EQ'ing to Harman curve doesn't give me pleasing results - why not?

[neRok]

So I measured my speakers (Kali IN-8 V2) in room, with UMIK-1 located roughly in the middle of my head position, and with it orientated vertically and with 90° cal file applied. In REW I then applied a HPF to Harman room curve and EQ'd to that line, and the resulting measurement got much better;


I then listened to it like this for a couple of hours. Initially it seemed good because there are no super-obvious loud spots in the response (and I hadn't listened to much music for a few days), but after a while I had the feeling it wasn't so great. When I came back the next day to listen to it with fresh ears, it immediately sounded off. Using EQ APO I toggled the EQ on and off, and with the EQ on it sounded/felt like a void in the music that was actually "sucking in" the music. Not all of the spectrum, but definitely in the important midrange. It is very strange. I wondered if it is phase issue, but nothing major looks different in it. I now wonder if it is simply too much EQ (dropped the levels too much)? But if I'm EQ'ing to the room curve, how could this be?

So the next thing I did is leave that EQ off, but I copied over and applied the 50Hz -4.3dB PEQ, and the 160Hz PEQ except I reduced it from -13dB to -10dB. I listened like this for a little while, and whilst there were some "loud spots" that still needed taming, the "void" was gone. I then copied the 288 Hz PEQ but also reduced it from -6.8Db to -3.8dB. I have listened it this way for a few hours now, and it sounds "better" - it being "less" EQ (not so strict to the room curve). I am now wondering why this is the case? More so, what is the real reason behind my experience?

I have some thoughts on this already. But my thoughts are impacted by an experience I once had and that is related to my ears, as my ears stick out a little (are more forward facing the usual). Many years ago I had a pair of cheapish Beyerdynamic open headphones, the DT440's. I really liked the sound of these (for what they were to me at the time), and so when they needed replacing I opted for the apparently better quality and closed model DT770's, because I was going to be using them on the train a lot. But I hated the sound of the 770's. Something was off about them, and it was like the headphone was inside a sea-shell strapped to my head. It was a very peculiar experience considering their reputation on the internet. In testing I realised that if I pinched the tip of my ear beneath the pad, thus flattening my ear, they sounded better. So I came to the conclusion that the geometry of my ears is affecting either the sound output of the headphone driver, or it's to do with the way my ears receive the sound. I gave away the 770's (the recipient liked them) and got some 880's, and they sounded good and proper, like I expected compared to the 440's.

So I'm wondering if my ears are impacting what I am hearing in my room in an unusual manor? Perhaps my hearing is more sensitive to sounds coming from in front of me, rather than reflections from behind me? The mic however is picking up all sounds, and I presume REW is treating them all "equally" (to suit a normal listener). But if I'm more sensitive to forward sounds, then this might be why I'm apparently over EQ'ing and creating a "void"? So now I wonder if I need to somewhat time-gate the reflections in REW, or if I can reduce the way late reflections are added in to the mix vs early reflections and direct sound? And then I wonder if perhaps I am more sensitive to forward sounds in the higher or mid frequencies vs the lower, and so any adjustments to REW measurements need to be on a slope?

Another thought I had was that maybe I just need to measure at various points around my heads position, and then average them, and EQ against that average? I've seen a few mentions of this method on the forum but haven't look in to it thoroughly, but perhaps it is more proper?

What do you think?

 

[Rednaxela]

Which speakers?

 

[neRok]

Kali IN-8 V2 with none of their toggles applied. I don't think it's some hearing preference related to the coax driver because I actually really like the point source sound and stereo imaging these give vs other speakers I've tried nearfield (like my old B&W bookshelfs).

 

[ZolaIII]

How about that you show us waterfall plots?

 

[amirm]

So the next thing I did is leave that EQ off, but I copied over and applied the 50Hz -4.3dB PEQ, and the 160Hz PEQ except I reduced it from -13dB to -10dB. I listened like this for a little while, and whilst there were some "loud spots" that still needed taming, the "void" was gone. I then copied the 288 Hz PEQ but also reduced it from -6.8Db to -3.8dB. I have listened it this way for a few hours now, and it sounds "better" - it being "less" EQ (not so strict to the room curve). I am now wondering why this is the case? More so, what is the real reason behind my experience?

This is a perfectly valid approach and many do this (only EQ up to a certain frequency). FYI there is no "harman curve" for speakers in a room in the way you are imagining. There is a target curve but that doesn't mandate specific filters for higher frequencies. Measurements do not represent what we hear above a few hundred hertz so care needs to be taken in full range equalization.

 

[neRok]

Edit: This is the EQ that REW generated to the Harman target curve. I I think I set its auto-eq range up to 10k Hz.


