Which speaker would you prefer?

[shadyJ]

Wouldn't those two additional assumptions vary depending on who measured them?

Are Stereophiles tests done in an anechoic chamber?

Stereophile uses quasi-anechoic windowed measurements, they do not have an anechoic chamber.

One thing about those measurements is that the high-end around 10 kHz and above tends to change the most rapidly as you move off-axis, so even a modest toe-in can reduce that brightness of that elevated high-treble. Another thing to keep in mind is that high-treble region is less reflected than lower frequencies, so if you want to hear that response, you need to be sure you are listening on-axis.

 

[<eargiant]

Stereophile uses quasi-anechoic windowed measurements, they do not have an anechoic chamber.

One thing about those measurements is that the high-end around 10 kHz and above tends to change the most rapidly as you move off-axis, so even a modest toe-in can reduce that brightness of that elevated high-treble. Another thing to keep in mind is that high-treble region is less reflected than lower frequencies, so if you want to hear that response, you need to be sure you are listening on-axis.

That's another great point - the relation between elevated treble and on-axis listening.

So are anechoic measurements more accurate than the quasi-anechoic method that Stereophile employs?

 

[shadyJ]

That's another great point - the relation between elevated treble and on-axis listening.

So are anechoic measurements more accurate than the quasi-anechoic method that Stereophile employs?

Anechoic is not necessarily more accurate. Anechoic just grants you accuracy down to a lower frequency, but even an anechoic chamber you are still limited to what low frequencies you can test at. This is the advantage to having large anechoic chambers, this allow you to test at lower frequencies. If you wanted an anechoic chamber that could be used down to 20 Hz, it would need to be the size of a football stadium (no such anechoic chambers like that exist, they would cost many millions of dollars to build). BUT you can get extremely accurate low frequency results by just doing groundplane measurements in an open area, so anechoic chambers are not needed for accurate low-frequency response measurements.

Back to your question though, one thing I will say is that anechoic (and quasi-anechoic) IS more accurate and better than responses that are derived from summing up close mic measurements of drivers which you sometimes see in reviews. Anechoic is also very useful for taking distortion measurements at low drive levels and also CSD measurements.

 

[Dennis Murphy]

That's another great point - the relation between elevated treble and on-axis listening.

So are anechoic measurements more accurate than the quasi-anechoic method that Stereophile employs?

Quasi-anechoic doesn't mean the graphs have room effects in the treble region. The program windows out room reflections. In the anechoic chamber, there aren't any reflections, so a longer window is possible with increased resolution. But the general shape of the curve would be the same. The differences between the two methods get more pronounced when you get into the bass region, since what you see on the anechoic chamber plot is the inherent bass response down to around 80 Hz, whereas the Stereophile plot splices a nearfield bass response onto the quasi-anechoic in the 200 - 300 Hz region. That's necessary because a measurement window that's short enough to avoid room reflections above 300 Hz won't be wide enough to capture the longer bass waves.

 

[<eargiant]

Anechoic is not necessarily more accurate. Anechoic just grants you accuracy down to a lower frequency, but even an anechoic chamber you are still limited to what low frequencies you can test at. This is the advantage to having large anechoic chambers, this allow you to test at lower frequencies. If you wanted an anechoic chamber that could be used down to 20 Hz, it would need to be the size of a football stadium (no such anechoic chambers like that exist, they would cost many millions of dollars to build). BUT you can get extremely accurate low frequency results by just doing groundplane measurements in an open area, so anechoic chambers are not needed for accurate low-frequency response measurements.

Back to your question though, one thing I will say is that anechoic (and quasi-anechoic) IS more accurate and better than responses that are derived from summing up close mic measurements of drivers which you sometimes see in reviews. Anechoic is also very useful for taking distortion measurements at low drive levels and also CSD measurements.

Quasi-anechoic doesn't mean the graphs have room effects in the treble region. The program windows out room reflections. In the anechoic chamber, there aren't any reflections, so a longer window is possible with increased resolution. But the general shape of the curve would be the same. The differences between the two methods get more pronounced when you get into the bass region, since what you see on the anechoic chamber plot is the inherent bass response down to around 80 Hz, whereas the Stereophile plot splices a nearfield bass response onto the quasi-anechoic in the 200 - 300 Hz region. That's necessary because a measurement window that's short enough to avoid room reflections above 300 Hz won't be wide enough to capture the longer bass waves.

