DSP Robotics Support

bass, treble c-weight filter, google, search.

Wasn't there a whole thread about LU filters like two weeks ago?

("DSP" plus "plain English" results in things like, "That block on the diagram makes stuff sound different." )

deraudrl wrote:Wasn't there a whole thread about LU filters like two weeks ago?

("DSP" plus "plain English" results in things like, "That block on the diagram makes stuff sound different." )

well, I mean someone can just use the module I put together and add that to signal of a limiter, that's simple enough. But also, check out the example of a limiter that's available here:

http://dsprobotics.com/support/viewtopi ... ls#p105394

but still.. I mean My last post before had no "tech talk". Use the esquire limiter example to derive what you will how LU works.

wlangfor@uoguelph.ca wrote:

deraudrl wrote:Wasn't there a whole thread about LU filters like two weeks ago?

("DSP" plus "plain English" results in things like, "That block on the diagram makes stuff sound different." )

well, I mean someone can just use the module I put together and add that to signal of a limiter, that's simple enough. But also, check out the example of a limiter that's available here:

http://dsprobotics.com/support/viewtopi ... ls#p105394

but still.. I mean My last post before had no "tech talk". Use the esquire limiter example to derive what you will how LU works.

Exactly. My post was directed at steph_tsf, who routinely does page-long esoteric data dumps in this subforum and now, inexplicably, wants "plain English".

deraudrl wrote:

wlangfor@uoguelph.ca wrote:

deraudrl wrote:Wasn't there a whole thread about LU filters like two weeks ago?

("DSP" plus "plain English" results in things like, "That block on the diagram makes stuff sound different." )

well, I mean someone can just use the module I put together and add that to signal of a limiter, that's simple enough. But also, check out the example of a limiter that's available here:

http://dsprobotics.com/support/viewtopi ... ls#p105394

but still.. I mean My last post before had no "tech talk". Use the esquire limiter example to derive what you will how LU works.

Exactly. My post was directed at steph_tsf, who routinely does page-long esoteric data dumps in this subforum and now, inexplicably, wants "plain English".

And I'm not a fan of baseless terms, thrown in without any background. So I understand steph_tsf's wish to have it explained in clear words. Loudness units describe in essence amplitude. Filters work on frequencies. There is no relation between the two that actually makes sense with the information we've been given so far.

Were would we end if anybody here just invents their own names for something they build? The common language on this forum is science. That's math, physics and whatever else is needed to explain a mechanism. Without an explanation it is just an annoyance.

wlangfor@uoguelph.ca wrote:bass, treble c-weight filter, google, search.

Kind of joke. I vaguely assumed that "LU" was "Loudness Unit". I was expecting a fresh open discussion about how to properly sum dB HL (Hearing Level) components, following the Fletcher-Munson "isophonic" contour curves. Kind of adaptive "loudness" compensation, done in digital precision. The Fletcher-Munson "isophonic" contour curves don't run parallel. There is thus a form of mild multiband compression / expansion requirement, hiding over there. Are there digital multiband compressors / expanders , dedicated to this (for classic music maybe)? How to read and interpret the Fletcher-Munson "isophonic" contour network, in case there are two or three different sounds playing together, very dissimilar in frequency? I have the impression that the wireless digital speakers branded "Sonos", try incorporating such kind of compensation, kind of adaptive "loudness". I am not pleased with the result. They sound boomy, sometimes, on some audio contents.

steph_tsf wrote:

wlangfor@uoguelph.ca wrote:bass, treble c-weight filter, google, search.

Kind of joke. I vaguely assumed that "LU" was "Loudness Unit". I was expecting a fresh open discussion about how to properly sum dB HL (Hearing Level) components, following the Fletcher-Munson "isophonic" contour curves. Kind of adaptive "loudness" compensation, done in digital precision. The Fletcher-Munson "isophonic" contour curves don't run parallel. There is thus a form of mild multiband compression / expansion requirement, hiding over there. Are there digital multiband compressors / expanders , dedicated to this (for classic music maybe)? How to read and interpret the Fletcher-Munson "isophonic" contour network, in case there are two or three different sounds playing together, very dissimilar in frequency? I have the impression that the wireless digital speakers branded "Sonos", try incorporating such kind of compensation, kind of adaptive "loudness". I am not pleased with the result. They sound boomy, sometimes, on some audio contents.

Though I'm sure the Fletcher Munson curves were taken into account and possibly an inspiration; LU derives its weighting wholly from A, C weighting and sometimes double precision to round the decimals in a more accurate way.

And in the regards of speakers, that's math regarding the size of the speaker and some dbspl. I've hoped that LU would be a good bridge to making some of the dbspl calculations conversion ready to a distance grid, but that's sci-fi atm. Maybe though; you can use dbspl in a manner that is resulting in the diminishing of the sound, the linear rarefaction. Assuming you can have a range of sound, then you can also have a rounded estimate of where, or what distance the sound fades out of human hearing's range.

And after all isn't that the same thing? And could not compression breakpoints be derivative from the same formula. Funny how instead relying on a 0-1 float results in a more palpable calculation based on the formula of diminishing sound. Compression = distance.

TBH, atm we gain nothing from double precision because it's only the alpha that can handle double precision with any certainty. Every time we use code, dsp, or a primitive that doesn't account for 16 or 32 extra decimals we'll be merely adding to the rounding, we lose what we gain after three re-calculations. Double precision is best for the last calculation due to this.

Maybe that's food for thought, but I gave up talking about such heady things a while ago.