Okay, here's the current circuit work--a version of W. Marshall Leach's two-sided common-base head amp. Two needed, one each for left and right channel:
The parts that carry DC and that involve bias setting the operation point are in blue. The red parts carry only signal.
Neither end of the battery is referenced to ground. The two amplifier transistors (stacked at left) have their bases referenced to AC ground through capacitors. Much smaller capacitors serve as high-frequency (radio-frequency noise) bypasses between the input on the emitters of the transistors and the transistor bases.
The transistor alone near the center is in a voltage-regulating circuit. Unfortunately, so simple a circuit working by itself can't hold the voltage stable enough against supply variations (as the battery runs down) so a second trick works against it. It tries to hold a steady voltage between its two ends; each end draws from the adjacent supply rail through an 820K resistor and feeds the amplifying transistor base through a 750K resistor ... but the 750K resistor forms one side of a voltage divider. The other side is 10x larger, 7.5M, to the opposite supply rail. The 10% pull from the opposite rail opposes the imperfect regulation, improving the control several times over, keeping the collector current in the resistors within about 0.6% of its center value. The highest collector current occurs with the battery a little more than half depleted, so the low-current extremes are at new-battery and exhausted-battery.
I haven't plotted the curve, but here are the test values: 7.99 volts, 97.8 microamps; 7.60, 99.0; 7.18, 100.6; 6.78, 101.8; 6.41, 102.7; 5.98, 103.3; 5.59, 103.2; 5.19, 102.7; 4.80, 101.2; 4.41, 98.5 . You'll notice that although it's a 9v battery (9.6v when absolutely new) I start at 8.0 volts. That's because I'm supplying the amplifier through a pilot-light LED, which produces a small but visible glow even at fifty microamperes, while dropping only 1.6 volts. (At its rated 20 mA, it's very bright, but a small dot that doesn't actually illuminate much.)
You'll notice that the currents are lower than the 120 uA to 150 uA I need for 20x amplification. I have to drop the 820k, 750k, and 7M5 resistors by about 30%, and I might need to drop the 3M0 and 13M resistors on the regulating transistor to keep the curve centered.
Since the curve changes direction, it must represent a function of at least second-order (quadratic). I suspect it's something close to a cosh (hyberbolic cosine), but I'm not going to try to combine three Ebers-Moll models with the network algebra. My algebra muscles aren't flaccid, but they're not ripped either, and it would take me several days to get it right. (This is something you do in full-sized notebooks, not half-sized.)
I've had to layer this on top of the work I'm doing, so I have a very messy workbench, especially with several varied copies of the design:
I'm re-using the battery supply I built to replace-if-necessary my UL-listed lab supplies, but I clipped in two extra batteries so I could go up near 9v if I need to:
I can't use the lab supplies for this because the supplies AND the scope are referenced to ground, and this amplifier needs a floating supply. If I'm going to keep using this, I'll change the voltage control to a coarse/fine dual-shaft arrangement and maybe make it possible to switch batteries in and out. And maybe build a couple more. It's a cute design.
Here's the amplifier on a breadboard. I'm afraid I've just blown up a part of one of the big photos:
The layout isn't quite what it will be in the final package because of the limitations of the breadboard. If you look very carefully on the left, you can see the regulating transistor behind its resistors in a kind of outrigger.
Here are the current and supply voltage readings. The front meter has current in microamps; the rear meter shows the supply voltage.
I promised, in my previous article, to tell you what this mysterious mixer-motor part is:
That white loopy thing on the end of the shaft is ... wait for it ... a flyball governor! The center opposite the shaft attachment presses against contacts that supply current to the field windings. The inner spring pulls the weights in; rotation pushes them out. The contacts are arranged so that as the weights move out the contact resistance decreases and more current flows into the field windings. It's a pecularity of this armature-field arrangement that the more current you put through the field windings, the lower the motor's speed/load curve and the more slowly it wll turn under a given load. (A little like running in lower gear.)
Here's the scope showing the amplification. Input signal above, output below. Note that the input is on a scale 10x as sensitive as the output. You can see the peak-to-peak voltage readings in the right-hand column.
These signals are ten to twenty times what the real signals will be. Because so little of this is shielded, traces of signals at those levels are very, very noisy, and if I run it off a 9v battery and pick the battery up the noise overwhelms the output signal.
The amplification and phase relationships are nice and flat from a little under 20Hz to about 33 kHz. Leach has a very nice design. My big concern is how much extra noise my voltage regulation introduces.