Edit2: The waterfall without EQ;

 

[Rednaxela]

Have a look at every post by @Floyd Toole that includes the word target.

Like so:

It will most probably answer your main question.

 

[Rednaxela]

FWIW, I do EQ above 500Hz, but not from a single point measurement but a MMM around the MLP.

What I’ve learned is that you need to really study the anechoic response of your speakers too and see how it correlates to the in-room measurement. You might see the same anomaly in both and want to gently correct it.

In the 1-5kHz area your ears are super sensitive, so be vary careful what you do there. The correction in your first post may have simply been too brutal.


But first try an MMM.

 

[moonlight rainbow dream]

AFAIK it is not a "target" curve. They took the JBL M2 (measures flat anechoic), stuck it in a bunch of different rooms, measured the in-room response and came up with an average. That is all. It is not some kind of magical formula to make your setup sound the best.

 

[ZolaIII]

View attachment 294518View attachment 294517View attachment 294516

Edit: This is the EQ that REW generated to the Harman target curve. I I think I set its auto-eq range up to 10k Hz.
View attachment 294520

Edit2: The waterfall without EQ;
View attachment 294521

Let's try with R60 decay times. It seems to me that somehow PEQ makes negative undesired effect in 100 to 200 Hz region.

 

[amirm]

Edit: This is the EQ that REW generated to the Harman target curve. I I think I set its auto-eq range up to 10k Hz.

I used the auto generated filters by REW and that was enough to not try again. As noted, you may want to make a correction in higher frequencies if the anechoic measurements of your speakers show an issue. Otherwise, fix the modal (low frequency region) and use the target curve to get the balance of bass vs treble to your liking.

 

[neRok]

Let's try with R60 decay times. It seems to me that somehow PEQ makes negative undesired effect in 100 to 200 Hz region.

R60 as in the RT60 screens? I'm not sure how to use them or what they mean, but I did something;



On the measurements without EQ, there is a broad raised flat area on the RT60 decay model (if that's what it's called) in that 100 to 200 Hz range that you suggested. And there is a spike on the RT60 page. After EQ that flat area is gone, but that spike is higher.

Have a look at every post by @Floyd Toole that includes the word target.

Ok I will. I'm probably at the stage now where I should read his book too, because it seems just trying things without a proper understanding is not getting me anywhere quickly.

 

[GXAlan]

Measurements do not represent what we hear above a few hundred hertz so care needs to be taken in full range equalization.

Is this true with averaging?

 

[GXAlan]

Have a look at every post by @Floyd Toole that includes the word target.

Post in thread 'What is your favorite house curve'
https://audiosciencereview.com/foru...s-your-favorite-house-curve.2382/post-1476897

Here’s a great one.

What we may need is an index of Floyd Toole’s comments since there are buried nuggets in ASR which “simplify and summarize” key points of his book.

 

[ZolaIII]

@neRok well it seems we indicated problem. Now the question is how to deal with it (RT60 spike with PEQ). What's very interesting it's that it doesn't show in RT60 decay times plot. Generally decey times are better with PEQ in place but with that nasty spike it actually sounds worse.
Hire are mine to mixed target (no +2 dB low bass boosted and - 8 dB & 20 KHz).

Idea is to get them even as much as you can.
How is the room and speakers placement done if you don't mind asking? So that we start from basics and try to get rid of it.
Edit: sorry for the long response time, still drinking first morning coffee.

 

[boxerfan88]

Below setting is what I used to generate the filters for my setup which turned out very nice.

Try this in the REW AutoEQ setting:

• Target Settings
  • Room Curve
    • LF Rise Start = 200Hz
    • LF Rise End = 100Hz
    • LF Rise Slope = 3dB
  • Target Level = set appropriately
• Filter Task
  • Match Range = 20Hz to (300Hz or 400Hz) { try not to EQ anything above 400Hz }
  • Flatness Target = 3dB
  • Allow narrow filters below 200Hz

After REW generates the PEQ filters, tweak the filters from there to your liking.

^ use the waterfall chart to check that the room mode ringing is well controlled.

Hope the above suggestion helps...

 

[Frank2]

I've played a lot with equalization and in my experience, the larger bandwidth overall shape of the curve is just as important as the smaller bandwidth 'peaks and valleys'. Your original curve in the OP has a 2-3 dB large bandwidth 'plateau' from 100 Hz to 1 kHz. This will give a warm midrange. That plateau is gone in the corrected curve and this may be audible as a midrange deficiancy to your ears and to your personal taste.

