• 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). There are daily reviews of audio hardware and expert members to help answer your questions. Click here to have your audio equipment measured for free!

Is this a good room correction strategy?

khrisr

Member
Joined
Jul 30, 2021
Messages
32
Likes
15
I just got a new set of speakers (KEF R3) which I've tried to tune using REW. I'm thinking of inserting a minidsp 2x4 HD between DAC (Oppo UDP-203) and amplifier (NAD C325BEE) to implement the corrections.

As a prototype I created convolution filter using REW + UMIK-2 for measurement using RTA pink noise. I can hear a subtle but significant improvement in the sound and I see it helps some of the music that I found it a bit harsher on the ears. Still needs tweaking but I'm new to all this so wanted to post my overall strategy here to check if I'm going in the right direction. Thanks for reading and any tips are much appreciated.

==FIG 1==

Room = 22 x 21x 7.5 with half partition wall with kitchen in right half
Left speaker = 29" from front wall/45" from left wall (half port bung in place to reduce boundary effect)
Right speaker = 29" from front wall/axially centered (port fully open)

room.jpg


==FIG 2==

In room response (Pink = left speaker Blue = right speaker)

response.jpg


==FIG 3==

Corrected L+R response (Blue = corrected Orange = original)

corrected.jpg


==FIG 4== Roon signal path

Screen Shot 2021-07-30 at 3.46.37 PM.png


Here are some questions I'm contemplating. Would be great to hear any recommendations -

(1) Will adding miniDSP to the signal chain degrade sound significantly? I see the chips have similar specs so I'm hoping they would be equivalent.

miniDSP => ADSP21489 => 115 DR 107 S/(N+D)
UDP-203 => AK4458VN => 115 DB SNR THD+N: -107dB

(2) Is having port bung only on one side recommended? I haven't heard of this but it does make the L and R responses much closer.

(3) I read about phase corrections in addition to filters but I'm not making much progress understanding how to do it. Is this needed?

(4) I only used negative EQ parameters but it appears the response has been normalized. I was getting clipping as a result so I turned on volume normalization in Roon. I don't understand how DSP works but is it effectively converting my filters to +ve EQ? I'm wondering if that might result in wasting amp power trying to fill in the response dip between 50-100 hz

:::: LEFT filters ::::
Filter 1: ON HSQ Fc 90.00 Hz Gain -1.80 dB Q 1.00
Filter 2: ON PK Fc 122.0 Hz Gain -2.00 dB Q 4.139
Filter 3: ON PK Fc 199.5 Hz Gain -3.00 dB Q 2.000
Filter 4: ON PK Fc 590.0 Hz Gain -3.40 dB Q 5.312
Filter 5: ON HSQ Fc 1000 Hz Gain -1.00 dB Q 1.00 (tweaked house curve)
Filter 6: ON HSQ Fc 3000 Hz Gain -0.50 dB Q 1.00 (tweaked house curve)
Filter 7: ON HSQ Fc 4000 Hz Gain -1.00 dB Q 0.500 (tweaked house curve)
Filter 8: ON PK Fc 1150 Hz Gain -1.50 dB Q 5.000 (tame presence brightness)

:::: RIGHT filters ::::
Filter 1: ON HSQ Fc 100.0 Hz Gain -2.00 dB Q 1.00
Filter 2: ON PK Fc 137.0 Hz Gain -5.00 dB Q 7.320
Filter 3: ON PK Fc 161.5 Hz Gain -2.00 dB Q 8.131
Filter 4: ON PK Fc 204.0 Hz Gain -4.50 dB Q 7.000
Filter 5: ON PK Fc 561.0 Hz Gain -5.50 dB Q 3.000
Filter 6: ON HSQ Fc 1000 Hz Gain -1.00 dB Q 1.00 (tweaked house curve)
Filter 7: ON HSQ Fc 3000 Hz Gain -0.50 dB Q 1.00 (tweaked house curve)
Filter 8: ON HSQ Fc 4000 Hz Gain -1.00 dB Q 0.500 (tweaked house curve)
Filter 9: ON PK Fc 1150 Hz Gain -1.30 dB Q 5.000 (tame presence brightness)

Cheers!
 

