Thermopile Amplifier Circuit Explanation The thermopile used has an output of 1mV/W. With a gain of 1000 this gives 1V/W (the setting we use most). There is a jumper to select one of two feedback resistor/capacitor combinations (Gain of 100 or 1000). The ceramic capacitors were chosen for a 10ms time constant for rolloff frequency around 16Hz which is much faster than the actual thermopile. The noise at the amplifier output is less than 1mV (1mW) with the R/C filter. The op-amp (AD8638ARZ) is a single supply, rail-to-rail output, auto zero amplifier. The typical offset at room temp is 3uV (I measured about 1uV). The output needs to go to 0V when the thermopile output is zero. This requires a negative voltage for the op-amp although it can be very small because of the rail-to-rail output, about 0.1V or more. An isolated switching power supply (wall-wart) was used for convenience. Most of the switching noise was filtered out using a high speed diode and capacitor to make a smoothed rectifier (with a very light load). Ferrite beads (tiny inductors) and more R/C filters further knocked the noise down to below 1mV (the resolution of my scope). A resistor and forward biased diode were used to to make the -0.7V negative rail. The circuit ground is floating and the positive rail is the positive side of the wall wart (after filtering). Note: The die-cast box is tied to the floating circuit ground via an aluminum standoff. This, along with the small circuit dimensions and a shielded thermopile cable, minimize noise pickup. Also, the op-amp can't handle more than 16V (V+ to V-). The max output voltage is close to the power supply voltage minus the two diode drops (about 1.4V). We're using a 9V supply so the max output voltage is about 7.6V (9V - 1.4V). Depending on the max output voltage one could use a 5-15V power supply. Note: The new version has the correct PCB layout for diodes 1 & 2 (old version was too small). The new version also uses five 22uF low ESR/ESL caps in parallel instead of one generic 100uF cap. This should cut down on the power supply ripple voltage (not that it was noticeable).