Eq-ing Concepts and Biquad Understanding

[Trdat]

I just want to get a few things down pat with using biquads, well with any Eq-ing. I understand the goal is to attenuate peaks and dips and most of the time we get these around the crossover over region. But, lets say you get good SPL support across the frequency response around the crossover point does the general frequency response apart from the crossover area and the roll off's need any Eq-ing?

Why I ask this is because once you take a measurement say from 1 metre or closer you get a tremendous amount of peaks and dips so where do you start?

Are we fixing every dip we see? How do we identify what needs Eq-ing?

Of course I am talking in the digital domain with DSP such as Hypex Design Filter.

 

[pierre]

Good question. Answer tonight after work.

 

[ppataki]

This thread might easily end up in a rabbit hole since I am sure people will have their own concept about this

I am sharing my personal experience here, not necessarily following the recommended ways of EQ-ing:
Since I have one-way systems for me it is imperative that I EQ the full range, from 20Hz to 20 000Hz.
That is going against the concept of just EQ-ing below the Schroeder frequency (usually 300-500Hz). I observed that the flatter the frequency response is, the better the system sounds (subjectively). And here flat means that all the peaks are eliminated for the full frequency range and also all the dips are filled above the Schroeder frequency.
Now what happens to the dips below the Schroeder frequency?
My strategy with those is the following: I set my target curve below the lowest dip and apply cutting EQ filters only throughout the entire frequency range.
Then I apply a second or even a third round of EQ-ing to further flatten the response (boosting filters - if needed - are then applied to above 300-500Hz only)

Now there is one comment I would like to add further: I have EQ-d rooms and systems where it absolutely did no harm whatsoever to boost the dips below the Schroeder frequency.....(that goes totally against all the recommendations btw)
So all in all I firmly believe this is up to experimentation - even if you do something that is against certain 'recommendations' it might actually end up sounding good
Once you are satisfied with your results then you can apply a low shelf filter to boost the lows and a high shelf filter to tame the highs - all up to taste
And then finally you might also consider a high-pass filter to limit the excursion of your drivers

 

[Trdat]

This thread might easily end up in a rabbit hole since I am sure people will have their own concept about this

I am sharing my personal experience here, not necessarily following the recommended ways of EQ-ing:
Since I have one-way systems for me it is imperative that I EQ the full range, from 20Hz to 20 000Hz.
That is going against the concept of just EQ-ing below the Schroeder frequency (usually 300-500Hz). I observed that the flatter the frequency response is, the better the system sounds (subjectively). And here flat means that all the peaks are eliminated for the full frequency range and also all the dips are filled above the Schroeder frequency.

That helps a lot. Appreciate the answer but if we take an example such as the one attached, I have posted a MW in a 3 way measured from 1 meter away with no delay just crossovers set. Crossove points are 400 and 2200hz.

Are you saying every single dip and peak needs to be attenuated according to the average SPL? Or are there certain parts of this frequency repsonse that would take priority? Or do we just use a low Q to attenuate a larger cross sectional area?

Can you see my point, I am trying to understand from such an archaic frequency response which is usually quite normal but when I say archaic from the point of view of an amatuer trying to work out where to start, what actual peak and dip do I EQ?

 

[Holmz]

That helps a lot. Appreciate the answer but if we take an example such as the one attached, I have posted a MW in a 3 way measured from 1 meter away with no delay just crossovers set. Crossove points are 400 and 2200hz.

Are you saying every single dip and peak needs to be attenuated according to the average SPL? Or are there certain parts of this frequency repsonse that would take priority? Or do we just use a low Q to attenuate a larger cross sectional area?

^Yeah^… lower Q

Can you see my point, I am trying to understand from such an archaic frequency response which is usually quite normal but when I say archaic from the point of view of an amatuer trying to work out where to start, what actual peak and dip do I EQ?

View attachment 247258

Usually one would start with the peaks on the left side, about where you wrist is if you are holding an iPad.
(Specifically the ones in the 20-200Hz range.)

And if there was some raising treble in the 5-15 kHz range, then there.

That 400-2.6k section you’ve shown looks a bit benign, and not worth a lot of fretting over, IME.

Additionally one needs to squint their eyes, or use a bit more smoothing, to go after then general trend… and not get too excited about the narrow dips.
Maybe a narrow peak that is massive would benefit, but in the general sense… it should be smooth, just not the fine nit-picking comb sense.

 

[ppataki]

That helps a lot. Appreciate the answer but if we take an example such as the one attached, I have posted a MW in a 3 way measured from 1 meter away with no delay just crossovers set. Crossove points are 400 and 2200hz.

Are you saying every single dip and peak needs to be attenuated according to the average SPL? Or are there certain parts of this frequency repsonse that would take priority? Or do we just use a low Q to attenuate a larger cross sectional area?

Can you see my point, I am trying to understand from such an archaic frequency response which is usually quite normal but when I say archaic from the point of view of an amatuer trying to work out where to start, what actual peak and dip do I EQ?

View attachment 247258

First, I would add smoothing to this curve - either Var smoothing or 1/12 octave smoothing (I always use the latter) and then I would apply EQ accordingly (designate a low enough target curve and then apply cutting)
Some others here would recommend applying a Frequency Dependent Window instead of smoothing - that is also an option to consider

 

[abdo123]

My strategy with those is the following: I set my target curve below the lowest dip and apply cutting EQ filters only throughout the entire frequency range.

That's what i used to do like two years ago, I've learned alot since then.

