In fact, with a bit of ingenuity, the lamp itself can serve as sensor: no need to use a photoresistor:The output characteristic of the optocoupler with miniature incandescent lamp isn't so good reproducible as in its LED counterpart. It's enough to shake the device and the value of its output resistance may be changed due to the filament's flexible suspension, the lamp being simultaneously a source of light and heat.
This drawback is particularly annoying in the dual optocouplers where the close matching of both resistive outputs is very important.
But these so-called obsolete optocouplers have an excellent ability to convert the RMS value of input AC current to the output resistance and then to a DC current.
In comparison with traditional thermo-couples these devices have smaller thermal inertia, higher output signal level and wider dynamic range of input AC voltage.
In my VK-8 RMS converter the optocoupler's lamp is alternately supplied with the AC and DC currents being correspondingly proportional to the unknown AC voltage and a regulated DC voltage. The lamp heating produces light which acts upon a photoresistor, the converter's automatic system maintaining its resistance always the same.
Therefore, the produced DC voltage can serve an exact (0,3%) RMS equivalent of the input AC voltage lying within 0,1-1V and having frequency range of 10Hz-1MHz.
As for the optocoupler own distortion, it can be made sufficiently low, if the voltage across its photocell is kept below 100mV and its resistance doesn't exceed 1-2kohm.
The optocoupler linearity also has the tendency to be better for the devices which require higher LED currents to reach the same output resistance. My experiments show that Tesla 3WK163-43 optocouplers exhibit slightly lower distortion in the equal working conditions.