Attachments

  • Screen Shot 2021-07-30 at 3.44.51 PM.png
    Screen Shot 2021-07-30 at 3.44.51 PM.png
    235.9 KB · Views: 51
  • Screen Shot 2021-07-30 at 3.45.38 PM.png
    Screen Shot 2021-07-30 at 3.45.38 PM.png
    230.1 KB · Views: 50
  • Screen Shot 2021-07-30 at 3.46.37 PM.png
    Screen Shot 2021-07-30 at 3.46.37 PM.png
    181.5 KB · Views: 53
  • Screen Shot 2021-07-30 at 3.46.37 PM.png
    Screen Shot 2021-07-30 at 3.46.37 PM.png
    181.5 KB · Views: 50
Last edited:

Daverz

Major Contributor
Joined
Mar 17, 2019
Messages
1,129
Likes
1,190
I think you should be able to get much better results with REW. I would play around a bit more with the target level and allow "narrow filters below 200 Hz".

Since Roon can do the convolution, another option would be DRC software like DRC-FIR, Acourate, Audiolense, or Focus Fidelity.

If you post your impulse response (as a stereo float32 wav of left and right impulse response) and mic correction somewhere, I can run DRC-FIR on it.
 

b4nt

Addicted to Fun and Learning
Joined
May 29, 2021
Messages
803
Likes
266
I wouldn't compensate left/right speakers using a DSP, I would perhaps first add an accoustic damper on the wall, at the left side....
 

raindance

Addicted to Fun and Learning
Joined
Sep 25, 2019
Messages
783
Likes
698
Your levels appear to be set way too low to evaluate any response. Also looks like you have a great big suckout at 80Hz from SBIR, perhaps.
 
OP
K

khrisr

Member
Joined
Jul 30, 2021
Messages
32
Likes
15
Your levels appear to be set way too low to evaluate any response. Also looks like you have a great big suckout at 80Hz from SBIR, perhaps.

I live in a very quiet area with ambient noise about 25 dB so I thought I'd test at my normal listening volume. Does the accuracy suffer if levels are low? And yes the left wall is a big problem. This was the best I could get with placement.
 
OP
K

khrisr

Member
Joined
Jul 30, 2021
Messages
32
Likes
15
I think you should be able to get much better results with REW. I would play around a bit more with the target level and allow "narrow filters below 200 Hz".

I do have narrow filters turned on. I think the reason REW is not compensating more is I set max boost to 3 dB. I was getting poor results initially so I thought I'd constrain the adjustments. I'll try again after giving some rest time to my ears.
 

b4nt

Addicted to Fun and Learning
Joined
May 29, 2021
Messages
803
Likes
266
Did you orient your speakers to the listeners? Did you check distances between speakers and speakers/listeners?

Maybe suspend anything on that wall and check what other DSP corrections could be. Not the same for left/right, I do not think this is acceptable.
 
OP
K

khrisr

Member
Joined
Jul 30, 2021
Messages
32
Likes
15
Did you orient your speakers to the listeners? Did you check distances between speakers and speakers/listeners?

Maybe suspend anything on that wall and check what other DSP corrections could be. Not the same for left/right, I do not think this is acceptable.

The speakers are pointing straight out. I read that KEF recommends no toe in so I did it that way. Is that not good practice? Also I measured at various distances +- 2 ft from listening position + moving the speakers around +- 1 ft. Thats about as much room I have to play with unfortunately.

