A collection of speaker target responses in .csv/.txt format

[Le Concombre]

Can you add legends? I’m having trouble seeing which curve is the target and which are the real responses.

In #30 there's no Target per se. Interwoven lower curves are L & R convolved Vector Averages ; upper ones L & R convolved RMS Averages. As a matter of fact I have not bothered finalising a Target Curve for the VAs. If I was to I'd take Harman's Reference Room 12 dB up to 160 Hz, lower 160 HZ value by 3 dB and then compute the OP's attached Dr Toole's.
But Maybe it would not suit you ; maybe in your room a straight line above 160 would be fine, maybe HRR 10 would be a better fit to your natural bass response...

I can describe my making-off a bit more... This target thing might be a trap of misplaced rationalisation while i chose a "Trust your feelings Luke" attitude, after a lot of reading (I'll spare you quoting Toole, Olive and al but I don't invent anything) and hands-on software manipulations of course (not sure commercial packages can do what I describe below ).

I liked the amplitude and the bass response obtained with convolution filters done by aiming Synthesis Arcos Target for RMS but missed the precision, 3D soundstage etc yield by convolution filters done by aiming HRR 10 or 12 for the Vector Averages.

I don't follow the tutos that invite you to limit the use of VA to Phase correction in RePhase while the Frequency Response should be corrected in REW following the "usual suspects" targets designed for steady state. Besides HRR 10 quoted here and there in Harman's funded literature I haven't found any published target to apply to the Direct Sound (assuming it's represented by the VA produced in REW).

I created an hybrid of HRR 12 and B&K. It later proved to be wrong (see above : the 160 to 20 K delta should be about 3 dB with a 3 dB lower starting point at 160, not a 6 dB delta from 160 to 20 k).

In REW, I manually (using Qs of 0.40 0.67 1.41 2.87 4.36, that is from 3 to 1/3 of octave width) designed frequency response correction of the VAs following that hybrid target

Produced the filters in RePhase

Convolved the RMS averages in REW, created new eq filters aiming Synthesis Arcos for the Left channel (cleaner)

Looked at the result on VA in RePhase. Visually it looked like a 3dB slope from 160 to 2K then flat to 20K ; thus I DECIDED that a 3dB slope from 160 to 2K then flat to 20K was the target for VA above 160

And I tailored the Right channel to that target in RePhase. Up to the dip starting (going towards Low Frequencies) at 260. DECIDED to ignore it since it does not show on RMS

Saw it looked a bit ugly (see #30) above 300 or so when convolving RMS Right and comparing it to Synthesis. DECIDED that I don't care for what matters is MATCHED Direct sound for L and R. (Well, to be thorough : the shrill zone for me is 700 +- 100 Hz, heard the 1/3 octave 2 dB bump @ 643 that showed and that one I decided to correct after I checked that it was a rather positive smoothing of the Right VA too.) Confirmed it was a nice fit to Synthesis for LF while preserving natural bass response (see #31) of that Right channel

Tweaked eq filters in REW for the Left channel to fit Synthesis on RMS average, (at 99% I kept the eQ points initially created, in the end R is corrected by 26 filters and L by 20, with no Q narrower than 4.36 and no amplitude larger than + or - 6 dB.

Saw, when back from RePhase, that though fitting very nicely the Synthesis target, it did not fit the Right channel. DECIDED to push 1 dB 1 octave wide around 100 to create that match of the L to R.

Obtained what I present in #30 :

a very close to Synthesis target steady state response albeit optimised for Direct sound above Schroeder
Matched L & R steady state below Schroeder
Matched L & R Direct sound above Schroeder

 

[ernestcarl]

Did as you suggested and retook all measurements after buying phonoabsorbing bases for the speaker stands. Also found out that previous measurements had been taken with an high frequency boost activated in the speakers.
This is the end result (Var smoothing applied only after averaging). Don't know if it's good, wrong, right... it sounds good to my ears. Tried Toole trained listeners tool curve too. I have both EQ settings saved, sometimes one sounds better, sometimes the other does (mainly genre dependent I'd say).

