Should we (I) get into speaker testing & measurement

[Julf]

perhaps create a metric by which to express the overall performance of a speaker.

That would be very hard to do, as speakers interact differently with different rooms, and things like size, efficiency and power rating are issues for some people, but not others.

 

[NTK]

Excellent, perhaps I can at least contribute some ideas.



NTK has done the field separation, and very impressively.
He has actually written code while all I did was work on the maths.
I believe his code would do the core job.
He would be better placed than I am to add interfaces and other essential bits but I believe I could do so, especially if he advises.
I believe the author of REW measurement software is a member here too.
I always planned to use REW so it would be nice to have his advice available in case of issues in that area.
The Klippel NFS does one trick that I haven't worked out fully - that is to factor out the reflections off the speaker of the reflections from the room.
I haven't studied this too hard because it's their patented bit so it's protected even if I did work it out in detail*.
I asked NTK (does he have a real name?) his comments on it but I haven't had a response yet, probably Xmas and all.
But I think it will have minimal effect unless you try to measure a bulky speaker in a small room.
That may be an impressive stunt for Klippel to demonstrate but hardly a problem for typical use.

Best wishes
David

*Also because it looked quite difficult!

Dave: My name is Steve. I have just finished my first (very slow) reading of Klippel's patent. As you know, patents are not the most readable writings (and definitely not the most enjoyable) . I'll have to re-read at least another time to maximize my understanding (especially on the math). My plan is to post a summary of it -- primarily to help me with my understanding and let others who have read it critique/correct me, secondarily so others who don't want to read can have some idea what it is about.

I am more than happy to help as much as I can with this endeavor. As I mention in the other thread, I currently do not have any room at my current dwelling to do any mechanical hardware work. However, I do have some preliminary ideas on the robotic design. I will do a write-up. Give me a week

 

[Verausci]

That would be very hard to do, as speakers interact differently with different rooms

True, but the metric would be an assessment of the inherent performance of the speaker itself at x meters in an anechoic environment, not its performance when applied to a room divergent from anechoic conditions.

and things like size, efficiency and power rating are issues for some people, but not others.

Performance I mean in the sense of the speaker's ability to most accurately reproduce the signal being given. It could be the case where there is one small speaker and one large, both capable of reproducing the input signal to, for example, 0.75 units (1 unit being exact signal reproduction). In such a case, the choice of speaker would depend on the distance the listener intends to sit from the speaker.

Of course there are so many stages where signal alteration would occur before even reaching the speakers (quality of the DAC, power amp, etc), not to mention the microphone and ADC being used. Nonetheless I think a good theoretical assessment of a speaker would be to take a recording of some audio (a music track for example) from a speaker and 'null' the recording against the original digital file. The lesser the amount of resultant audio, the greater the speaker's performance. Though of course I think this would be very tricky to implement!

 

[Sal1950]

If I want a $300 bookshelf speaker there is no option in Revel line.

Not too far off Amir, the M16 is only $450

 

[amirm]

Not too far off Amir, the M16 is only $450

That's for each one, not a pair.

 

[Dave Zan]

Hi Steve, nice to have a name to put to the ID.

...Klippel's patent. As you know, patents are not the most readable...My plan is to post a summary of it -- primarily to help me...secondarily so others who don't want to read can have some idea what it is about.

I didn't find it an easy patent even by the usual standard of patent obscurity (and I worked in the Patent Office for a couple years).
I also find, like you, that when I try to help others understand then it helps me too, maybe even more.
So, hopefully we can at least bounce a few ideas around.

I am more than happy to help as much as I can with this endeavor. As I mention in the other thread, I currently do not have any room....to do any mechanical hardware work.

I noticed your comment in that thread, that's partly why I offered to do some of the mechanics, despite it made more sense for it to be done closer to Amir.
Even better that Amir says he has someone to do it.

...let me respond more tomorrow.

So, how much space do you have?
A CTA 2034 far field scanner would be simpler if you have the space for it, preferably 6 m or more.

Best wishes
David

 

[Wes]

My vote is that you do testing of the components that matter the most to consumers.

That means the transducers - speakers (or test different masterings / releases on CD, SACD, hiRes downloads, etc.)

 

[Sal1950]

If I want a $300 bookshelf speaker there is no option in Revel line.

Not too far off Amir, the M16 is only $450

That's for each one, not a pair.

You said "speaker" as in the singular. If you meant "a pair" that should have been specified.

 

[laurelkurt]

It is not just the low frequency response but the rest of the measurements in order to arrive at the right data set to then predict listener preference.

That can be done using semi-manual system with an automated turntable and manual or automated microphone placement. Harman measures 70 points for example. But even this setup goes well above $10,000 and still not optimal.

To get the high throughput that I have now on electronics, an automated system that performs all of these measurements quickly and far more accurately is needed. Otherwise the scattering of speaker measurements is already out there.

