• 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.

REW, A Graph and What It All Means

Joined
Jul 17, 2018
Messages
25
Likes
3
#1
I tried REW for the first time following a combination of guides. Ultimately, I took a measurement, had REW EQ it and applied what it said to do in JRiver through the parametric EQ. I did this twice. The first time, I took separate measurements for each speaker, averaged out what was found and then applied the REW eq. The second time, I had both channels measures simultaneously. I was surprised that with both attempts, there were few areas of change. (6 frequencies were identified by the EQ when averaged and 10 when taken simultaneously.) I preferred the sound of the eq using the simultaneous measurements rather than the averaged, for what that's worth. Further, I didn't, ultimately, reduce two frequencies as much as recommended after adjusting the numbers more to my liking. All this leaves me with a few questions (after my before graph which is the simultaneous measurement - I ran it twice with a minimal deviation near the end of the range):

BEFORE.jpg

1. Is there a preferred way to measure: individually, simultaneous, multiple points in a room (I did them only from my listening position).?

2. With eq, there was less pronounced bass (I reduced the reduction here by 1/2) and a less bright high end (I reduced the reduction here as well, but by 1/4). Is that to be expected or are there different ways to apply the filter (I used what the guide suggested) that change the sound more or less rather than simply aiming for flat?

3. I expected a much larger range of frequencies to adjust. Not because I could hear that it would be necessary, but by virtue of reading what others had done vis-a-vis REW. Does the above graph indicate that there wasn't a whole lot of eq-ing to be done?

4. Where would I go from here assuming I feel further room correction is necessary/desired/warranted?
 

GrimSurfer

Addicted to Fun and Learning
Joined
May 25, 2019
Messages
869
Likes
917
#2
I have nothing to offer on measuring methodology, as more experienced folks may wish to answer.

There's a bit of EQing to be done based on your graph. The aim is to flatten the nodes while maintaining a consistent slope.

The nulls cannot be tamed by EQing. Room treatment could reduce the nulls at 90 and 330 Hz. The null at 31 Hz will likely be a bitch to resolve unless you get some sort of tuned bass trap. These can be bulky, expensive, and difficult to locally procure.
 

amirm

Founder/Admin
Staff Member
CFO (Chief Fun Officer)
Joined
Feb 13, 2016
Messages
21,297
Likes
28,765
Location
Seattle Area
#3
2. With eq, there was less pronounced bass (I reduced the reduction here by 1/2) and a less bright high end (I reduced the reduction here as well, but by 1/4). Is that to be expected or are there different ways to apply the filter (I used what the guide suggested) that change the sound more or less rather than simply aiming for flat?
Yes, that is to be expected especially if you made the response flat. The cure is to have the "target curve" be sloping down. Do this to taste and music you listen to.

Even with above, you may need to give it a bit of time to get used to it. Without EQ you have massive boost of some bass notes and taking them down naturally makes the sound less bass heavy. The reward is much tighter bass with more detail elsewhere in the spectrum.
 

amirm

Founder/Admin
Staff Member
CFO (Chief Fun Officer)
Joined
Feb 13, 2016
Messages
21,297
Likes
28,765
Location
Seattle Area
#4
3. I expected a much larger range of frequencies to adjust. Not because I could hear that it would be necessary, but by virtue of reading what others had done vis-a-vis REW. Does the above graph indicate that there wasn't a whole lot of eq-ing to be done?
My preference is to fixe just a few peaks in low frequencies and resist the temptation to fix more. Make sure you set the averaging to 1/12 or 1/24 octave for low frequencies so that you can see all the peaks.
 

amirm

Founder/Admin
Staff Member
CFO (Chief Fun Officer)
Joined
Feb 13, 2016
Messages
21,297
Likes
28,765
Location
Seattle Area
#5
1. Is there a preferred way to measure: individually, simultaneous, multiple points in a room (I did them only from my listening position).?
I would not do the averaging when you have the option to run both speakers at once. As to whether to do it with one or two speakers, there are different schools of thought on that. One says bass is mono so having both speakers on is more real as far as what you are hearing. The other says measure each speaker and correct the peaks. I tend to want to do it one speaker at a time.
 

