• WANTED: Happy members who like to discuss audio and other topics related to our interest. Desire to learn and share knowledge of science required as is 20 years of participation in forums (not all true). Come here to have fun, be ready to be teased and not take online life too seriously. We now measure and review equipment for free! Click here for details.

Room EQ, do's and dont's

RayDunzl

Major Contributor
Central Scrutinizer
Joined
Mar 9, 2016
Messages
6,251
Likes
2,296
Location
Riverview, FL
#81

Krunok

Major Contributor
Joined
Mar 25, 2018
Messages
2,476
Likes
1,133
Location
Zg, Cro
#82
Time aligning the drivers (using linear phase XO) and linearizing the phase was audible.
Are you saying that if I manage to measure phase correctly at my LP and linearize it I would get nice step response and the result will be hearable?
 
Joined
Feb 5, 2019
Messages
96
Likes
55
#83
...
Doing it manually, fix something, remeasure, see what you get, try again, etc. You can fiddle with it for hours. ...
For me, doing it manually is half the fun, and educational. Yet it leads to the slippery slope of continuous improvement which quickly reaches a point where the manual approach just doesn't cut it anymore.

When first starting out, the flaws of room acoustics are huge and easily audible. One typically finds frequency response variations of 12 dB or more, strong ringing in CSD, etc. These are easy to fix manually with room treatment: rearrangement, tube traps, bass traps, acoustic foam, diffusers, etc. This improvement is truly an eye-opening, jaw-dropping :D jump up and down for joy, night vs. day thing (unlike many other audio improvements called "night and day" that range from nonexistent to splitting hairs).

Now one has a much better sounding system in a big obvious way. The cleaner sound reveals the next level of subtle imperfections that were masked by the bigger flaws he just fixed. The positive experience leads one to pursue it further, only to find that low hanging fruit has been picked. The next step will require more equipment (mics, digital EQ). Dive in and make that next improvement, it can still be done manually.

Rinse & repeat each iteration gets more finicky, at some point you must rely on software analysis and automation no matter how much you enjoyed doing it all manually. That's not necessarily a bad thing.
 

mitchco

Active Member
Joined
May 24, 2016
Messages
169
Likes
338
#84
Are you saying that if I manage to measure phase correctly at my LP and linearize it I would get nice step response and the result will be hearable?
No, what I am saying is that the individual divers in the speaker need to be time aligned first. This means an active system using a linear phase digital XO where the software, either manually or automatically, calculates the acoustic center of each driver and aligns them through the use of digital delay. Then, if required, linearizing the phase after... As mentioned in the previous post, linearizing the phase of a passive system, but not having the drivers time aligned, to my ears, has little audible effect.
 

andreasmaaan

Major Contributor
Patreon Donor
Joined
Jun 19, 2018
Messages
2,624
Likes
1,718
#85
Rinse & repeat each iteration gets more finicky, at some point you must rely on software analysis and automation no matter how much you enjoyed doing it all manually. That's not necessarily a bad thing.
Could you be more specific? I was very much with you to this point, but am struggling to think of anything in particular that requires software analysis/automation (unless you include things like gating an impulse response in this category).
 

Krunok

Major Contributor
Joined
Mar 25, 2018
Messages
2,476
Likes
1,133
Location
Zg, Cro
#87
No, what I am saying is that the individual divers in the speaker need to be time aligned first. This means an active system using a linear phase digital XO where the software, either manually or automatically, calculates the acoustic center of each driver and aligns them through the use of digital delay. Then, if required, linearizing the phase after... As mentioned in the previous post, linearizing the phase of a passive system, but not having the drivers time aligned, to my ears, has little audible effect.
I managed to correct phase (red and green is lef and right before correction, blue and violet is after correction). As you said, difference in sound is quite subtle but it does exist.



IR response didnt' improve at all. Are you saying I cannot really hope to correct IR response as long as I have passive XO?
 

