Harman curve for loudspeakers

[dasdoing]

I had tried it several times

Denis Sbragion's DRC windowing, or the REW one?

 

[thewas]

Denis Sbragion's DRC windowing, or the REW one?

Denis Sbragion's DRC, actually it was the first DRC product I tried a long before I discovered REW.

 

[dasdoing]

Denis Sbragion's DRC, actually it was the first DRC product I tried a long before I discovered REW.

his software is so complex you can get crazy studying it. I use (and unuse) it since 2011 and only in 2020 I started understanding it. You realy have to dedicate weeks of work to it

 

[thewas]

his software is so complex you can get crazy studying it. I use (and unuse) it since 2011 and only in 2020 I started understanding it. You realy have to dedicate weeks of work to it

That is true, I used it back then with some predefined settings of http://www.alanjordan.org/DRCDesigner/DrcDesignerHelp.html

 

[dasdoing]

That is true, I used it back then with some predefined settings of http://www.alanjordan.org/DRCDesigner/DrcDesignerHelp.html

I tried the designer, and I tried Align 2 https://sl3zxapwn7jdkocakforzltw6e--www-ohl-to.translate.goog/about-audio/audio-softwares/align2
they both try to simplify the software, but at the end they are limiting

just for you to have an idea, a tweeked configuration here for my system will look like this:

drc --BCInFile=/home/studio/Downloads/l.pcm --PSOutFile=next_generation-l-filter.pcm --PLMaxGain=4.0 --PSPointsFile=/home/studio/drc_create/pa-44.1_extended.txt --PTType=N --PLStartFreq=26 --RTStartFreq=10 --ISPEStartFreq=10 -
-EPStartFreq=10 --MPStartFreq=10 --MPUpperWindow=132 --EPUpperWindow=132 --RTUpperWindow=132 --MCFilterType=M --MCPointsFile=/home/studio/drc_create/EMM-6-drc.txt --MPEndFreq=22050 --EPEndFreq=22050 --ISPEEndFreq=22050 --RTEndFreq=22050 --EPWindowE
xponent=1 --MPWindowExponent=1 --ISPEWindowExponent=1 --RTWindowExponent=1 /home/studio/Downloads/align2/files/drc/44.1KHz/insane-44.1.drc --MPLowerWindow=115000 --RTLowerWindow=115000 --EPLowerWindow=3500 --MPPFFinalWindow=115000 --EPPFFlatType=L
--PLType=L --PSFilterType=L

yes, I use insane.drc as a starting point....then I tweak basicly looking at excess group delay, which is pretty good at showing audible pre-delay above 1kHz

 

[Phorize]

I’ve learned quite a lot from the posts in this thread and the references. Thanks for weighing in everyone.

 

[dasdoing]

man, I realized Denis Sbragion's DRC always basicly only corrected direct sound above 1000Hz. I was stupid enough to always interpret the resulting curve as an error, and ajusted the target acordingly.
ever since this topic I searched for the best solution to window out the direct sound above 1000Hz and I am pretty convinced that Denis Sbragion's DRC windowing is more advanced then REW's FWD filtering. for two reasons: 1) https://www.diyaudio.com/forums/ful...-driver-range-line-array-191.html#post4620292 2) when you read his documentation on his software you realize how much thought has been put into every aspect of it.
So I tried out his software for the first time with a flat target.
resulting curves (L and R) are this (ignore the dips around 250Hz...I still will adress that...thses filters are "beta")
View attachment 102743

i thought: "wtf, if flat is too bright these curves bascily flat but with that boost around 3k will be even worse"

this is 5 cycles FDW in REW (ignore stuff below 1000Hz):

View attachment 102744

if you consider the diferences in REW windowing and DRC windowing pointed out in the diyaudio post linked above the direct sound is basicly flat.

so how does it sound? I can't belive what I am hearing, for real. shit is seams totaly balanced. I am amazed

*note that my room has a lot of absorbtion panels....."normal rooms" would result in a totaly diferent 500ms curve and would probably be falling more like the Harman curve

so I am hearing this second time right now. It is funny how a flat curve is way too bright but this isn't. I can only explain it with the equal loudness curves.

 

[celroid]

There's something that I wish to understand, I read about the harman curve from headphonesty. They posted this set of images here:

How come the harman curve above looks nothing like the one below, where it looks more flat?

It's making me believe that these graphs are somehow 'distorted' to give the look of the harman curve in the 1st image. And I've read that ideal speaker curve is also shaped like the harman curve and that's why the slight downward slone. What's curious is that I've never seen a speaker in amir's measurements with such a large downward slope as in the 1st image.

What's going on here? Can someone shed some light about headphone and speaker harman curves and why they don't look the same in amir's graphs?

 

[Jimbob54]

There's something that I wish to understand, I read about the harman curve from headphonesty. They posted this set of images here:
View attachment 108009View attachment 108008

How come the harman curve above looks nothing like the one below, where it looks more flat?