zermak

Active Member
Joined
Jun 2, 2019
Messages
108
Likes
46
Location
Italy
#6
You can pick up your target house curve based on your preferences. I am using the Harmann one and I am happy with it for my desktop speakers.
About averaging both channels or not it is up to you but if you have the patience and the time then you can EQ both channels own their own for best performance (I guess). I have different EQ-ing because my L channel is near a corner.
Again. I suggest you to take more measurements moving the mic around your listening position and average the responses for the single channel to cover the area in which you move your head. How many positions is up to you. I would try with at least 3 measurements for channel: +/-X centimeters from middle position on both sides. You can go even futher moving the mic on front and behind the middle position and so on. Just don't mess up the channels and average the right measurements (L or R) :)
I think there is something about measurements on DIRAC Live and the other room corrections systems.
 

RayDunzl

Major Contributor
Central Scrutinizer
Joined
Mar 9, 2016
Messages
7,170
Likes
3,594
Location
Riverview, FL
#7
Using the "Overlays" button, you can show the before and after on one graph.

I'd like to see that...


With eq, there was less pronounced bass
That's probably typical.


Example:
JBL LSR 308, both speakers playing, flat target (I apply slope elsewhere when desired), single measurement at the listening position:

1570732572265.png


My automated software takes measurements from a single point, not averaging, and corrects full range.

Here are the left and right before/after:

1570732834097.png
 
Joined
May 24, 2016
Messages
282
Likes
733
#8
@Jefrpol Looking good. Have a look at http://seanolive.blogspot.com/2009/11/subjective-and-objective-evaluation-of.html and open up the presentation. Check out slides 24 and 25 for the money shots. A downward flat, but tilting response from 20 Hz to -10 dB @ 20 kHz measured at the LP is preferred by most listeners and is one variation of the so called Harman Target response. Note on slide 25 that our ears/brain perceive that as a flat or neutral or response. A flat in-room response at the LP is not the preferred target and is perceived as overly bright.

Sean Olive has followed up with many perceptual listening studies and objective measurements after to confirm this preferred response (and headphones too). If interested in the methodology etc., these two presentations, while discussing headphones, also discuss loudspeakers and the preferred in-room response:

The Perception and Measurement of Headphone Sound Quality: Do Listeners Agree on What Makes A Headphone Sound Good?
and
Perception and Measurement of Headphone Sound Quality: Is There a Preferred Target Response.

I use DSP room correction products like Acourate and Audiolense. I have written a number of articles on both products you can read here along with what is in my sig.

My preference is for the 20 Hz to -10 dB at 20 kHz straight line which sounds neutral to my ears. My suggestion is to start with this target and give it a listen. The make any adjustments for preferences, room to dead/live or the type of music content you listen to.

Btw, room correction products like Audiolense and Acourate offer partial correction and if the speakers are well designed for both on and off axis response, you should only need to correct below 500 Hz. However, both these products also offer to do the heavy lifting under 500 Hz, but apply a "tilting tone control" to frequencies above 500 Hz to adjust to ones preference if the sound is a tad too bright or dull.

Good luck and happy listening!
 

MZKM

Senior Member
Joined
Dec 1, 2018
Messages
377
Likes
292
Location
Land O’ Lakes, Florida
#9
1) I do a cubic grid (3x3x3; so 27) with about a foot between the lateral measurements and about 1/2 a foot between the vertical measurements. I also use the 90° position and respective calibration file. You don’t have to do that much, 10 is more than enough. I usually just do psychoacoustic filtering.
2) Set a target curve. Harman’s is something like this:

3) The 10dB dips are ~90Hz and ~300Hz aren’t great.

4) Tinker, including speaker placement, toe-in, seat placement, etc.
 
Joined
Sep 16, 2019
Messages
85
Likes
68
#10
@mitchco

That study involves auto-EQ on a speaker with a non-flat response and a directivity issue. I wonder what would have been the result if they had made it flat anechoically and limited the in-room correction (using the same slope) to around 500Hz. Sean Olive himself has said, as little as a week ago, on his Facebook page it's better to use neutral loudspeakers and leave them as-is.

1570777769731.png
 

mi-fu

Senior Member
Patreon Donor
Joined
Jun 7, 2018
Messages
422
Likes
352
Location
New York
#11
Actually I have a question regarding the volume used during the measurement.

