Complaint thread about speaker measurements

[Blumlein 88]

On listening tests, it is become a major bottleneck to testing. Do you all want me to hold up reviews in general and speakers in the specific until I listen?

Formal listening tests are far in the future so let's not push for rigor there.

I'd say no don't let listening tests become a bottleneck. I'd say only do extensive listening to notable speakers or if something unexpected turns out a stellar result.

 

[hardisj]

With this approach you won't find many people sending in their speakers ...

That may be true. But you'd be surprised. Like I stated above, I tested a lot of speakers, wide open. My disclaimer was "there's always a chance it'll get damaged". But people were more excited to see the results than they were afraid of something being damaged.

And, reallly, 96dB @ 1m equivalent isn't all that demanding. Really. I did this kind of testing. A whole lot. Never a problem.

 

[mhardy6647]

On listening tests, it is become a major bottleneck to testing. Do you all want me to hold up reviews in general and speakers in the specific until I listen?

Formal listening tests are far in the future so let's not push for rigor there.

I'd like to see you listen. Umm... I'd like to hear you listen.... I mean... you know what I mean!

 

[hardisj]

Most passive speakers don't have a high-pass crossover though, no? So excursion due to bass/sub-bass can potentially get pretty large. If intending to use full-range, may be IMD can provide some useful information for the potential buyer?

I really, truly think you guys are WAY overthinking this. If a speaker can't do a 96dB sine sweep then you probably don't even want it. Certainly not if you're sitting more than 6 feet away (that would be 90dB at 1 meter, which is practically where most speakers' sensitivity lies anyway).

 

[LTig]

I suppose we are free to put anything we want on the Wish List.

IMD and THD are related.

It might be interesting to see if speakers with similar THD manage to create substantially different IMD.

I would expect so, when comparing 2-ways with 3-ways. The smaller the frequency range of a driver the less IMD it produces. Look the IMD graphs on the Neumann homepage and compare KH120 with KH310, AFAIR the KH310 is much better in the midrange.

 

[spacevector]

I really, truly think you guys are WAY overthinking this. If a speaker can't do a 96dB sine sweep then you probably don't even want it. Certainly not if you're sitting more than 6 feet away (that would be 90dB at 1 meter, which is practically where most speakers' sensitivity lies anyway).

I asked about IMD. What has 96dB got to do with it?

 

[hardisj]

I asked about IMD. What has 96dB got to do with it?

Sorry. I meant IMD. I took your post as to come from the perspective of damaging the driver.

IMD might be useful for the woofer. Moreso in a 2-way where the woofer is driven near Fs or below. But honestly, I’m still not sure I would spend extra time doing that test. THD and FR will tell the story quite well. IMD would be extra but I don’t see it as entirely necessary for a loudspeaker. I’d put that in the “if there’s extra time”.

Personally, I’d be more interested in long-term compression testing.

 

[folzag]

On listening tests, it is become a major bottleneck to testing. Do you all want me to hold up reviews in general and speakers in the specific until I listen?

I would be fine with that. I think listening tests are useful as we try and get a real-life sense for the score that @MZKM computes.

Long-term you could probably drop them most of the time, like was done with the headphone amps, but for here at the beginning we need a map from the objective to the subjective.

 

[spacevector]

If there are no max SPL measurements then we will likely end up with dubious comparisons where, say, a massive Neumann KH 420 would be considered equivalent to a tiny Neumann KH 80 simply because they both have an excellent frequency response. In reality there is a ~14 dB max SPL difference between these two speakers, but it would be impossible to tell from the current set of measurements.

Come on now, Amir does post a photo with the pink panther(s). Noone could look at both reviews and not conclude that one can get louder than the other.

 

[spacevector]

Personally, I’d be more interested in long-term compression testing.

Why though? Do you think most users suffer the consequences of long-term compression when listening at home? Just trying to understand why you think this is a more revealing test than (say) IMD.

 

[oldsysop]

To put some context: Listening and writing up thoughts takes a whole lot of time, especially doing so with any semblance of serious critical analysis. Most people who write reviews (like me) spend weeks or sometimes months listening to a speaker, not a few hours =]. Not that you actually need that long, but still.

Even with a manual turntable setup, I find measuring takes a lot less time than listening and taking notes and writing a written review that flows decently, not to mention properly positioning speakers and swapping components and whatnot. Imo best to leave the listening tests informal fo the stated purpose of the site. Listening before measuring is probably ideal, but could be a big bottleneck to measurements if aiming for a speedy pace. It can be a lot more convenient to measure several speakers back to back, for example, than to commit to listening before measuring every time. And you need to look at the measurements to make sure everything is kosher, so you can't just measure a bunch and then ignore the data until after you've listened.

Amir is Amir for his measurements and his dedication.
Your formal or informal listening doesn't interest me in the least.
To be clear, what Amir hears matters to me the same as what Paul McGowan hears, that is, nothing.
I say it with respect.

 

[hardisj]

Why though? Do you think most users suffer the consequences of long-term compression when listening at home? Just trying to understand why you think this is a more revealing test than (say) IMD.

simply put? Because compression directly impacts frequency response and that’s more of a known than distortion. I want to know how lower sensitivity speakers perform when they’ve been played for a longer period of time at a typical volume level.

