New set of graphs for all speakers reviews

[Haint]

Does bass extension correlate with a higher score? If so, I really want to hear some larger speakers that go down to 20hz in room... Confirm my decisions and all that.

Probably a useless exercise IMO. Even if you did hit the jackpot with positioning that was both great for mains and that could deliver bass as evenly as a couple subwoofers in better locations, I don't expect you'd be able to tell a difference between the 2 scenarios.

 

[stevenswall]

Probably a useless exercise IMO. Even if you did hit the jackpot with positioning that was both great for mains and that could deliver bass as evenly as a couple subwoofers in better locations, I don't expect you'd be able to tell a difference between the 2 scenarios.

Bass under 100hz is pretty easy to get flat with GLM on my Genelecs or using a MiniDSP.

Also, many listen without subwoofers, and I love when speakers are extended... Though even for mine that are measuring flat to 19hz or so in a room, I'm considering subs.

Also, extended speakers (the Phantom Reactor comes to mind) are great if someone doesn't have the space, spouse, or desire for subs. Shows how low regular speakers can go when engineers apply themselves.

 

[Haint]

Bass under 100hz is pretty easy to get flat with GLM on my Genelecs or using a MiniDSP.

Also, many listen without subwoofers, and I love when speakers are extended... Though even for mine that are measuring flat to 19hz or so in a room, I'm considering subs.

Also, extended speakers (the Phantom Reactor comes to mind) are great if someone doesn't have the space, spouse, or desire for subs. Shows how low regular speakers can go when engineers apply themselves.

That is true. I was assuming you were talking about large speakers and already had subs.

 

[stevenswall]

That is true. I was assuming you were talking about large speakers and already had subs.

Nope. Love large speakers and have some that are full range finally, but there's some stuff I'd like to be 10-20dB higher when stuff gets really low that makes the 8260's struggle in my living room. (They are nearfield currently. Makes me want to cry how much better they sound just being nearfield compared to in an untreated living room.)

 

[pierre]

This is truly wonderful work.

My only concern is, based on what I have seen here at ASR, I find the manufacturer data from Harmon and KEF to be not credible at all. I’d ditch those until we get some reliable independent data from @amirm or someone else. The KEF and Harman data in my view is simply not to be taken at face value. So I’d say either leave those out or put an asterisk next to the calculations based on KEF or JBL/Harman published data and reference it as being based on vendor data that is not independently verified or independently corroborated.

Origin of measurements should be explicit on the website. KEF LS50 for example has 2 measurements 1 from ASR and 1 from Princeton. The JBL 4367 is identified as Vendors/JBL.

This is also explicit on the speaker page:

On the quality of the publish data, we will see overtime if they are accurate or not. I guess they will be (or very close too).
Why would they cheat? They have great speakers, built the way they are as a consequence of proper research. Possibly
we get a golden sample for measurements.

 

[CrustyToad]

Excellent work! Curious to see more measurements added to your database.

A small point from a plotting point of view, better to avoid jet/rainbow type of colormaps. They do not have monotonously changing luminosity and therefore create false impressions....
See for instance here


viridis or inferno cmaps are much better for such purposes.

 

[QMuse]

I have found a few bugs. Scores now work as expected. Only remaining issues is on SP curve:
i still have an error increasing with frequency that i do not explain yet: that's for neumann speaker
control is what has been exported by the Klippel, computed are mine computation from
SPL H & V.

Code start here.

View attachment 54230

Interesting error. Graphically it seems as there is a center rotation point at app 200Hz around which graphs are rotated by a fixed angle.

 

[pierre]

Does bass extension correlate with a higher score? If so, I really want to hear some larger speakers that go down to 20hz in room... Confirm my decisions and all that.

I think that yes, mostly. When all parameters gets good, the main difference is on low freq extension (LFX) and then you see the larger speaker take the lead. The Olive formula doesn't take max SPL has a parameter: medium size speakers like Kii 3, D&D 8C, Genelec 8260 should be very good.

 

[stevenswall]

I think that yes, mostly. When all parameters gets good, the main difference is on low freq extension (LFX) and then you see the larger speaker take the lead. The Olive formula doesn't take max SPL has a parameter: medium size speakers like Kii 3, D&D 8C, Genelec 8260 should be very good.

Especially for their size if SPL isn't a consideration. The M2 is huge and doesn't seem to go as deep anechoically as it should with a 15" woofer vs the 8260 10" woofer.

I wish there were a secondary scoring that came into effect if speakers were similar but one was 1/3 or 1/10 the volume to boost things that were excellent but also did it in a smaller footprint.

 

[thewas]

The M2 is huge and doesn't seem to go as deep anechoically as it should with a 15" woofer vs the 8260 10" woofer.

