Electronic – Is it possible to have “in-out” communication in a single 56k modem

While I cannot provide a complete answer to your question, I think this info should be helpful for you in understanding the nature of the line requirements of the Plain Old Telephone Service (POTS) which normally drives the old style rotary phone. I think your project is quite intriguing and hope you will post more about your progress.

The POTS line, with all phones on-hook, should measure around 48 volts DC. Taking a phone off-hook creates a DC signal path across the pair, which is detected as loop current back at the central office. This drops the voltage measured at the phone down to about 3 to 9 volts. An off-hook telephone typically draws about 15 to 20 milliamps of DC current to operate, at a DC resistance around 180 ohms. The remaining voltage drop occurs over the copper wire path and over the telephone company circuits. These circuits provide from 200 to 400 ohms of series resistance to protect from short circuits and decouple the audio signals.

To ring your telephone, the phone company momentarily applies a 90 VRMS, 20 Hz AC signal to the line. Even with a thousand ohms of line resistance, this can still pack a bit of a shock so be careful when you are probing around trying to find a POTS line.

POTS Line Characteristics: Bandwidth: 180 Hz to 3.2 kHz

The low end is rolled off early to stay away from the 60 Hz power line noise region. The high end cut off is more critical. Voice on the telephone network is digitized at 8 kHz sampling rate which means that any signal above 4 kHz will be aliased back as noise in the voice band. Most voice CODECs roll off at about -25dB at 4 kHz with a -3dB down point around 3.2 kHz. The phone company decided years ago that the 180 Hz to 3.2 kHz range would be sufficient for speech intelligibility while allowing them to multiplex many calls over coax and twisted pair.

Signal to Noise: Approximately 45 dB

This is not as easy to quantify because noise comes in many forms, such as electrical interference from fluorescent fixtures or hiss from the many amplifier stages in the voice path. Speech correlated noise can be introduced from non-linear speech coding and compression algorithms. Crosstalk from other conversations is another form of noise. The phone company uses 8 bit mulaw nonlinear coding which yields about 12 bits of dynamic range. The bottom line is that you can never count on more than about 45 dB signal to noise ratio.

If you intend to use the mic input mic input to your sound card, it will have a current source or pullup resistor to supply a bias to the microphone to which it was intended.

It might not be good for the speaker to have DC flowing through it.

If your phone is connected to the telco, be very careful. The speaker may run at quite a CM voltage and the DC offset might be -48V (depending which way round it is wired).

Telephones wires are very good at attracting lightning. During a storm high common mode voltages may be on the earpiece.

Stricly speaking you should use an audio transformer to couple the earpiece signal into the sound card, this provides DC isolation and some protection against lightning. Adding a TVS across the output of that transformer would be a good idea too.

The ac signal on the earpiece might be quite small. Only a few hundred mV. For a mic input, this will be fine (and can be potted down at the output of your transformer) but for line levels, which are usually 1Vrms it might be very quiet.

Not all phones work the same way, so measure the voltage first using a scope and judge the divsor as required.

The audio transformer should also be protected against DC usign a DC blocking (aka coupling) capacitor.

This circuit protects against discharge and DC currents, though use with caution as a mistake will cost you a PC!