• 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

Krunok

Major Contributor
Joined
Mar 25, 2018
Messages
2,476
Likes
1,133
Location
Zg, Cro
#1
I kindly ask members to share their knowledge and experiences with manual room EQ.

I'm in no way expert with this, here's how I do it..

I design my filters with rePhase based on measurements done with REW with calibrated mic.

In the 20Hz-600Hz region I make RTA measurements at the area around my head at LP (app 40cm left and right, 20cm front and back and 20cm up and down).
When doing corrections in 20Hz-300Hz I don't mind using high Q filters (up to 30) and I correct up to +/-12dB.
In 300-600Hz region I'm using Qs up to 5 and I correct up to +/-5dB.

In the 600-20kHz region I make corrections based on sine sweep measurements taken app 1m from the speaker.
I'm using Q up to 2 and I correct up to +/-2dB.

Finally I enter crossover data and correct minimum phase filters that they are as close to 0 as possible.

My filters have 65536 taps and are made for 96khz sampling rate, they run with BruteFIR convolver on Volumio.

Any comments and guidelines from you experts are wellcomed as are any comments and experiences from those of you who did it and I thank you in advance for them! :)
 

Krunok

Major Contributor
Joined
Mar 25, 2018
Messages
2,476
Likes
1,133
Location
Zg, Cro
#3
Is it possible for REW to make the filter and skip the step with rePhase?
REW can't make FIR filters, only PEQ, so if your application allows for PEQs than yes.

What you can also do is import REWs filter in rePhase and generate FIR filter without modifying it. I haven't tried that, but I think it's possible.
 
Joined
Feb 5, 2019
Messages
96
Likes
55
#4
I kindly ask members to share their knowledge and experiences with manual room EQ.
I'm in no way expert with this, here's how I do it...
I'm no expert either, but here's my process. Maybe I'll learn some new ideas and improvements here.

I do it manually using basic math. First take a frequency sweep. I used to use the 1/3 octave warble tones from the old Stereophile test CDs, but now I use REW, much faster and more accurate. Either way, I now have a FR graph taken from the listener position. I take all measurements from the listener position.

Next I eyeball best fit my house curve line on top of it, about -1.5 dB / octave, and look for the biggest or sharpest deviations from this line.

For each such deviation, find the frequency range and the logarithmic center, and the center amplitude. For example, it might be a -4 dB dip from 220 to 350 Hz. The center ratio is sqrt(350/220), call it 1.261. So the center freq is the same however you want to compute it, 220 * 1.261, or 360 / 1.261, it is 277 Hz. This ratio is about 1/3 octave on each side; that is, log base 2 of 1.261 is .335 or about 1/3. So I apply a parametric EQ centered at 277 Hz, width 1/3 octave on each side, +4 dB.

I do this for the strongest 2 or 3 deviations, then re-run the sweep and repeat.

I don't worry about phase because I use parametric EQ with gentle amplitudes and slopes, which minimizes phase distortion to be inaudible in most cases. When deviations get less than 3-4 dB, I declare victory. Try not to let the cure be worse than the disease.

PS: I don't resort to EQ until I've exhausted what I can do with room treatments. I've gotten creative with room arrangement, built floor-to-ceiling tube traps, multiple layers of RPG foam, etc. This fixes all the biggest problems. Then use digital EQ to fine-tune the rest.
 

Krunok

Major Contributor
Joined
Mar 25, 2018
Messages
2,476
Likes
1,133
Location
Zg, Cro
#5
I'm no expert either, but here's my process. Maybe I'll learn some new ideas and improvements here.

I do it manually using basic math. First take a frequency sweep. I used to use the 1/3 octave warble tones from the old Stereophile test CDs, but now I use REW, much faster and more accurate. Either way, I now have a FR graph taken from the listener position. I take all measurements from the listener position.

Next I eyeball best fit my house curve line on top of it, about -1.5 dB / octave, and look for the biggest or sharpest deviations from this line.

For each such deviation, find the frequency range and the logarithmic center, and the center amplitude. For example, it might be a -4 dB dip from 220 to 350 Hz. The center ratio is sqrt(350/220), call it 1.261. So the center freq is the same however you want to compute it, 220 * 1.261, or 360 / 1.261, it is 277 Hz. This ratio is about 1/3 octave on each side; that is, log base 2 of 1.261 is .335 or about 1/3. So I apply a parametric EQ centered at 277 Hz, width 1/3 octave on each side, +4 dB.