What I do is correct the small bandwidth peaks and valleys first (room mode valleys are not really correctable) to get a smoother curve, and then change the overall curve with high bandwidth or shelving filters to my personal liking. For these large bandwidths, a 1 dB correction can already be very audible in changing the overall character of the sound.

 

[Gruesome]

with it orientated vertically and with 90° cal file applied

With the microphone pointed towards the midpoint of the speakers maybe you'll get more direct sound contribution? (Although the Umik-1 is omnidirectional.)

 

[neRok]

Try this in the REW AutoEQ setting:

I was checking out your settings and found a mistake in my config, but not a big one. I had Harman curve set in REW settings, and also "add room curve" ticked on REW EQ screen. So the slopes of the LF and HF were roughly "doubled", but the resulting target was the same from 200Hz to 1kHz. At 100Hz the difference was <1dB, including comparing to your settings. The major difference with yours is the "flat" (unadjusted) treble, and that your "bass boost" stops at 100Hz, whereas the Harman target continues to gain the lower it goes.

With the microphone pointed towards the midpoint of the speakers maybe you'll get more direct sound contribution? (Although the Umik-1 is omnidirectional.)

I'm looking back through some old measurements at the moment because I think I did test horizontal vs vertical mic and found the difference in resulting measurements was minimal (that's the conclusion I remember). I haven't got to those measurements yet though because I have gotten distracted by other data I have found.

How is the room and speakers placement done if you don't mind asking? So that we start from basics and try to get rid of it.

Your RT60 results look much different to mine! My room is roughly 3m wide by 4.2 long (with 2.4m ceilings), but it isn't symmetrical.
Here's how it sits now (I only made up those speaker stands the other day, hence I've been attempting to EQ again);

Here's my room layout;

Here's some makeshift room treatment I've got going on that seems to have really improved the decay times above 1kHz (more info in the quoted thread, but it's just some boxes with a blanket over the top that dropped the decay times). That's interior door 2 in the pic.

Like I said, I've just been going over some old measurements. This time last year I had B&W speakers sitting on my desk, and the desk was rotated 45° counter-clockwise in to the corner (the top-left corner on room plan). I happened to be testing some things out with the room and the mic. Here's a test I did comparing the speaker response only (no sub, no EQ), and comparing the difference of having those interior doors open vs closed. The response looks basically the same;

On the waterfall and spectogram tabs it is evident that the decay time rises slightly when having the doors closed, which I guess is expected. But looking at the RT60 tab just now, and with the doors open that 200Hz is already there and is quite large, but with the doors closed it gets even higher! Surely this means something, but what?

This suggests to me it is a room problem, not a speaker or speaker placement problem? (because both speakers and placement were different on this old tests). And there are a few things that resonate in the room that could be the culprit. Firstly the ceiling in the top-left corner does creak a bit at low frequencies (under 50Hz), whilst the window has a few resonances* above that, and maybe it is the problem at 200Hz? But wouldn't this show up as distortion? Because nothing stands out on the distortion graph for either of these old tests ("door closed" test shown).
(*The problem with the window is that it is a tall sliding window and has a locking mechanism in the frame that must be made out of thin rods that vibrate easily to the music.)

Other possible room problems are the cupboard door that I indicated, but I have recently taken to packing cardboard in the frame to stop it, and so it shouldn't be part of my recent EQ attempts. Possibly the "backboard" of my desk makes noise, because it does vibrate a bit when the music is very loud, but I'm not testing at "very loud", so I don't think so. Neither of the internal doors seem to be a problem, nor is the external door particularly. But there is a road outside that external door+window that has a bit of "slow" (local) traffic noise, which maybe gets in to the results? Again I would think that would appear as distortion? And considering vehicles are different sizes, weights, speeds, etc, I wouldn't think they would constantly appear as a problem at ~200Hz.

Edit: Just remembered one other thing I've wondered about, and that's the height of my speakers within my room. Even though the B&W were smaller (standard dome tweeter with 6.5" woofer), I had the tops of speakers at about the same height as the Kali are now (also when I had the Kali on desk for a while, I had them at about this height). This speaker height is roughly level with my ears (the coax tweeter is roughly at my eyesight level), and whilst I'm not super tall, that's the height my head is at when sitting on my PC chair. And this does put the woofer roughly close to the mid-height of the room (the middle of the woofer 1125mm above the floor). I have wondered if this could be a problem? The lowest "axial height" bar in REW room sim is at 71Hz, and so multiples are at 142Hz and 213Hz.

 

[alex-z]

The "Harman Curve" aka JBL Synthesis target is not based upon a single response. There is broad variation in the bass and treble region, because of variables such as listening volume, hearing loss, and the reflection arrival times of your room. Also just outright personal preference.