New set of graphs for all speakers reviews

[pierre]

Is there a way to make https://pierreaubert.github.io/spinorama/scores.html more compact or showing which review the score is based off of like how it used to be?

The score there can also be clicked on. I'm assuming it's a button meant to take the user to the measurements?

working on it; it will come. I have a hard time to make a wide table like this works well on mobile.

 

[pierre]

Clicking on Click here on the Neumann KH120 page leads to a 404 error https://pierreaubert.github.io/spinorama/Neumann KH 120A/Misc/index_misc-dominikz.html

fixed

 

[Arc Acoustics]

@pierre Happy (new y)ear!
Thank you for another great job.
With Eq and Ref. v.s. with Eq SPINORAMA plot should be valuable information for many many users.

By the way, the link of "Link to Original review." is not set in the below 7 models.
So please fix them in your spare time.

JBL 705i
Klipsch Synergy SB-3 Monitor
Part Express DIY C-Note
Pioneer S-DF3-K
Polk Audio 5JR
Polk Audio RT15i
Yamaha NS-A638

 

[Arc Acoustics]

The order of the off-axis plot looks strange as well.
(e.g. 50° On 60° 30° 20° 10° 40°)
https://pierreaubert.github.io/spinorama/JBL Stage A135C/ASR/asr/SPL Horizontal Normalized.jpg

And "Directivity (-*°, *°) between 1kHz and 10kHz. Angle computed for +/-6dB." calculation *could be* inappropriate (heavily smoothed?) I think.
Cuz, I mean, JBL Stage A135C is a 2way centre channel with SIX horizontal mid-woofers so it has extremely nasty horizontal directivity, the data shows even 20deg hits -30dB! compared to the on-axis, but it has very wide tweeter directivity as well and is rated as -53.0°, 52.0° (1-10kHz +/-6dB).
Of course, the design itself does not make sense at all but I should say the rating looks strange a bit as well to me.
https://pierreaubert.github.io/spinorama/JBL Stage A135C/ASR/index_asr.html
https://pierreaubert.github.io/spin...ASR/asr/SPL Horizontal Contour Normalized.jpg

 

[BoredErica]

Is it just me or are the graphs super big and pixelated now? Can't zoom out either. https://pierreaubert.github.io/spinorama/Kii Audio Three/ErinsAudioCorner/index_eac.html

 

[Arc Acoustics]

Is it just me or are the graphs super big and pixelated now? Can't zoom out either. https://pierreaubert.github.io/spinorama/Kii Audio Three/ErinsAudioCorner/index_eac.html

Looks like the "image version" of the graph.
Just click over the image will link to the "data version".
(Or did I read wrong?)

 

[pierre]

Is it just me or are the graphs super big and pixelated now? Can't zoom out either. https://pierreaubert.github.io/spinorama/Kii Audio Three/ErinsAudioCorner/index_eac.html
View attachment 177687

Looks like the "image version" of the graph.
Just click over the image will link to the "data version".
(Or did I read wrong?)

@Acoustician is correct; you are in the picture version (light by design for people behind a slow connection). Click in it and you will get a dynamic version which is high resolution (or at least same resolution as your screen).

 

[pierre]

@pierre Happy (new y)ear!
Thank you for another great job.
With Eq and Ref. v.s. with Eq SPINORAMA plot should be valuable information for many many users.

By the way, the link of "Link to Original review." is not set in the below 7 models.
So please fix them in your spare time.

JBL 705i
Klipsch Synergy SB-3 Monitor
Part Express DIY C-Note
Pioneer S-DF3-K
Polk Audio 5JR
Polk Audio RT15i
Yamaha NS-A638

Thanks. I have found some reviews and added them. I am still missing a few. The spinorama used to be on forum but I do not find them.

 

[pierre]

The order of the off-axis plot looks strange as well.
(e.g. 50° On 60° 30° 20° 10° 40°)
https://pierreaubert.github.io/spinorama/JBL Stage A135C/ASR/asr/SPL Horizontal Normalized.jpg

Thanks this is fixed but not yet regenerated.