Also yes I applied different amounts of correction to each speaker. The left speaker has a big dip between 50-100 hz so I thought I'd compensate by going half port + more EQ. Does having asymmetrical filters cause problems? I was getting audible phase problems in my initial attempts but I was also trying to boost small ranges of frequencies so I thought it was related to that.
 

b4nt

Addicted to Fun and Learning
Joined
May 29, 2021
Messages
803
Likes
266
Swap left and right speakers, recheck? Using a DSP, I would first compensate spearkers imperfections, then only room imperfections. Would not know exactly how to do that, as the room itself has an impact.

And according to my room or needs, I wouldn't care that much about manufacturer recommandations. Unless for minimal distance from the wall and so, like for rear events...

And as you are willing to tweak with that wall, I would recommend you to Google for DIY decorative accoustic pannels. Cause you should get reflexions from there (maybe less important depending on speakers orientation).
 
OP
K

khrisr

Member
Joined
Jul 30, 2021
Messages
32
Likes
15
Swap left and right speakers, recheck? Using a DSP, I would first compensate spearkers imperfections, then only room imperfections. Would not know exactly how to do that, as the room itself has an impact.

Got it. I assumed the difference is because one speaker is closer to the boundaries but yea I'll check it out.

Btw I was playing with the room simulator in REW and it seems if I can move the whole setup 2 ft further out from the front wall the dip should improve significantly. I'll see if I can carve out that additional space. And will shop for panels too. Thanks for the tips!
 

Daverz

Major Contributor
Joined
Mar 17, 2019
Messages
1,129
Likes
1,190
I do have narrow filters turned on. I think the reason REW is not compensating more is I set max boost to 3 dB. I was getting poor results initially so I thought I'd constrain the adjustments. I'll try again after giving some rest time to my ears.

What does REW's predicted response look like?

What equalizer setting are you using in REW? What are the target settings? You can copy your target settings by right clicking on the Target Settings pane.

You have filters going out to 4000 Hz, so I don't understand why it is not flattening the big hump at 600 Hz, the wide hump from 100 to 300 Hz and the peak at 40 Hz. It's making things worse between 50 and 90 Hz, probably because of the wide (low Q) filters used.

Also, you're response is pretty good over 700 Hz or so, so no reason to EQ out so far. I would suggest limiting the EQ to under 300 Hz. Try taming that peak between 500-600 Hz first by moving the speakers and listening position around and treating the wall.

I find REWs automatic EQ interface to be tricky to use. You might find it easier just to eyeball the main response peaks and humps with the cursor, turning on the predicted response, and then creating a few PK filters by hand, adjusting the gain and Q by trial and error until you get a reasonable predicted response.
 
Last edited:
OP
K

khrisr

Member
Joined
Jul 30, 2021
Messages
32
Likes
15
Your "corrected" response looks worse than the original.

What does REW's predicted response look like?

What equalizer setting are you using in REW? What are the target settings? You can copy your target settings by right clicking on the Target Settings pane.

You have filters going out to 4000 Hz, so I don't understand why it is not flattening the big hump at 600 Hz, the wide hump from 100 to 300 Hz and the peak at 40 Hz. It's making things worse between 50 and 90 Hz, probably because of the wide (low Q) filters used.

Also, you're response is pretty good over 700 Hz or so, so no reason to EQ out so far. I would suggest limiting the EQ to under 300 Hz. Try taming that peak between 500-600 Hz first by moving the speakers and listening position around and treating the wall.

So why its boosting and not cutting is a mystery I'm trying to figure out. If I move the corrected plot down by -4.4 dB (see below) it approximately matches the response predicted by REW so I assumed its normalizing it somehow (Roon volume levelling?)

corrected_denorm.jpg


Target type is None and level 38.7 dB
 

Daverz

Major Contributor
Joined
Mar 17, 2019
Messages
1,129
Likes
1,190
So why its boosting and not cutting is a mystery I'm trying to figure out. If I move the corrected plot down by -4.4 dB (see below) it approximately matches the response predicted by REW so I assumed its normalizing it somehow (Roon volume levelling?)

View attachment 144588

That doesn't look like normalization. The peak at 40 Hz doesn't budge.

