We then induce a voltage in the secondary winding that is a bit higher than we need, so that we’ll still have a high enough voltage to drive the linear voltage regulator when the battery voltage starts to drop (refer to the circuit in Figure 1). T1 are T2 are turned on and off by an astable multivibrator. We’ve used a 4047 low-power multivibrator for this, which has been configured to run in an astable free-running mode. The complementary Q outputs have a guaranteed duty-cycle of 50%, thereby preventing a DC current from flowing through the transformer. The core could otherwise become saturated, which results in a short-circuit between 6V and ground.This could be fatal for the FETs. The oscillator is set by R1/C1 to run at a frequency of about 80 kHz. R2/R3 and D1/D2 make T1 and T2 conduct a little later and turn off a little faster, guaranteeing a dead-time and avoiding a short-circuit situation. We measured the on-resistance of the BS170 and found it was only 0.5 Ω, which isn’t bad for this type of FET. You can of course use other FETs, as long as they have a low on-resistance. For the transformer we used a somewhat larger toroidal core with a high AL factor. This not only reduces the leakage inductance, but it also keeps the number of windings small. Our final choice was a TX25/15/10-3E5 made by Ferroxcube, which has dimensions of about 25x10 mm.
This makes the construction of the transformer a lot easier. The secondary winding is wound first: 77 turns of a 0.5 mm dia. enamelled copper wire (ECW). If you wind this carefully you’ll find that it fits on one layer and that 3 meters is more than enough. The best way to keep the two primary windings identical is to wind them at the same time. You should take two 30 cm lengths of 0.8 mm dia. ECW and wind these seven times round the core, on the opposite side to the secondary connections. The centre tap is made by connecting the inner two wires together. In this way we get two primary windings of seven turns each.The output voltage of TR1 is rectified by a full-wave rectifier, which is made with fast diodes due to the high frequency involved. C4 suppresses the worst of the RF noise and this is followed by an extra filter (L1/C5/C6) that reduces the remaining ripple. The output provides a clean voltage to regulator IC2. It is best to use an LM317HV for the regulator, since it has been designed to cope with a higher voltage between the input and output. The LM317 that we used in our prototype worked all right, but it wouldn’t have been happy with a short at the output since the voltage drop would then be greater than the permitted 40 V.
If you ensure that a short cannot occur, through the use of the usual 6k81 resistors in the signal lines, then the current drawn per microphone will never exceed 14 mA and you can still use an ordinary LM317. D7 and D8 protect the LM317 from a short at the input. There is virtually no ripple to speak of. Any remaining noise lies above 160 kHz, and this won’t be a problem in most applications. The circuit can provide enough current to power three microphones at the same time (although that may depend on the types used). When the input voltage dropped to 5.1 V the current consumption was about 270 mA.
The reference voltage sometimes deviates a little from its correct value. In that case you should adjust R4 to make the output voltage equal to 48 V. The equation for this is: R4 = (48–Vref ) / (Vref / R5+50µA). To minimise interference (remember that we’re dealing with a switched-mode supply) this circuit should be housed in an earthed metal enclosure.