It's making me believe that these graphs are somehow 'distorted' to give the look of the harman curve in the 1st image. And I've read that ideal speaker curve is also shaped like the harman curve and that's why the slight downward slone. What's curious is that I've never seen a speaker in amir's measurements with such a large downward slope as in the 1st image.

What's going on here? Can someone shed some light about headphone and speaker harman curves and why they don't look the same in amir's graphs?

Where is the second chart from (and when?).

First chart is only headphones- no speaker curve on it.

Headphones need a totally different response to speakers to sound balanced due to how the sound hits our ears and interacts with them so the 2 will never look the same . Hence my first question as to where that cyan line and second chart come from

 

[celroid]

Where is the second chart from (and when?).

First chart is only headphones- no speaker curve on it.

Headphones need a totally different response to speakers to sound balanced due to how the sound hits our ears and interacts with them so the 2 will never look the same . Hence my first question as to where that cyan line and second chart come from

The're both from headphonesty

 

[celroid]

I would like to see the harman preferred headphone frequency response along with the preferred speaker frequency response on the same graph. Has anyone done this?

Even more awesome would be if a real measured speaker was put on that graph, and a real headphone. So that there would be 4 lines in total. This would put things in perspective for me

 

[sigbergaudio]

if you consider the diferences in REW windowing and DRC windowing pointed out in the diyaudio post linked above the direct sound is basicly flat.

so how does it sound? I can't belive what I am hearing, for real. shit is seams totaly balanced. I am amazed

*note that my room has a lot of absorbtion panels....."normal rooms" would result in a totaly diferent 500ms curve and would probably be falling more like the Harman curve

A flat curve like this will sound very thin to almost everyone, and your bass falls off very early as well. Your speakers should measure flat in an anechoic chamber, not in your room. In your room you should be looking at ~1dB/octave drop. For those who have response down to 20hz and have a subwoofer and/or EQ, it's relatively common to boost even more between 20-100hz, perhaps around 6dB. So you would see a ~6dB drop from 100-10khz, and a 10-15dB drop from 20hz-20khz depending on preference.

 

[Ilkless]

There's something that I wish to understand, I read about the harman curve from headphonesty. They posted this set of images here:
View attachment 108009View attachment 108008

How come the harman curve above looks nothing like the one below, where it looks more flat?

It's making me believe that these graphs are somehow 'distorted' to give the look of the harman curve in the 1st image. And I've read that ideal speaker curve is also shaped like the harman curve and that's why the slight downward slone. What's curious is that I've never seen a speaker in amir's measurements with such a large downward slope as in the 1st image.

What's going on here? Can someone shed some light about headphone and speaker harman curves and why they don't look the same in amir's graphs?

As the author of the article - first graph shows the FR at the eardrum 'reference point' of the dummy head Harman uses, so it's reference speaker response + the response changes induced by the presence of a dummy head, ears and torso.

Second shows target curve of their loudspeaker reference, and the cyan line shows the response that is yielded when you subtract the contribution of the dummy from the measured response for headphones. The second graph is meant to show how closely the response that emerged from their headphone testing panel tracks that from their loudspeaker tests when you subtract the contribution, so the comparison is equal. The reason you do so is that in loudspeaker testing the loudspeaker output interacts with the listener's torso, head and ears naturally, and induces certain FR, phase and timing changes. Headphones/earphones bypass this interaction and hence the changes need to be baked into their FR to recreate this interaction. Yet, these changes very between individuals, so you can't just use an individual to measute it. Instead, this is typically calibrated using a dummy head and torso as a rough surrogate for the human being to capture the broad characteristics of the interaction (this is to make things standardised and replicable for R&D). So all the cyan line is saying is what the Harman headphone response is when subtracting for said calibration so you can have their relative tonal balances compared. And it tracks it very closely, with slightly less bass and treble, but similar shape.

BTW, Sean Olive has since clarified that the 8kHz peak in their in-ear target (green curve in first image) is an artefact of their test methodology back in 2017. Improved testing now indicates the in-ear target curve actually tracks the over-ear target very closely.

Where is the second chart from (and when?).

First chart is only headphones- no speaker curve on it.

Headphones need a totally different response to speakers to sound balanced due to how the sound hits our ears and interacts with them so the 2 will never look the same . Hence my first question as to where that cyan line and second chart come from

Both charts were from Harman.

 

[Jimbob54]

As the author of the article - first graph shows the FR at the eardrum 'reference point' of the dummy head Harman uses, so it's reference speaker response + the response changes induced by the presence of a dummy head, ears and torso.

Second shows target curve of their loudspeaker reference, and the cyan line shows the response that is yielded when you subtract the contribution of the dummy from the measured response for headphones. The second graph is meant to show how closely the response that emerged from their headphone testing panel tracks that from their loudspeaker tests when you subtract the contribution, so the comparison is equal. The reason you do so is that in loudspeaker testing the loudspeaker output interacts with the listener's torso, head and ears naturally, and induces certain FR, phase and timing changes. Headphones/earphones bypass this interaction and hence the changes need to be baked into their FR to recreate this interaction. Yet, these changes very between individuals, so you can't just use an individual to measute it. Instead, this is typically calibrated using a dummy head and torso as a rough surrogate for the human being to capture the broad characteristics of the interaction (this is to make things standardised and replicable for R&D). So all the cyan line is saying is what the Harman headphone response is when subtracting for said calibration so you can have their relative tonal balances compared. And it tracks it very closely, with slightly less bass and treble, but similar shape.