View attachment 91841

Looks good — although I would not necessarily completely fill out wide or deep troughs in the bass without considering loss of headroom or potential additional distortion at louder listening volumes. Also, it’s usually easier to focus A-B testing one section at a time. Say, PEQs 1-10 first, then 12-16, and lastly 17-19.

 

[Le Concombre]

This might look bass heavy but just sounds right. I once advocated a change of convolution filters based on different targets depending on the program. I'm just happy with the one set presented here. Maybe because it respects the most my room's natural bass response? Please note it's only 3 dB above the much touted RR1 bElow 50 Hz while being much leaner in the mud region, especially between 100 et 250 Hz. RR1 might be suited for pop and such or to act as a loudness compensator but is improper for acoustic sounds reproduction IMO while Synthesis Arcos, while beefier below 50 Hz, sounds considerably more natural most probably because of its profiling of the mud region. Works for me with any kind of music.

Very B & K down to 120 Hz

 

[Le Concombre]

The target I used to finalize Vector Averages corrections

 

[Le Concombre]

The screen capture presents 5 L+R couples : Synthesis ARCOS (kindly posted as HATS by staticV3 ), "Trained Listeners only" that is a lowered bass level ARCOS, Toole's as pointed by thewas_ with 3 bass options : ARCOS bass, "Trained Listeners only" bass and extrapolated bass following the general 0.5 dB/octave slope of the Toole's published curve (attaached here as "Toole in India". Those neat looking RMS come at the expense of the perfect matching of Vector Averages. The "4 Direct sound" I attached above is just for info ; we most certainly have all read that Toole recommends a straight line. Here's what I get for Toole's with "Trained Listeners only". From Listening sessions I think so far that that mismatch is OK as long as Vector Averages variations are kept smooth (no narrow peak should appear for the sake of making RMS look too good) and most importantly : the perceived balance corroborates VA reading : though RMS are perfectly matched in level, it's better to boost R as per a better VA matching.
Thanks to HQPlayer (Embedded) it's easy to switch around the Trained Listeners Target with a Higher Highs option and a Lower and Higher Lows options, thus having at hand a precise if limited Tone Control at matched levels

 

[ernestcarl]

Hmmmn.

Quick question:

There is a difference when measuring individual left and right channels in REW vs summing L+R at the same time -- the bass response will be boosted, of course.

When using a Harman target curve (with boosted bass) would you say it makes more sense to follow the trail of the ~4-6dB bass incline using the summed L+R channels or with each channel individually?

You can see that there's an additional +3.5dB bass boost in the blue curve. It seems intuitive to follow the target for the total summed response since much bass in music is already set in mono anyway. As far as I recall, I've been using the L+R sum more often than not since this seems to give a better overall balance esp. when I'm already using some form of loudness compensation DSP to boost bass at lower listening levels.

----

Wait... had to delete and check something out.

Huh!

Okay...

Interestingly, the difference is way more exaggerated (I've double checked) when using the MMM and is actually very subtle (about a single dB!) when doing single-point measurements. Perhaps it's just the MMM method that's apparently making the boost bigger than it is...

Eh, since I'm primarily using MMM when EQing, I guess my question still makes sense to ask.

 

[tuga]

In order to use REW's automatic EQ feature, I started collecting and converting every speaker target that I could find into the necessary .txt format.
In this thread I want to share my work so far, as well as ask you guys for target responses that I may have missed.

You seem to have missed the Brüel & Kjær curve which I personally find a lot more neutral than Olive's (and more in line with the trained listeners' from Olive's test, which in my view has too small a sample to be relevant and is using untrained listeners who generally prefer a boom-n-tizz presentation):

 

[staticV3]

@tuga Here is the converted Brüel & Kjær target: https://drive.google.com/drive/folders/1FxaZLMRZjZ2xtWtZ4xGT_HLf72S1c8Wc?usp=sharing
For some reason I can no longer attach the .txt file as it is too large, so a Gdrive link will have to do.

