EQing with the Convolution method gave me bad results, but I don't understand why...

[Marcus Aseth]

This is very strange... I've followed this video-tutorial

which I remember worked for me in the past, but this time, it produced strange results.
There are two issues:
1) First issue is that during the Trace Arithmetic step, during the inversion (1/A) operation I set the Regularization to 8% (max 5db) as the video shows, but when I then go and apply the resulting Convolution into Equalizer APO it require me a preamp correction of -10db. That is strange right?
2) The second and most troubling factor is that, without any Convolution, when I pause the audio, the audio just stops. With the resulting Convolution though, when I pause the audio, I get a "metallic" trailing sound - something that last only few milliseconds, but you can definitely hear it. Not only that, if I do a measurement in REW with a relatively fast sweep, Length of 256k, during the sweep on the high frequencies you can clearly hear a metallic sound, I have a pair of Kali LP6 v-2 and I know they are not supposed to sound like that...
You guys have any idea what I might have done wrong, or if maybe some update to REW are at the root of these problems?

 

[teyhyrh4r]

I am playing around with inversion for weeks now and it's very complicated. the main reason is the phase. you can't simple invert the phase, even with regularization.
I will post my findings soon in a nother forum, could post here to if there is intrest.
So far I can tell you, if you are not interested in phase correction you can create the minimum phase version of your meassurements.
if you invert those you wont touch the phase. you can zero the left window of the filter created in this manner and you wont have any delay.

 

[Marcus Aseth]

So far I can tell you, if you are not interested in phase correction you can create the minimum phase version of your meassurements.
if you invert those you wont touch the phase. you can zero the left window of the filter created in this manner and you wont have any delay.

Unfortunately I'm only proficient at following tutorials step by step without messing it up, but I really don't understand any of this
Anyway did I even needed the convolution method to begin with?! I've ended up EQing with REW the traditional way, letting it generate the EQ filters

 

[teyhyrh4r]

Unfortunately I'm only proficient at following tutorials step by step without messing it up, but I really don't understand any of this
Anyway did I even needed the convolution method to begin with?! I've ended up EQing with REW the traditional way, letting it generate the EQ filters

naaah, you dont need it. If you are at the beginning of this journey you really shouldn't mess with this. this is experimental territory, let me phrase it like this

 

[ddude003]

Although you may not _need_ it, the math theory of convolution goes back quite a ways... Experimental, not so much... IR and FIR creation tools and convolvers have been around for quite some time... Maybe you could think of Serkan's method(s) as a DYI of things that have been lurking in various tools that are laid bare with his method(s) revealed in his videos and instructions... And hardware vector processing units have become consumer commodity and affordable over the past 20 years and embedded in intel and other cpus... With these hardware extensions in most every computer today it has become pretty easy to perform the math on most laptops and desk side computers... And today, if you have a Nvidia GFX card you have hundreds, if not thousands, of processing units to perform this elegant math on...

Or... Just get yourself a Pultec or GML 8200 style equalizer plugin...

 

[teyhyrh4r]

Although you may not _need_ it, the math theory of convolution goes back quite a ways... Experimental, not so much... IR and FIR creation tools and convolvers have been around for quite some time... Maybe you could think of Serkan's method(s) as a DYI of things that have been lurking in various tools that are laid bare with his method(s) revealed in his videos and instructions... And hardware vector processing units have become consumer commodity and affordable over the past 20 years and embedded in intel and other cpus... With these hardware extensions in most every computer today it has become pretty easy to perform the math on most laptops and desk side computers... And today, if you have a Nvidia GFX card you have hundreds, if not thousands, of processing units to perform this elegant math on...

Or... Just get yourself a Pultec or GML 8200 style equalizer plugin...

Experimental I call the aprouch to create FIR filters in REW, since I am not aware of this beeing a standard aprouch.
I myself have studied Denis Sbragion's documentation https://drc-fir.sourceforge.net/doc/drc.html#sec36 for 12 years now and still have a hard time doing it in REW. it's a lot of trail and error

 

[teyhyrh4r]

I must say though that I never heard of this Serkan guy. It's the guy from the video, yes? I will have a look

 

[teyhyrh4r]

ok, now watched the video. he is not creating FIR filters at all as he is simply minimum phasing his filters at the end. there is no convolution needed for this. the resulting filters can be imported as EQ in EQ-APO for example.

 

[ddude003]

Having worked professionally on signal and image processing since the early 80's I really don't know how to respond to that... Other than you may be looking at some other videos... And BTW, I have never heard of you...

 

[teyhyrh4r]

Having worked professionally on signal and image processing since the early 80's I really don't know how to respond to that... Other than you may be looking at some other videos... And BTW, I have never heard of you...

it's all good. let's work

 

[ddude003]

So lets answer the questions posed by the OP then...
1) The preamp correction of -10db is not strange... It means that Equalizer APO is applying -10db to correct an issue with gain staging being too high and that would/is causing clipping... Somewhere along the way of doing the room sampling and making correction filters and then convolving them creates an output signal of approximately +10db... Maybe your initial room sampling is done at above 75db? Or another step in your process has created the additional +10db... At any rate, Equalizer APO believes its too hot and clips, therefore the additional correction of -10db... I am not a user of Equalizer APO so this is what I think is most likely happening...

2) This "metallic trailing" sound issue could be caused by a filter that is longer than need be and/or post ringing, noise or distortion... Not knowing the exact steps or possible missteps in your measurement and filter creations process is is difficult for me to diagnose the exact cause...

 

[AudiocRaver]

Experimental I call the aprouch to create FIR filters in REW, since I am not aware of this beeing a standard aprouch.
I myself have studied Denis Sbragion's documentation https://drc-fir.sourceforge.net/doc/drc.html#sec36 for 12 years now and still have a hard time doing it in REW. it's a lot of trail and error

Hey teyhyrh4r, I first started messing with Denis Sbragion's DRC back in 2012, EQing my son's system. Wow, that takes me back. Fun, educational, and overwhelming.

Like ddude003 says, the science has been around for a long time. But for someone new to that area of audio like I was, working with a command line tool like Denis's DRC (there are GUI "assistant" shells now), it sure can seem experimental. Give me tools like Dirac Live and REW any day.

 

[teyhyrh4r]

Hey teyhyrh4r, I first started messing with Denis Sbragion's DRC back in 2012, EQing my son's system. Wow, that takes me back. Fun, educational, and overwhelming.

Like ddude003 says, the science has been around for a long time. But for someone new to that area of audio like I was, working with a command line tool like Denis's DRC (there are GUI "assistant" shells now), it sure can seem experimental. Give me tools like Dirac Live and REW any day.

I am talking about correcting, amplitude and phase. I am not aware of a non-experimental aprouch of doing it in REW, and Dirac gives you very little control. Acourate software would be the state of the art for this