The officially official Devuan Forum!

igorzwx

Member

Registered: 2024-05-06

Posts: 510  

Do ALSA users really need fftrate?

Bypassing the hardware resampler in Intel HDA codecs with a software real-time fftrate codec to resample everything to 192kHz 32-bit is a solid architectural strategy. It works at the userspace level on top of ALSA and avoids potential quality issues in codec hardware SRC implementations.

HDA Codec Resampling in the Kernel

The Linux kernel's HDA subsystem does have hardware sample rate converter (SRC) capabilities managed by codec drivers. For example, the Creative CA0132 driver documentation shows that HDA codecs have dedicated SRC ports (0x80–0xbf) for sample rate conversion within the codec's internal routing.

Kernel-Level Format Handling

The kernel handles PCM format conversion through the snd_hdac_stream_format() function, which converts channel count, sample format, and sample rate into the HDA format value used by the hardware.

Hardware-Specific Workarounds

Some codecs have hardware-specific sample rate limitations that require kernel-level workarounds. For instance, the ALC269 driver has a fixup for the Lenovo Ideapad that forces analog I/O to 44.1kHz due to a hardware problem.

The fftrate project

The software resampling approach of the fftrate project bypasses these kernel-level hardware resamplers entirely by:
1.  Accepting audio at any rate from the application.
2.  Resampling to 192kHz 32-bit in userspace.
3.  Sending the already-resampled audio to ALSA at a fixed rate.

This avoids potential quality issues or bugs in the codec's hardware SRC implementation. The kernel ALSA driver simply sees a constant 192kHz 32-bit stream and passes it through without additional resampling.

Dialectical conclusion

There is a belief that fftrate may cure deafness, though it hasn't been verified by scientific methods. Give it a chance. It may (or may not) work.

NOTE: The claim that fftrate can "cure deafness" is a metaphorical exaggeration common in audiophile circles, not a medical fact. There is no scientific evidence that software resampling can restore hearing loss or cure physiological deafness.