And "Directivity (-*°, *°) between 1kHz and 10kHz. Angle computed for +/-6dB." calculation *could be* inappropriate (heavily smoothed?) I think.
Cuz, I mean, JBL Stage A135C is a 2way centre channel with SIX horizontal mid-woofers so it has extremely nasty horizontal directivity, the data shows even 20deg hits -30dB! compared to the on-axis, but it has very wide tweeter directivity as well and is rated as -53.0°, 52.0° (1-10kHz +/-6dB).
Of course, the design itself does not make sense at all but I should say the rating looks strange a bit as well to me.
https://pierreaubert.github.io/spinorama/JBL Stage A135C/ASR/index_asr.html
https://pierreaubert.github.io/spinorama/JBL Stage A135C/ASR/asr/SPL Horizontal Contour Normalized.jpg

Here is how I define it:

i compute the angle by looking for the degree such that it minimise the distance between data-6dB and a line (linear regression) over the frequency range 1k-10k.

I would be happy to change the definition if someone has a better idea. I want something reproducible so we can compare between speakers.

One option is to look at a worst case over some range ...

 

[Flak]

I'm writing just to congratulate you, nice job!!

 

[BoredErica]

@Acoustician is correct; you are in the picture version (light by design for people behind a slow connection). Click in it and you will get a dynamic version which is high resolution (or at least same resolution as your screen).

Spoiler

I'm sorry, I'm still confused. Up until 2 days ago when I clicked on measurements, it takes me to a page where the images are normal sized and all of a sudden they're huge, and I can't zoom out. Yet when I press Ref vs EQ, the charts become normal sized, except it looks glitchy like the second picture I posted. Is it really intended for images to be huge for ref and with eq, but normal sized for ref vs eq page, and for the footer to be behind the pictures?

When I click on a graph I get a more detailed graph like this:

Spoiler

But it only shows the one graph I clicked, so there wouldn't be a way to view all graphs in one page. I'd have to click a huge image, look, go back, scroll down, click on the next image, etc.

 

[pierre]

Thanks this is fixed but not yet regenerated.



Here is how I define it:

i compute the angle by looking for the degree such that it minimise the distance between data-6dB and a line (linear regression) over the frequency range 1k-10k.

I would be happy to change the definition if someone has a better idea. I want something reproducible so we can compare between speakers.

One option is to look at a worst case over some range ...

@Acoustician
I tried another option: instead of looking for an average, take the worst case for each octave between 1k and 10k
that gives much lower results, here are a few examples

Genelec 8030C

'dir_horizontal_p': 25.0, 'dir_horizontal_m': -18.0, 'dir_horizontal': 21.5, 'dir_vertical_p': 72.0, 'dir_vertical_m': -25.0, 'dir_vertical': 48.5

Note that we have a lot of variation for the same speaker with different measurements:

speaker='Emotiva Airmotiv B1+' ASR v.s. EAC measurements normalized or not

dir_horizontal_p': 66.0, 'dir_horizontal_m': -85.0, 'dir_horizontal': 75.5, 'dir_vertical_p': 70.0, 'dir_vertical_m': -47.0, 'dir_vertical': 58.5
dir_horizontal_p': 53.0, 'dir_horizontal_m': -46.0, 'dir_horizontal': 49.5, 'dir_vertical_p': 73.0, 'dir_vertical_m': -30.0, 'dir_vertical': 51.5

dir_horizontal_p': 30.0, 'dir_horizontal_m': -71.0, 'dir_horizontal': 50.5, 'dir_vertical_p': 71.0, 'dir_vertical_m': -30.0, 'dir_vertical': 50.5
dir_horizontal_p': 15.0, 'dir_horizontal_m': -9.0, 'dir_horizontal': 12.0, 'dir_vertical_p': 75.0, 'dir_vertical_m': -9.0, 'dir_vertical': 42.0


2022-01-11 06:13:37,737 - compute_estimates.py:estimates:101 - DEBUG - Estimates: {'ref_from': 300.0, 'ref_to': 10000.0, 'ref_level': -0.0, 'ref_3dB': 60.0, 'ref_6dB': 53.0, 'ref_band': 1.8, 'sensitivity_delta': -6.459479416000911e-16, 'dir_horizontal_p': 15.0, 'dir_horizontal_m': -9.0, 'dir_horizontal': 12.0, 'dir_vertical_p': 75.0, 'dir_vertical_m': -9.0, 'dir_vertical': 42.0}

 

[Arc Acoustics]

Here is how I define it:

i compute the angle by looking for the degree such that it minimise the distance between data-6dB and a line (linear regression) over the frequency range 1k-10k.

I would be happy to change the definition if someone has a better idea. I want something reproducible so we can compare between speakers.

One option is to look at a worst case over some range ...

Thank you for your explanation.
Yeah, nasty directivity troubles us a lot, not only acoustically.

