• 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!

Eq-ing Concepts and Biquad Understanding

arvidb

Member
Joined
Dec 1, 2022
Messages
96
Likes
114
Location
Sweden
Wow, this was a confusing thread! This is my understanding so far (and I'm far from being an expert so please correct me and/or fill in any details I've missed):

I think there are two kinds of measurements being confused in this thread: nearfield measurements of the speaker itself, and measurements for room correction.

1) Speaker measurements: This is done nearfield (e.g. at 1 m distance from the speaker) and the purpose is to correct the speaker itself over its entire frequency range. Mainly done by the speaker manufacturer (or the DIY:er). The measurement must be done in an anechoic chamber, with a Klippel scanner, or using time gating while having the speaker suspended far enough from all reflecting surfaces. The results could be used to improve the design (time aligning drivers, adjusting directivity), design crossovers, and to get the overall response curve for the speaker (to be applied later; you can find such data here for many commercial speakers).

... and, once you have a speaker that sounds good all by itself:

2) Room correction: Measurements done at the listening position(s) with the speaker in its intended position in the room. Measurements are averaged over multiple positions* (e.g. using the moving microphone method) and only low frequency peaks are corrected.

Why only peaks? ―Because the nulls (troughs) cannot be corrected: if you increase the output at those frequencies, the reflected waves that cancel out the sound will also increase in amplitude and you will still get the nulls (together with a lot more distortion from your poor speakers).

Why only low frequencies? ―Because our ears and brain are very good at resolving direction of the sound at higher frequencies; in essence our brain already does this filtering for us. The speakers themselves already sound good, remember? While it is possible to equalise the average, omnidirectional room response at higher frequencies as well, your speakers won't sound right if you do!

(Also, placing the target level in REW below the average level – much below the level calculated by "Calculate target level from response" – to get rid of the troughs is at best pointless: you will compensate by increasing volume later, so it is equivalent to trying to boost the troughs. Except now you also have filters all over the place doing who-knows-what to phase etc...

*) It's also possible to do more advanced room correction taking phase into account and using FIR filters (i.e. not biquads). This requires frequency-dependent gated sweep measurements, I think, and might also be able to actively correct room mode ringing? See e.g. Barnett, "State-of-the-Art of Digital Room Correction" (and watch the video linked in part 2, if you are really interested).
 

levimax

Major Contributor
Joined
Dec 28, 2018
Messages
2,348
Likes
3,462
Location
San Diego
Wow, this was a confusing thread! This is my understanding so far (and I'm far from being an expert so please correct me and/or fill in any details I've missed):

I think there are two kinds of measurements being confused in this thread: nearfield measurements of the speaker itself, and measurements for room correction.

1) Speaker measurements: This is done nearfield (e.g. at 1 m distance from the speaker) and the purpose is to correct the speaker itself over its entire frequency range. Mainly done by the speaker manufacturer (or the DIY:er). The measurement must be done in an anechoic chamber, with a Klippel scanner, or using time gating while having the speaker suspended far enough from all reflecting surfaces. The results could be used to improve the design (time aligning drivers, adjusting directivity), design crossovers, and to get the overall response curve for the speaker (to be applied later; you can find such data here for many commercial speakers).

... and, once you have a speaker that sounds good all by itself:

2) Room correction: Measurements done at the listening position(s) with the speaker in its intended position in the room. Measurements are averaged over multiple positions* (e.g. using the moving microphone method) and only low frequency peaks are corrected.

Why only peaks? ―Because the nulls (troughs) cannot be corrected: if you increase the output at those frequencies, the reflected waves that cancel out the sound will also increase in amplitude and you will still get the nulls (together with a lot more distortion from your poor speakers).

Why only low frequencies? ―Because our ears and brain are very good at resolving direction of the sound at higher frequencies; in essence our brain already does this filtering for us. The speakers themselves already sound good, remember? While it is possible to equalise the average, omnidirectional room response at higher frequencies as well, your speakers won't sound right if you do!

(Also, placing the target level in REW below the average level – much below the level calculated by "Calculate target level from response" – to get rid of the troughs is at best pointless: you will compensate by increasing volume later, so it is equivalent to trying to boost the troughs. Except now you also have filters all over the place doing who-knows-what to phase etc...

*) It's also possible to do more advanced room correction taking phase into account and using FIR filters (i.e. not biquads). This requires frequency-dependent gated sweep measurements, I think, and might also be able to actively correct room mode ringing? See e.g. Barnett, "State-of-the-Art of Digital Room Correction" (and watch the video linked in part 2, if you are really interested).
Sometimes troughs can be filled.... the only way to tell is by testing... add 6 dB boost at trough and measure again... if no change then forget boost and try moving things around... if boost fills trough then no reason not to limited by headroom of system.
 

Holmz

Major Contributor
Joined
Oct 3, 2021
Messages
2,018
Likes
1,241
Location
Australia
Wow, this was a confusing thread! This is my understanding so far (and I'm far from being an expert so please correct me and/or fill in any details I've missed):

I think there are two kinds of measurements being confused in this thread: nearfield measurements of the speaker itself, and measurements for room correction.

