Novel current source: the Tail Lengthener

[Sch3mat1c]

Contemplating regulators one day, I thought of the general form of some typical circuits. Of course you are all aware of the series/parallel and voltage/current identities which occur in basic electronics analysis.

The typical forms are:
Pass voltage regulator: these start with a not-very-good voltage source (most often a somewhat resistive power supply) and have a dynamic (active) resistance (such as a transistor) in series with the load. To regulate voltage, the error amplifier reads the voltage in parallel with the load. Input, output and ground make this a three terminal regulator.
Shunt voltage regulator: a not-very-good constant current source (most often a series resistor from a voltage source) drives the load with excess current. A dynamic resistance is in parallel with the load, shunting excess supply current as needed. This regulator needs only two terminals.
Pass current regulator: the most common type, a voltage source supplies a dynamic resistance in series with a current sensor and the load. This is a two terminal device.
So the one remaining layout is...
Shunt current regulator: a current supply (perhaps a resistor) drives a load in parallel with a dynamic resistance. The current sensor is in series with the load, so there are input, output and ground pins, a three terminal regulator.

What does this look like? Here are two representative schematics.



On the left, you essentially have the Io pin driving an emitter follower (assume for a moment that Rc = infinity) driving a constant resistance load, Rt, the "long tail" resistor. As such, the incremental resistance of the Io pin will be around hFE * Rt, which is a modestly constant current, but has no offset so it's no better than a resistor. The B-E resistor Rc adds that offset by shunting base current away until about 0.7V drop across the resistor, thus setting the output current. The result is an I/V curve which looks like a typical open collector, except the knee bends over at a rather high voltage (maybe 1/3 the supply), and the linear range has a bit steeper slope. Maybe more like a pentode, if you're into that. The downside is, a large amount of current is wasted, particularly at high voltages (if the knee falls at 1/3 +V, the shunt current near +V must be twice the output current!), but this behavior is understood as the nature of any shunt regulator.

This circuit is remarkably simple, using only two resistors and one transistor -- simpler, even, than any standard [series] current source using BJTs. It suffers from many downsides, including: wide knee (due to finite transconductance), poor incremental resistance and awful tempco (current will drop by about half from R.T. to 125C).

The circuit on the right is a somewhat more refined approach. The voltage developed by the current sense/set resistor Rc is compared to a TL431 bandgap voltage reference (2.50V) using a differential pair. For the diffamp, I've chosen a tail resistor Re, which is more than sufficient as the emitter voltage doesn't vary much beyond 1.8V (above common) in use. The differential pair outputs a constant current which appears across Rb, which biases the shunt transistor, which itself outputs a constant current to Rt, shunting the requisite tail current as needed. Because transistors are themselves native CCSs, PSRR is reasonable.

I call this connection the Tail Lengthener, because you can take a regular tail resistor, put the sense resistor and circuit in series (at the load), add a supply connection and have a much "longer" tail as a result.

Tim

[Duo]

This looks like the perfect way to waste more power in class A tube circuitry.

What a grand idea!