Krunok

Major Contributor
Joined
Mar 25, 2018
Messages
2,476
Likes
1,133
Location
Zg, Cro
#88
This is how I managed to get phase measured in REW:

I took 6 sweep measurements from various positions around LP for left and right speaker. In REW, under All SPL/Controls I applied 1/12 smoothing, Time align and calculated vector average. Then I went to check IR response and applied gating of 3.3ms as that is when the first reflections appears. After that I exported measurement as text, imported it into rePhase and corrected the phase response.
 

Krunok

Major Contributor
Joined
Mar 25, 2018
Messages
2,476
Likes
1,133
Location
Zg, Cro
#89
So, I did some blind tests yesterday together with my 2 sons and the results were interesting.

Here is the phase response of left and right speakers after phase correction (gating set to 3.3ms):



Here is left speaker amplitude response before vs after phase correction:



.. and my right speaker amplitude response before and after phase correcetion:



Although amplitude response is practically identical I managed to differentiate phase corrected filters as better with 80% accuracy, my younger son with 90% accuracy and my older son with 100% accuracy. Subjectively all three of us agreed that female vocals sounded more realistically and that soundstage got wider and deeper with LF sounds being more "natural". So I'll keep the phase corrected filters. :)

P.S. amplitude gain settings remained the same on both set of filters, only phase gain was adjusted to make phase response linear. I can't explain the difference in LF amplitude respone around 30Hz.
 
Last edited:
Joined
Feb 5, 2019
Messages
96
Likes
55
#90
I've got a mic dilemma. I got the UMik-1 recommended here. Now I've measured my room with 3 different mics, each with very different responses. Each mic produces consistent results when re-measured. So I don't know which to trust.
A picture's worth 1,000 words. Here are the non-EQed raw response curves measured from the same listener position, each using that mic's calibration curve. Pink is the Umik-1, Blue is the Zoom H4, Yellow is the Rode NT1-A. Note 2 dB per division to accentuate the differences.
Mag3.6-micCompare.png

My gut says trust the UMik-1 since its calibration data is supposedly individually measured. But it shows a sharp dip at 72 Hz that neither of the other 2 mics show. All 3 mics agree on the 180 Hz dip. The NT1A is a nice mic that sounds great recording music and measures the lowest noise & distortion of these 3 mics. But that doesn't mean it has flatter frequency response! BTW, the UMik-1 showed exactly the same response from L and R channels, but the other 2 mics each showed variations in the 2-4 dB range.
PS: here it is with manually applied parametric EQ, within 2 dB of my desired slope from 100 Hz on up. Grey is before EQ, red is after. But, if the UMik-1 is right (and the other 2 mics are wrong), I have a trough from 50 to 90 Hz to deal with, too big to EQ. So I'll be revising room arrangement & treatment:
Mag3.6-190304-fr.png
 
Last edited:

andreasmaaan

Major Contributor
Patreon Donor
Joined
Jun 19, 2018
Messages
2,624
Likes
1,718
#91
I've got a mic dilemma. I got the UMik-1 recommended here. Now I've measured my room with 3 different mics, each with very different responses. Each mic produces consistent results when re-measured. So I don't know which to trust.
A picture's worth 1,000 words. Here are the non-EQed raw response curves measured from the same listener position, each using that mic's calibration curve. Pink is the Umik-1, Blue is the Zoom H4, Yellow is the Rode NT1-A. Note 2 dB per division to accentuate the differences.
View attachment 23085
My gut says trust the UMik-1 since its calibration data is supposedly individually measured. But it shows a sharp dip at 72 Hz that neither of the other 2 mics show. All 3 mics agree on the 180 Hz dip. The NT1A is a nice mic that sounds great recording music and measures the lowest noise & distortion of these 3 mics. But that doesn't mean it has flatter frequency response! BTW, the UMik-1 showed exactly the same response from L and R channels, but the other 2 mics each showed variations in the 2-4 dB range.
The other two mics are not measurement mics, so there’s no reason to assume they should have a flat response, partucularly not in the low bass and high treble. I’d trust the UMik more here.
 