1) Speaker measurements: This is done nearfield (e.g. at 1 m distance from the speaker) and the purpose is to correct the speaker itself over its entire frequency range. Mainly done by the speaker manufacturer (or the DIY:er). The measurement must be done in an anechoic chamber, with a Klippel scanner, or using time gating while having the speaker suspended far enough from all reflecting surfaces. The results could be used to improve the design (time aligning drivers, adjusting directivity), design crossovers, and to get the overall response curve for the speaker (to be applied later; you can find such data here for many commercial speakers).

... and, once you have a speaker that sounds good all by itself:

2) Room correction: Measurements done at the listening position(s) with the speaker in its intended position in the room. Measurements are averaged over multiple positions* (e.g. using the moving microphone method) and only low frequency peaks are corrected.

Why only peaks? ―Because the nulls (troughs) cannot be corrected: if you increase the output at those frequencies, the reflected waves that cancel out the sound will also increase in amplitude and you will still get the nulls (together with a lot more distortion from your poor speakers).

Why only low frequencies? ―Because our ears and brain are very good at resolving direction of the sound at higher frequencies; in essence our brain already does this filtering for us. The speakers themselves already sound good, remember? While it is possible to equalise the average, omnidirectional room response at higher frequencies as well, your speakers won't sound right if you do!

(Also, placing the target level in REW below the average level – much below the level calculated by "Calculate target level from response" – to get rid of the troughs is at best pointless: you will compensate by increasing volume later, so it is equivalent to trying to boost the troughs. Except now you also have filters all over the place doing who-knows-what to phase etc...

Tak !


*) It's also possible to do more advanced room correction taking phase into account and using FIR filters (i.e. not biquads). This requires frequency-dependent gated sweep measurements, I think, and might also be able to actively correct room mode ringing? See e.g. Barnett, "State-of-the-Art of Digital Room Correction" (and watch the video linked in part 2, if you are really interested).

I thought that a gated measurement was almost certainly NOT correcting the room, and gating out the speaker, inside of a room.
 

FrantzM

Major Contributor
Forum Donor
Joined
Mar 12, 2016
Messages
4,337
Likes
7,733
Hi Allow me to be lazy. There is much material to digest in this thread. I promise to read it all.:facepalm:

M situation is more prosaic: I need to input the Harman Curve as a set of biquads to Audyssey MultEQ-X. How do I do that? I have the values / set of frequencies-dB pairs, e.g.:
0 Hz--------- 2.5 dB
5 Hz---------2.5 dB
8 Hz---------2.5 dB
16 Hz--------6.6
...
...
...
20 KHz ------ -7.1 dB

That would be quite helpful. Thanks in advance.

Peace,
 

voodooless

Grand Contributor
Forum Donor
Joined
Jun 16, 2020
Messages
10,227
Likes
17,807
Location
Netherlands
Hi Allow me to be lazy. There is much material to digest in this thread. I promise to read it all.:facepalm:

M situation is more prosaic: I need to input the Harman Curve as a set of biquads to Audyssey MultEQ-X. How do I do that? I have the values / set of frequencies-dB pairs, e.g.:
0 Hz--------- 2.5 dB
5 Hz---------2.5 dB
8 Hz---------2.5 dB
16 Hz--------6.6
...
...
...
20 KHz ------ -7.1 dB

That would be quite helpful. Thanks in advance.

Peace,
The wonky way: import the target curve in REW, and use the AutoEQ to generate the biquads.
 

arvidb

Member
Joined
Dec 1, 2022
Messages
96
Likes
114
Location
Sweden
I thought that a gated measurement was almost certainly NOT correcting the room, and gating out the speaker, inside of a room.
I re-watched part of the video (beginning at 46 minutes or so) and it seems he's using frequency-dependent windowing ("FDW") to include room response at low frequencies and try to exclude it at high frequencies. So in essence he's measuring for both room correction and speaker correction simultaneously. Since the speaker is not in a quasi-anechoic environment I don't know how well it will work; a kind of "poor man's quasi-anechoic measurement" perhaps? (Plus the bonus low frequency room response.) Edit: I really don't know nearly enough to make any judgements about this - anyone who reads please do your own research.

FDW seems to perhaps be unrelated to the "excess-phase correction" he's also doing (i.e. correcting for ringing using FIR filters). I can't say I understand the method described in the video or how/if FDW comes into play during filter generation. Sorry for the continued confusion. :)

The whole thing seems contentious: turns out he's a member here and his thread about this got locked after 17 pages of discussion. :)
 
Last edited:

levimax

Major Contributor
Joined
Dec 28, 2018
Messages
2,348
Likes
3,462
Location
San Diego
I re-watched part of the video (beginning at 46 minutes or so) and it seems he's using frequency-dependent windowing ("FDW") to include room response at low frequencies and try to exclude it at high frequencies. So in essence he's measuring for both room correction and speaker correction simultaneously. Since the speaker is not in a quasi-anechoic environment I don't know how well it will work; a kind of "poor man's quasi-anechoic measurement" perhaps (plus the bonus low frequency room response).