 

[jhaider]

The one overall gripe I have would be to set up a checklist for measurements to display, so key data such as polar maps does not get inadvertently left out. Here's my current one for my future speaker reviews, since V-CAD grew spinorama functionality, that you could use as a template:

1) Impedance/phase angle
2) On-axis FR
3) Spinorama
4a) Horizontal polar map, raw
4b) Horizontal polar map, normalized to design axis (on axis if not specified) if useful
5) Horizontal off axis FR 0ª-15º-30ª-45º, line salad format
6) Vertical off axis FR, 0,º ±10º, ±20º, ±30º, line salad format
7) Listening window FR
8) Listening position sound power
9) Additional measurements, such as
-effect of any grill
-effect of any switches or controls

Dynamic compression would be really cool if you could figure it out, but I'm not convinced you can make stable compression measurements in your setup. It's not fair for us to expect you to empty your garage and keep it empty except for the speaker measurement robot.

The whole purpose of the spinorama is to present all important information we know matters in a simple and easy to read graph.

That assertion goes beyond the literature. It is wrong for obvious and non obvious reasons. The obvious reasons relate to output, compression, etc. The non-obvious reasons concern FR: issues over space are often easier to see in a polar map.

Let me give you an example from the Harbeth measurements. The horizontal polar plot shows a severe dispersion disruption at the crossover that isn't immediately obvious in the spinorama. The DI kink does not look notably worse than speakers with even horizontal dispersion and vertical dispersion issues, even though we know that two ears located roughly on the same horizontal plane are more sensitive to horizontal response errors than vertical response errors.

That dispersion disruption may not show up in the early reflection averages, but to someone acclimated to higher performance it will likely be heard as midrange coloration. Also, as a general rule resonances are easier to see on a polar map than a spin. They are usually present in both, but it's the difference between seeing a clear ridge of energy and noting a subtle bump in the DI when there's a dip in the FR. Lastly, and this may be an acclimation thing, but DI makes overall coverage pattern ("dispersion") harder to visualize than a polar map.

There's alot of noise about possible differences in the graph if we measure with a different reference axis, but I think that's 98% irrelevant.

There's a solid argument for doing it correctly. If only because that is...correct. That is not up for interpretation. It directly stated. See ANSI/CTA-2034, at 11.

Why go to great expense and effort to be incorrect?

2. Amir is not working for speaker manufacturers and have no obligations to them to present data in the most flattering way possible.

While that is true, it is not an excuse for ignoring manufacturer instructions.
When Harman did their room correction study, one thing they made clear was that they followed the manufacturers' instructions. That is basic step 1 for any set up.

 

[Costas EAR]

The score that @MZKM computes, replaced any need for listening test from amir.

For listening test poems, go to TAS. Plenty of poetry, oh, i'm so tired of these...

 

[MZKM]

The score that @MZKM computes, replaced any need for listening test from amir.

For listening test poems, go to TAS. Plenty of poetry, oh, i'm so tired of these...

I would not say that. The probability accuracy of the formula is pretty high, but not 100%, so it's not a golden egg.

 

[ctrl]

My last attempt to get Amir to verify the accuracy of his measurements.

In the upper part of the following diagram you can see six speaker measurements of Amir. Between 5kHz and 10kHz the ripple still looks too much like a pattern to me.

In contrast, measurements from my developed DIY loudspeakers.

The red curve shows a metal tweeter in a DIY wave guide measured at 180cm distance. In green the measurement at a distance of 30cm (with very slightly changed crossover). Nice to see, the ripple decreases when the microphone is closer to the speaker.

The blue curve shows the same DIY wave guide, but with a ring-dome tweeter. Since the wave guide has a decisive influence on the frequency response, the ripple is very similar to the red curve.

The orange curve shows a different DIY speaker. The tweeter used has its wave guide already integrated. The ripple of the measurement shows no similarity at all to the blue and red curves.

 

[spacevector]

I could do two or three reviews a day if all I had to do was post the spinorama graphs. Getting bogged down with give me this and that graph is adding a lot of time and effort to capture, document and post the results in the review.

Perhaps you can post only the most revealing graphs (say Spinorama) and the raw data. The membership generates all 'jewelry graphs' thereafter.

 

[folzag]

The score that @MZKM computes, replaced any need for listening test from amir.

Numbers without units are meaningless. Until we have a grasp of the qualitative differences between 5.5, 6.5, and 7.5, the numbers are meaningless.

 

[HammerSandwich]

Numbers without units are meaningless.

Q?

 

[Blumlein 88]

Numbers without units are meaningless. Until we have a grasp of the qualitative differences between 5.5, 6.5, and 7.5, the numbers are meaningless.

Quantity has a quality all of its own. Bigger is better.

But I get your point. Is 7.5 for say a $3000/pr speaker make paying $30,000 for 9 a good value? Is 5.5 for $300 a great bargain? How much better and in what way does 7.5 sound better/different than 5.5 in this case.

And the answer is: trained listeners rate something 7.5 less often than do civilians, or audio journalists or audiophiles. So they all rank speakers in the same order from good to bad, but the difference in the orders is dependent upon who was doing the ranking. Which now that I think of it raises a question that hasn't occurred to me. When Olive says he has .86 correlation which group of listeners is that with? I assumed it was trained listeners only. And perhaps he says that which I simply don't remember reading at the moment.