Sorry for the delay, I was traveling and did not have access to a PC to type a real message on.
>You expressed your doubts that this has a discernable effect on audio. Does it? According to the manual it looks as though it should not, as the two filters have cutoffs above 20 kHz. Even without knowing the precise phase response, it seems very unlikely that it will be meaningfully different in the audio band. The setting appeared to me to be dictated by marketing rather than any actual effect. This is not the case with all DACs, I have certainly seen some which give you the (completely useless) option to hard code a high shelf EQ band into the reconstruction filter, but it doesn't look like this is the case here. >It my be, that this has no discernable effect on audio, but I for once want to least have the option to decided this for myself. >http://gerrit.rockbox.org/r/#/c/467/ That is a fair point. I almost rejected that commit as well, but I didn't fully understand what it did so I deferred to the device maintainer. Looking at it now, that option is also somewhat pointless, since the slow rolloff option essentially duplicates what our EQ can do while the fast roll off looks uniformly better. >In the same line as 44.1/48kHz playback, software balance, parameteric eq, surround filters and Doom. Its not like those things. The mixer sampling rate, software balance, and EQ have no correct setting. They depend on the hardware that the device is connected to and the input files that user intends to play. Therefore we cannot anticipate what to use, but a user can, and so they should reasonably be configurable. In contrast, reconstruction filter coefficients have an optimal value that minimizes distortion and frequency error. This value will be correct for all users, all headphones, and all source files as it does not depend on any of those things. My recommendation was that you determine the correct value, and use it, because you, the device maintainer, most completely understand the hardware. Most likely the datasheet hints at what it should be, probably by defaulting to the setting the designers intended you to use. >When it comes to audio we generally try to expose all hardware capabilities, if only to make audiophiles happy, because that's our "core business". I'm all for exposing capabilities (although I also dislike how many settings we have), but not for settings that are actually useless, and also extremely confusing/misleading. I have a doctorate in a related field and it took me several hours of googling through AES conference proceedings to fully grasp what the underlying concept of this setting was. This is not like setting crossfeed or tone controls or whatever, its more analogous to exposing DMA settings or PLL dividers to a user: extremely technical and not a great idea to screw around with (although in this instance Wolfson seems to have realized that and completely neutered the setting). Mike