Master Preference Ratings for Loudspeakers

[pierre]

Since you can now calculate the post-EQ scores, would you be interested in doing it for every single speaker if I provided you with the same type of EQ corrections that were posted in the Vantoo T0 thread?

if you give me eq as a set of iir filters, I can. I am trying to compute optimal eq automatically but this is more complicated than I anticipated. Progress are slow but I can compute the score from a parametric eq and a measurement automatically.

 

[flipflop]

if you give me eq as a set of iir filters, I can. I am trying to compute optimal eq automatically but this is more complicated than I anticipated. Progress are slow but I can compute the score from a parametric eq and a measurement automatically.

I've laid out some principles to ensure that the resulting EQ profiles have a practical use and that all speakers get an equal treatment:

• There's no change at all below 300 Hz because the room is dominant at those frequencies.
• The EQs will also be based around the SPL @ 300 Hz in order to avoid a sudden jump or fall in the frequency response.
• Max frequency is 20 kHz because that's as high as the measurement data goes. Even if some of the filters affect frequencies higher than 20 kHz, it should be of no concern.
• No boost to large (5+ dB) dips. This seemed reasonable when I first thought of it, but I've actually struggled to think of a good reason to not go all-out with the EQ. I guess it could put too much stress on the built-in amps in active speakers.
• It's the LW that's being EQed. Like any of the other criteria, this is open to discussion.

My approach is to use the automatic EQ function in REW to get the FR within ±1 dB of the target and then make manual adjustments to get it as close as possible to flat using the remaining filters available.

IIR filters for Adam S2V are attached to this post.

 

[pierre]

I've laid out some principles to ensure that the resulting EQ profiles have a practical use and that all speakers get an equal treatment:

• There's no change at all below 300 Hz because the room is dominant at those frequencies.
• The EQs will also be based around the SPL @ 300 Hz in order to avoid a sudden jump or fall in the frequency response.
• Max frequency is 20 kHz because that's as high as the measurement data goes. Even if some of the filters affect frequencies higher than 20 kHz, it should be of no concern.
• No boost to large (5+ dB) dips. This seemed reasonable when I first thought of it, but I've actually struggled to think of a good reason to not go all-out with the EQ. I guess it could put too much stress on the built-in amps in active speakers.
• It's the LW that's being EQed. Like any of the other criteria, this is open to discussion.

My approach is to use the automatic EQ function in REW to get the FR within ±1 dB of the target and then make manual adjustments to get it as close as possible to flat using the remaining filters available.

IIR filters for Adam S2V are attached to this post.

The file looks like an impulse response. Can you generate the IIR filters? I can also do a conversion ( I wanted to be able to take a FIR filter too).

 

[q3cpma]

if you give me eq as a set of iir filters, I can. I am trying to compute optimal eq automatically but this is more complicated than I anticipated. Progress are slow but I can compute the score from a parametric eq and a measurement automatically.

Sorry if that has already been asked, but why do your scores and the ones from @MZKM differ so much, even relatively? I compared both when I saw that the F208 was higher than the 8341A on your website.

 

[flipflop]

The file looks like an impulse response. Can you generate the IIR filters? I can also do a conversion ( I wanted to be able to take a FIR filter too).

Sorry about that.
I hope this is what you need:

 

[MZKM]

Sorry if that has already been asked, but why do your scores and the ones from @MZKM differ so much, even relatively? I compared both when I saw that the F208 was higher than the 8341A on your website.

If you look, he has ASR & Vendor versions for the measurements of a few speakers, so I believe the high score for the F208 is using the latter.

 

[pierre]

Sorry about that.
I hope this is what you need:

SPK FLT
-----------------
NBD ON 0.34 0.24
NBD LW 0.32 0.19
NBD PIR 0.29 0.20
SM PIR 0.71 0.93
SM SP 0.92 0.95
LFX 43 43
LFQ 0.77 0.74
-----------------
Score 5.5 6.6
-----------------

Where SPK is speaker's measurement from ASR and FLT stand for filter. Hope this helps.
I will put the source code on github soon so you can easily run it yourself.

Nice increase in the score. You may loose ~3dB (to prevent potential clipping in D/A).

 

[pierre]

I've laid out some principles to ensure that the resulting EQ profiles have a practical use and that all speakers get an equal treatment:

• There's no change at all below 300 Hz because the room is dominant at those frequencies.
• The EQs will also be based around the SPL @ 300 Hz in order to avoid a sudden jump or fall in the frequency response.
• Max frequency is 20 kHz because that's as high as the measurement data goes. Even if some of the filters affect frequencies higher than 20 kHz, it should be of no concern.
• No boost to large (5+ dB) dips. This seemed reasonable when I first thought of it, but I've actually struggled to think of a good reason to not go all-out with the EQ. I guess it could put too much stress on the built-in amps in active speakers.
• It's the LW that's being EQed. Like any of the other criteria, this is open to discussion.

