Eq-ing Concepts and Biquad Understanding

[Trdat]

Here is how I would approach this:

- in REW set the target curve to 78dB

- Set the Filter Tasks like this and click

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!

And then do I click on Generate measurement from predicted?

And does the predicted response show on the graph?

 

[ppataki]

And then do I click on Generate measurement from predicted?

And does the predicted response show on the graph?

If you need just one round of EQ-ing then you can export the filters into a convolution file by clicking on File --> Export --> Export filters impulse response as WAV

The Generate measurement from predicted will generate the predicted result as a new measurement - useful to see what your graphs will look like after applying the filters
You can also click on the Generate measurement from filters to generate the correction curve itself

 

[Trdat]

The Generate measurement from predicted will generate the predicted result as a new measurement - useful to see what your graphs will look like after applying the filters

I am confused here, it is not popping up. Am I meant to see the corrected curve on top of the frequency reponse? Undert the graph it says predicred response but its not there.

You can also click on the Generate measurement from filters to generate the correction curve itself

This as well, where does the correction curve pop up on the graph?

Apologize I am sure its simple enough but just cant seem to see what is happening.

 

[Trdat]

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.

I watched this. I can give this a try, doesn't seem to difficult at all.

 

[Dumdum]

The higher the freq the more boosting is fine due to the energy in music being higher in the low end vs the top end…

 

[ppataki]

I am confused here, it is not popping up. Am I meant to see the corrected curve on top of the frequency reponse? Undert the graph it says predicred response but its not there.

This as well, where does the correction curve pop up on the graph?

Apologize I am sure its simple enough but just cant seem to see what is happening.

I have created this little animation where you can see the steps:

https://drive.google.com/file/d/1dNINAQ2vxLL8HgBiekBVffqyFgY3Jjzu/view?usp=share_link

At the end I am exporting the filters as a convolution file (only 1 channel) but in reality you shall do that with both your channels' filters saved as a stereo convolution file.

 

[Trdat]

I have created this little animation where you can see the steps:

https://drive.google.com/file/d/1dNINAQ2vxLL8HgBiekBVffqyFgY3Jjzu/view?usp=share_link

At the end I am exporting the filters as a convolution file (only 1 channel) but in reality you shall do that with both your channels' filters saved as a stereo convolution file.

Great, thanks so much I got it.

The question that remains is that if I want to, or "shall I say" have to add the filters in manaully into the Hypex Design filter I suppose I save the Eq filter at the top and just take each frequency and add it in exactly the same as the biquad into the filter designer right?

I won't be using a convolution but that is good to know how to use that, it would come in handy.

 

[ppataki]

Great, thanks so much I got it.

The question that remains is that if I want to, or "shall I say" have to add the filters in manaully into the Hypex Design filter I suppose I save the Eq filter at the top and just take each frequency and add it in exactly the same as the biquad into the filter designer right?

I won't be using a convolution but that is good to know how to use that, it would come in handy.

Yes, indeed
Once you clicked on the Match Response to Target button and the optimization has finished just click on the EQ Filters button on the top and you will see the filters that you will need to manually type into Hypex

 

[Ruhled]

Yes, indeed
Once you clicked on the Match Response to Target button and the optimization has finished just click on the EQ Filters button on the top and you will see the filters that you will need to manually type into Hypex

View attachment 248241

This seems like way, way too much eq to my eyes. Too many filters and some are over 20db! This isn't going to work too well imho.

I've always gotten a better result with maybe 3-6 filters and no more than 10db correction or even less. Instead of a 1db flatness target try maybe 3 to 5db instead.

We should strive to correct the biggest issues and not everything under the sun because the measurement will never be perfect enough to do that correctly. Especially when you move your mlp inches.

Iirc op's response measurement tipped up in the treble. I would hit that with a shelf for starters and then go from there. That will probably save a lot of filters right off the bat. Maybe go for half the biggest dip rather than the whole thing too.

*Edited to add more details

 

[levimax]

This seems like way, way too much eq to my eyes. Too many filters and some are over 20db! This isn't going to work too well imho.

I've always gotten a better result with maybe 3-6 filters and no more than 10db correction or even less. Instead of a 1db flatness target try maybe 3 to 5db instead.

We should strive to correct the biggest issues and not everything under the sun because the measurement will never be perfect enough to do that correctly. Especially when you move your mlp inches.

Iirc op's response measurement tipped up in the treble. I would hit that with a shelf for starters and then go from there. That will probably save a lot of filters right off the bat. Maybe go for half the biggest dip rather than the whole thing too.

