Consider the problem of determining that there is a modem at the other end of the phone. If a modem dials a number that does have a modem connected to it, then it does not detect an answer tone, and therefore does not initiate its side of the negotiation sequence. The session sounds like this:
How does the modem tell the difference between an answer machine and a modem answering the phone? The modem "hears" either the following two sounds:
The modem answer tone is shown below:
Its spectrum is calculated below (by dividing it into segments of 16ms, and calculating the Fourier series coefficients for each segment):
What do you think the specification of the answer tone is? Try to match it using the waveform applet. In what ways are the sounds you can generate with that applet and the answer tone different?