIP fax without optimization fails at 20% — four times the rate. Dialogic's own research found that a single-page fax over normal IP conditions achieves only about 80% success. Historically, fax-over-IP systems have been unable to exceed 85% overall transmission success rates.
Real-World Failure Patterns
If you have experienced VoIP fax problems, you have likely seen one of these specific failure patterns:
Garbled or corrupted pages: Packet loss mid-transmission corrupts the modulated signal the receiving machine is trying to decode. Pages arrive with missing lines, distorted text, or blank sections. With ECM disabled, the fax "succeeds" — the machine hangs up normally — but the document is unusable.
Connection-then-disconnect: The call connects, 1–1.5 pages transmit successfully, and then the VoIP line drops. This is typically caused by echo cancellation: the canceller detects fax tones, mistakes them for a feedback condition, and terminates the call.
Long fax failures: Short faxes succeed; faxes over 5 pages consistently fail. This is PCM clock drift — the jitter buffer gets corrupted over longer transmissions. The workaround (splitting faxes into batches of 5 pages) is a tell-tale sign of a VoIP fax reliability problem.
Fast busy / no answer: The machine dials and gets a fast busy signal. This can happen when the VoIP carrier's capacity is exhausted or when the signaling pathway doesn't support fax call types.
Stutter tone interference: Certain VoIP systems use a stutter dial tone to signal voicemail. A fax machine connected through an ATA hears the stutter tone and interprets it as an error condition, refusing to proceed.
Why Cloud Fax is Structurally Different
Cloud fax services like mFax do not send fax audio over VoIP at all. They use a fundamentally different architecture that sidesteps every problem described in this article.
How VoIP Fax Works
Your document becomes audio tones. Those tones become VoIP packets. Those packets cross an unpredictable internet path where they can be lost, delayed, and compressed. The receiving machine tries to reconstruct the audio and decode the document.
How Cloud Fax Works
Your document is uploaded as a file over HTTPS. The cloud service's own infrastructure — using dedicated PRI lines or managed SIP trunks — transmits the fax to the PSTN on your behalf. The "last mile" connection to the phone network is handled by infrastructure optimized for fax.
The key difference is where the PSTN connection happens. In VoIP fax, it happens on your internet connection — subject to whatever jitter, packet loss, and codec compression your provider applies. In cloud fax, it happens on the service provider's dedicated infrastructure, where every relevant parameter is controlled.
This is why services like mFax deliver a 98% success rate. It is not because they have better VoIP — it is because they bypass VoIP entirely for the fax transmission itself.
The process is also simpler for the end user: upload a PDF or snap a photo with the mFax mobile app, enter the fax number, and send. No ATA configuration, no codec settings, no T.38 negotiation. The fax goes out over dedicated fax infrastructure regardless of what internet connection you are on.
For organizations with serious fax reliability requirements — healthcare, legal, insurance — mFax Business adds HIPAA-ready compliance, virtual fax numbers, and team accounts starting at about $9/mo (billed annually). There are no rigid fixed tiers: you build your own plan, sizing the exact seats and pages your team needs and paying only for what you use.
The Bigger Picture: POTS Is Going Away
There is one more reason to understand VoIP fax problems: traditional analog phone lines are disappearing. AT&T, Verizon, and other carriers are actively decommissioning copper PSTN infrastructure. Businesses that relied on POTS for reliable fax are being forced to migrate.
The migration path is not "use a VoIP line instead of your POTS line." The migration path is cloud fax — services that receive and transmit faxes through their own carrier-grade infrastructure and deliver documents digitally. This bypasses both VoIP fax reliability problems and the hardware costs of maintaining physical fax machines.
Fax is not going away — healthcare, legal, and government sectors still rely on it heavily. But the infrastructure that made fax reliable (dedicated copper circuits) is going away. The replacement is purpose-built cloud fax infrastructure, not VoIP passthrough.
For a broader look at the underlying technology, see our guide to fax over IP (FoIP) and how it compares to traditional transmission.
Summary: Why VoIP Faxing Is Trouble
VoIP fax problems are not bugs — they are features of the technology working exactly as designed. Packet networks are designed to tolerate packet loss. Compressed codecs are designed to discard information. Voice processing features are designed to improve voice calls. None of these were designed with fax in mind.
Here is the core incompatibility in one sentence: fax requires a continuous, lossless, precisely-timed audio channel, and VoIP provides none of those things by default.
The five failure mechanisms — packet loss, jitter, codec compression, voice processing features, and latency/clock drift — are all consequences of using a protocol designed for analog circuits on a network designed for digital packets. T.38 mitigates several of them, but only when both ends support it, which is not guaranteed.
If you need reliable fax, the practical answer is cloud fax. Try mFax.to — fax from your phone or browser, delivered over infrastructure purpose-built for the job.
Also see: VoIP Fax Problems: Troubleshooting Guide | VoIP Fax Machine: What to Expect | Analog vs. Digital Fax