Should we (I) get into speaker testing & measurement

[Mashcky]

I have been following this thread closely and have little to add except my votes/encouragement for the following ideas:

• Speaker measurements are perhaps the greatest good ASR could contribute to consumer audio and the Klippel-or-nothing is a reasonable conclusion given time and energy constraints on one person.
• Attempt a fundraising campaign to raise the additional $60k ($70k?) beyond what Amir is interested in contributing.
• Trial of advertisements on ASR to supplement a campaign but priority to not ads if possible. Bias is real but can be compensated for to some degree by awareness of potential bias (see research on pharmaceutical contributions to doctors, etc). However, the illusion of independence is valuable to readers.

 

[North_Sky]

Basic loudspeaker's measurements are important, frequency response on and off axis.
We need to know that in addition to impedance across the full audio spectrum, overall distortion level, phase, power response, bass limit, ...etc., for best match to its dedicated amplifier of choice.

So I voted yes. And that goes for subwoofers too.

 

[amirm]

Do you have any opinion about the speaker measurements data presented on this site:

https://www.soundstagenetwork.com/index.php?option=com_content&view=article&id=16&Itemid=140

The measurements are useful but not really there. They have their own notions of what a "listener window is" as far as the angles they measure. Such measurement is not tested against listener preference so you can't easily judge how good the speaker is.

My goal with this project is to 100% comply with the research so that we can use its predictive power (as specified in CEA-2034A/Harman research). We can't do this with soundstage measurements.

 

[Blumlein 88]

@amirm
Could you get hold of the Klippel results for one or preferably two or three speakers, and do a review/write up as an example of what such testing shows us vs the kind of measurements we normally see?

Even if you aren't allowed to reveal the speaker it would be educational and generate some additional interest. It could be a measurement review of Mystery speaker A, Mystery speaker B and Mystery speaker C.

 

[Vovgan]

My goal with this project is to 100% comply with the research so that we can use its predictive power

That, coupled with your skills at explaining the meaning of measurements to lay people like me, would surely become the best resource to go to, but having read this thread I had a feeling that you’d parked this idea of starting measuring speakers, hence my desire to find the next best option(s).
However, if you haven’t parked the idea, please give us a feeling when can we expect to see things starting.
Thanks!

 

[Ilkless]

Is it possible to get a short loan from Klippel to churn out a few detailed datasheets (eg. of your own Revels), so that potential contributors on the fence can really see how effective and efficient the Klippel is so everyone knows what they stand to get, without having to rely on Klippel's goodwill for the long term.

 

[RayDunzl]

Is it possible to get a short loan from Klippel

I would think "No".

 

[Ilkless]

I would think "No".

Ah. Or some other method to get sample datasheets of speakers that are well-known/owned by several members on ASR. I would think that samples would make the cause much more persuasive, especially to those who aren't already in the ASR community. For instance, it appears Warkwyn (MISCO's consulting arm) has a NFS set up. I wonder how much it'd cost to send one/a few well-regarded speaker that hasn't been measured at a relatively authoritative source (eg. current PMC and ATC standmounts).

 

[nhunt]

I would think "No".

Well, I think Amir mentioned in this thread that there is a company in the US who will do measurements as a service. I think it was a 1000 dollars plus some more service charges per unit or something along those lines.

 

[amirm]

@amirm
Could you get hold of the Klippel results for one or preferably two or three speakers, and do a review/write up as an example of what such testing shows us vs the kind of measurements we normally see?

You can get a lot of data depending on what options you pay for. My main interest though is in generating the so called "spin data" as standardized now in CEA-2034A standard. Here is a Revel F22e as measured such:

The black curve is on-axis measurement. Note how it goes down to 40 Hz or so. The typical measurement you see in magazines like stereophile consists of two measurements pieced together. On that goes from say 100 Hz and goes to 20 kHz. And another that goes from 100 Hz down. The former is measured as you could. But then it is "gated" to truncate any room reflections. Doing so eliminates the resolution of the graph in low frequencies. So another measurement is made with a microphone close to the woofer for low frequencies. Problem with this method is that there is a amplification of sound due to speaker front surface artificially amplifying the amount bass energy there really is. Here is JA's measurement of F228Be showing this issue: https://www.stereophile.com/content/revel-performa-f228be-loudspeaker-measurements

I have circled the hump which as you can tell, does NOT exist in the proper measurements Klippel performs where it eliminates the reflection mathematically. It is then able to use the same measurement distance as for high frequencies, avoiding the hump in stereophile measurements.

With respect on-axis performance then, we have the full response from a few hertz to 20+ kHz allowing us to examine whether it is flat or not. The flatter it is, the more accurate the speaker. And the more accurate, the better listener preference is.

Next is the "listening window." Here, it is better look at a difference between it, and on-axis window as expressed in orange: the "Early Reflection DI." Here, we want a smooth curve with no dips and peaks and we have that with F228 speaker, telling us this speaker performs well in room with much less messing around with acoustic products and such. Its reflections will be similar to direct sound and thus, better fused into one by the brain.

In contrast to this very useful measurement, we have a bunch of ad-hoc measurement points in stereophile:

It is not easy to characterize these dips and peaks as easily as we can with the orange line. And at any rate, there is no study correlating them to listening preference.

If I get this system, I will do a much more extensive write-up than this brief.

 

[amirm]

That, coupled with your skills at explaining the meaning of measurements to lay people like me, would surely become the best resource to go to, but having read this thread I had a feeling that you’d parked this idea of starting measuring speakers, hence my desire to find the next best option(s).

