• 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!

How to roughly correct room response with an uncalibrated mic.

kokoko3k

Member
Joined
Nov 29, 2021
Messages
8
Likes
4
This is not a way to make your room magically sound good, but if you're fighting with a bunch of standing waves and are not able to properly cut them by ear, this will be useful.

Even if I was waiting for a Sonarworks Xref20R4 to arrive, i was too impatient, and yesterday an idea came to my mind.
What if I find the sweet spot of the room?
Can I derive the right equalization by the difference between it and the real listening point, given that all the other variables (room,speakers,mic itself) remain the same?

Of course you need at least a mic that is able to capture the frequencies to be corrected with a good SNR, this was indeed my case.

So, what i did was:
play and record a sine sweep in the sweet spot -> sweet_sine.wav
play and record a sine sweep in the bad spot -> bad_sine.wav
Imported the wavs in audacity and normalized them to 0.0dB

At this point maybe rew could do the work by itself by subtracting responses an equalizing the resulting one, but i wasn't able to do that, but one thing crossed my mind, the software i was looking for was AutoEq
Even if it was meant to correct headphones/earbuds, it has a nice feature that given two measurements, is able to make an headphone sound like another.

I've exploited that function to make my bad spot to sound like my sweet spot.

So, I imported the wavs in RoomEqWizard and exported the measurement as text csv, comma separated fields, frequancyt step: 0.09, no smoothing at all:
sweet_sine.wav -> /koko/sweeps/sweet_sine.csv
bad_sine.wav ->/koko/sweeps/bad_sine/bad_sine.csv

AutoEq is surprisingly able to read the csv exported by rev with no modifications, given that you export as comma as field separator.
Given that compensation/zero.csv is a file provided by AutoEq when youi don't need to compensate anything, in the end the command given was:
Code:
python ./autoeq.py --input_dir  /koko/sweeps/bad_sine/bad_sine.csv --output_dir /koko/sweeps/MyResults --sound_signature /koko/sweeps/sweet_sine.csv --show_plot --equalize --compensation compensation/zero.csv --parametric_eq
Which produced the following:
Code:
Preamp: -6.5 dB
Filter 1: ON PK Fc 16 Hz Gain 7.1 dB Q 0.85
Filter 2: ON PK Fc 68 Hz Gain -20.4 dB Q 1.61
Filter 3: ON PK Fc 155 Hz Gain -12.9 dB Q 2.04
Filter 4: ON PK Fc 280 Hz Gain -12.4 dB Q 4.96
Filter 5: ON PK Fc 625 Hz Gain -15.2 dB Q 0.96
Filter 6: ON PK Fc 664 Hz Gain 11.4 dB Q 6.58
Filter 7: ON PK Fc 1592 Hz Gain -11.9 dB Q 2.77
Filter 8: ON PK Fc 1874 Hz Gain 3.8 dB Q 3.62
Filter 9: ON PK Fc 2241 Hz Gain -10.5 dB Q 2.22
Filter 10: ON PK Fc 2347 Hz Gain 6.5 dB Q 1.38
Filter 11: ON PK Fc 3645 Hz Gain 4.1 dB Q 4.53
Filter 12: ON PK Fc 4588 Hz Gain -12.9 dB Q 2.60
Filter 13: ON PK Fc 5154 Hz Gain 7.3 dB Q 2.85
Filter 14: ON PK Fc 5647 Hz Gain 6.8 dB Q 1.39
Filter 15: ON PK Fc 5816 Hz Gain -4.3 dB Q 2.15
Filter 16: ON PK Fc 16565 Hz Gain -0.2 dB Q 2.15
Filter 17: ON PK Fc 17596 Hz Gain 6.1 dB Q 0.11

I admit, i've had to strip equalizations at higher frequencies, because they were simply wrong, probably because the SNR of the microphone wasn't good enough and i couldn't push the volume higher at the time i made the measurement (also, i discovered an error in the measurement of bad_sine.wav, probably the microphone moved during the sweep!)
I've also corrected the "power" of the equalization by dividing all the cuts by a factor of two; it was just to extreme, can't say why.
But at least i was able to identify the center frequencies of the standing waves and their widths, which was exactly what i need.

Hoping this could be helpful for someone else, it took me the entire day :)
 
Last edited:

f1shb0n3

Senior Member
Forum Donor
Joined
Aug 29, 2020
Messages
305
Likes
459
Location
Seattle Area
Your approach to make the "bad" seat sound like the "good" one will only work if your speakers have exceptional directivity and both seats are fairly close. Your "good" seat has issues to fix too probably with room modes and such, so definitely consider getting a calibrated mic now that you've learned the basics and got that far. UMIK-1 is not an expensive investment at $80 from MiniDSP or a third-party measured and calibrated one from Cross-Spectrum Labs for $110.
Calibrated measuring mic is one of the most important investments in objective audio with the highest ROI. It gets you improvements that can be actually heard and easily A/B tested.
 

f1shb0n3

Senior Member
Forum Donor
Joined
Aug 29, 2020
Messages
305
Likes
459
Location
Seattle Area
If you are interested in exploring how far an uncalibrated mic can take you here's an idea - calibrate yours. Borrow a calibrated mic from someone, measure a sine with your mic and the calibrated one at the same position and using your method generate a calibration curve for your mic referenced to the calibrated measurement. No idea if it will work to "calibrate" your mic, but this is my understanding of how mic calibration is done in the first place.
 

dominikz

Senior Member
Forum Donor
Joined
Oct 10, 2020
Messages
487
Likes
1,693
If you are interested in exploring how far an uncalibrated mic can take you here's an idea - calibrate yours. Borrow a calibrated mic from someone, measure a sine with your mic and the calibrated one at the same position and using your method generate a calibration curve for your mic referenced to the calibrated measurement. No idea if it will work to "calibrate" your mic, but this is my understanding of how mic calibration is done in the first place.
Perhaps it will also be interesting to see this post, comparing in-room measurements made with a Rode NT2A large-diaphragm condenser in omni mode vs those made with a Cross-Spectrum Labs calibrated Dayton EMM-6 measurement microphone.
In short - in my experience condenser omni mics will give pretty solid results for low-frequency in-room measurements even if not calibrated. Of course, you can't be a sure your measurements are valid until you compare them to those made with a calibrated mic.
 