Spoiler: Quick comparison

 

[sfdoddsy]

I've tried the other curves and much prefer something closer to Toole and B&K. The others sound way too bass heavy to me.

 

[Le Concombre]

Hmmmn.

Quick question:

There is a difference when measuring individual left and right channels in REW vs summing L+R at the same time -- the bass response will be boosted, of course.

When using a Harman target curve (with boosted bass) would you say it makes more sense to follow the trail of the ~4-6dB bass incline using the summed L+R channels or with each channel individually?

View attachment 96923

You can see that there's an additional +3.5dB bass boost in the blue curve. It seems intuitive to follow the target for the total summed response since much bass in music is already set in mono anyway. As far as I recall, I've been using the L+R sum more often than not since this seems to give a better overall balance esp. when I'm already using some form of loudness compensation DSP to boost bass at lower listening levels.

----

Wait... had to delete and check something out.

Huh!
View attachment 96925
Okay...

Interestingly, the difference is way more exaggerated (I've double checked) when using the MMM and is actually very subtle (about a single dB!) when doing single-point measurements. Perhaps it's just the MMM method that's apparently making the boost bigger than it is...

Eh, since I'm primarily using MMM when EQing, I guess my question still makes sense to ask.

MMM gives remarkably reproductible results but I recall reading it's not perfect below 100 HZ or so, especially if boundaries are close. I don't use it any longer since correcting Time Domain and thus needing Impulse Responses

 

[Le Concombre]

I've tried the other curves and much prefer something closer to Toole and B&K. The others sound way too bass heavy to me.

View attachment 97040

curves is the right word : BK and furthermore Toole are rather descriptions of anticipated Steady State responses of well behaved (directivity...) loudspeakers in our rooms.
The real target is flat line with parabolic terminaison above Schroeder.
The trick is what AVERAGE should we flatten if at all... I eQ above Schroeder for my approach is to keep low end as close to untreated as possible at least for my bass heavier convolution set. If I flatten (low Qs small Amplitudes) the Vector Average FDW 4 obtained in REW the convolution filters make the Steady State very Toole up to 10K ; If I flatten (low Qs small Amplitudes) the RMS Average FDW 3 obtained in REW the convolution filters make the Steady State very BK up to 10K. In both instances I have to boost back the treble (13800 1.41 2.7 dB --given as mere illustration) to meet either curve, with some discrepancies like a 3 dB boost around 2K . REW's author recommends to correct amplitude on RMS because (if I understood correctly my readings) because of phase and measurement distances causing amplitude artefacts occurring in the right end of the spectrum. What puzzles me is that VA FDW4 looks like my speakers's response in anechoic chamber...
Well, as of above Schroeder, let's my Steady State responses look more BK/Trained listeners than Toole then...
As of the bass response, I take my flattened Average of what happens above Schroeder frequency as seen from a Frequency Dependant Window of 3 cycles (meaning at 300 Hz I'm looking at events occurring before 3/300 = 10ms while above 1200 Hz it's Direct Sound since my first reflections occur round 2.5 ms and that 3/1200 = 2.5 ms) and then adjust it to RMS average at Steady State (no FDW) which was a move up of Steady State by 8.4 dB if I recall correctly (SS bass is # 8.4 dB above the 200/300 Schroeder region where there's overlapping ; btw I don't dig the "sitting high on comb filter" verbiage some use to reach the same goal...) and then I apply 3 Low End extensions : HATS as given by @staticV3 , Trained listeners and flat.
Maybe you should consider having 3 bass extensions to switch as a tone control, BK flat is fine with classical where Trained rarely hurts even with Classical but BK is bass shy IMO for most pop very well suited by the JBL HATS (or Arcos) bass extension. What you need to avoid is the beefing of the 100/200 region by following 1dB/octave - RR1 type "target" some recommends

 

[Senior NEET Engineer]

IMO target curves are anti science. Adjusting steady state response will have detrimental impact on the direct and early reflected sound. Parametric EQ will let you adjust sound signature of speaker independently from the room.