In a nutshell, we are doing first-class evaluation of electronics. Do we want to invest in first-class evaluation of speakers? One where the prediction results for listener preference is almost a given?

YES! But to be honest, I just spent $7,200 on a Kef R900 based system this year, so I won't be shopping any time soon. Soundstage does their measurements at NRC, Audioholics does theirs outdoors, and J Atkinson has his own methods. I gather you're proposing something much more involved with this system. I'd love to see you pull this off, but I have very limited SSDI income. I can only afford to make very occasional (small) patreon donations, which I'll try to keep up with.

 

[Celty]

I'd actually have more interest in headphones testing.

 

[mhardy6647]

I'd be happy (just) to see a bevy of impedance and phase angle curves for the loudspeakers that John Atkinson/Stereophile doesn't (don't) test!

https://www.stereophile.com/content/polk-lsiiim703-loudspeaker-measurements

FWIW, first on my list would be:

https://forum.polkaudio.com/discussion/188411/sound-vision-reviews-the-legend-l800/p1

 

[Juhazi]

Yes, that Legend L800 is interesting. The double mids and tweeters with a wedge look like giving wide dispersion, but actually the effect is just the opposite with huge interferences horizontally. Only benefit I can see is wider (but ragged) dispersion of highest treble above 10kHz

https://audioxpress.com/news/polk-a...ip-speaker-line-with-improved-speaker-designs
"this obviously aims at achieving better crosstalk cancellation designs, something that Polk Audio actually patented nearly 40 years ago, with its original Stereo Dimensional Array (SDA) loudspeaker, and for which the new updated patent now provides a new improved version addressing the “Head Related Transfer Function” issue, not by implementing complex filtering, but primarily by reforming the front of the loudspeaker enclosure shape to more closely mimic the impact of acoustic waves “crosstalking” around a human face. "

IMO that's total BS!

 

[Julf]

True, but the metric would be an assessment of the inherent performance of the speaker itself at x meters in an anechoic environment, not its performance when applied to a room divergent from anechoic conditions.

Sure, as long as you measure directivity across the frequency range, but how do you distill that into one number?

Performance I mean in the sense of the speaker's ability to most accurately reproduce the signal being given.

And my point is that that is pretty had to describe in a single number.

 

[Juhazi]

Yes, loudspeakers are extremely complicated devices technically and acoustically.

Basics compiled by Joh Atkinson, Stereophile mag. Then add personal opinions of how important each measurement is!
https://www.stereophile.com/features/99/index.html
https://www.stereophile.com/features/100/index.html
https://www.stereophile.com/features/103/index.html

 

[Verausci]

I'd actually have more interest in headphones testing.

RTINGS do some pretty comprehensive headphone tests: https://www.rtings.com/headphones

 

[aarons915]

I'd be particularly interested in using the measurements to determine the performance of vintage speakers as opposed to modern speakers.

Additionally, I think it would be interesting to compare the performance difference of these £10,000 speakers to £1000 speakers; perhaps create a metric by which to express the overall performance of a speaker. This could then be used to create some sort of price to marketing ratio or price to performance ratio to objectively assess whether the speaker is worth the price being asked for it. Not only this, but it could also be used to evaluate the effects of different design decisions and methods.

This has been done for awhile now with Harman's preference algorithm where they correlate aspects of the Spinorama to listening tests, so it could be done with the Klippel system.

 

[mhardy6647]

FWIW, I think that JA/Stereophile, to their (his!) credit, do a good job of measuring loudspeaker performance, and - to some extent - even correlating that to subjective impressions of the sound of a loudspeaker under test. To wit, JA will correlate what he measures with the effusive prose of the speaker's reviewer.

Yes, that Legend L800 is interesting. The double mids and tweeters with a wedge look like giving wide dispersion, but actually the effect is just the opposite with huge interferences horizontally. Only benefit I can see is wider (but ragged) dispersion of highest treble above 10kHz

https://audioxpress.com/news/polk-a...ip-speaker-line-with-improved-speaker-designs
"this obviously aims at achieving better crosstalk cancellation designs, something that Polk Audio actually patented nearly 40 years ago, with its original Stereo Dimensional Array (SDA) loudspeaker, and for which the new updated patent now provides a new improved version addressing the “Head Related Transfer Function” issue, not by implementing complex filtering, but primarily by reforming the front of the loudspeaker enclosure shape to more closely mimic the impact of acoustic waves “crosstalking” around a human face. "

IMO that's total BS!

View attachment 44122

Well, in fairness, the intent (minimization of driver crosstalk) of the L800 (and of the earlier SDA loudspeakers from Polk) has to be taken into account when considering their unusual aesthetics/design. In the context of what they're tryin' to do -- they do know what they are doing.

Frankly, having had a chance to hear the much smaller (but not inexpensive) L200 bookshelf/stand mount loudspeakers from the new Polk "Legend" line at an in-home, real-world demo... I am really interested in some objective assessment of what sort of load they really present to an amplifier (i.e., how easy, or hard, they really are to drive). The L200s sound good if a little artificial (to my ears and taste).