OP
K

kokoko3k

Member
Joined
Nov 29, 2021
Messages
8
Likes
4
Wops, i missed the replies.
f1shn0n3: I was waiting for the Sonarworks Xref20R4 to arrive and now it did, and I repeated the measurements with it.
I have the old and the new resulting equalizations (the light blue line in the screenshots)

This is the resulting equalization from the measurement made with the xref20r4:

a1.xref.png


This is the resulting equalization from the "difference" between the good spot and the bad spot (gains are globally tweaked manually):

a1.autoeq.png


Don't mind the last 3 filters in the higher frequencies, because i copied them manually from the first because they sounded right to my ear.
Given that, something still matches closely while something else is completely off.

Good seat recording were made in the equilateral triangle vertex, bad seat recording were made near the wall, 2mt away from the good spot.
It is entirely possible that some room mode was present in both of the recording.

f1shb0n3 said:
Your approach to make the "bad" seat sound like the "good" one will only work if your speakers have exceptional directivity and both seats are fairly close.
Thanks, care to explain further that?
 
Last edited:

Blumlein 88

Grand Contributor
Forum Donor
Joined
Feb 23, 2016
Messages
14,625
Likes
23,695
What microphones do you have available for you to use?
 
OP
K

kokoko3k

Member
Joined
Nov 29, 2021
Messages
8
Likes
4
What microphones do you have available for you to use?
Sonarworks Xref20R4 arrived and made the new measurements and correction,and im satisfied by now.
When i had the idea of making the bad spot to sound like the good one (first post) i used a Samson Q7
 

f1shb0n3

Senior Member
Forum Donor
Joined
Aug 29, 2020
Messages
305
Likes
459
Location
Seattle Area
Wops, i missed the replies.
f1shn0n3: I was waiting for the Sonarworks Xref20R4 to arrive and now it did, and I repeated the measurements with it.
I have the old and the new resulting equalizations (the light blue line in the screenshots)

This is the resulting equalization from the measurement made with the xref20r4:

View attachment 179625


This is the resulting equalization from the "difference" between the good spot and the bad spot (gains are globally tweaked manually):

View attachment 179626


Don't mind the last 3 filters in the higher frequencies, because i copied them manually from the first because they sounded right to my ear.
Given that, something still matches closely while something else is completely off.

Good seat recording were made in the equilateral triangle vertex, bad seat recording were made near the wall, 2mt away from the good spot.
It is entirely possible that some room mode was present in both of the recording.


Thanks, care to explain further that?
These graphs are non-standard and not readable for me. Have you tried measuring with REW?

The reason why you can't translate "good" seat to "bad" seat is because speakers don't project even tonality off-axis, as seen on Amir's measurements, as physics of cone drivers work. On top of that the room interacts differently at different seats, making measurements even less correlated.

Btw I wonder why would you want to do this in the first place - you found a sweet spot right, make it the listening position :)

You can get very far in improving your system by ear - repositioning, toe-in, height, a little room treatment to remove echo, bass/treble controls, etc. Once I do all that I like to use room correction systems for final adjustment and customization - Denon's Audyssey XT32 is good, but I got best results with Dirac Live for stereo. These would remove room mode effects and allow you to customize tonality to your liking - boost bass, reduce treble, whatever you prefer and can easily test and compare.
 
OP
K

kokoko3k

Member
Joined
Nov 29, 2021
Messages
8
Likes
4
These graphs are non-standard and not readable for me. Have you tried measuring with REW?

The reason why you can't translate "good" seat to "bad" seat is because speakers don't project even tonality off-axis, as seen on Amir's measurements, as physics of cone drivers work. On top of that the room interacts differently at different seats, making measurements even less correlated.

Btw I wonder why would you want to do this in the first place - you found a sweet spot right, make it the listening position :)

You can get very far in improving your system by ear - repositioning, toe-in, height, a little room treatment to remove echo, bass/treble controls, etc. Once I do all that I like to use room correction systems for final adjustment and customization - Denon's Audyssey XT32 is good, but I got best results with Dirac Live for stereo. These would remove room mode effects and allow you to customize tonality to your liking - boost bass, reduce treble, whatever you prefer and can easily test and compare.
The graphs i posted are not measurements, but the correction curves i load into the equalizer when i listen.
The first one has been calculated by REW after the measurement made with the new Xref20R4
The second one has been calculated by AutoEq when i firstly tried to make the bad listening point to "sound" like the good one

And no, I cannot move the listening position, because in the night I listen to movies/series lying on the bed before going to sleep; however i found the sound to improve by keeping the speakers upside down so that the tweeter is exactly at ear level and by slightly angling the speakers to the ceiling.
Angling them to the ceiling means that they are not pointing to the ears, but maybe this helped in reducing reflections?
Repositioning and angling has been completely done by ear; I don't know how nor why , but it works.
 
Top Bottom