another MAutoEqualizer question :)

[Cooker]

I made a test and not sure how to evaluate result. With pink noise the analyzer shows very similar result with x plug-in analyzer but with a sine wave thats even more flat with x plug-in I can't get a flat line with mautoeq. Why?

Also seems it doesn't recognize S signal (as in mid-side) when analyzing, which I think isn't a good thing if used for mastering.

[MeldaProduction]

What do you mean by flat? With a sine you should have just one line with some sidelobes and since MAutoEq uses smoothing (you can change this) it becomes a hill.

I'll check the stereo version. Anyway the analyzer merges both channels. If you need just to equalize stereo or mono content separately, you can just put it into "stereo" mode for example and then it will be processing just stereo. Note that it is the same thing to use a plugin, who processes stereo and mono separately with different settings, and to use 2 instances, one which processes just mono content followed be one processing just stereo (or conversely).

[Cooker]

What do you mean by flat? With a sine you should have just one line with some sidelobes and since MAutoEq uses smoothing (you can change this) it becomes a hill.

I'll check the stereo version. Anyway the analyzer merges both channels. If you need just to equalize stereo or mono content separately, you can just put it into "stereo" mode for example and then it will be processing just stereo. Note that it is the same thing to use a plugin, who processes stereo and mono separately with different settings, and to use 2 instances, one which processes just mono content followed be one processing just stereo (or conversely).

Sorry, I mean sine sweep (and I can't share the sweep since its from another product I own, but I can pm link if you want?). By flat I mean like pink noise when averaged it looks close to a straight line with 3 db slope across whole freq. range. but mautoeq averages something like U shape with sine sweep.

And thanks for the tip, I simply assumed left-right would also work if an ms plug-in was inserted before it. I also noticed it has mono+stereo setting so can I assume this setting would be healthier if inserted on (single track) stereo source than left-right?

Edit: Ok I found another sine sweep that doesn't look flat with mautoeq. Search as "MDA plug-in" on google and it has something called TestTone in the free bundle.

Choose LOG. Swp. as mode and adjust freq. range.

[MeldaProduction]

Sorry, I mean sine sweep (and I can't share the sweep since its from another product I own, but I can pm link if you want?). By flat I mean like pink noise when averaged it looks close to a straight line with 3 db slope across whole freq. range. but mautoeq averages something like U shape with sine sweep.

And thanks for the tip, I simply assumed left-right would also work if an ms plug-in was inserted before it. I also noticed it has mono+stereo setting so can I assume this setting would be healthier if inserted on (single track) stereo source than left-right?

Edit: Ok I found another sine sweep that doesn't look flat with mautoeq. Search as "MDA plug-in" on google and it has something called TestTone in the free bundle.

Choose LOG. Swp. as mode and adjust freq. range.

Ok, I checked and you're right. Though I tried Voxengo Span to have some competition and it created an opposite hill in maximize mode . I also tried pink noise and that one is flat, which is good. Maybe there is some bad handling of some fft bins or more likely it is caused by the log scale conversion, I'll check. Anyway right now it should work fine with real-world signals.

About the mono stereo stuff - I'm not sure if I understand. Anyway left-right mode is always the healthiest one since there is no encoding/decoding involved. With eq the results of LR and MS mode should be approximately the same, but MS also performs some encoding/decoding, so there may be some round-off noise added. Though this is irrelevant in most cases, the point is that if you are processing stereo track, there is no reason to change the mode.

[MeldaProduction]

So the cause seems to be that the analysis averages the data in nonlogarithmic (not dB) mode and since the resolution in the lowest frequencies is poor (using 8192 FFT right now) and the sweep stays there longer, it contributes to the long-term average more as the LF sines are projected into separate bins.

Fortunately in real world this is not so important, since most signals have natively more random character, which compensates for this the same way it does when analysing pink noise. Also since both comparisons and targets are affected by the analysis the same way, it will have no effect on the automatic equalization.

Though I'll keep digging, just in case...

[Cooker]

First, thanks for sharing the info.

The idea last night that MS approach came a little more logical to me was more from a mastering perspective but I guess best thing to do is to make listening tests (I didn't make any yet specifically about this).

And yeah I didn't worry a lot when pink noise was flat and thought maybe it was your preference as analyzing approach, also if a signal like sine-sweep is wanted as reference you always give the option to draw it. But I'd be more comfortable if you looked into it deeper, your product seems unique at how it works but I "feel" it also handles matching a little different compared to others (but I can be totally wrong too) and I thought this has something to do with the analyzer (as this was the reason I felt for a need to test).

Last questions if you don't mind, some prefer to analyze short part(s) of song like chorus while I prefer the whole song (but this also has to do with the type of song, like non-pop type songs). When analyzing long time material, is there a change needed/preferred for the analyzer section of mautoeq? also will oversampling option of eq will be added later?

[MeldaProduction]

My pleasure. And don't worry about the sweep, it is just a result of insufficient FFT precision, which fortunately should have minimal effect on realworlds signals.

About the long analysis - personally I usually analyze only short sections, but that's because I don't have enough time . Usually it is fine to jump through a few sections of the songs. But if you want to analyze it entirely, then go for it, not special settings needed. I think it won't help it so much, but at least it will give you more reliable analysis.
About oversampling - it is actually not there just because there is no place for it in the GUI . But in eq there is only a single reason for using upsampling - in 44kHz peak filters go to 0dB at 22kHz, but it no important effect on audio quality. Though I'll add it to to-do list, just in case somebody wants it.

[Cooker]

Thank you!