Anyone who’s spent much time searching the web for interesting circuits is likely to have found at least one TL431 based audio amplifier, the circuit is based on the principle that any comparator can be used in linear mode if it’s rolled off with enough negative feedback. Although the TL431 is often referred to as a programmable or adjustable zener, it is, in fact, a comparator with its own 2.5 V reference all neatly wrapped up in a TO92 package.
The problem with the TL431 amplifiers to be found on the web is that they simply roll it back with large nfb and leave it at that, which results in the very low gain, to make matters worse some such circuits make a bit of a hash of biasing the control input. The circuit presented here takes care of the low gain by adding an AC shunt to the feedback path and using an electret mic for the input — the 2.5 V set on the control input at stable operating condition suits an electret mic perfectly. The first prototype had a 35 ohms loudspeaker as a load (RL), this gave good results although the TL431 ran a bit warm with a Vcc of 12 V. An old 130-ohm telephone earpiece is likely to present a less stressful load. AC shunt C2 (100 µF) has to be a quality component in terms of its ESR specification — don’t just use a scruffy capacitor lying about as you may experience RF sensitivity. It was necessary to add a series resistor (R3; about 100 ohms) or in extreme cases an inductor (L1; 100 – 220 µH). Components C1 & R1 are entirely optional to selectively feed some un-shunted feedback to reduce noise; 1.5 kΩ & 5.6 nF are as good a place as any to start off with.
Initial set-up depends on the current drawn by the electret mic and the value for RL — anywhere between 200 and 2,000 ohms is good. R2 allows the TL431 cathode to swing despite the AC shunt, 1.2 kΩ was found to be satisfactory, P1 can be a 47 kΩ trimpot and is used to set the voltage drop on RL. In the case of moving-coil speakers a compromise between voltage swing and pre-biasing, the cone should be sought, with a resistive load adjust for 0.5 Vcc, once the operating point is determined P1 can be measured and replaced by an equivalent fixed resistor.
The circuit has a couple of handy features, firstly it works very well on the end of a twisted-pair — the output can be tapped off at the wiper if RL is a pot at the power supply end, secondly by salvaging the JFET from an old electret mic (some common types of JFET will work but not quite as well), just about any piezoelectric element can be used as the transducer. Brass disc sounders give a good output (handy as vibration sensors if glued to a structure); even the quartz discs from clock crystals give some output, a phono crystal cartridge gives a high output and the piezoceramic pellet from a flightless cigarette lighter gives a huge output… the range of possible applications is awesome!
A surprising application is the ability to test the microphonic sensitivity of ordinary capacitors! Disc ceramic types don’t need to be tapped very hard to produce an output but rolled metalized foil types produce some output too.