Float 16 in C++

freddyliu

Hi,

I was using the Bela board for some low-latency audio processing and had two questions:

Since the bela records audio with 16-bit bit depth, is there a way to save the audio files in 16 bit? Right now I'm reading the received signal using C++ floats and when I write they are written in 32-bits.
Is there a way to perform computation using 16 bit floats? Right now because of the low latency requirements my program is getting some underruns while performing operations with 32 bit floats. Since the audio is recorded with 16bits anyways.

Thanks in advance for the help!

giuliomoro

What files are you writing? If you use the AudioFileWriter class, it will write files as 16-bit integers https://github.com/BelaPlatform/Bela/blob/dev/examples/Audio/file-recorder-stream/render.cpp .

Is there a way to perform computation using 16 bit floats?

No. 16-bit floats are actually very low resolution and not suitable for audio because some bits are wasted on the exponent and they are not supported in hardware. 16-bit int is what you'd want but I don't think you'd get better CPU performance.