From Avisynth wiki
Revision as of 02:55, 14 December 2015 by Raffriff42 (Talk | contribs)
(diff) ← Older revision | Latest revision (diff) | Newer revision → (diff)
Jump to: navigation, search
Re: behavior for invalid preset files - the following is probably true:
If SuperEQ encounters an error in the preset file, all bands after the invalid line will be set to the last valid decibel value;
but if the file can not be parsed at all, all bands may be set to +20 dB.

but I have not tested it very much.
Raffriff42 17:56, 16 September 2015 (CEST)

I guess i need to run it through a debugger to verify the above. I can't make sense of the code itself. I see code like 'my_eq.bands[n] = ((-readval)+20);'. Indicating that the first band in Loudness.feq (4dB is changed to 16dB). It's probably my misunderstanding.
Anyway a couple of things are still missing:

1) definition of equalization. I guess here it means:
Since equalizers, "adjust the amplitude of audio signals at particular frequencies," they are, "in other words, frequency-specific volume knobs (source: Seems to apply here too (well at frequency bands instead of particular frequencies).
2) Why are the bands (1-18) chosen the way they are. Guess we need to ask Sh0dan. I see that indeed the lower frequencies of the bands are C notes, and the higher frequencies F# notes (see and So they are spaced half an octave (sqrt(2)) apart. Why? Does this make sense?
3) About your pictures (the white noise examples): Why does a value of 0dB in a band correspond with -44dB in the spectrum analysis?
4) The lowest bands C-1, C0 and C1 are missing (same for F#). Why?

Admin 01:12, 19 September 2015 (CEST)

Definition of equalization - it originally meant to flatten the frequency response (going way way back to analog telephone lines, if not before) but now "equalizing" can also mean any audio effect that tweaks the frequency response, such as making a musical instrument sound "sizzling," "thumping" or "honking."
Does it make sense? From an old-time audio engineering POV, no. They would want 1/3 octave equalizers, since that's what they're used to. Here's a professional 1/3 octave equalizer; note the center frequencies are industry standard.
Audio engineers also don't want such sharp separation in the frequency bands. They want some interaction between the bands, which gives them more control. There's a standard Q factor for 1/3 octave graphic equalizers, which I don't recall.
Very sharp analog filter cutoffs are prone to phase anomalies, but SuperEQ seems OK here, probably because of the extremely high (16383) FIR filter order. Having little experience with an equalizer made this way, I made some tests (like subtracting EQ'd and not-EQ'd sweep tone) and did not find anything that would make me avoid this filter. It seemed fine, actually. It's just different!
Why does a value of 0dB in a band correspond with -44dB in the spectrum analysis? Spectrum analysis is not perfect - there a tradeoff between frequency and time resolution. Some low-level noise/artifacts are to be expected.
The lowest bands C-1, C0 and C1 are missing (same for F#). Why? I am confused by the question. One, I have never heard equalizer bands referred to by musical notes, and Two, the lowest band's effect extends to zero(!)
Raffriff42 03:03, 19 September 2015 (CEST)
EDIT Oh, you're asking me to answer these in the article...will do, but I'll have to get citations and stuff.
Raffriff42 13:30, 19 September 2015 (CEST)
Would be nice if you can expand it a little bit (only when relevant of course). Could you enlarge the pics by 250% (so they are cleary visible to the reader)? (Ok, -44dB is the noiselevel i guess.) How do people choose the equalizer bands? Are different choices made or are the bands usually the same? Wrt the question. I noticed that the lowest frequency of every band is a the frequency of a C-note. Octaves range from Cj to C(j+1), as least according to and . I don't know anything about this stuff, just saying what i read. ;) Admin 23:09, 20 September 2015 (CEST)
Personal tools