Full disclosure -- I am a Polk fanboy from way, way back. A pair of Monitor Series Model 7A was (were) my gateway drug into 'modern' hifi ca. 1978. Still have 'em. Still like 'em.

polk_brochure_back_pg by Mark Hardy, on Flickr

DSC_5417 by Mark Hardy, on Flickr

Oh, and Matt Polk and I share an alma mater, as well -- so there is that.

 

[napilopez]

FWIW, I think that JA/Stereophile, to their (his!) credit, do a good job of measuring loudspeaker performance, and - to some extent - even correlating that to subjective impressions of the sound of a loudspeaker under test. To wit, JA will correlate what he measures with the effusive prose of the speaker's reviewer.



Well, in fairness, the intent (minimization of driver crosstalk) of the L800 (and of the earlier SDA loudspeakers from Polk) has to be taken into account when considering their unusual aesthetics/design. In the context of what they're tryin' to do -- they do know what they are doing.

Frankly, having had a chance to hear the much smaller (but not inexpensive) L200 bookshelf/stand mount loudspeakers from the new Polk "Legend" line at an in-home, real-world demo... I am really interested in some objective assessment of what sort of load they really present to an amplifier (i.e., how easy, or hard, they really are to drive). The L200s sound good if a little artificial (to my ears and taste).

Full disclosure -- I am a Polk fanboy from way, way back. A pair of Monitor Series Model 7A was (were) my gateway drug into 'modern' hifi ca. 1978. Still have 'em. Still like 'em.

polk_brochure_back_pg by Mark Hardy, on Flickr

DSC_5417 by Mark Hardy, on Flickr

Oh, and Matt Polk and I share an alma mater, as well -- so there is that.
View attachment 44149

I haven't heard the L800s, but I do find the approach to reduce interaural crosstalk interesting. It should in theory at least provide abetter phantom center by minimizing the 2KHz-ish dip inherent in stereo listening.

As for the L200, I measured their horizontal response about a month ago. It's a good speaker.

 

[Juhazi]

L200 in conventional and it's horizontal off-axis is not good at all.

The clain that SDA-PRO of L800 works because it has twin mids and twin tweeters at same separation as our ears it the total marketing BS I meant. Just think about it a little... We need two widely separated speakers to get stereophonic illusion with our two ears, 25cm separation of a single speaker simply just messes everything up. This messing up might be heard as "enhanced space", a bit like what several dsp settings of HT receivers can do for stereo signal. L800 are supposed to be used as L/R main speakers in multichannel setup, so double mids and tweeters are playing same signal (L or R channel). Polk has Hifi line for stereo listening, with pretty good and sane looking multiway speakers.

https://polkaudio.com/discover/sda-technology

versus this is how stereo hearing/imaging really works https://en.wikipedia.org/wiki/Stereophonic_sound
- we need two receptors, but two sources with same 25cm separation will just make interferences that make haphazard phantom effects in treble.
- with 2-channel stereo both time and amplitude difference are used, mostly just spl difference because monophonic mics are used in studio.

Appendix
There is more info about SDA-PRO in manual here https://d3vqw2nv1topde.cloudfront.net/assets/Product Documents/Legend/Polk_Legend_Owners_Manual_English_L800_MP.pdf

so, they say they are sending ipsilateral signal in reverse phase to other of tweeters (or mids too?) . They do this with a single cable that connects L and R speakers. This is said to make appropriate cancelling just in front of the head of the listener. All this without any dsp. But how do they handle the varying frequencies?...

 

[ScottG]

I didn't read through the whole thread, but this just looks like an automated near-field polar with cepstral editing.

The automated polar is nothing new, though their technique is different.

Note: getting closer to the source entails losses in measurement accuracy (with combing artifacts and non-coincident sources/drivers), so the Klippel machine is making measurements closer and further out to compensate with probably some modest gating at further distances for higher freq.s (which also results in some measurement accuracy loss).

http://www.audiomatica.com/wp/wp-content/uploads/audiomaticaISEAT200911.pdf

As for the cepstral editing:

http://musicanddesign.speakerdesign.net/Cepstral.html


The fact is that you can do this with a more typical turn-table with the typical gating + near-field, add-in cepstral editing for those low freq.s in a good environment (sealed with repeatable low-freq. behavior), and get results that are at least as good (higher resolution). You could get half of your vertical at the same time you are doing your horizontal by making an array of mic.s vertically (an "arc") - you could do this cheaply with inexpensive mics. with a good calibration file.

https://outline.it/outline-products/measurement-systems/et-250-3d/

I'd suggest EASERA and an audio interface with a lot of inputs (for those vertical mic.s):
http://easera.afmg.eu/index.php/es-software-en.html

Soundeasy has cepstral editing:
http://www.bodziosoftware.com.au/