I’ve been developing 22-33KHz square-wave oscillators for the Harmonic VCO. And I’ve only recently discovered that the constant current sources (CCS) I’ve used before have some temperature effects. Not something that’s a problem supporting differential pairs or gain cells, but definitely a problem where a precise current is needed to set musical pitch. Stable, drift-free operation is required for that.
So, this weekend, I developed a “stable platform,” a CCS that is rock solid. An original design I’ve not seen anywhere else, in bottom half of work page below:
The innovation here was recognizing that if BJT junctions were used to set up voltage to current interfaces, then such a scheme could also work with BJTs of opposite polarity, esp. with one transistor diode connected. This formulation should lead to balancing Vbe temperature coefficients. In fact, the new CCS design behaves quite well, the opposing NPN + PNP BJTs offer first-order temperature compensation, which was borne out by early testing. The design actually has self-compensating negative feedback, offering extremely good thermal response, and repeatability. Particularly as the BJT designs were essentially complementary, using 2N5087 (PNP) and 2N5088 (NPN). Another nuance was to thermally connect the two transistors, which was done merely by strapping them together with a Ty-Wrap cable tie (see Breadboard, below). I am still seeking a metal dual TO-92 heatsink, but these are apparently not easily obtainable anymore. I may have to have them manufactured. On a whim, I also simulated the design on iCircuit, an iOS simulation tool. Whoa! Using default transistors having β=100, the design simulates extremely closely with measured currents and voltages! The 2N5088/2N5087 combination are quite complementary, and with β ≥︎ 300, offer practically “magic” performance, e.g. act very nearly as ideal BJTs.
This CCS building block already looks versatile, suitable for numerous applications. In fact, by reversing the roles of the transistors, and using a negative power-supply voltage, a mirror-image Constant Current Sink can be constructed.
So, this weekend, I developed a “stable platform,” a CCS that is rock solid. An original design I’ve not seen anywhere else, in bottom half of work page below:
Work
The innovation here was recognizing that if BJT junctions were used to set up voltage to current interfaces, then such a scheme could also work with BJTs of opposite polarity, esp. with one transistor diode connected. This formulation should lead to balancing Vbe temperature coefficients. In fact, the new CCS design behaves quite well, the opposing NPN + PNP BJTs offer first-order temperature compensation, which was borne out by early testing. The design actually has self-compensating negative feedback, offering extremely good thermal response, and repeatability. Particularly as the BJT designs were essentially complementary, using 2N5087 (PNP) and 2N5088 (NPN). Another nuance was to thermally connect the two transistors, which was done merely by strapping them together with a Ty-Wrap cable tie (see Breadboard, below). I am still seeking a metal dual TO-92 heatsink, but these are apparently not easily obtainable anymore. I may have to have them manufactured. On a whim, I also simulated the design on iCircuit, an iOS simulation tool. Whoa! Using default transistors having β=100, the design simulates extremely closely with measured currents and voltages! The 2N5088/2N5087 combination are quite complementary, and with β ≥︎ 300, offer practically “magic” performance, e.g. act very nearly as ideal BJTs.
This CCS building block already looks versatile, suitable for numerous applications. In fact, by reversing the roles of the transistors, and using a negative power-supply voltage, a mirror-image Constant Current Sink can be constructed.
Breadboard
Simulation
Comments
Post a Comment