On 03/31/2014 03:38 AM, Len Ovens wrote: not at all. analog parallel filters can have some benefits wrt signal-to-noise ratio, but you need to get the phase alignment right. for digital, it's mostly an unnecessary complicaton, unless you are going for effects that really require parallel processing (such as upwards compression). unless you know very well what you are doing, parallel signal paths are full of dangerous pitfalls. a while ago, someone on some mailing list brought up the concept of parallel equalising (as in, split the signal, filter one half, recombine it with the original). reasoning was, hey, parallel compression is what the pros do, so parallel eq'ing must be great too, right? makes no sense, and produces something like this (the blue line is what you set your eq to, the red line is what you get when combining it with an unprocessed signal): http://stackingdwarves.net/download/4th-order%20highpass.png http://stackingdwarves.net/download/4th-order%20lowpass.png http://stackingdwarves.net/download/Parametric%20Mitra-Regalia.png morale: anything that introduces group delay shouldn't be used in parallel processing unless you know what you are doing. hth, jörn -- Jörn Nettingsmeier Lortzingstr. 11, 45128 Essen, Tel. +49 177 7937487 Meister für Veranstaltungstechnik (Bühne/Studio) Tonmeister VDT http://stackingdwarves.net

On Sun, Mar 30, 2014 at 06:38:14PM -0700, Len Ovens wrote: They almost always have a series connection of individual sections. Putting the sections in series means that they are independent of each other, the gains (in dB) just add, and the order doesn't matter. Parallel arrangements are used in e.g. 1/3 oct band equalisers. Putting filter in parallel requires very careful design because strange things may happen in the regions between the center frequencies due to phase differences. For the same reason, combining an equalised signal with the original may produce unexpected results. Things are different *within a single EQ section*. For example implementing a parametric or shelf filter as the sum of the input and a filtered version of the same has some advantages, even if you get exactly the same response (including phase) as when using another scheme. For one, if the gain is set to zero, you get a bit-exact copy of the input as output. This is not the case for e.g. a biquad set for a flat response - this will act as two filters that just happen to cancel each other. But even then such sections should be put in series. My 4-band LADSPA (also available as an LV2) works that way. The advertising blurb should at least tell you what the thing is doing... If you put filters in series with nothing non-linear in between them then the order doesn't matter, provided 1) you use a sample format with nearly unlimited dynamic range such as floats, and 2) none of the filters has gross defects. The second condition is not satisfied with all LADSPA and LV2 plugins, there are some that will happily add noise at -50 dB relative to the signal for some combinations of inputs and settings. Parallel arrangements have their place, e.g. for multiband compressors used for mastering. Regarding these you should be aware that such processing makes nonsense of whatever careful EQ you have done before. Multiband compression amounts an EQ that is changing all the time, and you usually have no idea of what exactly it is doing. If the mixing and mastering are done by the same person (as they usually are in our context), everything done in mastering could have been done better during the mix, e.g. by compressing tracks instead of the mix. Ciao, -- FA A world of exhaustive, reliable metadata would be an utopia. It's also a pipe-dream, founded on self-delusion, nerd hubris and hysterically inflated market opportunities. (Cory Doctorow)