DonH56

Major Contributor
Technical Expert
Patreon Donor
Joined
Mar 15, 2016
Messages
2,477
Likes
2,334
Location
Monument, CO
#92
The other two mics are not measurement mics, so there’s no reason to assume they should have a flat response, partucularly not in the low bass and high treble. I’d trust the UMik more here.
Exactly right. My UMIK-1 essentially overlies my Earthworks M30. In general recording mics do NOT have flat response by design. In your plots, the Zoom 4 is not at all consistent with the other two, but the NT1 is at least in the ballpark of the UMIK-1. Knowing nothing else I'd attribute the difference in the bass dip to either mic placement or differences in the environment (e.g. where you were standing, something else moved, etc.) I use a boom stand to hold the mics and, while the smaller measurements mics do OK in a regular mic clip, my larger diaphragms like the NT1 often need a suspension (spyder) for isolation. You cannot hold a mic and expect consistent results, and placing them on books or the couch seat is also a big no-no if you want an accurate reading. (Not saying you did any of that; I do not know.)
 
Joined
Feb 5, 2019
Messages
96
Likes
55
#93
I used my NT1As with the same pole with suspension spyder that I use for recording. As I mentioned, each mic's measurements are repeatably self-consistent so I think I've got the placement etc. down.
If I really trust the UMik-1's calibration data, I can now use it to create equally accurate calibration data for the other mics :)
 

DonH56

Major Contributor
Technical Expert
Patreon Donor
Joined
Mar 15, 2016
Messages
2,477
Likes
2,334
Location
Monument, CO
#94
Make sure you are using the cal file for your UMIK-1... I have made that mistake once or twice. Vexing.
 

Krunok

Major Contributor
Joined
Mar 25, 2018
Messages
2,476
Likes
1,133
Location
Zg, Cro
#95
. But, if the UMik-1 is right (and the other 2 mics are wrong), I have a trough from 50 to 90 Hz to deal with, too big to EQ. So I'll be revising room arrangement & treatment:
View attachment 23086
I'd also trust the UMIK-1.

Btw, if you aim for 68dB level in the range of 50-90dB that is not too big to EQ. ;)
 
Joined
Feb 5, 2019
Messages
96
Likes
55
#96
That would be good enough, but in my view it is too much to EQ. The red line shows the effect of +6 dB @ 64 Hz. The room sucks up energy at that freq and I'm only getting half the boost I apply (I added 6 dB but only measured a 3 dB increase). After this +6 dB boost it's only about 3 dB below 35-40 Hz, but getting it will take another +6 dB, for a total of +12 dB which is way too much for EQ. This situation calls for changes to room treatment.

I measured different speaker-wall distances to see if that would help. I discovered that without EQ my current speaker positioning already has the flattest FR curve. The good news is I did that original speaker positioning subjectively by ear, so at least my ears work. The bad news is the solution won't come that easily.

The 72 Hz null is so narrow I'm thinking bass trap instead of extra tube traps because I think bass traps have tighter Q than tube traps. Other suggestions?
 

Krunok

Major Contributor
Joined
Mar 25, 2018
Messages
2,476
Likes
1,133
Location
Zg, Cro
#97
That would be good enough, but in my view it is too much to EQ. The red line shows the effect of +6 dB @ 64 Hz. The room sucks up energy at that freq and I'm only getting half the boost I apply (I added 6 dB but only measured a 3 dB increase). After this +6 dB boost it's only about 3 dB below 35-40 Hz, but getting it will take another +6 dB, for a total of +12 dB which is way too much for EQ. This situation calls for changes to room treatment.
Is this response of both speakers playing? Can you post measurements of each speaker?

Btw, to me it is acceptible to apply amplitude correction up to 12dB, but no more than that. Dip at 74Hz looks like is related to the dimensions of your room. I would suggest you use correction at 70Hz with lower Q (3 or 4) instead at 64Hz (you probably used Q>5).
If that doesn't work you can try changing phase at 70Hz of one speaker and see what will happen.
 
