This device connects to an ordinary wire-line telephone circuit and produces a ring tone on a speaker. It is customizeable for the hard of hearing.
This design is made for a particular client. However, this client has the most common type of hearing loss found in the elderly, namely increasing with frequency starting at about 500 Hz so that it is extreme at 5 kHz. The sound needed is therefore below 200 Hz. Higher frequencies are filtered out so that the sound is not offensive to people with normal hearing, even at high volume.
Conceptually, the design is in three parts: synthesizer, ring detect, and bias. See schematic.
For the purpose of synthesizing the sound, the built-in pulse-width modulator (PWM) of the Arduino Uno (AU) is too slow, so direct control of a binary output ("bit-banging") is used instead, using the micro-second timer. The output is filtered in hardware and passed to a single-ended driver.
There are 2 power supplies: one for the AU and filter and one for the driver. The choice of a single-ended driver avoids adding a third supply. I thought that it wouldn't be good to try to use only one supply because, if low voltage, the output sound would be too soft and, if high voltage, the regulator of the AU would run too hot. Anyone who builds this with a single supply is invited to write me and tell me how it worked out.
The entire ring tone is stored as a list of pulse-widths in an array. For individual notes a square envelope was found too offensive, so a linear rise (attack) of 50 ms and fall (decay) of 200 ms was used, giving an agreeable sound. This is easily built into the stored array.
The hardware filter is a simple design. (I am not an electrical engineer.) Improving on it would be easy and probably should be done in order to get stronger filtering. The design shown has the virtue of using a limited number of different parts.
The 250 ohm pot. is the volume control. There seems to be no need for it to be accessible outside the box.
Driver bias is enabled in software so that the driver transistor can be cool when not in use. I put a heat-sink on it, but it is probably not needed. The transistor seems always to be cool to the touch. Its duty-cycle is very short. I chose a ring length of 2 s because the ring signal is about 2 s here. The device returns to its initial state between rings. Bias is measured using the A-to-D of the AU. The 10 micro-F capacitor is charged by a binary output to increase the bias. When it is high enough, the output is set low and the cap. slowly discharges so that the bias current goes to zero when the device is idle. The bias stays high long enough so that there is no distortion to a 2 s ring. DC bias current is run through the speaker.
Ring detection is by a voltage divider. Notice that both sides of the telephone line are isolated from the AU. This is to keep digital noise from the AU from contaminating the voice channel. In North America a ring signal of about 100 V at 10 Hz is common. The ring pot. will need to be adjusted where this voltage is different. The idea is to present about 5 V RMS to the AU. CMOS protection diodes keep it from harming the AU. The diodes have a limit of about 1 mA. This design easily falls below that.
Ring signal filtering is done in software: Several high samples in a row are needed to start the synthesizer cycle.
A physically large speaker is needed to produce a strong low sound. This may be the hardest part to obtain for some users. High fidelity is not needed, so second-hand speakers are a good choice. Ask friends. Auto wreckers sometimes have them. However, beware cracked speaker cones. They will introduce unwanted high frequencies. 12" (30 cm) diameter works well in the 100-200 Hz range.
All the NPN transistors can be 2N2222 or any similar. The PNP driver is a TIP115 or similar. The circuit was built on a blank shield and one extra board, most of it on the shield. The extra board was only for the high current and high voltage connections and circuit. An RJ-11 socket was mounted on that board. The enclosure was a plastic Hammond box with no ventilation. Panel mounted sockets brought in DC power from modular supplies (wall warts). A drill press is handy for precise location of mounting holes for the AU. The hole for the RJ-11 was simply filed into a square.
All parts are obtainable from commercial electronics stores.