Re: [Audacity-devel] Equalization results in "pre-echo" ??? Bug?
A free multi-track audio editor and recorder
Brought to you by:
aosiniao
From: <Mar...@ao...> - 2005-05-25 00:00:42
|
Checking the archives, this message appears to have gone missing, so I re-post. Sorry if it's a duplication but the archive seems empty from 7-9 May. In a message dated 06/05/2005 00:54:13 GMT Standard Time, Mar...@ao... writes: Markus The short answer is 'no'. I didn't see the gross effects with the FFT filter. The long answer is... I have tested the FFT filter with a series of tones, white noise, impulses (at a 44100Hz sampling rate) and the file mentioned earlier. I do not see the same effects but I do see some others, not entirely desirable. I normalised the amplitude of the test signals to -12dB to avoid any clipping. Then I tested the generated test signals with a 'brick wall' filter with a cut-off of approx. 5kHz and set the gain to +12dB and cut to -12dB (I feel this is a rather low value for the possible attenuation, but perhaps this was discussed before I started lurking on this list). The impulse gave me an approximation to a sin(x)/x response (plus an impulse due to the non-infinite attenuation) of approx. 384 samples long (I imagine due to an FFT size of 256 and a 50% overlap) which was asymmetrical, for reasons I don't quite understand; perhaps the overlapping window is being applied in one direction and not the other. The leading part always seemed to be longer as I moved the impulse about, although I only tried a few positions. I tried the same thing in 'CoolEdit 2000' and got a symmetrical response wherever I put the impulse, again only trying a few. (Incidentally, Cooledit allows me to set the FFT size to a range of values from 128 to 24000, not just powers of 2.) The tones gave me filtering but a visible (in spectrum view) and audible artifact at the cut-off frequency which varied in amplitude depending on the tone frequency, being visibly higher for a tone near the cut-off frequency (but audibly masked somewhat by the tone). White noise results were inconclusive, due to the random nature of the signal, I suspect. The speech file definitely did not give the 'pre-echo' effect reported previously, and I have not seen them elsewhere if the FFT filter. For the speech file (I used a lower cut-off frequency of about 1kHz, due to the low sampling rate) spurious energy at the cut-off frequency was not apparent to me at this time, but it may well be signal dependant, judging by other trials. HTH Martyn ----------------------------------------------------------------------- In a message dated 05/05/2005 22:05:12 GMT Standard Time, me...@me... writes: Thanks Martyn, do you see these effects with the FFT filter, too? Markus Mar...@ao... schrieb: > I have tried to reproduce these effects on 1.2.3 Windows version of > Audacity using the _Ifl_EqualizationTest_NormNr.mp3 file, I set the > equalisation to 'clear' and applied it. I get some horrible aliassing > components, very visible on the 'Spectrum' view and very audible. The > 'echo' is also clearly visible on this view, being about 250ms long. > Similar things occur when I try some of the presets. > > To narrow down the problem, I generated a tone, using the presets, and > then repeatedly amplified the first and last few seconds by -24 dB to > make them very small to create a test signal. I then applied the > 'flat equalization'. Zooming in on a section of the waveform so that > you can see about one cycle reveals that there is something very odd > going on - the waveform has a regular zig-zag appearance > superimposed. Looking at either end of the tone section also reveals > some very strange effects, like bits of the waveform being read out of > a buffer backwards or something. This definitely looks like a bug to me! > > Martyn > > In a message dated 04/05/2005 15:16:46 GMT Standard Time, > l_d...@ad... writes: > > > I'll check later when I get home. Reproducing the problem is the > > first step. > > > Here are some files that demonstrate the problem: > > http://cleanspeech.sf.net/misc/_Ifl_EqualizationTests.zip > > _Ifl_EqualizationTest.mp3 is the original 32kbps file > > _Ifl_EqualizationTest_NormNr.mp3 has been normalized and de-noised > > _Ifl_EqualizationTest_Norm_Nr_EqFlag.mp3 has had flat equalization > applied to it. It shows the "soft pre-echo". > > HTH > > > > > > ------------------------------------------------------- > This SF.Net email is sponsored by: NEC IT Guy Games. > Get your fingers limbered up and give it your best shot. 4 great > events, 4 > opportunities to win big! Highest score wins.NEC IT Guy Games. Play to > win an NEC 61 plasma display. Visit http://www.necitguy.com/?r=20 > _______________________________________________ > Audacity-devel mailing list > Aud...@li... > https://lists.sourceforge.net/lists/listinfo/audacity-devel > > ------------------------------------------------------- This SF.Net email is sponsored by: NEC IT Guy Games. Get your fingers limbered up and give it your best shot. 4 great events, 4 opportunities to win big! Highest score wins.NEC IT Guy Games. Play to win an NEC 61 plasma display. Visit http://www.necitguy.com/?r=20 _______________________________________________ Audacity-devel mailing list Aud...@li... https://lists.sourceforge.net/lists/listinfo/audacity-devel |