• WANTED: Happy members who like to discuss audio and other topics related to our interest. Desire to learn and share knowledge of science required. There are many reviews of audio hardware and expert members to help answer your questions. Click here to have your audio equipment measured for free!

Speaker comparison, making meaningful room measurements?

txbdan

Active Member
Joined
Apr 21, 2020
Messages
213
Likes
198
Hi all,

I currently have 9 pairs of bookshelf speakers in the basement and i'm conducting a little shootout in order to find which I like best. As part of the journey, I've decided to pick up the miniDSP USB mic and perform some measurements using REW to help objectify things and help map what I hear to what i measure.

Here's an example of three in-room measurements. The mic is horizontal at tweeter level, at listening position about 8ft away. I'm in my basement which is a decently large room and seems pretty acoustically dead. insulation on the ceiling, foam tiles on the floor, and racks of drying clothes behind me, and lots of crap in general. :)

The three speakers here are Elac Debut 2.0 b5.2, Ascend CBM-170 SE, and Q Acoustic 3020i. By ear, the Elacs sound flat, the Ascends seem hyped more trebly, and the 3020is are somewhere in the middle.

may6_500pm_comparisonlegend.jpg


Questions:
  1. I see the room modes down below 300Hz. Should i be focusing mainly on 300Hz and up?
  2. How much smoothing should be used?
  3. The differences are there, but somewhat subtle considering how different these speakers sound. I guess i can see there's more going on between 2-3k with the Ascends, for example. Is this degree of nuance about right?
  4. Should i attempt to "normalize" the levels by using the dB offset in the graph options to account for speaker sensitivity?
  5. Should i attempt to use gating to minimize room interaction?
  6. Does this approach in general make sense? What should i do better?
Thanks!
 

richard12511

Major Contributor
Forum Donor
Joined
Jan 23, 2020
Messages
4,335
Likes
6,700
Hi all,

I currently have 9 pairs of bookshelf speakers in the basement and i'm conducting a little shootout in order to find which I like best. As part of the journey, I've decided to pick up the miniDSP USB mic and perform some measurements using REW to help objectify things and help map what I hear to what i measure.

Here's an example of three in-room measurements. The mic is horizontal at tweeter level, at listening position about 8ft away. I'm in my basement which is a decently large room and seems pretty acoustically dead. insulation on the ceiling, foam tiles on the floor, and racks of drying clothes behind me, and lots of crap in general. :)

The three speakers here are Elac Debut 2.0 b5.2, Ascend CBM-170 SE, and Q Acoustic 3020i. By ear, the Elacs sound flat, the Ascends seem hyped more trebly, and the 3020is are somewhere in the middle.

View attachment 62317

Questions:
  1. I see the room modes down below 300Hz. Should i be focusing mainly on 300Hz and up?
  2. How much smoothing should be used?
  3. The differences are there, but somewhat subtle considering how different these speakers sound. I guess i can see there's more going on between 2-3k with the Ascends, for example. Is this degree of nuance about right?
  4. Should i attempt to "normalize" the levels by using the dB offset in the graph options to account for speaker sensitivity?
  5. Should i attempt to use gating to minimize room interaction?
  6. Does this approach in general make sense? What should i do better?
Thanks!

Use the RTA in REW while generating pink noise and moving the mic all around the listening area to generate an average. One mic position measurements are not very reliable.
 
OP
T

txbdan

Active Member
Joined
Apr 21, 2020
Messages
213
Likes
198
Interesting, i'll give that a shot. I'd imagine i'd have to do a controlled set of movements to keep the comparison controlled across the different speakers. Also, for the in-room measurement, i'm using both speakers (L & R). Is that right? I did do some single speaker measurements at 1m. Would that be better for comparison? Maybe it'd excite the room less.
 

QMuse

Major Contributor
Joined
Feb 20, 2020
Messages
3,124
Likes
2,785
Hi all,

I currently have 9 pairs of bookshelf speakers in the basement and i'm conducting a little shootout in order to find which I like best. As part of the journey, I've decided to pick up the miniDSP USB mic and perform some measurements using REW to help objectify things and help map what I hear to what i measure.