Is there a preferable SPL (in terms of db) that I should use? Or should I use the volume that I mostly listen to music to?
 

MZKM

Senior Member
Joined
Dec 1, 2018
Messages
377
Likes
292
Location
Land O’ Lakes, Florida
#12
Actually I have a question regarding the volume used during the measurement.

Is there a preferable SPL (in terms of db) that I should use? Or should I use the volume that I mostly listen to music to?
If your system has loudness compensation/control, then you need to take into account how that affects the sound.
For a regular setup, do it around the typical average listening SPL.

I do wonder at what SPL are target curves from some companies derived at, like is Harman’s target for 80dBC or 90dBC or...?
 

audimus

Senior Member
Joined
Jul 4, 2019
Messages
328
Likes
239
#13
Actually I have a question regarding the volume used during the measurement.

Is there a preferable SPL (in terms of db) that I should use? Or should I use the volume that I mostly listen to music to?
The filters calculated are relative numbers relative to the target, so the absolute levels are not that critical. And the target should be set to smooth out the curve not set so high that the filters are being used for gain. And it is unrelated to your listening level except you don’t want to hurt your ears with it being too loud during measurements.

You just need to make sure they are loud enough for the measurement to be accurate. Take a noise floor measurement and aim for about 30-40 db over the peaks on that floor. Typically around 70db to 75db should work comfortably with a calibrated mic unless you are in a noisy room.

You should also adjust the SPLs of all speakers to be around the same SPL for taking measurements.
 

audimus

Senior Member
Joined
Jul 4, 2019
Messages
328
Likes
239
#14
I do wonder at what SPL are target curves from some companies derived at, like is Harman’s target for 80dBC or 90dBC or...?
Does not matter what the target curve is at. It is more of a definition of the slopes than absolute values. You adjust the overall gain to get the reference levels or listening levels and the filter corrections will be relative to that.
 

audimus

Senior Member
Joined
Jul 4, 2019
Messages
328
Likes
239
#15
1. Is there a preferred way to measure: individually, simultaneous, multiple points in a room (I did them only from my listening position).?
I have tried both multi-position and RTA and prefer the pink noise RTA method by sweeping the mic slowly in loops in multiple planes across the seating area and flat target curves for multi-channel HT and a 5 fixed-position frequency sweep measurement like Anthem ARC with a “best fit” target curve for two channel music. For the latter, the idea is to use a target curve that more or less follows the local median slope of the measurement with VAR smoothing (think of it as virtual large grain smoothing of the measured curve). This way, the filters are just smoothing the curve around that target than applying too much gain boost or cut. This way, you can do a full range EQ without artifcats.

If the bass management is downstream, I set the crossovers for each channel and then do the measurement and correction for each channel separately with cross-overs in place (as large or full range speakers). I don’t average between speakers. The filters are applied to each channel separately.
 

MZKM

Senior Member
Joined
Dec 1, 2018
Messages
377
Likes
292
Location
Land O’ Lakes, Florida
#16
Does not matter what the target curve is at. It is more of a definition of the slopes than absolute values. You adjust the overall gain to get the reference levels or listening levels and the filter corrections will be relative to that.
Our hearing changes with level, lower level signal need more bass and treble to sound as neutral; this is the whole purpose of loudness adjustment on stereo's and stuff like Audyssey's Dynamic EQ on Denon/Marantz receivers.
 

mitchco

Active Member
Joined
May 24, 2016
Messages
282
Likes
733
#17
@TimVG No disagreement here. I don't do Facebook. I did mention in my post to keep eq applied below 500 Hz ;) But folks should realize that sophisticated DSP correction systems like Audiolense and Acourate use frequency dependent windowing (FDW) that users can control so only below 500 Hz the direct and reflected sound is being modified, but beyond that, only the direct sound that is being modified, like a tilting tone control. So no issues. Also some of us use constant directivity waveguides, aside from requiring eq compensation, the whole point with constant directivity if you eq on axis, you already know what the off axis response is going to be :) Wrt to room correction, I really like @j_j 's presentation on Acoustic and Psychoacoustic Issues in Room Correction.