Joined
Feb 5, 2019
Messages
96
Likes
55
#98
I wanted that particular filter to cover the range from 45 to 90 Hz. That's 1/2 octave on each side, which is Q=1.414, centered at 64 Hz.
All of my parametric EQ settings are 1/3 octave (on each side) or wider. That is Q=2.145 or less. I'd go narrower if I had to, but I try to avoid high Q because I'd rather have a few ripples in the response than phasey bloated sound from steep filters. Don't let the cure be worse than the disease!
For similar reasons, as a general rule, I avoid EQ more than +/- 6 dB. If the correction must be that big, in most cases you'll get better results from room treatment which fixes the root of the problem, rather than EQ which just slaps a band-aid over it.

This response is measured from listener position with both speakers playing. I think you're right, 72 Hz is a room mode not some kind of speaker phase issue. These planar speakers have a near flat-line phase response. The mic's L and R responses are identical. But I can play the tone with just one speaker to see what that looks like.
 

Krunok

Major Contributor
Joined
Mar 25, 2018
Messages
2,476
Likes
1,133
Location
Zg, Cro
#99
I wanted that particular filter to cover the range from 45 to 90 Hz. That's 1/2 octave on each side, which is Q=1.414, centered at 64 Hz.
That doesn't take into account that deepest point in that regions is at 74Hz sou you would do better to move your fitler frequency closer to it. For that reason I suggested you use 70Hz.

All of my parametric EQ settings are 1/3 octave (on each side) or wider. That is Q=2.145 or less. I'd go narrower if I had to, but I try to avoid high Q because I'd rather have a few ripples in the response than phasey bloated sound from steep filters. Don't let the cure be worse than the disease!
Amplitude non-linearities are more affecting SQ than phase non-linearities. Besides, you can always correct the phase once you are done with amplitude response. That is the beauty of FIR filters.

This response is measured from listener position with both speakers playing. I think you're right, 72 Hz is a room mode not some kind of speaker phase issue. These planar speakers have a near flat-line phase response. The mic's L and R responses are identical. But I can play the tone with just one speaker to see what that looks like.
It's ok to measure both speakers response to check LF response up to Schroeder frequency but you can't make usable correction based on such response. As you will be applying correction to each speaker separately you have to measure them separately as well.
 
Joined
Feb 5, 2019
Messages
96
Likes
55
That doesn't take into account that deepest point in that regions is at 74Hz sou you would do better to move your fitler frequency closer to it. For that reason I suggested you use 70Hz. ...
Alternatively, I could try center 72 Hz, 1/7 octave each side (Q=5). That would cover 65 to 80 Hz. Then add my current center 64, Q=1.4 filter on top of it, +6 dB each, to flatten the entire range 45 to 90 Hz and bring it up to level. But that still would be +12 dB of boost, which demands 16x more power from the amp in this frequency range, and consequent phase distortion from a steep filter, and a -12 dB overall level reduction to avoid digital clipping.

However, perhaps the phase distortion isn't really distortion. REW says this 72 Hz null is minimum phase, so the phase shift of the filter should actually correct the phase shift of this room null. In that light, this narrow filter might be desirable!

Amplitude non-linearities are more affecting SQ than phase non-linearities. Besides, you can always correct the phase once you are done with amplitude response. That is the beauty of FIR filters.
I'm using a Behringer DEQ2496 for the EQ. I'd like to stay within its capabilities. My goal isn't perfection, but something around the knee of the results vs. effort curve. Actually, saying that is self-delusional considering the hours I've spend building giant tube traps, measuring & arranging the room, etc.

It's ok to measure both speakers response to check LF response up to Schroeder frequency but you can't make usable correction based on such response. As you will be applying correction to each speaker separately you have to measure them separately as well.
I'll try that tip: measure each speaker independently. My DEQ2496 can apply different curves for L and R. Since these are room modes, I don't expect much difference but it's worth a try -- surprises keep it interesting!
 
Top Bottom