I do this for the strongest 2 or 3 deviations, then re-run the sweep and repeat.

I don't worry about phase because I use parametric EQ with gentle amplitudes and slopes, which minimizes phase distortion to be inaudible in most cases. When deviations get less than 3-4 dB, I declare victory. Try not to let the cure be worse than the disease.

PS: I don't resort to EQ until I've exhausted what I can do with room treatments. I've gotten creative with room arrangement, built floor-to-ceiling tube traps, multiple layers of RPG foam, etc. This fixes all the biggest problems. Then use digital EQ to fine-tune the rest.
Interesting.. :)

Can you maybe post pre and post measurements done in REW?
 

Krunok

Major Contributor
Joined
Mar 25, 2018
Messages
2,476
Likes
1,133
Location
Zg, Cro
#7
Joined
Feb 5, 2019
Messages
96
Likes
55
#8
Yes, that's the response measured at the listener position, both speakers playing, with a pair of mics, averaged. The individual mics were within 3 dB of each other, and some of that deviation is the mics themselves. Even a matched pair will typically have 2 dB variations.

The region you indicated used to be worse... I've improved it but it's still the weakest area. However, keep in mind the Y axis spacing is 2 dB. I'd really like to fix the 190 Hz null but it is stubbornly resistant to room arrangement, treatment, and EQ.
 

Krunok

Major Contributor
Joined
Mar 25, 2018
Messages
2,476
Likes
1,133
Location
Zg, Cro
#9
Yes, that's the response measured at the listener position, both speakers playing, with a pair of mics, averaged. The individual mics were within 3 dB of each other, and some of that deviation is the mics themselves. Even a matched pair will typically have 2 dB variations.

The region you indicated used to be worse... I've improved it but it's still the weakest area. However, keep in mind the Y axis spacing is 2 dB. I'd really like to fix the 190 Hz null but it is stubbornly resistant to room arrangement, treatment, and EQ.
I believe that region can easilly be corrected by adding the filter at app 280Hz, +3dB, Q=2 (or something close to these values).

Can you pls post that graph with 1/12 smoothing?
 

Krunok

Major Contributor
Joined
Mar 25, 2018
Messages
2,476
Likes
1,133
Location
Zg, Cro
#10
Here is response of my speakers, same scale and smoothing. My room is "normal" livingroom, so not treated in any way.

 
Joined
Feb 5, 2019
Messages
96
Likes
55
#11
I believe that region can easilly be corrected by adding the filter at app 280Hz, +3dB, Q=2 (or something close to these values).
You mean 180 Hz? Exactly. The graph here already has +6 dB @ 180 Hz, 1/4 octave applied! The room seems to have a strong null there. But if there's a way to fix it that I haven't already tried, I'm all ears. :)
However, I try to avoid narrow EQ because it feels like squashing the signal, and I try not to make the cure worse than the disease.

Here is response of my speakers, same scale and smoothing. My room is "normal" livingroom, so not treated in any way.
...
Lucky you! My room was FAR worse than that. Before the tube traps, bass response was much worse, like 12 dB peaks and troughs. And before the RPG foam, upper bass to midrange was also bad. Most of the improvement I got came from the room treatments. In the various places I've lived over the years, I've never had a room that measured as good as yours without treatment. Most were MUCH worse. It took a lot of effort to make them sound good.

PS: I'm curious what you're getting for distortion.
 

Krunok

Major Contributor
Joined
Mar 25, 2018
Messages
2,476
Likes
1,133
Location
Zg, Cro
#12
You mean 180 Hz? Exactly. The graph here already has +6 dB @ 180 Hz, 1/4 octave applied! The room seems to have a strong null there. But if there's a way to fix it that I haven't already tried, I'm all ears. :)
However, I try to avoid narrow EQ because it feels like squashing the signal, and I try not to make the cure worse than the disease.
No, I mean 280Hz. I wouldn't fight that dip at 180Hz but I would try to raise a wider region with center at app 280Hz with a low Q filter (say with Q=2).

If you send me export of that measurement in txt format I can try to see how would that look when I make such filter in rePhase.

Edit: it would actually be better if you can send me exports of left and right speaker measured seprately. Do you have separate filters for left and right speaker?
 