BTW, Sean Olive has since clarified that the 8kHz peak in their in-ear target (green curve in first image) is an artefact of their test methodology back in 2017. Improved testing now indicates the in-ear target curve actually tracks the over-ear target very closely.



Both charts were from Harman.

Makes sense now- couldnt understand why the cyan didnt have the rolling peaks and troughs of the first chart- that explains!

 

[andreasmaaan]

BTW, Sean Olive has since clarified that the 8kHz peak in their in-ear target (green curve in first image) is an artefact of their test methodology back in 2017. Improved testing now indicates the in-ear target curve actually tracks the over-ear target very closely.

This is new to me, thanks

 

[celroid]

As the author of the article - first graph shows the FR at the eardrum 'reference point' of the dummy head Harman uses, so it's reference speaker response + the response changes induced by the presence of a dummy head, ears and torso.

Second shows target curve of their loudspeaker reference, and the cyan line shows the response that is yielded when you subtract the contribution of the dummy from the measured response for headphones. The second graph is meant to show how closely the response that emerged from their headphone testing panel tracks that from their loudspeaker tests when you subtract the contribution, so the comparison is equal. The reason you do so is that in loudspeaker testing the loudspeaker output interacts with the listener's torso, head and ears naturally, and induces certain FR, phase and timing changes. Headphones/earphones bypass this interaction and hence the changes need to be baked into their FR to recreate this interaction. Yet, these changes very between individuals, so you can't just use an individual to measute it. Instead, this is typically calibrated using a dummy head and torso as a rough surrogate for the human being to capture the broad characteristics of the interaction (this is to make things standardised and replicable for R&D). So all the cyan line is saying is what the Harman headphone response is when subtracting for said calibration so you can have their relative tonal balances compared. And it tracks it very closely, with slightly less bass and treble, but similar shape.

BTW, Sean Olive has since clarified that the 8kHz peak in their in-ear target (green curve in first image) is an artefact of their test methodology back in 2017. Improved testing now indicates the in-ear target curve actually tracks the over-ear target very closely.



Both charts were from Harman.

Thanks, now I finally understand why the ideal curve for headphones and speakers was so different.

 

[dasdoing]

A flat curve like this will sound very thin to almost everyone, and your bass falls off very early as well. Your speakers should measure flat in an anechoic chamber, not in your room. In your room you should be looking at ~1dB/octave drop. For those who have response down to 20hz and have a subwoofer and/or EQ, it's relatively common to boost even more between 20-100hz, perhaps around 6dB. So you would see a ~6dB drop from 100-10khz, and a 10-15dB drop from 20hz-20khz depending on preference.

I don't think you understand what I did there, maybe read it again? the direct sound IS (kind off-ish) the anechoic response in my LP

btw: my room has a lot of absorbtion, 6 corners, sidewalls, ceiling, 3/4 of backwall....Harman test subjects would create a totaly diferent curve in my room

 

[ZolaIII]

The Harman headphone curve is the response of a flat loudspeaker (Salon2 IIRC) in a moderately reflective room (intended to be a representation of domestic listening), measured by a calibrated head-and-torso simulator that acts as a surrogate for a human listener.

Harman hedaphone curve isn't response of anything especially not the speakers. This is how speakers simulated on headphones looks like:
https://github.com/jaakkopasanen/AutoEq/blob/master/compensation/loudspeaker_in-room_flat_2013.png
Speakers cannot go so deep down (much higher distortion) and even if they could that energy is permanently lost (fade out) in that area (physical distance). That's what subs are for (to try to compensate a bit).
To sum it up Harman curve on speakers would be equal to something alike tube or something else which will add distortion (boost) in lows. And pretty much that's what it does (Harman curve) with headphones (if you have good hearing abilities).

 

[dasdoing]

This is how speakers simulated on headphones looks like:
https://github.com/jaakkopasanen/AutoEq/blob/master/compensation/loudspeaker_in-room_flat_2013.png

it is a curve that sounds equal to speakers meassuring flat at LP of the room they used.....meaning it is too bright because they equalized the whole room.
the best way to come to a curve for natural sound would be to do the same in a anachoic chamber.
the harman heaphone curve without the bass boost is probably the closest https://github.com/jaakkopasanen/AutoEq/blob/master/compensation/harman_over-ear_2018_wo_bass.png

 

[ZolaIII]

the harman heaphone curve without the bass boost is probably the closest

It is but for the hedaphones if you have a great hearing and still with a little shave off (needed) in highs. Speaker will have a roll of in lows anyway (distance related one). So it would still be a try to compensate that with speakers (Harman 0 dB bass).