FDW seems to perhaps be unrelated to the "excess-phase correction" he's also doing (i.e. correcting for ringing using FIR filters). I can't say I understand the method described in the video or how/if FDW comes into play during filter generation. Sorry for the continued confusion. :)

The whole thing seems contentious: turns out he's a member here and his thread about this got locked after 17 pages of discussion.
Earlier I recomended MMM exactly because you don't have to worry about windows, or averaging, or smoothing (I do var before corrections but the MMM measurements are quite smooth to begin with). I always get exactly the same FR meaurements as I so with averaged windowed sweeps .... only difference is it takes 3 minutes instead of 3 hours. MMM is a great place to start for FR EQ.
 
OP
Trdat

Trdat

Addicted to Fun and Learning
Forum Donor
Joined
Sep 6, 2019
Messages
967
Likes
396
Location
Yerevan "Sydney Born"
The wonky way: import the target curve in REW, and use the AutoEQ to generate the biquads.
When you say wonky? Are you implying the dodgy way or because of its difficulty of uploading biquads in manually into the source?
 
OP
Trdat

Trdat

Addicted to Fun and Learning
Forum Donor
Joined
Sep 6, 2019
Messages
967
Likes
396
Location
Yerevan "Sydney Born"
(Also, placing the target level in REW below the average level – much below the level calculated by "Calculate target level from response" – to get rid of the troughs is at best pointless: you will compensate by increasing volume later, so it is equivalent to trying to boost the troughs. Except now you also have filters all over the place doing who-knows-what to phase etc...
Are you recommending that the target level not be to far below the measurement?
*) It's also possible to do more advanced room correction taking phase into account and using FIR filters (i.e. not biquads). This requires frequency-dependent gated sweep measurements, I think, and might also be able to actively correct room mode ringing? See e.g. Barnett, "State-of-the-Art of Digital Room Correction" (and watch the video linked in part 2, if you are really interested).
But Audiolense just measures from the MLP does Audiolense perform a frequency dependant gated sweep in its correction process before adding FIR's? I use Audiolense I should know, lol but I dont understand the theory behind it.
 

voodooless

Grand Contributor
Forum Donor
Joined
Jun 16, 2020
Messages
10,227
Likes
17,807
Location
Netherlands
When you say wonky? Are you implying the dodgy way or because of its difficulty of uploading biquads in manually into the source?
Wonky because it's kind of a curious way of doing it. But it does work, I just helped @FrantzM do it. The REW Auto-EQ is horrible though... It works much better by hand.
 
OP
Trdat

Trdat

Addicted to Fun and Learning
Forum Donor
Joined
Sep 6, 2019
Messages
967
Likes
396
Location
Yerevan "Sydney Born"
Wonky because it's kind of a curious way of doing it. But it does work, I just helped @FrantzM do it. The REW Auto-EQ is horrible though... It works much better by hand.
Auto Eq, you mean the automated process when you click "match response to target" or when you use a the file to add into the DSP?

I matched response to target and manually entered the biquads in the Hypex design filter and it wasn't bad. Are you suggesting there is a better way to appraoch this compared to pptakai's instructions?
 

voodooless

Grand Contributor
Forum Donor
Joined
Jun 16, 2020
Messages
10,227
Likes
17,807
Location
Netherlands
Auto Eq, you mean the automated process when you click "match response to target" or when you use a the file to add into the DSP?
Yes, I mean the match response to the target thingy.
I matched response to target and manually entered the biquads in the Hypex design filter and it wasn't bad. Are you suggesting there is a better way to appraoch this compared to pptakai's instructions?
No, I don't know of a better way.
 
OP
Trdat

Trdat

Addicted to Fun and Learning
Forum Donor
Joined
Sep 6, 2019
Messages
967
Likes
396
Location
Yerevan "Sydney Born"
Yes, I mean the match response to the target thingy.

No, I don't know of a better way.
Sorry a little confused if the wonky way is match response to target than the better way should be the one you mentioned of being by hand. You said that you helped FranzM and it worked okay.

How would you do it by hand? To figure out exactly where a biquad is necessary and then add a biquad manually?
 

voodooless

Grand Contributor
Forum Donor
Joined
Jun 16, 2020
Messages
10,227
Likes
17,807
Location
Netherlands
How would you do it by hand? To figure out exactly where a biquad is necessary and then add a biquad manually?
I used the magnificent concept of eyeballing ;)
 
OP
Trdat

Trdat

Addicted to Fun and Learning
Forum Donor
Joined
Sep 6, 2019
Messages
967
Likes
396
Location
Yerevan "Sydney Born"
So, I took measurement and gated it and then attempted the EQ in REW and later took another measurement from the MLP. Is this process an option(taking two different measurements) or is it still better to choose one of the options(MLP vs 1m gated)?

And is the MMM method considered a MLP measurement without gating yes...?
 
Top Bottom