[ADUHF]

The above image shows the combined result of 5 different sound power curves added together. To get the average response, you would just scale the amplitude in the curve at the bottom down by 1/5th. Or divide all of its dB values by 5.

The response curves of different headphones, or different EQ curves could also be combined and then averaged in the same way.

The analysis panel in Equalizer APO's Configuration Editor is limited to +/-100 dBs though. Which potentially limits the number of curves that can be combined at one time in a single project. There may be some other workarounds though for combining more curves together.

 

[bigshot]

I spoke with a headphone designer and he quoted the 3 to 5 dB range to me. He was working on high end headphones and his goal was a sample variation of no more than 1 dB.

An average of a bunch of different headphones plus an average of people's preferences equals chaos. Better to lock down at least one parameter and calibrate to the particular set of headphones you happen to own.

What the hell makes you think you'll come anywhere within shouting distance of +/-100dB. That is absurd.

 

[ADUHF]

I spoke with a headphone designer and he quoted the 3 to 5 dB range to me. He was working on high end headphones and his goal was a sample variation of no more than 1 dB.

An average of a bunch of different headphones plus an average of people's preferences equals chaos. Better to lock down at least one parameter and calibrate to the particular set of headphones you happen to own.

What the hell makes you think you'll come anywhere within shouting distance of +/-100dB. That is absurd.

I probably should have started a separate topic for this. Since markanini has already been down a similar path with his approach though, I thought he might have a few insights to offer on this.

I think you may have missed the point though on some of the above, bigshot (probably because I haven't explained it that well). So I'll try to post another example or two of what I might use this for, when I can. Perhaps in a separate thread.

If you're combining multiple response curves together that have a somewhat similar shape though, like the above sound power curves, or the response curves of several different headphones with similarly-shaped FRs, like on this diffuse field graph...



...then the values can begin to add up fairly quickly in certain spots. And with enough curves being combined, you could easily exceed a range of 100 dB, or more in the final combined response curve.

The sound power curves of well-extended loudspeakers will have about a +10 to +15 dB rise from the treble to the bass, for example. And if you combine ten or more of those curves together, then you could easily get close to or even exceed a range of 100 dB in the final resulting combined response curve (as shown in my previous sound power example above).

I've already figured out a possible work-around for this though. Which involves simply moving the decimal point one digit to the left for all of the plotted points in each response curve in the stack. This would effectively scale all the response curves in the stack down to 1/10 of their original amplitude. So you could fit 10 times as many curves into a single "stack". And not have to worry about the final result exceeding the +/-100 dB limit in the Analysis Panel.

If you're only combining ten curves together, then this would also save you the trouble of dividing or scaling the final resulting curve by the number of curves in the stack. Which is pretty cool!

I know this is probably getting rather confusing though, so maybe I'll start a different discussion for this where I can post some more examples, and perhaps go into a few more details.

 

[ADUHF]

There is one other thing that the stacking method allows, btw, that cannot be as easily accomplished with a spreadsheet, or some other simple mathematical method. And that is that it allows different response curves to be combined together which do not have the same quantity or configuration of points in their plots.

Why is that important, you might ask?

Well, let's say you want to combine the response curves of several different headphones together, like the ones in the graph in my last post, for example. But the plots you've created for each headphone have different quantities or configurations of points. (Which is something you can easily do in Equalizer APO's configuration Editor, using it's variable graphic EQ option.) The stacking method above still allows the curves to be added or combined together in the Analysis Panel, regardless of how many points each curve has, or what the exact configuration of the frequency bands is in each curve.

This is a feature that I take advantage of all the time, when creating EQ curves for my headphones, either from scratch, or using various reference curves (including the response curves of my own headphones) in Equalizer APO's Configuration Editor...

Some examples of different "variable" graphic EQs in Equalizer APO's Configuration Editor from my previous EQ projects:

But I hadn't really thought of using an approach like this to compute an average response curve from multiple response curves before, which is a bit different. It could be a potential time-saver though, because it means that each curve in the stack can have as many, or as few points as necessary to achieve an accurate plot for that particular speaker, or headphone, etc... Which could also improve the accuracy and precision of the final result.

All of the sound power curves in my earlier example above have the exact same quantity, and configuration of points or frequency bands in their curves. And the only things that differ are the amplitudes or decibel levels of the points (which are displayed in the columns on the righthand side of each plot)...

