We want an environment where we can play speech recordings and re-record to determine distortions added by the equipment. We'd like it to be reproducible, meaning if we repeated the test 10 times, we'd get similar results each time. The recording apparatus is very small, like 1 cm x 5 cm x 10 cm and the cone of the speaker to play the recordings is like 4 cm in diameter.
We're thinking a mini anechoic room, something as simple as a double-walled cardboard box with some bumpy foam inside. Suspend the recording apparatus with clips on the ends of thin wires. The speaker can be embedded in the foam with just the front exposed. We can put the box in a quiet room to limit external noise.
Is there a limit to how small the box can be, or should the box be large enough to accommodate the lowest wavelength of sound in speech?
Since all the sound energy eventually degrades into heat, would we expect the internal temperature to increase significantly? If so, should we look into foam with a low thermal expansion rate?