Target type is None and level 38.7 dB

Is there a target curve?

If the target level is 38.7 dB, then the top of the predicted response should follow that target line. So something is wrong. I would not bother using the exported filter impulse response in Roon until you have a better handle on what REW EQ is doing.
 
OP
K

khrisr

Member
Joined
Jul 30, 2021
Messages
32
Likes
15
That doesn't look like normalization. The peak at 40 Hz doesn't budge.

Well … I was actually not trying to cut the 40 Hz peak. My strategy was to cut everything above 100 hz to make a more flattish curve with downward slope. I need to tune the filters more but I’ve been at it for 3 days so just checking in case I’m going down a non-optimal route.

If this DSP approach turns out to be not good enough the alternative I suppose would be to ditch it and just go straight to 2 sub woofers and see if I can blend them into the nulls. But that’s at least another $1k budget item :[ Assuming there is still some gaps I can then go for targeted acoustic treatments (traps, diffusers etc) to fine tune further. But before all that … try and see if software DSP is an easy win. How’s that order of precedence?
 
OP
K

khrisr

Member
Joined
Jul 30, 2021
Messages
32
Likes
15
If the target level is 38.7 dB, then the top of the predicted response should follow that target line. So something is wrong. I would not bother using the exported filter impulse response in Roon until you have a better handle on what REW EQ is doing.

Just realized I copied the measurements with both L+R channels firing. That’s why the levels measure higher. I’ll try to recheck my measurements after tweaking the placement. The room sim indicates moving further out and right of the corner might improve things.
 

Daverz

Major Contributor
Joined
Mar 17, 2019
Messages
1,129
Likes
1,190
Well … I was actually not trying to cut the 40 Hz peak. My strategy was to cut everything above 100 hz to make a more flattish curve with downward slope.

What I meant was that the normalized curve should just be shifted up or down equally across the spectrum.
 

Jim Matthews

Major Contributor
Forum Donor
Joined
Mar 25, 2021
Messages
1,063
Likes
1,296
Location
Taxachusetts
Got it. I assumed the difference is because one speaker is closer to the boundaries but yea I'll check it out.

Me, too.

I'm not reading the raw curve as having peaks, at all.
That "dip" between 40 and 100 cycles looks suspicious.

A cheap way to test my theory would start with blocked ports...
 
Last edited:

Hipper

Addicted to Fun and Learning
Joined
Jun 16, 2019
Messages
753
Likes
608
Location
Herts., England
I don't see much point in producing graphs that go below 30Hz or above 20kHz in your case.

You suggest you can't move the speakers anymore, Can you move the listening chair a bit, forwards or backwards? Can you use some sort of diagonal positioning? Maybe the two speakers either side of the corner and the chair facing the corner?

I think you should concentrate on the bass first - 0-500Hz say. Use graphs with that range and 'no smoothing'. Concentrate most of your efforts to getting this as flat/smooth as possible. Lower peaks, ignore very narrow dips but try and fill the broader dips to some degree. Once this is done the best you can, look at the mids and highs but alter them with broad adjustments.

I wouldn't fuss about phase - that changes as you adjust the frequency response (FR). Lowering decay times is sound improving (within reason). Again often adjusting FR will improve decay times but the best for this is room treatment, notably absorbent panels, bass traps etc..

I know nothing about using port bungs.

Daverz suggests you can run some DRC software with Roon - this sounds like a good idea. I know nothing about this either!

I use REW to measure and manually import the filters to a Behringer DEQ2496 but I'm old fashioned - I use a CD play and DAC. I did modernise a bit and also listen to headphones from a dedicated PC with JRiver Media Centre on it. That has a Parametric EQ and and Graphic EQ.

When I measure individual speakers with REW, say the left speaker, I place the microphone exactly where my left ear would be, pointing at the speaker. I use the measurement sweep not pink noise. If measuring just 0-500Hz you can set the measurement sweep to only this range.
 
Top Bottom