 

[ernestcarl]

MMM gives remarkably reproductible results but I recall reading it's not perfect below 100 HZ or so, especially if boundaries are close. I don't use it any longer since correcting Time Domain and thus needing Impulse Responses

This is true. Although, I did measure away from boundaries.

 

[ernestcarl]

IMO target curves are anti science. Adjusting steady state response will have detrimental impact on the direct and early reflected sound. Parametric EQ will let you adjust sound signature of speaker independently from the room.

What are you EQing against for comparison's sake if not after some kind of 'imagined target'. It doesn't have to be specific to a T and unchanging. You need to take account the on-axis and off-axis responses, yes -- better yet, have all those measured gated, nearfield, and well away from walls -- so whatever changes you make on-axis, you can see and predict how it will affect ALL the other curves too. Some speakers are (most aren't) more amenable to EQ than others. For speakers that have particularly bad off-axis responses, then heavy absorption is necessary if you want to EQ the on-axis completely flat to a target -- many people are already quite conservative anyhows. Spatial averaging like MMM is also the method being recommended here most of the time... Any kind of imagined target we try to conform and adjust to shouldn't be that much different from the speaker's natural response -- e.g. why force a significant amount of bass boost extension or HF shaping to a speaker that's incapable of reproducing said frequencies properly anyway? That would be the naive approach.

 

[Le Concombre]

Attached :
My speakers measurements in the huge world class Cabasse's anechoic chamber
My eQ work on R channel:
Thick blue line represents native Steady State (RMS Average in REW of non FDW measurements, could be MMM that looks much alike)
Thick red line represents native RMS as FDWindowed at 3 cycles
Green line represents the Full Range convolution of native RMS as FDWindowed at 3 cycles
Purple line represents the Full Range convolution of native Steady State

The levels are modified for presentation ; in the process the above 300 Hz part is lowered to avoid boosting the bass region

 

[Le Concombre]

Toole's ideal Steady State (extended flat left from the lowest value, 64, plotted by Toole) and B&K are close indeed. Yet I'm pleased that Toole's is a better fit.
I have summed L &R and applied a 1/12 smoothing on those curves.
Please note that the only "target" applied was a flattening of the responses up to 300 Hz, upper frequencies are left untouched
Thus, it's a boosts and cuts of the low range only, with pros and cons : pro being no eQ above Schroeder frequency, the cons being much bigger amplitude moves with much higher Qs than when taking the approach I described a few posts above, while pretending to Digitally Correct the Room by modifying the speakers responses might very well be, as stressed by Barry Diament, a non-sense and thus should be approached more lightly than here while the point of balancing L &R in my asymmetrical room is still a very good reason to apply eQ...

 

[ernestcarl]

Attached :
My speakers measurements in the huge world class Cabasse's anechoic chamber
My eQ work on R channel:
Thick blue line represents native Steady State (RMS Average in REW of non FDW measurements, could be MMM that looks much alike)
Thick red line represents native RMS as FDWindowed at 3 cycles
Green line represents the Full Range convolution of native RMS as FDWindowed at 3 cycles
Purple line represents the Full Range convolution of native Steady State

The levels are modified for presentation ; in the process the above 300 Hz part is lowered to avoid boosting the bass region

I don't use convolution at the moment. Nor do I have the off-axis curves for my speakers from Amir's Klippel NFS.

However, even my own primitive method of using just basic PEQs full-range has proven to be very beneficial.

I've posted my 'speaker corrections' before in other threads -- although, I've always been able to make small improvements/refinements so here it is yet again.

*Title is wrong. Measurements were taken nearfield at 30cm from the speaker's horn throat and not at 2m!

Missing is the 30 deg vertical -- forgot to save it a while back and I have no motivation to re-measure. The speaker is the Sceptre S8 which uses a coaxial driver with a HF horn waveguide and so one can expect the vertical curves to be similar to that of the horizontal as well.