Part of the issue is that there are so few loudspeakers designed using ANSI/CTA 2034-A standard. I have mentioned elsewhere that a single on axis flat response in an anechoic chamber doth not make a good loudspeaker as it does not account for the off axis response. But if you have not read the Sean's presentations (and AES papers) I linked to above, it is worth a read as there is a preferred target response for both loudspeakers and headphones. Maybe one day more manufacturers will start using the 2034 standard.

@MZKM Yes, you want to calibrate between 75 dB and 85 dB SPL C weighting and slow integration at the LP. Relatively speaking, our ears are the most flat through this range. As an ex recording/mixing engineer, in the industry, most final mixes and masters are mixed at reference level (i.e. 83 dB SPL) for setting the bass level relative to the treble. Our ears are nonlinear and the frequency response changes with SPL. I use a loudness control in JRiver Media Center and calibrate the SPL so that below 83 dB SPL it kicks in to compensate for our dropping sensitivity in the bass as the level goes down. Works a treat. For more on the magic of 83 db SPL see Bob Katz's article: https://www.digido.com/portfolio-item/level-practices-part-2/
 
Joined
Jul 17, 2018
Messages
25
Likes
3
#18
Appreciate all of the responses.

@RayDunzl I'll look to take a measurement with the eq applied to see what changes are being made and to what extent. After applying the changes through JRiver, I was more curious to hear the differences in sound with the eq on and off and it didn't occur to me to measure the room again.

As for using the Harman curve as a target, how would you set REW to apply that? For my first try with REW, I simply set it for full-range speakers and that was it.

In regard to measuring, I used the 90 deg calibration file. For two channel, should I have used 0-deg?

As for volume, I used the spl meter to set it for 85db at the suggestion of one guide.
 

audimus

Senior Member
Joined
Jul 4, 2019
Messages
328
Likes
239
#19
Our hearing changes with level, lower level signal need more bass and treble to sound as neutral; this is the whole purpose of loudness adjustment on stereo's and stuff like Audyssey's Dynamic EQ on Denon/Marantz receivers.
What does that have to do with room correction measurements done with a mic? This was in reference to what SPL to use for calculating correction filters.
 
Joined
Sep 16, 2019
Messages
85
Likes
68
#20
@TimVG No disagreement here. I don't do Facebook. I did mention in my post to keep eq applied below 500 Hz ;) But folks should realize that sophisticated DSP correction systems like Audiolense and Acourate use frequency dependent windowing (FDW) that users can control so only below 500 Hz the direct and reflected sound is being modified, but beyond that, only the direct sound that is being modified, like a tilting tone control. So no issues. Also some of us use constant directivity waveguides, aside from requiring eq compensation, the whole point with constant directivity if you eq on axis, you already know what the off axis response is going to be :) Wrt to room correction, I really like @j_j 's presentation on Acoustic and Psychoacoustic Issues in Room Correction.

Part of the issue is that there are so few loudspeakers designed using ANSI/CTA 2034-A standard. I have mentioned elsewhere that a single on axis flat response in an anechoic chamber doth not make a good loudspeaker as it does not account for the off axis response. But if you have not read the Sean's presentations (and AES papers) I linked to above, it is worth a read as there is a preferred target response for both loudspeakers and headphones. Maybe one day more manufacturers will start using the 2034 standard.

@MZKM Yes, you want to calibrate between 75 dB and 85 dB SPL C weighting and slow integration at the LP. Relatively speaking, our ears are the most flat through this range. As an ex recording/mixing engineer, in the industry, most final mixes and masters are mixed at reference level (i.e. 83 dB SPL) for setting the bass level relative to the treble. Our ears are nonlinear and the frequency response changes with SPL. I use a loudness control in JRiver Media Center and calibrate the SPL so that below 83 dB SPL it kicks in to compensate for our dropping sensitivity in the bass as the level goes down. Works a treat. For more on the magic of 83 db SPL see Bob Katz's article: https://www.digido.com/portfolio-item/level-practices-part-2/
Hi Mitch

Have you ever tried correction in Audiolense or Acourate, and making some (semi-)anechoic mesurements of the speaker(s) afterwards? It would give some insight to how valuable in the in-room method is compared to correcting based on true anechoic information.

Bw
 
Top Bottom