Here's an example of three in-room measurements. The mic is horizontal at tweeter level, at listening position about 8ft away. I'm in my basement which is a decently large room and seems pretty acoustically dead. insulation on the ceiling, foam tiles on the floor, and racks of drying clothes behind me, and lots of crap in general. :)

The three speakers here are Elac Debut 2.0 b5.2, Ascend CBM-170 SE, and Q Acoustic 3020i. By ear, the Elacs sound flat, the Ascends seem hyped more trebly, and the 3020is are somewhere in the middle.

View attachment 62317

Questions:
  1. I see the room modes down below 300Hz. Should i be focusing mainly on 300Hz and up?
  2. How much smoothing should be used?
  3. The differences are there, but somewhat subtle considering how different these speakers sound. I guess i can see there's more going on between 2-3k with the Ascends, for example. Is this degree of nuance about right?
  4. Should i attempt to "normalize" the levels by using the dB offset in the graph options to account for speaker sensitivity?
  5. Should i attempt to use gating to minimize room interaction?
  6. Does this approach in general make sense? What should i do better?
Thanks!

Those 3 FR curves are so similar so I think you would have a really hard time recognising them in a blind test.
 
OP
T

txbdan

Active Member
Joined
Apr 21, 2020
Messages
213
Likes
198
Those 3 FR curves are so similar so I think you would have a really hard time recognising them in a blind test.

Yeh, that's kinda what's confusing me. The speakers sound quite different. I'm even adjusting the volume per their sensitivity measured with a pink noise tone when listening. Do the differences i hear really look so small on a plot? Or are my measurements not accurate? I don't have enough experience to know what a 3dB bump at 2kHz sounds like for example.
 

hardisj

Major Contributor
Reviewer
Joined
Jul 18, 2019
Messages
2,907
Likes
13,908
Location
North Alabama
Yeh, that's kinda what's confusing me. The speakers sound quite different. I'm even adjusting the volume per their sensitivity measured with a pink noise tone when listening. Do the differences i hear really look so small on a plot? Or are my measurements not accurate? I don't have enough experience to know what a 3dB bump at 2kHz sounds like for example.

The differences will be more than just the single mic measurement. The RTA doesn't easily show the 'spatial' characteristics of what you are hearing.

Also, your scale is 10dB. That's why the contours look similar. Having said that, with that knowledge, it changes perception: the graphs don't look alike. The green vs blue result is nearly 5dB apart > 600Hz. I suggest changing to 5dB scale on the y-axis.
 

thewas

Master Contributor
Forum Donor
Joined
Jan 15, 2020
Messages
6,741
Likes
16,175
As said a single frequency response in a room doesn't tell much as for example if you move your mic for just few millimeters you can measure a significantly different one although when we move our head so our tonal sound perception doesn't really change, an omnidirectional microphone "hears" very differently than a ears and a brain. That's why as written above for listening position measurements more descriptive are several measurements in a volume around our position or the moving microphone method. But also even in that case you could get 2 averaged frequency responses that are almost identical but sound different to us because we just measure the sum of sound coming from all directions which our ears and brain though weigh differently for different angles.
 

NTK

Major Contributor
Forum Donor
Joined
Aug 11, 2019
Messages
2,656
Likes
5,819
Location
US East
Frequency response measurements in a room is not a reliable indicator of speaker quality. What you measure with a mic in a room is a complex summation of direct and reflected sounds.

That is why Amir spent over 100k USD for the NFS (and other build anechoic chambers) to get anechoic data and to generate the spinorama curves. You cannot deconvolute in-room measurements to get the directivities of the speakers, and directivities is a major determinating factor of the sound quality of a speaker--if you believe in Dr Toole's research.
 

hardisj

Major Contributor
Reviewer
Joined
Jul 18, 2019
Messages
2,907
Likes
13,908
Location
North Alabama
You might want to have a look at this video I made a couple months ago which briefly touches on why you need to make a "moving mic measurement" to give you the proper response at your listening position. Not saying you need to go build this. But it will give you some insight on the benefit and the idea.

