TL431 Linear Power Supply

After messing about with a TL431 shunt regulator whilst working on a previous project, I realised how versatile it is. These devices have been around for decades and will no doubt remain in production for many years to come. I remember them from my TV repair days where they were used in the feedback loop of switch mode power supplies (and still are), but here I'm simply using it like a programmable zener diode in a conventional linear power supply. Although I've never seen or used it in this configuration before, it works perfectly. The circuit is very conventional, having the usual arrangement of mains transformer, bridge rectifier, smoothing capacitor and output transistors driven by a regulator, but with the addition of crowbar protection just in case an output transistor should fail short circuit, resulting in the full unregulated voltage being fed to the load. The schematic is shown below


Although earlier I likened the TL431 to a programmable zener diode, and yes it can be thought of in that way, this little TO-92 package is so much more. In fact it's a voltage comparator, except with the inverting input connected to a 2.5V internal reference and the non-inverting input externally accessable on a pin labelled Ref. This pin will always want to be 2.5 volts in order to equal the internal 2.5 volt reference, and the cathode pin (strangley known as K) will do whatever it needs to do (within limits) to achieve this balance. The Ref pin is fed from the junction of R1 and R2 which form a potential divider connected across the supply to be monitored which in this case is the output terminals. This sample voltage is compared to the internal 2.5 volt reference and an error voltage is then produced on the cathode which will vary up or down accordingly to maintain regulation. There is a simple equation to work out the value of the resistors which is Vout = R1 R2 + 1 x 2.5

On the schematic above, R1 is 10K and R2 is 2K8. R2 is made up from a fixed value of 1K8 in series with a 1K potentiometer which is used to set the output voltage. With the values shown, this can be anything between 11 and 16 volts, although the crowbar will trip and blow the fuse at about 15.5V. In fact, slowly turning up the output voltage until the fuse blows is the method used to check if the crowbar is working correctly, albeit at the expense of a sacrificial fuse or two! The great thing about the TL431 is that the inherent voltage drop across the pass transistor base-emitter junctions (and also a small amount across the fuse and internal cabling at higher currents) is automatically compensated for because R1 and R2 are connected directly to the output terminals, essentially incorporating these components within the regulation feedback loop. If the TL431 cathode has to go a little bit higher to overcome those losses, then that's what it will do. Incidentally, the preset pot ‘fails safe’ in the event of it becoming intermittent with age. If the wiper lifts off the track, the output voltage goes down not up

Veroboard layout

The crowbar circuit is fairly standard with the main component being a thyristor (or Silicon Controlled Rectifier) which most of the time just sits there doing nothing until it receives a voltage on its gate. When this happens, it will turn on and become pretty much a short circuit which will cause the fuse to blow. Here, the gate is fed from a 15V zener which conducts and develops a voltage across the 470R resistor when the output supply rises above about 15.5V. The 100n capacitor is there to suppress any voltage spikes that might cause false triggering. When the thyristor is fitted to the veroboard, fold the anode (A) and cathode (K) legs underneath and solder them directly to the terminal block (not shown on the diagram). This will prevent the veroboard tracks taking the full short circuit current if the thyristor fires. A crowbar seems quite a brutal way of shutting down a power supply, but the fuse is a quick blow type and the circuit can handle it. With a fuse you have peace of mind knowing that the supply is completely disconnected without any chance of it working again without human intervention

If you haven’t got one to hand already, the most expensive part will be the mains transformer. I used a 15 volt 15 amp toroidal as it was discontinued and on special offer, but ideally, a secondary closer to 18 volts would have been better. I wanted to keep the voltage across the pass transistors as low as possible to keep their power dissipation to a minimum. 15V is about as low as you can go whilst still maintaining regulation, though using a shunt regulator instead of a series regulator does help in this respect. With the transformer I used, the output current is about 10amps before regulation is lost, but if the situation allows, a higher current can be achieved by reducing the output voltage slightly. The primary fuse needs to be an anti-surge (time delay) type due to toroidals having a large switch on current. The transformer secondary feeds into a 35A bridge rectifier and smoothing capacitor. The capacitor is actually made up of four separate 10,000uF 40V 105 capacitors in parallel which are mounted on a matrix board and connected together on the underside with thick copper wire. This approach makes for a lower profile and surprisingly costs less than a single capacitor! For the main high current cable runs I used 2.5mm (30 amp) stranded wire, with 3 individual runs of 1.5mm (21 amp) stranded wire from the emitters to the output

Three TIP3055 pass transistors are used so the current can be shared between them. The 0.1 ohm (10 watt) emitter resistors are there to balance out any slight differences in tolerance. These power transistors along with the TIP122 darlington driver are all mounted on the same heatsink (please forgive the 'lazy' way I drew a darlington on the schematic!). Construction is down to personal choice and whatever you can lay your hands on, but I was lucky enough to obtain a nice enclosure from a scrap electrosurgery unit at my place of work, which I stripped out and re-purposed. The front panel was overlayed with a sheet of smoked black semi-transparent acrylic sheet, cut down to the required shape and size. I chose this so that the low cost LED voltmeter I purchased online could shine through from behind, which I think looks much nicer than if it were panel mounted. For convenience, a socket was added to the veroboard to feed the voltmeter, but effectively it's just connected across the output terminals. The current capability of this circuit depends on the power handling of the components used and can be scaled up or down accordingly
. The PSU has been used with a 2m VHF mobile transceiver with no reports of hum or noise, indicating that it has good RF immunity

Internal view