Thanks. I definitely have NOT parked the idea. I am working it behind the scenes, seeing how it can be done. I have upped how much I will personally invest for example. And removed the dependency on donations to get this started.

Every morning for example, I open by piggy bank and count the pennies in there.

 

[MZKM]

You can get a lot of data depending on what options you pay for. My main interest though is in generating the so called "spin data" as standardized now in CEA-2034A standard. Here is a Revel F22e as measured such:

The black curve is on-axis measurement. Note how it goes down to 40 Hz or so. The typical measurement you see in magazines like stereophile consists of two measurements pieced together. On that goes from say 100 Hz and goes to 20 kHz. And another that goes from 100 Hz down. The former is measured as you could. But then it is "gated" to truncate any room reflections. Doing so eliminates the resolution of the graph in low frequencies. So another measurement is made with a microphone close to the woofer for low frequencies. Problem with this method is that there is a amplification of sound due to speaker front surface artificially amplifying the amount bass energy there really is. Here is JA's measurement of F228Be showing this issue: https://www.stereophile.com/content/revel-performa-f228be-loudspeaker-measurements

View attachment 38051

I have circled the hump which as you can tell, does NOT exist in the proper measurements Klippel performs where it eliminates the reflection mathematically. It is then able to use the same measurement distance as for high frequencies, avoiding the hump in stereophile measurements.

With respect on-axis performance then, we have the full response from a few hertz to 20+ kHz allowing us to examine whether it is flat or not. The flatter it is, the more accurate the speaker. And the more accurate, the better listener preference is.

Next is the "listening window." Here, it is better look at a difference between it, and on-axis window as expressed in orange: the "Early Reflection DI." Here, we want a smooth curve with no dips and peaks and we have that with F228 speaker, telling us this speaker performs well in room with much less messing around with acoustic products and such. Its reflections will be similar to direct sound and thus, better fused into one by the brain.

In contrast to this very useful measurement, we have a bunch of ad-hoc measurement points in stereophile:

It is not easy to characterize these dips and peaks as easily as we can with the orange line. And at any rate, there is no study correlating them to listening preference.

If I get this system, I will do a much more extensive write-up than this brief.

Interesting tidbit, John Atkinson doesn’t measure pink noise, he measures an impulse response, and uses an FFT to convert that to a ~ 200Hz-25000Hz measurement. This means he gets SPL measurements at intervals, and the software plays connect the dots; so while pretty close, it’s not 100% a true frequency response.

 

[mitchco]

As mentioned before, one can download the CTA 2034 A standard for free. Lots of good reading plus considerable references.

I am not only interested in the spins but also the predicted in room response, which is part of the standard reports to be produced:

"Unlike previously published standards, this standard describes how to measure and report the directivity of a loudspeaker, whether it stands by itself or is mounted in or on a wall or ceiling. It also describes how to use this directivity data to estimate the in-room frequency response that more recent research has shown correlates well to subjective listening preferences of consumers."

Here is an example from the standard (page 37) of the predicted in-room response from anechoic data and the measured in-room response comparing the two. Amazingly accurate:

 

[Audiocrusader]

Another great option:

 

[BDWoody]

If I get this system, I will do a much more extensive write-up than this brief.

Are you still teasing yourself with the idea that you may not do this? Seems like the ball has started rolling down the hill...which is great by the way. I certainly hope so.

You have created quite a special place for yourself in the audio world with this site, and being able to really get into these measurements and the equipment and everything that would follow would be fascinating, and I'm guessing quite impactful!

Imagine, measuring boxes that aren't going to pretty much sound the same, assuming they aren't broken...

 

[Audiocrusader]

May be of interest to some:

https://www.audioxpress.com/article/book-review-how-to-design-and-specify-loudspeaker-drivers

 

[amirm]

Another great option:

As I thought I already commented, this box won't do the job we are talking about. It is more of a data acquisition than automation of speaker measurements. It is akin to using a PC and a USB microphone with REW.

 

[Audiocrusader]

As I thought I already commented, this box won't do the job we are talking about. It is more of a data acquisition than automation of speaker measurements. It is akin to using a PC and a USB microphone with REW.

The book I linked to in my last post has a detailed section of pros/cons of all loudspeaker measurement suites including Kippel and Loudsoft. Worth a read for anyone serious about this stuff. I think it's a bit more comprehensive than REW.

Fine QC may offer what you're talking about:

 

[DDF]

But then it is "gated" to truncate any room reflections. Doing so eliminates the resolution of the graph in low frequencies. So another measurement is made with a microphone close to the woofer for low frequencies. Problem with this method is that there is a amplification of sound due to speaker front surface artificially amplifying the amount bass energy there really is

Restating this in technical terms (same meaning) Stereophile measures don't modify the near field woofer measurements with the baffle diffraction signature at the listening distance. So the real response below 300Hz is highly suspect. My method to doing this is linked here:
https://www.diyaudio.com/forums/mul...ing-crossovers-measurement-3.html#post2595287
Dead link edit, new link for my presentation FWIW: http://www.audio.claub.net/software.html

The other drawback to JA's measures is that the impulse windowing causes the frequency response to be averaged in ~ 200 Hz sliding windows, obscuring tight peaks and dips. Certainly still better than nothing.

 

[Audiocrusader]

Retail price in CAD on all Loudsoft products:

https://solen.ca/manufacturer/loudsoft/