 

QMuse

Major Contributor
Joined
Feb 20, 2020
Messages
3,124
Likes
2,785
Yeh, that's kinda what's confusing me. The speakers sound quite different. I'm even adjusting the volume per their sensitivity measured with a pink noise tone when listening. Do the differences i hear really look so small on a plot? Or are my measurements not accurate?

The difference you're hearing is more related to the directivity characteristics of those speakers which are not represented in your measurements. Some of them spread sound more widely than others and sme of them are doing that spread more smoothly than others. Directivity characteristic is visible from Early reflections and power sound responses in spinorama chart, your in-room response is pretty much direct on-axis sound mixed with some reflections, so you cannot see it there.
 
OP
T

txbdan

Active Member
Joined
Apr 21, 2020
Messages
213
Likes
198
Frequency response measurements in a room is not a reliable indicator of speaker quality. What you measure with a mic in a room is a complex summation of direct and reflected sounds.

That is why Amir spent over 100k USD for the NFS (and other build anechoic chambers) to get anechoic data and to generate the spinorama curves. You cannot deconvolute in-room measurements to get the directivities of the speakers, and directivities is a major determinating factor of the sound quality of a speaker--if you believe in Dr Toole's research.

Certainly. To be clear, my goal isn't to accurately measure the speaker, its to put some data behind what i hear. In other words, everything is relative so depending on which two speakers i'm comparing, or what i ate for breakfast that day, my opinions seem to change. hah I'm trying to find some correlations between my opinions and reality. Basically trying to check myself.
 
OP
T

txbdan

Active Member
Joined
Apr 21, 2020
Messages
213
Likes
198
You might want to have a look at this video I made a couple months ago which briefly touches on why you need to make a "moving mic measurement" to give you the proper response at your listening position. Not saying you need to go build this. But it will give you some insight on the benefit and the idea.


Ha, great idea.
 

QMuse

Major Contributor
Joined
Feb 20, 2020
Messages
3,124
Likes
2,785
hah I'm trying to find some correlations between my opinions and reality. Basically trying to check myself.

You cannot do that with in-room measurements unless some of the speakers would show non-linear behaviour. Even without EQ all these 3 speakers are showing similar FR so the difference is in their directivity patterns which cannot be measured in-room.
 
OP
T

txbdan

Active Member
Joined
Apr 21, 2020
Messages
213
Likes
198
You cannot do that with in-room measurements unless some of the speakers would show non-linear behaviour. Even without EQ all these 3 speakers are showing similar FR so the difference is in their directivity patterns which cannot be measured in-room.

I see. So that's why you need to measure multiple positions and average them or use the RTA moving method? Which i guess is how measurement for room correction work? I've only used Audessey before.

Say i had a MiniDSP or something and want to measure the room so i could create a compensation EQ to meet some target response. If i did that with different speakers, i'd get different compensation EQs, right?
 

QMuse

Major Contributor
Joined
Feb 20, 2020
Messages
3,124
Likes
2,785
I see. So that's why you need to measure multiple positions and average them or use the RTA moving method? Which i guess is how measurement for room correction work?

No. What I meant was that you cannot measure directivity characteristics of the speaker in room unless you have equipment like Klippel NFS.

Say i had a MiniDSP or something and want to measure the room so i could create a compensation EQ to meet some target response. If i did that with different speakers, i'd get different compensation EQs, right?

In the 20-400Hz range room dominates the response so all 3 speakers would have similar measurements. In the 400-20kHz range response is dominated by the speaker so diferences among them would be shown clearly.

In my opinion it is possible to make quite decent room EQ curve (in the 20-400Hz range) that would work well for a number of speakers in the same room.
 
Top Bottom