There is technically full reason believe that that an 8052 type processor can be used to send FSK signals such as those of caller ID. The 8052 could generate the output pulses as square waves which could be shaped with an RC low pass filter and connected to a small transformer to couple the signal the output line. Caller ID uses a 1200 Hz burst (1 cycle) to represent a "1" and a 2200 Hz burst (2 cycles) to represent a "0" for a net baud rate of 1200 bps.

I implemented a very similar communications system for a military project that used a very similar protocol. It was called a 2 wire interface and used a burst of 2560 Hz (4 cycles) to represent a "1" and a burst of 3200 Hz (5 cycles) to represent a "0" for a net baud rate of 640 bps. It was implemented using a state machine run via interrupts on TIMER2 generating the output pulses via the HIGH SPEED OUTPUT mode. The timer 2 output was also wired external on the chip back into the INT1 external interrupt to permit a seperate interrupt to manage the count of pulses generated in an output burst. This code has been implemented successfully on both a Dallas DS80C320 at clock of 7.3728 MHz (this processor works at 4X clocking) and on a Cygnal C8051F126 at clock of 22.1184 MHz (which is a processor with 1X clocking). Even the slower Dallas part has more than enough bandwidth to process the control state machine and the code was written in Keil C51.

So from this experience I would give a positive response that it should be quite possible to do the CID FSK generation using a 8052 type processor.

Likewise I believe that it should also be quite possible to decode CID signals off the receiving end of the line using an 8052 type processor through the use of the TIMER2 in the input capture mode. A small opamp with a high gain would be needed to pickup the signal off the incoming wire (maybe via a transformer - maybe not) and then the opamp would clip at the output to +5/GND rails and present the signal to the T2 input capture pin as a square wave.

Now I realize that development of such algorithms is not for the faint hearted but it is totally feasible.

Michael Karas