Originally Posted by harrisu
Yep. That's more like it. I also realized after many iterations that flatter response below 5K sounds much more natural and has no boominess to it. Here is the curve I use.
and here is my Sub response
Not sure what having a point at 0 with -17 would give you. I just limit the curtain so there is no correction like following for Sub.
Please note that this is for minidsp and not monolith but I think Dirac should operate the same way.
I spent the past week continuing to refine target curves and your advice about flatter midrange was very helpful. I applied that logic to the curve I posted earlier and continued to get better results, until I settled on something for a while. I called this curve, RoomFeel Rolloff 10
. But as I listened to a wider variety of music, I couldn't help but feel like something was off on some genres like rock and metal. They had plenty of bass but they sounded awfully dull.
So I got curious and actually loaded your target curve exactly as-is. I called this curve, harrisu
. This had less bass, as expected, and I found the treble slightly bright on female vocals on my system, but the dullness of rock and metal was fixed compared to RoomFeel Rolloff 10, and the overall balance of instruments was significantly improved on tracks with female vocals and a lot of other instruments and chaos going on. The vocals sounded drowned out in RoomFeel Rolloff 10 compared to harrisu, where they were front and center as they should be. And being able to have an instant comparison, it was obvious how muffled rock and metal sounded on RoomFeel Rolloff 10 compared to harrisu. I must have killed too much treble on RoomFeel Rolloff 10.
So I abandoned RoomFell Rolloff 10 and started making refinements to the harrisu curve. For bass, with an 80 Hz crossover, which clearly looked like the best crossover to use for harrisu, I found it easiest to just run the bass trim hot to get the bass levels that I wanted -- that way I don't run into issues with Dirac running out of headroom to boost. I ended up running the subwoofer trim about 4 dB hot to get the bass to where I wanted it -- I achieved this by just using the gain knobs on the subwoofers themselves and measuring the results in REW, loading the target curve in REW as well to compare.
I had to spend a lot more time on treble. Knowing how dreadful rock could sound if I rolled off treble too much now, I tried adding heavy rolloffs to harrisu starting at around 8 kHz to get female vocals to sound natural and not overly sharp and ear piercing. Rolling off as late as 8 kHz, instead of something like 1 kHz, allowed me to still tame female voice harshness while not destroying the energy in rock music. Even at 8 kHz and beyond I had to be careful not to destroy the sounds of high frequency details coming from things like cymbals.
I finally got something that sounded pleasant for female vocals in music without completely dulling the sound of rock, and while the steep and heavy rolloff after 8kHz I settled on made me nervous, I was willing to stick with it if it really did sound best. So I loaded the updated curve called harrisu Rolloff 3
to my LCR speakers and spent some time listening to this for a few days.
And then my next issue became apparent: while music sounded fine across a variety of genres, voices in movies sounded veiled and muffled. I didn't realize how much high frequency content exists in human speech, even male voices. Things like the sound of an "S" or "Th" extend up to 8-10 kHz and harrisu Rolloff 3 practically removed the details in these sounds and made it sound like everyone was talking behind a heavy curtain. But of course, if the levels at 8-10 kHz are too high then voices sound unbearably sharp because these sounds are overemphasized.
So I spent another several iterations trying to find the right balance in voices, across a wide variety of content, and landed on a separate curve with a different treble rolloff... just for voices in the center channel called, harrisu Center 5
. The rolloff was much more shallow compared to the one I settled on for LR/music. Again, running a different curve for center/movies vs LR/music made me slightly uncomfortable but I was willing to do it if it really did sound best.
This combination did sound pretty good overall for a while. But I got curious and, after learning everything I did while experimenting with the center channel target curve, I listened to some music and found that it actually did have the same issues that voices in movies used to have, for certain songs -- details in "S" and "Th" sounds, among plenty of others, were significantly lowered and did make it difficult to make out what words the singer was saying sometimes. But simply running harrisu Center 5 on the LR speakers still sounded a tad too bright, so I took the most natural looking curve I could come up with that was in between my C and LR curves and came up with a new LCR curve, called harrisu LCR 9
. This curve is practically the same as the C curve up to 10 kHz, since I didn't want to ruin the balance I achieved for voices -- it just rolls off slightly steeper after 10 kHz. I loaded harrisu LCR 9 on my LCR speakers and found that I finally ended up with one curve that sounds right for both music and movies. It has been a long week!
Here are the curves I'm using now. As I mentioned, I'm running the same harrisu Sub curve but I'm simply running the sub trim hot to get more bass. We'll see if I don't start tweaking something again tomorrow! But I'm hoping to stick with this for a while so a have a baseline of sound that I'm familiar with before Dirac BMM drops in a few months or so, so I can write about how BMM compares to standard Dirac.
Originally Posted by harrisu + pink soda LCR 9
Originally Posted by harrisu Sub