Issue 156 July 2003
Stealth Telephone Screener
Mad Dash for Flash Cash Grand Prize Winner

ANALOG PROCESSING

The analog output from the PWM is filtered by a second-order, Sallen-Key low-pass filter with a 4-kHz cut-off frequency. This provides adequate filtering for the 32-kHz PWM pulse rate and AC coupling for the output.

The input to the A/D converter is low-pass filtered at 3.5 kHz after passing through a variable-gain stage. For normal VOX operation, the RXGAIN signal is high, which allows Q4 to turn on and provides a gain of 24.5 for the caller’s voice. When recording a new greeting using a locally connected phone, RXGAIN is low, providing a gain of approximately three. This keeps the audio signal within the range of the ADC. 

POWER

A 9-V battery, feeding a linear regulator with ultralow-quiescent current, supplies the power. The brownout timer is disabled to minimize the standby power consumption, which results in a total draw of approximately 5 µA while waiting for a call. 

When active, the entire system draws 40 mA for about 40 s for each call that rings through or 20 mA for 15 s if it hangs up. A typical alkaline battery will last for a few thousand calls. Hopefully, you don’t get that many calls too often!

Battery voltage is monitored through a resistive divider (R6, R7), which uses 1-MW resistors to avoid putting a significant load on the battery because the current is drawn continuously. But, this large resistance causes an error of as much as 50 mV in the measured voltage because of the 100-nA A/D pin input current. This much error is not a concern, because the low battery threshold is not critical. 

CONSTRUCTION

With the exception of a small portion of the phone-line interface, building the Telephone Screener was a mostly straightforward process. The one connection between the top terminal of J1 and K1 is subject to high voltage when the phone line is on-hook, so the corresponding wiring should be kept clear of any other circuitry.

As with any mixed signal device, the analog and digital power and grounds should be kept separate, and any high-current paths (e.g., the speaker, EEPROM, and relay coil) should have separate power runs. I still ran into a problem when high-current spikes from the EEPROM were getting into the audio, but adding a filter (R10, C5) to its V_DD pin tamed the noise.

There mustn’t be any exposed metal parts that are connected to the circuitry, because these would be a shock hazard with the nonisolated interface. This meant that I had to take extra care during debugging to make sure that the ICSP programming port and the phone line weren’t plugged in at the same time.