*Edited to add more details

I use the REW filter generator a lot and you can't think about it the same way you would if you were trying to do it manually. What ever algorithm REW uses generates a lot of overlapping filters that when combined seem to work out pretty well.

 

[RayDunzl]

Yes, indeed
Once you clicked on the Match Response to Target button and the optimization has finished just click on the EQ Filters button on the top and you will see the filters that you will need to manually type into Hypex

View attachment 248241

Did you forget:

Equalizer
Target Settings
--- Calculate target level from response <<<<<<
Filter Tasks
Modal Analysis
Resonance

 

[Trdat]

I use the REW filter generator a lot and you can't think about it the same way you would if you were trying to do it manually. What ever algorithm REW uses generates a lot of overlapping filters that when combined seem to work out pretty well.

I just tried it, adding biquads manually into Hypex design filter and it sounds okay. Actually surpsrisingly simlar to my passive left/right speaker while before it was just very different.

 

[Trdat]

Did you forget:

Equalizer
Target Settings
--- Calculate target level from response <<<<<<
Filter Tasks
Modal Analysis
Resonance

Are you recommending to also fill out the filter modal analysis and resonance sections?

 

[Trdat]

Yes, indeed
Once you clicked on the Match Response to Target button and the optimization has finished just click on the EQ Filters button on the top and you will see the filters that you will need to manually type into Hypex

Great thanks.

I manually added them to the Hypex filter and its worked out okay. I will keep playing around with it in REW see if different simulations give better results.

 

[ernestcarl]

I will keep playing around with it in REW see if different simulations give better results.

Indeed, it's simpler to take measurements without EQ at first and see what the prediction in REW says before finally programming them into your speaker's DSP amp.

If you have more time, try to look at EQ from multiple perspectives... not just a single-point MLP sweep or nearfield quasi-anechoic, but also measure different distances and positions/angles, averaging (e.g. MMM) -- maybe even see what frequency dependent windowing does. The same set of EQ filters can be saved as a file and is easy enough to load into a different group of measurements. This is probably obvious, but measuring at different distances and listening positions in the room makes one realize that any optimal EQ linearization applied at a single point in space will always potentially compromise the response elsewhere.

 

[RayDunzl]

Are you recommending to also fill out the filter modal analysis and resonance sections?

No.

Wondered if you forgot this:

--- Calculate target level from response <<<<<<

Because all the filters are cuts, and the smallest cut is 7.3dB, which is over 1/2 of the voltage swing around that frequency.

Just looks a little odd.

Carry on.

 

[Trdat]

No.

Wondered if you forgot this:

--- Calculate target level from response <<<<<<

Because all the filters are cuts, and the smallest cut is 7.3dB, which is over 1/2 of the voltage swing around that frequency.

Just looks a little odd.

Carry on.

So "calculate target level from response" is essentially taking in consideration a roll off and a tilt that you put in that section?

 

[Trdat]

Indeed, it's simpler to take measurements without EQ at first and see what the prediction in REW says before finally programming them into your speaker's DSP amp.

If you have more time, try to look at EQ from multiple perspectives... not just a single-point MLP sweep or nearfield quasi-anechoic, but also measure different distances and positions/angles, averaging (e.g. MMM) -- maybe even see what frequency dependent windowing does. The same set of EQ filters can be saved as a file and is easy enough to load into a different group of measurements. This is probably obvious, but measuring at different distances and listening positions in the room makes one realize that any optimal EQ linearization applied at a single point in space will always potentially compromise the response elsewhere.

My measurement taken was simply with the delays added. Then I took a few random measurements around the MLP and then vector averaged them and used that measurement in the EQ window. Later applying the biquads into the Hypex design filter.

I will look deeper into the MMM and using different averaging techniques in the EQ window which I am sure will give slighlty different results. I just don't know how to do the frequency dependant windowing, if my understanding is correct it takes out some of the room influence and we can see better the speakers response? Any link or explanation somewhere...?

If that is the case its good to learn this technique and also take it in consideration but I am prone to use the MLP with positions around that for the best idea of what needs to be corrected. Then again im not the expert so if people say otherwise I will try other techniques.

 

[Trdat]

--- Calculate target level from response <<<<<<

Because all the filters are cuts, and the smallest cut is 7.3dB, which is over 1/2 of the voltage swing around that frequency.

I think I might have just put together what this means. Using the match response to target makes to large of a correction and only uses cuts...?

What are you recommending to use a particular one? Or is there a method that uses both...?

 

[Destinyberry]

Good information!