DTMF Generator — Dial Pad Tones

A DTMF generator produces the dual-tone pairs a telephone network decodes as keypad digits. This page plays the standard 697/770/852/941 Hz row tones and 1209/1336/1477/1633 Hz column tones with two sine oscillators per key, synthesized in your browser with the Web Audio API. Nothing is dialled and nothing is uploaded.

How to use it

The page separates single tones from sequences: a 4x4 keypad for individual keys, and a Sequence Dialer for automated strings.

1. Press and hold a key. Mouse down or touch start begins both oscillators; release stops them. A 5 ms attack and a 10 ms release keep the edges click-free.
2. Read the dashboard. Active Key, Low Group (Row) and High Group (Col) update for every key, and the status bar spells the pair out, for example “Key 5 (770 Hz + 1336 Hz)”.
3. Or type. With focus outside the text field, the keys 0-9, *, # and A-D each play a fixed 160 ms burst.
4. Set the output level. The volume slider scales the pair from 0 to 100%, and each oscillator runs at half the slider value so the summed peak tracks the control. The default 50% leaves headroom in the browser mixer.
5. Dial a sequence. Type or paste digits into the Sequence Dialer, for example 1-800-555-0199#4. Hyphens and spaces are ignored; if no valid digit remains, the status line asks for DTMF characters.
6. Choose a speed and press Dial Sequence. Fast is 70 ms of tone with 50 ms between digits, Standard is 120/80 and Slow is 200/150. The preview highlights the digit being sent as the sequence plays.
7. Stop with the same button, which becomes Stop Sequence, and the status reports completion or a stop.

The Q.23 dual-tone matrix

ITU-T Recommendation Q.23, “Technical features of push-button telephone sets”, fixes the sixteen keypad combinations: one tone from the low group of rows and one from the high group of columns, sent simultaneously.

	1209 Hz	1336 Hz	1477 Hz	1633 Hz
697 Hz	1	2	3	A
770 Hz	4	5	6	B
852 Hz	7	8	9	C
941 Hz	*	0	#	D

Why two tones instead of one? A single-frequency detector can be tripped by a voice or a whistle, so a valid digit requires one tone from each group at the same time. The frequencies sit inside a 697-1633 Hz window that survives analogue telephone transmission, no DTMF frequency is an exact integer multiple of another, and the groups are separated by a 268 Hz guard band (941 Hz up to 1209 Hz). The difference product of any two DTMF tones lands at 936 Hz or below and any sum product at 1906 Hz or above, both outside the signalling band.

Timing and frequency tolerance

ITU-T Recommendation Q.24, “Multifrequency push-button signal reception”, defines what the receiving exchange must accept rather than what a keypad sends. Its timing minimums are 40 ms of tone and 40 ms of pause between digits, and RFC 4733 restates the same 40 ms figures when DTMF travels as an RTP event. Receivers tolerate small frequency errors: roughly plus or minus 1.5% is inside tolerance, while deviations of 3.5% or more must be rejected. At 697 Hz that is about 10.5 Hz, and at 1633 Hz about 24.5 Hz. This generator’s oscillators are exact to the sample grid, so any offset belongs to the far end or the recording path.

Speed preset	Tone on	Pause between digits	Step time	Digits per second
Fast	70 ms	50 ms	120 ms	8.3
Standard	120 ms	80 ms	200 ms	5.0
Slow	200 ms	150 ms	350 ms	2.9

All three presets exceed the 40 ms receive minimum, and the fastest stays inside the 8 to 10 digit-per-second ceiling that RFC 4733 notes for signalling rates. If a receiver drops digits, slow down before suspecting the decoder; if it misreads digits, check the level and the codec first.

Where DTMF still runs

DTMF remains the common language of IVR menus, voicemail and conference PINs, card-entry and door intercom systems, and two-way radio repeater control. On VoIP the tones travel in-band as audio or out-of-band as RFC 4733 events; a low-bitrate codec can distort in-band tones enough to break a decoder, while out-of-band events survive compression. The tones are audible and easy to record, so treat DTMF as signalling, not security.

Accuracy, limits and troubleshooting

• This page is a generator, not a telephone. There is no loop current, no 600 ohm line interface and no hybrid, so testing a real POTS line needs a telephone interface or an ATA.
• The summed peak equals the volume slider, so pushing the browser output and the OS mixer into clipping adds harmonics that a decoder can read as a different key.
• Frequency accuracy is sample-exact here; errors a real network sees come from transmission and vocoders, not this page.
• A held pair cannot talk off by itself, but loud music or speech leaking into a call microphone can confuse acoustic decoders.
• A, B, C and D are absent from consumer keypads and often ignored by switches; they appear here because the Q.23 matrix defines them.
• Keyboard capture is global on the page. Typing inside the sequence field is ignored, but a digit pressed outside it sounds a 160 ms burst.
• Browser autoplay policy requires a press before any tone, and if the status reads Audio unavailable, press a keypad button again to resume the audio context.
• Bluetooth and acoustic coupling can mangle in-band tones through latency, echo cancellation and compression; use a cable or an interface when a decoder must be trusted.

Related tools

• Tone Generator — Play single sines across the audio band for level and hearing checks.
• Spectrum Analyzer — See the two DTMF components as separate peaks and compare their levels.
• Frequency Check — Verify the octave and pitch you hear against a reference tone.