I currently re-samples everything in Roon and Audirvana to 24/48 before outputting the signal to my Lumin U2. I do this because I prefer letting Roon and Audirva do the conversions rather than doing it in my active speakers that perform all DSP on 24/48 signals. I simply think SQ is marginally better this way (using “Precise linear phase” re-sampling filter in Roon and similar in Audirvana).
However, since I also use the U2 for TV-sound sent over AirPlay from my Apple TV 4K, I have activated the re-sampling in the U2 in order for the signal from the Apple TV to be converted to 24-bit.
I am primarily using the Leedh volume control in the U2 with the DSP volume in the speakers fixed at – 15 dB (in order to avoid the lowest range in Leedh that is not sufficiently granular) and I normally listen at very low level, between -45 to -55 dB.
Now to my question:
Sending 32/48 signals from Roon and Audirvana to the U2 sounds audibly different compared to 24/48. The biggest difference is in the bass where 24-bit sounds rounder and 32-bit leaner and more staccato. A bit similar to comparing ported and sealed speakers in the bass region.
So, my question is: Does the Leedh attenuation work in 32-bit if incoming signal is 32-bit, or is the signal first converted to 24-bit since I have activated the re-sampling to 24-bit in the U2. That is, where is any re-sampling to 24-bit taking place in the U2. Before or after Leedh attenuation?