Great stuff, thank you both.

For the record, I believe Speaker A was measured in an anechoic chamber and Speaker B was measured in a quasi-anechoic.

So for this comparison if we only focus on the top end frequencies of the graphs I posted (even if one was done in a quasi-anechoic and the other was done in an anechoic) a reasonable informed person would probably surmise that Speaker A would have a more neutral top end. Is that a fair statement? In other words, for the upper frequencies it shouldn't matter if they were done in a quasi-anechoic vs an echoic chamber. Is that correct?

Or is there something else we should be looking at before we objectively conclude that Speaker B has an elevated top end?

 

[PENG]

Based on the graphs, most who then listen to Speaker A we will probably believe it is clearly superior and be predisposed to hearing a very neutral flat top end and most who listen to Speaker B after looking at the graph will also be predisposed and comment on it's elevated top-end that they can clearly hear.

Thanks, good to know I am not among the "most" then.

 

[shadyJ]

Great stuff, thank you both.

For the record, I believe Speaker A was measured in an anechoic chamber and Speaker B was measured in a quasi-anechoic.

So for this comparison if we only focus on the top end frequencies of the graphs I posted (even if one was done in a quasi-anechoic and the other was done in an anechoic) a reasonable informed person would probably surmise that Speaker A would have a more neutral top end. Is that a fair statement? In other words, for the upper frequencies it shouldn't matter if they were done in a quasi-anechoic vs an echoic chamber. Is that correct?

Or is there something else we should be looking at before we objectively conclude that Speaker B has an elevated top end?

Yes, if the speaker were measured with same mics in some positions with respect to the speaker, then anechoic and quasi-anechoic will show the same thing. Provided the speakers were measured competently, speaker A shows a more neutral top end.

One thing it would help to know before assuming totally comparable results between these graphs is where was the mic with respect to the speakers? Lets say speaker B was measured at 1 meter with mic facing the tweeter, and speaker A was measured at 2 meters with mic on midrange woofer height; the mic position at speaker B would inevitably capture more treble, and so the comparability of these measurements might not be 100%. Equipment can male a difference too, as some mics can have differing sensitivities in upper frequencies, and so they need a calibration file to correct against.

 

[AcuDefTechGuy]

Let's assume that sensitivity, impedance and distortion are nearly identical. Then what? Is that sufficient to confirm that Speaker B is less neutral? Has an elevated top end?

In reality I think they are probably equal from below 9kHz, which is the salient point.

Academically Speaker A is smoother from above 9kHz. But as pointed out, most music contents aren’t above 9kHz anyway.

 

[Beave]

I would prefer which ever of speaker A and speaker B is cheaper. That's it, that's the only determining factor, other than perhaps the color of the finish. But even there, I have no strong opinion one way or the other, so price is it.

 

[Beave]

In addition to the different measurement techniques discussed above, don't forget the following:

The measurements are both supposedly on-axis listening window, but the two measurers used slightly different axes (axises?). That accounts for some of the differences in the very high treble, which is very directional in general and especially so with this particular tweeter.

Also, the anechoic plot does not include the contributions of the rear port. The pseudo-anechoic plot does, but it is questionable in accuracy because there is something big that gets in the way of the little capsule mic he used, and that alters his measured response and level. His overall response plot may be off in the bass because of limitations of his technique in situations like this.

What's more, the anechoic plot and the pseudo-anechoic plot assume different spaces for the bass plotting, ie, 4pi space and 2pi space.

When all of these things are taken into account, I think you will all find that the two speakers sound more alike than you might think.

 

[<eargiant]

Yes, if the speaker were measured with same mics in some positions with respect to the speaker, then anechoic and quasi-anechoic will show the same thing. Provided the speakers were measured competently, speaker A shows a more neutral top end.

One thing it would help to know before assuming totally comparable results between these graphs is where was the mic with respect to the speakers? Lets say speaker B was measured at 1 meter with mic facing the tweeter, and speaker A was measured at 2 meters with mic on midrange woofer height; the mic position at speaker B would inevitably capture more treble, and so the comparability of these measurements might not be 100%. Equipment can male a difference too, as some mics can have differing sensitivities in upper frequencies, and so they need a calibration file to correct against.

