This is an amplifier built for a purpose: I wanted to build an amplifier for the tweeter in my Baby 2-way loudspeaker. Baby is slowly moving towards a commercial product, and we have not yet decided exactly how to do the amplification. There are 2 starting points – one is an old idea of my own, to use an IC/Opamp for the first section, and the other is a diagram I found on the Internet – a buffer circuit, that looked interesting. The Opamp remained, the buffer section has been reworked/improved compared to the original diagram. A couple of interesting features: there is no global feedback, you can use standard Darlington or complementary Darlington, also known as Sziklai configuration, there is no DC servo – DC offset is quite stable, and once adjusted (amplifier at working temperature ) it moves slowly between +/- 1 millivolt, not more. I can live with that, and this is no problem for the tweeter either.
Amplifier can be quite fast, depending on the selected transistors, and distortion is low, some would say very low for a non-feedback design. Input Opamp is of course important – I wanted to be able to use a power supply up to +/- 20 – 22 Volts, and OPA 604 can manage this. There are a few other Opamps that can handle this level of supply, and I will maybe try something in the future –but OPA604 is a good choice, and NOT responsible for the main part of the distortion. OPA604 exists also as a DIP package, socket mounted – easy to change.
This is the version with complementary Darlington, link to standard Darlington here: StdDarlington link. After some experimenting, I ended up with a rather classic tranistor selection – BC 327/337 and BD 139/140. Feel free to experiment with more exotic types. Normally, a transistor in a TO126 housing is chosen as bias control – easy to mount on the heatsink – but I stuck with BC327/337, very little power goes into them, and they do an excellent job in that place – so heatsink mounting is somewhat DIY – but it works. See picture! Power devices are japanese types - 2SB 834 and 2SD 880, 2 of each – and a 0.1 ohm ballast resistor – bipolar transistors accept wirewound power resistors, don't try with mosfets – they are extremely sensitive to resistors with just a shade of inductance in them, and go bananas with self-oscillation. Can be expensive.
Heatsink is large – bias is around 630 mA – quite substantial – this actually means Class A operation during normal listening level. You can use a regulated supply for the line section, and go straight from power supply/power capacitors to the output section, also with a higher voltage level. The power devices can take more voltage than the opamp. Except for a fraction less noise, there is no benefit in terms of power, when the opamp clips the signal, you can't go further. The regulation provides isolation for the line stage in terms of power variations due to large output currents, this can be an advantage, in this case a tweeter is normally an easy load with restricted current demands. I have tried both versions – both work fine. Depending on choice of transformer, you can get from 20 to 30 Watts output. Repeating myself, a tweeter is easy on your output devices, and music does not contain enormous amounts of energy in the tweeter region – I will show some measurements about this at a later stage.
Comments:
This version is made with an input capacitor – C6 – 2uF. Use a combination, 2uF, 2nF and 47pF or something like that, top quality here is OK!! Opamp is OPA 604, gain is about 20X, can be adjusted with trimmer X8 – a 10K Ohm trimmer. Note decoupling capacitors for Opamp power supply. J1 and J2: a low noise N-fet – have tried different types – not much difference. Same goes for LED – red, yellow or green. Q1 and Q4 – BD 139/140 on common heatsink. Q2 and Q3 – BC327/337 on main heatsink – important! Q7 and Q8: MJE 320/370, Q9,10,11 and Q12: 2SB 834 and 2SD 880; all on main heatsink, R30 and R31 are power resistors. R28 simulates loudspeaker – should be excluded.
Trimmers X9 and X10 allows you to adjust BOTH bias current and DC offset. Bias regulation is narrow – R1 and R6 can be adjusted/changed: lower value: more current – higher value: less current. Capacitors C15 and C16 are required to control self-oscillation in the complementary Darlington stage – you will need an oscilloscope to check, and yes – it is important. Complementary Darlingtons have a tendency to self-oscillation, and this will often lead to self destruction of output transistors. In this case, 400 pF did the job, with other output devices the value might be different.