My approach is to use the automatic EQ function in REW to get the FR within ±1 dB of the target and then make manual adjustments to get it as close as possible to flat using the remaining filters available.

IIR filters for Adam S2V are attached to this post.

Something that would improve the score a lot more is to decrease the max output and generate a higher -3db point. That would boost LFX, LFQ would stay the same and the others too. You only loose on max output and of course distorsion will increase in the bass but we are not super good at detecting bass distorsion.

The filter for the Adam is working well, thanks.

 

[flipflop]

Something that would improve the score a lot more is to decrease the max output and generate a higher -3db point. That would boost LFX, LFQ would stay the same and the others too. You only loose on max output and of course distorsion will increase in the bass but we are not super good at detecting bass distorsion.

The idea is to provide EQ filters that can easily be used in practice. Increasing the score by "gaming" the model is not my goal.
What you're suggesting sounds like it would effectively be a bass boost, which I want to avoid, because the bass region is highly room dependent and therefore something people should correct by themselves. There's no one-size-fits-all bass EQ, so it's better to leave it untouched IMO.

Would you be interested in displaying the post-EQ scores and the filters on your website? I'm not just doing this for personal gain.

 

[pierre]

The idea is to provide EQ filters that can easily be used in practice. Increasing the score by "gaming" the model is not my goal.
What you're suggesting sounds like it would effectively be a bass boost, which I want to avoid, because the bass region is highly room dependent and therefore something people should correct by themselves. There's no one-size-fits-all bass EQ, so it's better to leave it untouched IMO.

Would you be interested in displaying the post-EQ scores and the filters on your website? I'm not just doing this for personal gain.

If you have the energy to do all the ASR speakers sure. If not, we could try to automate your algorithm.

 

[flipflop]

If you have the energy to do all the ASR speakers sure.

I do. It's obviously going to take some time, though.
If no one else has anything to add, we should continue the conversation in private to avoid flooding this thread.

 

[pierre]

If you look, he has ASR & Vendor versions for the measurements of a few speakers, so I believe the high score for the F208 is using the latter.

Thanks for noticing that. Th current code output randomly 1 score if we have multiple measurements. I did patch it to output the ASR one if we have one.

 

[pierre]

The idea is to provide EQ filters that can easily be used in practice. Increasing the score by "gaming" the model is not my goal.
What you're suggesting sounds like it would effectively be a bass boost, which I want to avoid, because the bass region is highly room dependent and therefore something people should correct by themselves. There's no one-size-fits-all bass EQ, so it's better to leave it untouched IMO.

Would you be interested in displaying the post-EQ scores and the filters on your website? I'm not just doing this for personal gain.

That's not gaming the model and that's not a bass boost: you keep a flat target (or the one you want). That's what other systems like Audiolense and similar are doing. The tradeoff is that you are loosing max output.
I agree with you that it can and should be done in the room correction system. You can also convolve the room correction with the speaker correction.

 

[MZKM]

Just to update, my signature now takes you to a Google site I made which has the preference ratings spreadsheet, the speaker sensitivity spreadsheet I made a post about, and now a speaker selector (requires making a copy to your Google Sheets; I think it auto updates with new entries even after you make your copy, not sure), you can narrow by price, active/passive, score, score with sub, -6dB point, and sensitivity (for just price/score/type, I would recommend the one Pozz has on this forum, as it’s a lot more convenient).

 

[edechamps]

I added an "ideal subwoofer" preference rating calculation mode to Loudspeaker Explorer (it's under "Preference Rating", "Settings"). It uses a (customizable) -6 dB point of 20 Hz which I think is more realistic than @MZKM's 14.5 Hz and should make it easier to compare "with sub" ratings to "without sub" ratings with less risk of model extrapolation. In practice that just means that my "with sub" scores are 0.6 lower than @MZKM's.

In my previous post I presented prediction interval charts for the scores. As always, anyone can generate those using Loudspeaker Explorer (just hit "Select all" on the speaker selector). However, since these charts seem to have become quite popular, I can't resist the temptation to share them again for "ideal subwoofer" scores:

What I find really interesting on that second chart is that we can clearly distinguish two large clusters: those in the 5.0-6.3 range, and those in the 6.4-7.4 range. This would tend to put speakers into four "with sub" categories: bad (<5.0), mediocre (5.0-6.3), good (6.4-7.4), and excellent (>7.4).