The curves had to be plotted this way because I was originally using more of a spreadsheet-like approach to combine them together mathematically, with individual columns for each frequency band.

The stacking approach shown above does not require this though. So each response curve in the stack can have as many or as few points as needed for a precise plot. This potentially makes it much easier to combine the plots of different headphones (or loudspeakers, or EQ curves) together in a single project or graph. Even if the plots have different point configurations.

 

[bigshot]

When do you start considering how it actually sounds? I get the feeling you’re enjoying thinking about equalization mor than actually applying it to music.

 

[ADUHF]

When do you start considering how it actually sounds? I get the feeling you’re enjoying thinking about equalization mor than actually applying it to music.

This is a bit off topic. But I listen to music on my EQ-ed HPs almost every day. And sometimes post examples in various threads in the Head-Fi music forum. Particularly some of these...

https://www.head-fi.org/threads/pop-folk-chalga-other-balkan-music.887510/
https://www.head-fi.org/threads/latin-music-spanish-romanian-italian-mexican-brazilian-etc.867425/
https://www.head-fi.org/threads/r-b...wn-with-the-get-down-maybe-fonky-even.853482/
https://www.head-fi.org/threads/k-pop-j-pop-suggestions.915923/
https://www.head-fi.org/threads/popular-music-of-turkey-the-caucasus-central-asia.883699/
https://www.head-fi.org/threads/official-classic-rock-thread.901225/
https://www.head-fi.org/threads/disco-and-nu-disco.724632/

Balkan pop-folk (aka chalga, manele, tallava, laika, etc.) is one of my favorite genres at the moment. Though I also enjoy alot of Latin, R&B/hip-hop, Asian, classic/glam/alt-rock, and disco music. And other stuff as well, including some jazz, and classical, and country music. My background is mostly in the visual arts though. Including some film, and video. And especially animation, using both computers and also more traditional hand-drawn or cel-oriented approaches. So I have quite a bit of experience with graphing, image processing, and "point-pushing" from those disciplines. And am enjoying applying some of those skills to the art or subject of sound reproduction as well.

I've also dabbled a bit in music performance and composition. But that was mostly in days of yore, using stuff that looked something like this...

And this...

 

[bigshot]

How many different EQ curves do you use when you listen to music?

 

[ADUHF]

How many different EQ curves do you use when you listen to music?

When I'm listening for pleasure or entertainment, I normally use only one EQ solution or project at a time for a particular headphone. Or sometimes two, if I want to compare the difference between two different EQ solutions on the headphones.

 

[bigshot]

Then all this switching around with different EQ curves and adding, dividing and averaging different curves is just theoretical interest... You don't intend to actually listen to music that way? That makes sense because your mix and match approach to EQ doesn't seem to have any specific goal in mind. I guess you can train yourself to identify frequency bands by ear by just randomly diddling around.

 

[ADUHF]

Then all this switching around with different EQ curves and adding, dividing and averaging different curves is just theoretical interest... You don't intend to actually listen to music that way? That makes sense because your mix and match approach to EQ doesn't seem to have any specific goal in mind. I guess you can train yourself to identify frequency bands by ear by just randomly diddling around.

Although it probably doesn't seem like it at this point, there is a little bit of a method to my madness. I learn new things, and gain new insights from each new step in this process. And then try to apply what I've learned to my own headphones, and audio hardware and software. So there is a practical side to it as well. In addition to a more general quest for more knowledge.

I am in no major rush to find some of the answers I'm seeking though. And am enjoying exploring many of the different and sometimes unusual avenues that this subject has been taking me down. And I'm also ok with not knowing exactly where I'm going on some of this, or what the next steps will always be. If you want to characterize that as "diddling around", then so be it. LOL. Because it may not be all that far from the truth in some cases.

I need some easier and less time-consuming methods of computing the average responses from a variety of curves though. Including (ideally) curves with different numbers and configurations of points or bands. Because I don't particularly enjoy all of the number-crunching that's generally associated with that. A stack-based approach, like the one described above, might suffice for the time being though.

 

[bigshot]

When you decide to focus on your own specific target curve the suits both your ears and your particular equipment, you'll find that it is a long process to refine the curve to where you are satisfied with it. But once you arrive at that curve, you just implement it and go on to listening to music.