As been said before in time of (room-) EQ the low frequency limit is secondary also due to room gain, more important is the max reachable SPL. Few tune their loudspeakers to very deep bass with more disadvantages in harmonic and intermodulation distortions.
See this German article about the (non-)sense of 20 Hz loudspeakers (use google or deepl.com to translate)
https://www.hifi-selbstbau.de/grund...sinn-und-unsinn-einer-grenzfrequenz-von-20-hz

 

[MZKM]

I was looking at your scores and noticed that they differ quite a bit from mine, and looking at the components, all are a little different, but Smoothness of the PIR is the largest difference. I just checked with a few, and when changing between linear and log regressions, I get your scores, so that is the error (are you doing a linear regression without log spacing the frequencies?).

I also noticed things such as Flatness being different, yours being much tighter than what I am getting. The Zaph for instance from 80Hz-20kHz has a min SPL of 74dB (@ ~20kHz) and a max SPL of 85dB (@ ~810Hz), which is an 11dB window and thus +/-5.5dB, which is what I got, your data states +/-2.4dB (meaning a deviation window of 4.8dB).

 

[Snoochers]

Can you maybe add a quit note in the help section to describe what flatness, smoothness mean etc.

 

[pierre]

Can you maybe add a quit note in the help section to describe what flatness, smoothness mean etc.

yes this is on its way. I have a full tutorial underwriting in tutorial.ipny on github.

 

[vavan]

My only concern is, based on what I have seen here at ASR, I find the manufacturer data from Harmon and KEF to be not credible at all

my own in room measurements of f208 definitely do not look that nice

 

[Eetu]

Great work!

Concerning the credibility of vendor data: it's pretty interesting to compare JA's Stereophile measurements of the Reference 5 to KEF's own
Stereophile KEF Reference 5 Measurements
For example I can't find the 1-1.5khz dip-rise on the KEF chart. Am I missing something?

 

[pierre]

I was looking at your scores and noticed that they differ quite a bit from mine, and looking at the components, all are a little different, but Smoothness of the PIR is the largest difference. I just checked with a few, and when changing between linear and log regressions, I get your scores, so that is the error (are you doing a linear regression without log spacing the frequencies?).

I also noticed things such as Flatness being different, yours being much tighter than what I am getting. The Zaph for instance from 80Hz-20kHz has a min SPL of 74dB (@ ~20kHz) and a max SPL of 85dB (@ ~810Hz), which is an 11dB window and thus +/-5.5dB, which is what I got, your data states +/-2.4dB (meaning a deviation window of 4.8dB).

You are spot on on the SM and log stuff. I corrected it and I now have almost the same scores at you (<1% error, roundoff I think).
I will look tonight into the second part if work doesn't completely kill my brain today.

 

[pierre]

my own in room measurements of f208 definitely do not look that nice

I guess this is normal, your room in not an anechoic room. How much differences do you see between In room prediction and your measurement above 500Hz/1kHz? Below that, the room will dominate the measurement.

 

[vavan]

How much differences do you see between In room prediction and your measurement above 500Hz/1kHz?

see for example

 

[pierre]

I have found a few bugs. Scores now work as expected. Only remaining issues is on SP curve:
i still have an error increasing with frequency that i do not explain yet: that's for neumann speaker
control is what has been exported by the Klippel, computed are mine computation from
SPL H & V.

Code start here.

View attachment 54230

I have found what Klippel is using (that's the guess version):

def rms(spl):
avg = [(sp_weigths[column_trim(c)] * spl2pressure(spl[c]))**2 for c in sp_cols if column_valid(c)]
wsm = [sp_weigths[column_trim(c)]**2 for c in sp_cols if column_valid(c)]
return pressure2spl(np.sqrt(np.sum(avg)/np.sum(wsm)))

def rms_guess(spl):
avg = [(sp_weigths[column_trim(c)] * spl2pressure(spl[c])**2) for c in sp_cols if column_valid(c)]
wsm = [sp_weigths[column_trim(c)] for c in sp_cols if column_valid(c)]
return pressure2spl(np.sqrt(np.sum(avg)/np.sum(wsm)))


RMS for me was Lˆ2 norm
weigthed RMS: i have square the weigth too but Klippel did not.

wrms (klippel) = sqrt( sum( w*pˆ2) / sum( w ))
wrms (me) = sqrt( sum( wˆ2 pˆ2) / sum (wˆ2))

Looking at wikipedia : https://en.wikipedia.org/wiki/Lp_space#Weighted_Lp_spaces weigths are not squared and Klippel is correct.

WDYT?

 

[pierre]

see for example

I think I understand what you mean: the standard define a room response in an ideal room where you do not see reflections. In your case, in a real room, you have all the interaction with the room (and you see the normal modes in a room).

The standard defines Estimated In-Room Response as:

eir = 0.12*listening window + 0.44*early reflection + 0.44*sound power

which doesn't model room interaction.

EIR and your measurement should get close above Schroeder frequency.

It would be relatively easy to build a tool that display a correct Estimated In Room if you give it a description of your room and a speaker or 2.