But again most of my speakers are bookshelves with well behaving frequency response from the get go.

 

[antcollinet]

Not an expert here by any means.

But why are you measuring at 1m? Surely you should be measuring from the listening position.

 

[Trdat]

Usually one would start with the peaks on the left side, about where you wrist is if you are holding an iPad.
(Specifically the ones in the 20-200Hz range.)

And if there was some raising treble in the 5-15 kHz range, then there.

That 400-2.6k section you’ve shown looks a bit benign, and not worth a lot of fretting over, IME.

Additionally one needs to squint their eyes, or use a bit more smoothing, to go after then general trend… and not get too excited about the narrow dips.
Maybe a narrow peak that is massive would benefit, but in the general sense… it should be smooth, just not the fine nit-picking comb sense.

This is great thanks.

So below the Schroeder frequency is important, which we know due to room modes, SBIR and general bass problems. There is a lot of information to go on how to handle those types of dips so I will look into that.

Regarding the 5-15k, below I attached the tweeter measurement. So essentially I just work the dip between 10k and 13k and add a house curve to the tweeter roll off to my taste?

On the MW, that is exactly what I was trying to work out. So essentially those narrow dips or general dips dont need eq-ing unless there is a major dip or larger troughs have I got this correctly?

And yes, I think a smoothing would help but what smoothing is good to locate points of eq?

 

[Trdat]

Not an expert here by any means.

But why are you measuring at 1m? Surely you should be measuring from the listening position.

I was measuring at 1m away to set delays, which is what I was recommened around 4 to 5 times the lenght of the diameter of the woofer.

Many experts still maintain measuring should be nearfield for correction while many others opt for listening position. I am used to the listening position due to my experience with Audiolense although this correction will be done in Hypex Design Filter.

I will also measure at listening position once I understand a little more, the above is just being used as an example to try and identify what kind of a peak and dip actually gets eq'd.

 

[Trdat]

First, I would add smoothing to this curve - either Var smoothing or 1/12 octave smoothing (I always use the latter) and then I would apply EQ accordingly (designate a low enough target curve and then apply cutting)
Some others here would recommend applying a Frequency Dependent Window instead of smoothing - that is also an option to consider

I am confused about designating a low enough target curve and then apply cutting? Can you explain...?

Great, I will smooth 1/12 and take another look.

 

[antcollinet]

I was measuring at 1m away to set delays, which is what I was recommened around 4 to 5 times the lenght of the diameter of the woofer.

Many experts still maintain measuring should be nearfield for correction while many others opt for listening position. I am used to the listening position due to my experience with Audiolense although this correction will be done in Hypex Design Filter.

I will also measure at listening position once I understand a little more, the above is just being used as an example to try and identify what kind of a peak and dip actually gets eq'd.

Near field is what you'd choose to "anechoically" correct speaker characteristics - but won't help with room equalisation - since room effects vary by position.

 

[ppataki]

I am confused about designating a low enough target curve and then apply cutting? Can you explain...?

Great, I will smooth 1/12 and take another look.

If you can post here the full response using 1/12 smoothing I can better explain what I mean
I fully agree with @tonycollinet I would also measure at MLP (main listening position)

 

[voodooless]

I think the starting point should be to do a good measurement. By default there is no proper windowing, so you'll also see room reflections in your measurement. Proper windowing will remove those (mostly). Averaging will also help get a smoothing result:

http://www.ohl.to/audio/downloads/MMM-moving-mic-measurement.pdf

 

[Trdat]

If you can post here the full response using 1/12 smoothing I can better explain what I mean
I fully agree with @tonycollinet I would also measure at MLP (main listening position)

Okay Ill take measurement at MLP and post.

 

[Trdat]

If you can post here the full response using 1/12 smoothing I can better explain what I mean
I fully agree with @tonycollinet I would also measure at MLP (main listening position)

Unfortunately my mic stand is out of action but I do have a measurement from 1 meter away I think it will do for an example.

By the way this is only TW and MW no woofer.

 

[levimax]

I like the MMM method at the listening position. I get the same result much faster and easier than trying to window or average multiple measurements.

 

[LTig]

Absolutely yes!

 

[Holmz]

This is great thanks.}
,,,

View attachment 247259

May be the 11.x kHz could use 2-3 of a low Q boost.
Whether the whole10-20 kHz want a flatter lift is hard to say.

Usually there would be a bit of a downward slope from 20 to 20K.

Unfortunately my mic stand is out of action but I do have a measurement from 1 meter away I think it will do for an example.

By the way this is only TW and MW no woofer.
View attachment 247382

The null at 40 and peak at 60, and null at 80 are room modes.
I would think cutting the 60 back 6dB would help.

 

[ppataki]

Unfortunately my mic stand is out of action but I do have a measurement from 1 meter away I think it will do for an example.

By the way this is only TW and MW no woofer.
View attachment 247382

Here is how I would approach this:

- in REW set the target curve to 78dB

the Target type you can set to None and you can add your high-pass filter at the very end of the process

- Set the Filter Tasks like this and click on Match response to target

REW will apply cutting filters only and will try to flatten the response to your flat target curve at 78dB

Once done you can either export the results as a convolution filter or you can enter the filter values manually into an EQ plugin
Then you can apply a low shelf to set the bass to taste and a high shelf to do the same with the highs; finally you can use a high-pass filter to limit the very bottom end

Note: I would also try a 2nd variant of this approach by setting the target to 69dB

Edit: you shall perform these steps for each channel individually, making sure that the target curve has the same dB value!