I am using the current stable JRiver build and all tracks are analysed. I have read up on volume levelling and adaptive volume and concluded that volume levelling only seems like the right approach, at least in part because I have an RTi remote which is not yet programmed to drive the PC. The appears to work as intended except it seems to result in having to crank my amp up way too high producing an small but audible uptick in the noise floor in the room. For example, moderate evening music listening now requires an amp level in the region of +2 or so even with all volumes controls (jriver interval volume, saffire mixcontrol volume, hardware pot on the saffire pro 24) at 100%. My previous, non htpc based setup, would hit similar volumes at -20 or so for reference.
My DSP setup is currently
* output format = 5.1, upsample to 96kHz, no mixing
* volume levelling on
* adaptive volume off
* room correction off
* convolution on with normalise filter volume on
I have looked at contents of my library and can see the R128 peak level varies from -11.8 to +4.2 but I have no idea what, if anything, I can do with this information.
I have seen some threads that refer to the filter being used for convolution resulting in large cuts. The filter I have (R channel in a 2.1 setup shown) is not "done" yet (see attached) but my thinking is that I currently have typically -10 coming from volume levelling, -6 from the convolution "normalize filter volume", maybe another -10 or so from the filter itself (just looking at levels in the amplitude chart). The only one of those that is obviously under my control is the filter itself but that will inevitably involve some boosting (for the sub) so seems a minor contributing factor in total.
One thing I haven't done is the "reference level calibration" in room correction as I can't work out how to add room correction when convolution is active (it forces the sub not to play as far as I can see). I thought that setting room correction after convolution and marking the channels as full range would therefore leave the per channel routing (done in convolver) intact but I don't think it worked.
Thanks
Matt