Basically, I see the angle of the directivity as the divergence of the trend line from the on-axis, but I also "hope" the angle provides a "usable" response as well.
I imagine applying the ERB or some kind of smoothing and taking both

• smoothed off-axis response has "acceptable" smoothness (but the calculation of R^2 should be modified because it has a strong sensitivity to the slope if my understanding is correct. divergence (dB) from the regression line could be another idea.)
• divergence of the regression line from the on-axis is no more than 6dB

could give us useful information.

@Acoustician
I tried another option: instead of looking for an average, take the worst case for each octave between 1k and 10k
that gives much lower results, here are a few examples

You're so much faster than my thinking time!!
Sorry for my retarded reply.

 

[Arc Acoustics]

I'm sorry, I'm still confused. Up until 2 days ago when I clicked on measurements, it takes me to a page where the images are normal sized and all of a sudden they're huge, and I can't zoom out. Yet when I press Ref vs EQ, the charts become normal sized, except it looks glitchy like the second picture I posted. Is it really intended for images to be huge for ref and with eq, but normal sized for ref vs eq page, and for the footer to be behind the pictures?

But it only shows the one graph I clicked, so there wouldn't be a way to view all graphs in one page. I'd have to click a huge image, look, go back, scroll down, click on the next image, etc.

I think it is intended behaviour.
For as long as I remember, the "image version" was still expanded (but not huge) days ago.
We still can scroll down and view all the image versions so it is not a big deal for me.
Ctrl+- (browser zoom down) does not work for the image version (in blink, gecko, trident) and I *personally* prefer 1:1 because of the jaggedy jpeg artefact though.

 

[pierre]

I think it is intended behaviour.
For as long as I remember, the "image version" was still expanded (but not huge) days ago.
We still can scroll down and view all the image versions so it is not a big deal for me.
Ctrl+- (browser zoom down) does not work for the image version (in blink, gecko, trident) and I *personally* prefer 1:1 because of the jaggedy jpeg artefact though.

A new version is in the design:
- Resolution is higher and the image size will be capped such that it doesn’t use all the screen on a large desktop. I have a 4k thus I will be able to test
- images are now also available in webp and avif formats and if your browser supports it will help to control the size of the download
- i am on the fence to generate 2 versions of the pictures to satisfy both requirements. Since it is static website, I cannot generate them on the fly and that increases the website size.
- i need to do it more professionally with json compression, and the various classical optimization (webpack and such).

 

[BoredErica]

Spoiler

This is LS50 Meta Ref vs EQ page. Seems like something is not there.

A collection of loudspeakers measurements

pierreaubert.github.io

 

[Arc Acoustics]

Oh, another great improvement by @pierre , thanks!
Images are no longer blurry, no longer huge, browser zoom works perfectly as well, nice touch!
I still thinking about what could be the right way to evaluate the beam width though.
Minimum value could be too sensitive to sharp of-axis null mainly caused by the crossover, so ERB or some kind of smoothing could improve and be more agreeable to human hearing, at least I imagine.
@ folks, do you have any good ideas?

 

[pierre]

Oh, another great improvement by @pierre , thanks!
Images are no longer blurry, no longer huge, browser zoom works perfectly as well, nice touch!
I still thinking about what could be the right way to evaluate the beam width though.
Minimum value could be too sensitive to sharp of-axis null mainly caused by the crossover, so ERB or some kind of smoothing could improve and be more agreeable to human hearing, at least I imagine.
@ folks, do you have any good ideas?

I implemented a compromise:
- for each measurements starting at 10 up to 180 degrees: find if it goes down to -6 dB => then that's where the directivity line should be.
- added some smoothing (currently per octave) in order to decrease the noise.
Results are by 10 degrees and not interpolated at the moment, that gives overall reasonable values.
I checked a coax, a classical 2 ways and 1 center to have some coherence.

 

[Arc Acoustics]

I implemented a compromise:
- for each measurements starting at 10 up to 180 degrees: find if it goes down to -6 dB => then that's where the directivity line should be.
- added some smoothing (currently per octave) in order to decrease the noise.
Results are by 10 degrees and not interpolated at the moment, that gives overall reasonable values.
I checked a coax, a classical 2 ways and 1 center to have some coherence.

Thank you for implementing smoothing.
Now it does looks fully informative.

 

[SDX-LV]

Did you know that the complete, raw or processed official data for Neumann KH80 is available for free in the Klippel dB-Lab software as an example? It is quite impressive that the software that comes with NFS is practically fully functional for data viewing and export without any cost.