Previously, I used 20 PEQs. I cut it short to 19 so you could say it's a small improvement. LOL

Here is the stead state recorded at my listening couch.

BTW, I use a wrinkly blanket over the couch to minimize comb filtering reflections -- already have some random panels at the side walls.

As long as I place my mic precisely at the center, I don't even need to use FDW to get a clear view the phase.

I use acoustic timing reference with 'adjust clock with acoustic ref' in the analysis tab checked for my UMIK-1.

While the speakers already use internal DSP to flatten phase, even my basic, primitive speaker correction PEQs seem to provide a small improvement.

At the back of my mind is always distortion, so I was careful not to overdo my EQ work to avoid any unnecessary increase in that area. You could say it got better in some areas, and worse in others. That dip at around 800Hz is a bit deeper as seen from 3rd party anechoic measurements -- didn't EQ that much as there it increases 4th and 5th harmonics a bit too much for comfort.

From about 800Hz and up, the steady state response at all seats on my couch is pretty close to linear -- yes, it holds up very well without much change in sound character/timbre. The lower you go down around the transition zone, it becomes more uneven, but not too bad. I also keep a best 'average of all seats' preset which is an EQ focused in the transition zone and upper bass which is a balanced compromise setting for all seats.

It's awesome if you can easily save and recall all your DSP presets...

 

[Le Concombre]

All convolution sets but the one showing in Orange were made flattening (and lowering about 3 dB) RMS Average FDW 3 (above 200 Hz) first, then adjusting to 3 different bass levels with tweaks up to 300, then boosting 2.7 dB @ 13800 to meet Trained Listeners target. Comparison to the Orange (no eQ above 300 Hz) show tiny differences only, as intended, the most significant being a welcome 1 dB lowering of the 600-1000 Hz region where real harshness lies IMO.

For the Low End, Orange is heavily flattened while the 3 other options aim at taming bass region but with fewer eQ points, of lower Qs and of smaller amplitudes, hence they don't look as nice but it's not what matters, does it?

At 64 Hz (where Toole's ideal Steady State stops) Toole's is 2 dB above flat, Trained Listeners 1 dB above at 3 dB above flat while low end tailored on ARCOS (HATS) is 7.34 dB above flat. At 39 Hz (maximum amplitude) values are 0 (flat level), + 3.6, 5.25, 7.9 reducing the delta between the responses tailored to Trained Listeners and ARCOS to 2.6 dB (I love it with a lot of pop rock blues etc recordings

 

[Le Concombre]

Please note that our perception of Amplitude corrections are greatly affected (bettered) by Time domain corrections ; posted my results and recipe there : https://www.audiosciencereview.com/...ith-linear-phase-room-eq-filters.11121/page-2

 

[ernestcarl]

Please note that our perception of Amplitude corrections are greatly affected (bettered) by Time domain corrections ; posted my results and recipe there : https://www.audiosciencereview.com/...ith-linear-phase-room-eq-filters.11121/page-2

Sure, but within limits, perhaps... since the rule of diminishing returns apply here as well.

It's simply far more convenient to get speakers with excellent time-domain performance from the start (e.g. HEDD monitors) rather than having to learn to play the role of speaker designer & DSP expert.

Spectrograph wavelet ex.

These monitors are not completely phase flat across the board, but they're linear where it counts the most -- holding-up this quality midfield off-axis as well:

15 degrees off-axis sitting at my couch

Right now it's easier and cheaper to get speakers with a good frequency response (or at least good-enough to correct after the fact) than getting speakers with excellent time-domain response.

Call me a worry-wart, but one may end up doing things incorrectly, or apply excessive amounts of correction -- esp. to systems that may not be amenable to any more forced "correcting" -- I think there has to be trade-offs or compromises, too (e.g. added delay, correction only works well for a certain position, possible added artifacts...)

I think most speakers are "good-enough" here anyway -- not saying they're all excellent! -- rather just good-enough.