A question that comes up from time to time is should I sample rate convert before or after dithering?
Typically, SRC comes before dithering but I wondered if there would be any benefits to SRC after dither.
The above image depicts dither noise, which has been applied to a 1 KHZ sine wave with a 44.1 KHZ sample rate.
The above image depicts dither noise, which has been applied to the same 1 KHz sine wave with a 192 KHz sample rate.
As you can see, the noise is much less audible in the 192 KHz session.
However, does the process of SRC generate artefacts e.g. truncation.
To evaluate this notion, I bounced a 1 KHz sine wave at 192 KHz, 32-Bit Float Wave and imported it into RX7. I then analysed two versions of the file 1. SRC to 44.1 KHz followed by Dither to 16-Bit 2. Dither to 16-Bit proceeded by SRC to 44.1 KHz.
As expected, the SRC>Dither is identical to dithering from a 44.1 KHz session.
Dither>SRC does truncate, generating more audible artefacts. In particular, truncation noise is correlated and is far easier to hear then the uncorrelated noise of dither.
To conclude, always SRC before Dither, ideally the last two steps before distribution.