The Dual Tone Multi Frequency (DTMF) signal is composed by two frequencies as reported in the following table:
A decoder will give the correspondent digit after detecting the two carrier frequencies according to the table. In order to detect and distinguish the pair of frequencies sent, the common algorithms require usually the total power level of unwanted frequencies to be at least 20dB below the lowest frequency signal with a signal to noise ratio greater than 23dB.
Responding to the command AT+VTS, the module sends a command to the network infrastructure to generate on the other audio party the correspondent DTMF signal. The DTMF tone duration can be controlled partially by the module since it sends a "start playing tone" request and a "stop playing tone" request and these can be specified by the application controlling the mobile, except from time shifts introduced by the network. The network infrastructure generates this tone perfectly aligned with specifications requirement, without introducing problem during recognition.
The DTMF signal is generated by a separated source, typically a landline (corded) phone, and sent to the input lines of the module (Uplink path). The frequencies couples, sent on the voice channel,
are digitised, encoded and sent by the digital transmission system. In the receiving device the signal would be reconstructed, but since the digital transmission of the voice channel is compressed
and optimised for voice, this reconstruction depends on the kind of voice compression used for the transmission, and generally will not perfectly match the original signal.
There are four main types of compression for the voice channel and only the Full Rate
one has no distortion, while the other three offer a different trouble level (see figures):