On 2014-09-30 15:59, Alexander E. Patrakov wrote: > 30.09.2014 19:07, David Henningsson wrote: >> >> >> On 2014-09-26 14:17, Tanu Kaskinen wrote: >>> A summary of my proposal: >>> >>> To add LFE low-pass filtering, just hack the current remixer >>> implementation, no need for interface changes anywhere. >> >> Actually, I think this could be good enough, at least as a first step. >> >> Do you think it would be okay to: >> >> - Implement a LR2 or LR4 filter by "hacking the current remixer >> implementation" > > For me, it would be okay, but I would prefer a slightly more complex > plan, purely to avoid the blame of the form "you demand fully rewindable > code everywhere from others but just wrote an LFE-remixer that doesn't > handle rewinds correctly". The plan is: > > 1. Implement functions to calculate the coefficients of the required IIR > filters for LR2 and/or LR4. > > 2. Implement, in a separate file, some set of functions that would allow > one to apply an arbitrary matrix of IIR filters on a memblockq. Make > sure they are fully rewindable (that's the PR part), write a test. > > 3. Connect functions from (2) to the existing remixer implementation. If > it turns out that it is impossible to form the rewind requests > correctly, write a TODO item. > > (2), except the test, already sits on my laptop since the previous year > in the form of a virtual sink. I just need to refactor it from the > "parallel biquads" to the "chain of biquads" form (because both LR2 and > LR4 have pairs of identical poles which are inexpressible in the > parallel-biquads formulation) and remove the sink code. Reviving this thread due to renewed interest from management. In short, we should probably go ahead and write this filter. I can certainly try to do it, but I assume you will just complain about me breaking the math. :-) Did you end up doing anything more with your existing code? Need a hand anywhere? -- David Henningsson, Canonical Ltd. https://launchpad.net/~diwic