Last edited:

Krunok

Major Contributor
Joined
Mar 25, 2018
Messages
2,476
Likes
1,133
Location
Zg, Cro
#13
Lucky you! My room was FAR worse than that. Before the tube traps, bass response was much worse, like 12 dB peaks and troughs. And before the RPG foam, upper bass to midrange was also bad. Most of the improvement I got came from the room treatments. In the various places I've lived over the years, I've never had a room that measured as good as yours without treatment. Most were MUCH worse. It took a lot of effort to make them sound good.

PS: I'm curious what you're getting for distortion.
It's not that good as it looks. For example my spectrogram and decay graphs aren't as nice as yours.

With sweep measurement in distortion panel I'm getting app 0.5%, except at LF where it peaks at 5% at app 75Hz.
I prefer measuring distortion with RTA and single frequency tones as I get more consistent figures than with sweep.
 
Joined
Feb 5, 2019
Messages
96
Likes
55
#14
I don't use 1/12 octave smoothing because I don't want to apply filters that narrow, and such tight bumps don't necessarily reflect what we hear, especially above 6k. IME, 1/12 spacing can have you chasing ghosts. But here it is: green L, red R, blue is average.
Mag3.6-190216-fr-12octLR.png

Here's before/after the EQ, at 1/12 smoothing:
Mag3.6-190126-fr-compare2.png
 

Krunok

Major Contributor
Joined
Mar 25, 2018
Messages
2,476
Likes
1,133
Location
Zg, Cro
#15
I don't use 1/12 octave smoothing because I don't want to apply filters that narrow, and such tight bumps don't necessarily reflect what we hear, especially above 6k. IME, 1/12 spacing can have you chasing ghosts. But here it is: green L, red R, blue is average.
View attachment 22286
Here's before/after the EQ, at 1/12 smoothing:
View attachment 22287
I agree with you, I just wanted to check hom much would those 2 dips go down when you reduce the smoothing.

Are you using the same filters for left and right channel or different?

Is the first graph you posted with 1/6 smoothing average between left and right channel or a measurement taken while both speakers were playing?
 
Joined
Feb 5, 2019
Messages
96
Likes
55
#16
Same filters for both. First 1/6 oct graph is an average of L and R, which were both playing simultaneously for all of these measurements.
 

Krunok

Major Contributor
Joined
Mar 25, 2018
Messages
2,476
Likes
1,133
Location
Zg, Cro
#18
Same filters for both. First 1/6 oct graph is an average of L and R, which were both playing simultaneously for all of these measurements.
IMHO average of left and right doesn't really give a meaningful picture - you should do a measurement sweep where both speakers are playing. From what I have learned from experts up to 300Hz or so sound is dominated by both speakers response while north of 300Hz it is every speaker for itself so you should measure them separately.

Btw, in my experience it is very hard to do EQ with same filter for both speakers so I suggest you measure and correct them separately. After you have done that try to tune both speakers response in LF region.
 

Krunok

Major Contributor
Joined
Mar 25, 2018
Messages
2,476
Likes
1,133
Location
Zg, Cro
#19
I would also here like to stress a difference between Infinite Impulse Response (IIR) filters and Finite Impulse Response (FIR) filters.

IIR filters, also called parametric equaliser filters or PEQs, are generated by REW. Basically they consist of a number of pair values where one determines the frequency at which the filter is applied and the other is Quality factor (Q) which controls the width of the area around targeted frequency which is affected by the filter. Advantage of IIR filters is they are simple and require less processing power when applied. Disadvantage is they affect phase, especially when high Q values are used. Effect of that can be heard so avoid using filters with high Q values north of 300Hz.

FIR filters, generated by rePhase and some other tools, require more processing power when applied but are able to generate minimum phase filters and are more precise for room EQ than IIRs (PEQs). With FIR filters phase can be corrected independently of amplitude.

For that reason it is easier to make good "cure" with FIRs than with with IIRs. ;)
 
Joined
Feb 5, 2019
Messages
96
Likes
55
#20
Exactly! Phase distortion is related to the slope of the filter. Use gentle slopes (numerically small Q, wide frequency range) minimizes it.

Question: When the room boosts or cuts certain frequencies, it also induces phase distortion. Could it be possible that the phase distortion of a parametric EQ opposes this, so the filter corrects phase as well as amplitude? If so, a parametric EQ's "side effect" of shifting phase would actually be beneficial!
 
Top Bottom