Originally Posted by OJNeg
Hey luisdent, what's your reasoning behind boosting at ~7k? It seems to be that most plots show a peak around there (might be a ear canal resonance from what I understand). I'm sure you're familiar with the Linkwitz recommended EQ:
I've seen that page before (wanted to make a physical filter like he did), but I really don't know why his peaks are so large. The most common reason for peaks that large is improper insertion depth. You can see here from rin's post that the treble past 4khz starts to peak greatly with shallow depth and the er4s: http://rinchoi.blogspot.com/2010/05/how-deep-do-you-insert-you-er4.html (all er4 share a similar peak with depth issue)
Perhaps he has a shallower fit than some? Perhaps he is more sensitive to that region to begin with? I'm not sure. And most users will corroborate that the peaks usually go away with depth. This is why custom sleeves for the er4s so often don't sound as good. He also used a lot of his own reference material as a basis for his eq. He might prefer those areas cut more for other reasons...
With a deep fit giving you the best response, most graphing systems show there is a slight dip at the 5-8khz region.
In the headroom graph for the dip is closer to 5.5 khz
Rin's is closer to 8khz (grey line)
In the goldenears graph, the dip is closer to 7khz
My listening tests fit more closely to the goldenears graph. They also use the +6db bass rule, which is why they show a larger drop in the bass.
Anyhow, the 7khz is what sounded more natural. However, as gnarlsagan found, these points may not match everyone's perception. With the 2khz area for instance, he preferred to move the 2khz area up a bit to 2.4khz. I've gone back and forth. The curves I chose are a balance between covering the desired frequency range in an averaged way, and not overdoing anything.
In that example, there is a peak around 2.4khz indeed, but if you look at the graphs the "main" overall dip starts at 1khz and ends at 3khz. So, while 2.4khz might be the precise "peak", creating a single eq point based on that point will generate a boost from 1.4khz to 3.4khz for instance if the q width is reasonably accurate. Therefore, my 2khz peak covers the actual dipped region perfectly, but simply won't have the exact highest peak centered. However, this is where listening came into play. I found that centering the peak at 2.4khz personally caused to much emphasis with some things. Instead, using the 2khz overall dip center point provided a very close approximation of the overall problem area. And by not overdoing the cut, it's like I covered the fix 80% without accounting for the highest peak, which would still be reduced though. And I'd rather cover 80% without accounting for the small remaining peak than cover the peak, but have areas left unadjusted or adjust areas that shouldn't be adjusted... You could be more accurate and use multiple points to cover each part individually, but I chose to try and balance accuracy and simplicity. 5 to 6 eq points is reasonable. I originally started with 8 myself. I could probably easily make a more accurate eq with 12 or more. Or better yet, using an eq plugin in logic that lets you actually shape the eq better.
But my goal was to create an eq that anyone could use. This is a simple parametric eq. No shelving, no special filter shapes, just q and bandwidth values that nearly anyone could find an app to use with. Regarding precut, boosting and cutting don't make a difference, as we mentioned, as long as you match the precut to your greatest boost amount. Generating an eq of only cuts would require a lot more points and a lot more work and wouldn't make sense in this application, since there's no quality difference. And not every eq app is created equally. This is designed for apps that allow precise value input. For instance the denon app doesn't work well (although you can make a comparable eq with different values), whereas the accudio eq app's custom eq is excellent, allowing exact bandwidth and frequency values. There are mac audio units that are free and precise quality and windows apps that are free and precise as well.
So, back to the 7khz, feel free to move it around and see if any change works better for you. My choice was one of balance and "averaging" with a lean towards the ge and rin graphs. It's no secret around here that I like ge graphs. :-P But as you can see, with the bass rule taken into account, the ge and headroom and rin's graphs are actually all very similar.
So, while these peaks/dips might vary slightly depending how you want to account for them with a simple eq point, they are all subtle changes and designed to nudge the sound in the right direction. Hence the very small values. And an easy way to find which point you prefer is to try them for a bit with material you a familiar with. Shift the eq point up or down a bit and relisten for a while. When treble is more "flat" you should hear more naturalness and details should be easy to resolve as there is less masking. I find masking is the easier way to determine if something is relatively flat. Brightness can give the illusion of detail, but when you become familiar with how details relate to frequency response you can more easily identify when details are "actually" more resolvable. You can try tone testing, but that gets even more complicated, as you need to take into account target curves and have an easy to use tone system that not everyone has access to.
Anyhow, let me know if that didn't answer your question. :-)
Edited by luisdent - 3/19/14 at 12:22pm