Design Philosophy Back in the early 1970s new, marked transistors were fairly expensive, and I couldn't justify their cost for just messing about. However I was able to get hold of old broken radios and various bits of early discrete logic with comparative ease. I could then spend happy hours in front of the TV with a hot soldering iron, salvaging anything remotely usable. The very first bird trill circuit was built almost entirely from such scrap components. I think the only thing that was bought new was the zenner diode. Even the sounder was salvaged from a broken crystal earpiece. It was also built up piece by piece trying different oscillators, coupling methods, values etc. From a design/cost point of view it is probably woefully over-engineered, but as all the parts were free I could use transistors where others would use resistors, and an extra gain stage where it might be more economic to 'push' components. Circuit Description The tone oscillator is an astable comprising transistors Q5/Q6 and associated components, but the base bias resistors instead of going to the positive rail, go to an R/C network. This is so that the other oscillators can modulate it with low frequency level shifts. These vary the oscillator frequency and periodically stop it by changing the rate R13 and R14 can charge their respective capacitors. Trill is provided by Q7/Q8 and fed into the modulation network via C10. The trill effect is achieved by giving such a large level shift that the tone oscillator is stopped each half period. C10 also provides another function. Due to the relatively low impedance of R19/R20/Q8 it effectively decouples tone signals in R13 and R14 from each other and the rest of the modulation network, giving more predictable results. Getting the effect just right is quite critical, and R20 is variable so that it can be adjusted. Pitch change comes from Q3/Q4 and is coupled to the modulation network by R11. As it is a very slow oscillator coupled by quite a high value resistor it just gives a small regular frequency shift. R6 couples the astable Q1/Q2 to the modulation network. Due to the relatively fast rate and the charge and discharge behaviour of C10 over a number of trill periods, as well as the interaction with the pitch change astable, this sometimes gives a chirp effect rather than changing the trill pitch. Q15 - Q18 provide a bridge mode output stage that drives ceramic sounder X1. This, along with the R/C coupling from the tone oscillator, is done to ensure that the stage draws zero current when not actually producing any sound. D2 and D5 cancel the rectifying action of Q16 and Q18 repectively, preventing charge build up in C13 and C14 that would otherwise occur. Ideally there should be bleed resistors across these diodes to ensure the transistors cut off completely when there is no signal, but in practice it works quite happily without them. The actual drive circuits themselves are very similar to ttl logic line drivers, the output impedance of each half being only a few ohms. The astable Q13/Q14 provides a regular mute function. It shorts out the bias to Q10 in the voltage regulator, via Q12. This not only gives gaps in the sound, but also saves battery power. In the muted state the total power consumption is mostly that drawn by R24 and around 35uA. The 47M resistors will have to be bought from a decent supplier as they are not stocked by a lot of small shops. If you can tolerate the loss of efficiency, you can divide R25 - R28 values by 10 and multiply the values of C11 and C12 by 10, also making them tantalum electrolytic types. This was what I used for my very first attempt at this design. Q9/Q10/Q11 provide a very low power voltage regulator with extremely small drop-out voltage. The reason for the regulated supply is that most silicon transistors have a maximum reverse Vbe of 6V. The oscillator stages will try to push this beyond the supply voltage, due to charging characteristics of the capacitors. This causes partial zenner breakdown (typically around 10V) and means the behaviour of the circuit is supply depended - not good when battery operated. Setting the level to 6.5V seems to be a good compromise of performance against stability and battery life. At start up, Q10 is fed by R24 and starts to conduct. This then causes Q9 to conduct (the only PNP device in the whole design). Once its collector has reached about 6.5V D1 allows Q11 to conduct. This then robs Q10 of current and the circuit stabilises. The actual output is the zenner voltage of D1 plus Vbe of Q11. R21 is needed to ensure the D1 has enough current for zenner effect to take place, although at around 100uA it will still tend to produce a lower than normal voltage. R22 is only there to protect the base of Q11 from possible spikes. R23 provides similar protection for Q9. Setting up. With Q12 base shorted to its emitter adjust R20 for the most interesting combination of chirps and trills. Component Choices Although I have quoted BC548 and BC558 for NPN and PNP transistors respectively you can actually use almost any silicon transistors of the right polarity and reasonable hfe. In the original design the output darlingtons were made up of pairs of ordinary NPN transistors. However, the types shown here actually work out cheaper, unless you have a lot of old stock to use up ... like I had! The output circuit can drive almost all ceramic sounders or can just about drive a 50ohm miniature speaker. The ceramic units are far more efficient however, and give much longer battery life. Diodes can be any small signal silicon ones, although you might need to uprate D3 and D4 if you drive low impedance loads. BZX79 (D1) is quoted as a 500mW zenner. If you can get a lower rated device the voltage accuracy will improve. There used to be a 250mW type but I've not seen them for a long time. All capacitors except the supply decoupler need to be 'dry' types. The circuit is very high impedance which, while giving low power consumption, means the leakage from electrolytics would severely degrade performance. W Godfrey 1974-2007