Graphic Equalizer peak enhancement issue

jro45 · Post by **jro45** » Fri Mar 21, 2008 6:34 pm

I am another new user of GW, and I have been doing some experimentation with it just to learn the editing and effects tools. One issue has cropped up that was partially discussed in a 2005 response on this Forum. What I have found is that using the graphic EQ causes a substantial increase of some peaks even when using modest (+/- 2 db) gain settings with the master fader set at 0 db. Some peaks are typically increased by 6-8 db, but the average level changes nowhere near as much. If a track was originally recorded with peaks reaching 0 to -2 db or so, after EQ and maximizing a track to 0 db, the average level becomes substantially reduced compared to the original.

I was curious about this, so I tried using the graphic EQ with all of the sliders set at 0 db, and the master fader also set to 0 db. Some peaks in a CD track from Queen increased by 6.5 - 7.5 db, with the average increasing by only about 0.5 db.

I tried the same 0 db slider experiment on white noise. Originally the white noise has a fairly smooth time-amplitude spectrum, but after using the graphic EQ, substantial peak "hash" appears. In this case, peak levels go up by 8 - 8.5 db, and the average level increases by only 0.3 db. Give it a try.

This effect does not happen with the parametric EQ tried out with 5 center frequencies all set to 0 db. In fact, the peak an average levels before and after processing are identical.

Of course, this could be largely a phase shift issue with the graphic EQ, but I do know that analog equalizers can be made that show minimal or no phase shift when a frequency band gain/cut is set to 0 db (I've built some). Is this an inherent issue with the FFT, or something else?

I am using a Dell XPS 410, running GW v. 5.23 under Vista (GW works beautifully, by the way).

jro45 · Post by **jro45** » Sat Mar 29, 2008 12:53 pm

I had some time this past week to investigate the Graphic EQ function some more. Using a linearly swept sine wave, a number of peaks show up in the spectrum with all of the EQ sliders set at 0 db. The larger ones show up at 72 Hz (+2.9 db), 560 Hz (+2.1 db) and 5300 Hz (+2.7 db), with smaller peaks in between. These peaks cannot be minimized using the EQ sliders without affecting other parts of the spectrum.

The Graphic EQ fuction also shows a substantial phase shift of roughly -270 degrees/octave from 50 Hz to 10 kHz. This helps to explain the large enhancement of peaks in the waveform through the phase shifting of higher order harmonics.

Processing a track with all of the EQ settings at 0 db does result in a sound that is different, I suppose largely because of the gain peaks, but possibly also because of the huge phase shift and increased peak/average values campared to the unprocessed track.

These unintended modifications of a file by the Graphic EQ have led me to discontinue using it, and instead use only the Parametric EQ which shows none of the odd behavior of the Graphic EQ.

Beyond this issue, I find Gold Wave to be just a delight to use!

JackH · Post by **JackH** » Sat Mar 29, 2008 8:13 pm

Interesting discovery you've made. This might explain why I often get a pop/click when I use the graphic EQ on only part of a file. The noise happens at the beginning and end of the area I've selected.

Kummel · Post by **Kummel** » Thu Apr 03, 2008 5:35 pm

There is also a phase annoyance for the low/high pass and bandstop/bandpass filters.

jro45 · Post by **jro45** » Sat Apr 05, 2008 2:59 pm

Yes, there is some phase shift evident with the low/high and bandpass filters that increases with the number of second order filters chosen. However, it isn't all that bad even with 6 cascaded filters (12 poles). In the analog world, such high order filters would cause some very nasty phase shifts.

Both the Parametric EQ and the Spectrum Filter show little or no phase shift, even with non-zero gain/cut settings. These effects are practically ideal from that standpoint. They also process much faster than the Graphic EQ. A very minor drawback of the Spectrum Filter is a slight amplitude ripple with frequency above 12 kHz with the filter set for a flat response. Fortunately, this occurs in the upper end of the spectrum where it won't be noticed anyway. I particularly like the Spectrum Filter because it allows one to create very steep filters with no significant phase shift.

What would be nice is if the same algorithm used for the Spectrum Filter or the Parametric EQ were adapted to the Graphic EQ. The latter is a lot faster to adjust, but its major drawbacks keep me away from it for now.

Post by **GoldWave Inc.** » Thu Apr 10, 2008 12:57 am

The Equalizer filter uses a combination of lowpass, bandpass, and highpass filters. In places where the bands meet, there is some inconsistency in the gain. The Spectrum Filter algorithm may replace that combination of filters in a future version.

Chris

jro45 · Post by **jro45** » Thu Apr 10, 2008 5:32 pm

Thanks for the information, Chris. I'll be looking forward to the new algorithm for the Equalizer!

DewDude420 · Post by **DewDude420** » Thu Apr 10, 2008 7:25 pm

yeah, i forgot to write that if you take a sweep, run it unmodified through EQ, then take modified sweep and make a stereo wav with the unmodified as the other channel...it creates a really cool spirograph effect in the phase meter.....if you run it through adobe's auto-phase corrector..well..it's not pretty...but it's interesting....

yeah, i remember writing a reply and then not sending it and rushing out forgetting to shut laptop down, at some point during my commute it ran out of juice..oops.

I will add that sometimes, this "distortion" can have a more pleasing effect to one's ears. Chris, I say if you modify the way equalizer works...maybe leave an option to select old and new methods? Sometimes i can get just the right amount of bass because of the phase shifting and other things.