An unavoidable fact of life is that all parameters of all electronic components drift with temperature. Even the best voltage references, op amps, crystal oscillators, etc., have non-zero temperature coefficients. These effects can be handled by compensation methods, but in demanding applications, the only solution may be controlling the temperature of the component with a constant-temperature component oven.
Ovens tend to be relatively large (they must be at least big enough to hold the heated part) and thirsty, so they become to difficult to fit into miniature, power-efficient designs. The idea presented here can’t eliminate these problems, but it helps in minimizing both. It’s based on two ideas: First, multiplexing a power transistor so it alternates between temperature measurement and heating, and thereby control its own temperature. Second, close thermal coupling (e.g., via thermal epoxy) of the transistor to the “Component Under Thermostasis” (CUT). Thus, when the transistor thermostats itself, it also thermostats the CUT bonded to it.
To follow circuit operation, first assume Q1 is slightly above set-point temperature (see the figure). Then its VBE, and therefore A1 pin 3, will be less than the set-point voltage at A1 pin 2, and A1 pin 6 will sit at approximately 0 V. This holds the multivibrator S1 pin 11 low. S1 pin 14 also will be low, and S2 will be OFF, so no base drive flows to Q1. Q1 will gradually cool until the relationship at A1’s inputs reverses. A1 pin 6 then will go to +5 V and C1 will start to charge. About 700 µs later, S1 will switch ON, turning S2 on. This applies about 10 mA of base drive to Q1, and reverses the state of S3, which begins the discharge of C1 through R6. This state lasts for approximately 10 ms and deposits about 0.05 joules of heat in Q1. C1 then recharges and the cycle repeats until Q1 returns to set point, thus establishing a feedback loop that tends to hold Q1’s base-emitter junction at a constant temperature, which is roughly 55°C for the circuit values shown.
Just because Q1’s junction temperature stays constant doesn’t mean the CUT’s does! In fact, the inevitable thermal resistance of the bond between Q1 and the CUT will allow some “playthrough” of ambient temperature variations into the CUT’s temperature. D1 serves to compensate for this effect by effectively increasing Q1’s set-point temperature by about 1°C for every 14°C decrease in ambient temperature. For the typical case of an eightpin DIP package bonded to Q1, this does a good job of canceling heat loss and holding the CUT’s temperature “rock” steady. Even without any special effort at insulation, the small size of the Q1 + CUT assembly, and near 100% efficiency of Q1 as a heater, hold power demand to about 700 mW against a 25°C ambient. Warm-up takes less than one minute.