> 1. how does this help if we need reference sine tones from another source? I did a series of test tones in Reaper, all normalised to -18lufs. Here ist the test-audio: https://cloud.4ohm.de/s/c7ynWemXS9eXibH
Ok, so you're saying you want to take these Reaper-generated sine files as ground truth. I was just probing the system with known sines, which I trust for sure. > 2. at 1000Hz and levels above -45db, the lufs readout seems plausible. Below -45dbfs, the meter falls to -69lufs. I don’t understand. This is curious. For me that behavior kicks in around -32 dBFS. It goes away in double precision, so I think it has to do with "max(ma.EPSILON)". However, while EPSILON is indeed around -70 dB when interpreted as power, it should not suddenly take a large drop like that. It should just clip. In any case, this is a trivial one - we can just use "faust -double" (and probably should anyway). > 3. probably just some coding style, but does amp = 10^(level/20); equals amp = level : ba.db2linear; ? Yes, no difference expected. "Finger macro" - fewer characters to type :-) > 4. with amp = level : ba.db2linear; meter drops to -69lufs below -40dbfs at 1000Hz… Same glitch as in 2, right? So, it sounds like the filter polishing task is to make the Faust implementation match Reaper, but how do we know Reaper is correct? Since it's open-source, why don't we just copy its filters? Cheers, Julius On Wed, May 11, 2022 at 4:02 AM Klaus Scheuermann <kla...@posteo.de> wrote: > Hi Julius, > > Hm, I think the averaging time should not matter with continuous sine > tones coming in, right? > (According to specs, 0.4 is the value for ‘momentary’ and 3 for ’short > term’.) > > Some questions about your test-program: > > 1. how does this help if we need reference sine tones from another source? > I did a series of test tones in Reaper, all normalised to -18lufs. Here ist > the test-audio: https://cloud.4ohm.de/s/c7ynWemXS9eXibH > > 2. at 1000Hz and levels above -45db, the lufs readout seems plausible. > Below -45dbfs, the meter falls to -69lufs. I don’t understand. > > 3. probably just some coding style, but does amp = 10^(level/20); equals > amp = level : ba.db2linear; ? > > 4. with amp = level : ba.db2linear; meter drops to -69lufs below -40dbfs > at 1000Hz… > > Thanks for investigating with me :) > > Klaus > > > > > Klaus Scheuermann > kla...@posteo.de > +491716565511 > @schlunk:matrix.org > 4ohm.de > trummerschlunk.de > > > > On 11. May 2022, at 04:50, Julius Smith <julius.sm...@gmail.com> wrote: > > Hi Klaus, > > Could the averaging time be too small? I'm worried about the line > > Tg = 0.4; // 3 second window for 'short-term' measurement > > The comment seems to indicate it should be "Tg = 3;", i.e., 3 seconds of > averaging instead of 0.4 s. > > Below is an expansion of your test program that allows for more > exploration. > > Cheers, > Julius > > import("stdfaust.lib"); > > freq = hslider("[0] Test Sine Frequency (Hz) [unit:Hz]",1000,30,16000,1); > level = hslider("[1] Test Sine Level (dBFS) [unit:dBFS]",-10,-80,0,0.1); > avg = hslider("[2] Averaging Time (s) [unit:sec]",3,0.01,10,0.1); > > amp = 10^(level/20); > testSignal = amp * os.osc(freq); > > process = testSignal <: _,_ : lk2 : vbargraph("[3] LUFS S",-40,0) : > *(1e-7); > > //Tg = 0.4; // 3 second window for 'short-term' measurement > Tg = avg; > zi = an.ms_envelope_rect(Tg); // mean square: average power = energy/Tg = > integral of squared signal / Tg > kfilter = fi.highpass(1, 60) : fi.high_shelf(4, 1800); > > // 2-channel > lk2 = par(i,2,kfilter : zi) :> 10 * log10(max(ma.EPSILON)) : -(0.691); > > > On Tue, May 10, 2022 at 7:00 AM Klaus Scheuermann <kla...