Everything you are saying makes perfect sense.

I'm not sure what their protocols are but one speaker was measured by NRC and the other was measured by Stereophile's John Atkinson. How do their techniques compare (aside from the anechoic chamber)?

 

[Beave]

The NRC plot was taken on an axis between the midrange and tweeter.

The JA/Stereophile plot was taken on axis with the tweeter.

 

[shadyJ]

If I remember right, the NRC measures at 2 meters, and Atkinson measures at something like 1.2 meters. That can certainly make a difference, especially on large speakers. The farther you get from a speaker, the more its sound will cohere, so 2m is generally better, but 1m will be sufficient for things like near-field monitors and small bookshelf speakers. Atkinson close-mics the woofer and port, and that will always give him an artificially boosted bass response. To his credit, he usually mentions this. I think, but am not 100% sure, that NRC just uses a groundplane-based response correction for bass.

Something else to keep in mind for the off-axis measurements between these two is the reference axis they use, or the axis of rotation for the various angles of measurement. This can make a difference but probably not a huge one. I only bring this up to stress that no one's measurements are going to be 100% comparable to each other.

 

[Dennis Murphy]

In reality I think they are probably equal from below 9kHz, which is the salient point.

Academically Speaker A is smoother from above 9kHz. But as pointed out, most music contents aren’t above 9kHz anyway.

Most people (but not teenagers, who aren't people and can hear higher) would have trouble picking out a peak at around 15 kHz. But the elevated portion on graph 2 is of sufficient magnitude and wide enough so that it would be pretty easy to pick out. Most music does in fact have a good bit of content above 9 kHz. There are lots of overtones in the elevated portion, and it would definitely affect the sound of strings, winds, and even brass, although not in a way that a lot of people would find objectionable.

 

[Beave]

Well done, <eargiant.

 

[PENG]

But the elevated portion on graph 2 is of sufficient magnitude and wide enough so that it would be pretty easy to pick out.

Do you think post#32 (if factual) explained that elevated portion though? I assume it would because it looked too smooth otherwise, almost like the effect of a properly selected high shelf filter.

 

[PENG]

If I remember right, the NRC measures at 2 meters, and Atkinson measures at something like 1.2 meters.

If I remember right, the NRC measures at 2 to 3 meters depending on the size/types of speakers, then converted to 1 meter equivalent for the plots. To be more precise, I am referring to the Soundstage.com plots measured at the NRC chamber, not sure if it is NRC's own protocol as such.

 

[AcuDefTechGuy]

Most people (but not teenagers, who aren't people and can hear higher) would have trouble picking out a peak at around 15 kHz. But the elevated portion on graph 2 is of sufficient magnitude and wide enough so that it would be pretty easy to pick out. Most music does in fact have a good bit of content above 9 kHz. There are lots of overtones in the elevated portion, and it would definitely affect the sound of strings, winds, and even brass, although not in a way that a lot of people would find objectionable.

If that’s the case, then Speaker A is definitely the winner purely based on those 2 graphs because Speaker B has a very extended broad +5dB peak from 9kHz-20kHz - again assuming everything else being equal, including price (especially if Speaker A is $3,000 and Speaker B is $5,000).

 

[AcuDefTechGuy]

There are many speakers available today that have both very good measurements and sound quality. People don’t have to settle for one or the other.

But even the best measured speakers and subs available today will have worse measured specs (FR, Noise, THD, etc) than even a $300 AVR.

For example, even a $300 AVR has FR that looks like 10Hz-30kHz +/- 0.5dB and THD of 0.09%.

Bottom line, buy the speakers your ears favor the best. And don’t get ripped off.

 

[Dennis Murphy]

Do you think post#32 (if factual) explained that elevated portion though? I assume it would because it looked too smooth otherwise, almost like the effect of a properly selected high shelf filter.

The Stereophile plot shows an elevated high end on the tweeter axis. That's about all that can be concluded. Moving the mic down a little to between the tweeter and adjacent driver might reduce the peak marginally, but but it would still look basically the same unless there was a huge distance between the drivers, which I don't believe is the case for this speaker.