@posteo.de> > wrote: > >> Hi Julius, >> >> this is, of course, the way to go. >> >> I did some test with a series of sines, each at -18lufs, and found these >> filter settings to be quite close (except for the 30Hz): >> kfilter = fi.highpass(1, 60) : fi.high_shelf(4, 1800); >> >> Hz lufs >> 30 -16.693 >> 60 -18.111 >> 80 -18.204 >> 100 -18.211 >> 130 -18.177 >> 180 -18.133 >> 250 -18.113 >> 300 -18.099 >> 400 -18.099 >> 600 -18.169 >> 1000 -18.405 >> 2000 -18.241 >> 3000 -17.894 >> 4000 -17.784 >> 6000 -17.503 >> 8000 -18.083 >> 10000 -18.026 >> 12000 -18.035 >> 14000 -17.784 >> 16000 -18.083 >> >> What I don't quite understand is why the values read a little different, >> every time I do the test with same filter settings and same sines. (I am on >> faustide.grame.fr). >> >> For my application, this should work fine. Out of curiosity, why do I >> approximate the filters? So that it works on all samplerates? >> >> Full test code is this: >> import("stdfaust.lib"); >> >> process = _,_ : lk2 : vbargraph("LUFS S",-40,0); >> >> Tg = 0.4; // 3 second window for 'short-term' measurement >> zi = an.ms_envelope_rect(Tg); // mean square: average power = energy/Tg >> = integral of squared signal / Tg >> kfilter = fi.highpass(1, 60) : fi.high_shelf(4, 1800); >> >> >> // 2-channel >> lk2 = par(i,2,kfilter : zi) :> 10 * log10(max(ma.EPSILON)) : -(0.691); >> >> Thanks!! >> Klaus >> >> >> >> >> On Sun, 2022-05-08 at 13:17 -0700, Julius Smith wrote: >> >> Hi Klaus, >> >> To go after this, it would be useful to measure the discrepancy for >> some number of sinusoidal frequencies across the audio band, with at >> least one example including both single-channel and multichannel >> input. >> Based on the filter approximations used, I would predict a measurable >> discrepancy around 1 kHz (guessed transition-frequency tuning), and >> very high frequencies (due to bilinear transform frequency-warping). >> >> The high-frequency discrepancy should go away with oversampling, even 2x. >> >> Glad to hear noise is looking good! >> >> - Julius >> >> On Sat, Jul 3, 2021 at 1:08 AM Klaus Scheuermann <kla...@posteo.de> >> wrote: >> >> >> Hello everyone :) >> >> Would someone be up for helping me implement an LUFS loudness analyser >> in faust? >> >> Or has someone done it already? >> >> LUFS (aka LKFS) is becoming more and more the standard for loudness >> measurement in the audio industry. Youtube, Spotify and broadcast >> stations use the concept to normalize loudness. A very positive side >> effect is, that loudness-wars are basically over. >> >> I looked into it, but my programming skills clearly don't match >> the level for implementing this. >> >> Here is some resource about the topic: >> >> https://en.wikipedia.org/wiki/LKFS >> >> Specifications (in Annex 1): >> >> https://www.itu.int/dms_pubrec/itu-r/rec/bs/R-REC-BS.1770-3-201208-S!!PDF-E.pdf >> >> An implementation by 'klangfreund' in JUCE / C: >> https://github.com/klangfreund/LUFSMeter >> >> There is also a free LUFS Meter in JS / Reaper by Geraint Luff. >> (The code can be seen in reaper, but I don't know if I should paste it >> here.) >> >> Please let me know if you are up for it! >> >> Take care, >> Klaus >> >> >> _______________________________________________ >> Faudiostream-users mailing list >> Faudiostream-users@lists.sourceforge.net >> https://lists.sourceforge.net/lists/listinfo/faudiostream-users >> >> >> >> >> >> > > -- > "Anybody who knows all about nothing knows everything" -- Leonard Susskind > > > -- "Anybody who knows all about nothing knows everything" -- Leonard Susskind
tlufs2.dsp
Description: Binary data
_______________________________________________ Faudiostream-users mailing list Faudiostream-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/faudiostream-users