 

[MZKM]

uses a (customizable) -6 dB point of 20 Hz which I think is more realistic than @MZKM's 14.5 Hz

The Rythmik L12 is rated at -6dB @ 12Hz. Pipe organs go down to 16Hz and the opening to Edge of Tomorrow is as low as 10Hz if I recall.
14.5Hz allows for a max score of 10 (though getting perfect scores for the other portions is improbable).

 

[edechamps]

The Rythmik L12 is rated at -6dB @ 12Hz. Pipe organs go down to 16Hz and the opening to Edge of Tomorrow is as low as 10Hz if I recall.

By "realistic" I meant in the context of the Olive model. I doubt there were any speakers that could go down that low in Olive's sample. And even if there were, surely the material they used for the listening tests didn't go that low. This leads me to believe that feeding a LFX of 14.5 Hz into the model will result in excessive extrapolation. Or, to put it in a different way: I really doubt someone would rate a speaker 0.6 higher just because it can go down to 15 Hz instead of 20. Besides, most FR data only goes down to 20 Hz, anyway.

 

[MZKM]

By "realistic" I meant in the context of the Olive model. I doubt there were any speakers that could go down that low in Olive's sample. And even if there were, surely the material they used for the listening tests didn't go that low. This leads me to believe that feeding a LFX of 14.5 Hz into the model will result in excessive extrapolation. Or, to put it in a different way: I really doubt someone would rate a speaker 0.6 higher just because it can go down to 15 Hz instead of 20. Besides, most FR data only goes down to 20 Hz, anyway.

True, but having a theoretical max of 10 is nice to have

Besides, most FR data only goes down to 20 Hz, anyway.

My graphs don’t go that low either, as It’s worthless info for all speakers measured thus far don’t even come close.

 

[chaking]

Apologies up front if this has been covered or if I'm coming across as not appreciative of the great work here, certainly not my intention. In fact, I love the efforts to collect and present this data, and it's absolutely effective! And if this is the wrong thread, I will not be offended by removal

Couple questions:
1. Is there any inclination to include the # of drivers, size, material of drivers etc (more attributes) into this data table to draw more inferences?
2. The visualization of the data is always appreciated, but I also love to sort and filter data quickly in tabular form. The way this is presented (https://www.audiosciencereview.com/forum/index.php?pages/Audio_DAC_Performance_Index/ ) makes it somewhat difficult to navigate the data quickly as the filters take time to setup and the bar chart doesn't necessarily convey as much information to the user as the space it's using would indicate. Whereas in a classic table with sorts and filters you can quickly manipulate it and it's fairly efficient with screen real-estate. Has presenting it in tabular form with basic sort/filter controls been considered? I do see the Google Sheets list, however there doesn't appear to be sort/filter that I can see.

1 offer:
If the ideas have come up but effort isn't available, I would be glad to throw my time in to help out standing up a flexible schema/db, table, or otherwise.
-- I've used the following page for ideas around tables in the past: https://medium.com/nextux/design-better-data-tables-4ecc99d23356

Thanks for all the great work!

 

[MZKM]

Apologies up front if this has been covered or if I'm coming across as not appreciative of the great work here, certainly not my intention. In fact, I love the efforts to collect and present this data, and it's absolutely effective! And if this is the wrong thread, I will not be offended by removal

Couple questions:
1. Is there any inclination to include the # of drivers, size, material of drivers etc (more attributes) into this data table to draw more inferences?
2. The visualization of the data is always appreciated, but I also love to sort and filter data quickly in tabular form. The way this is presented (https://www.audiosciencereview.com/forum/index.php?pages/Audio_DAC_Performance_Index/ ) makes it somewhat difficult to navigate the data quickly as the filters take time to setup and the bar chart doesn't necessarily convey as much information to the user as the space it's using would indicate. Whereas in a classic table with sorts and filters you can quickly manipulate it and it's fairly efficient with screen real-estate. Has presenting it in tabular form with basic sort/filter controls been considered? I do see the Google Sheets list, however there doesn't appear to be sort/filter that I can see.

1 offer:
If the ideas have come up but effort isn't available, I would be glad to throw my time in to help out standing up a flexible schema/db, table, or otherwise.
-- I've used the following page for ideas around tables in the past: https://medium.com/nextux/design-better-data-tables-4ecc99d23356

Thanks for all the great work!

I use Google Sheets and there is no way to use filters in a published sheet. However, if you go to my signature and go to Speaker Selector, you can copy a spreadsheet which I have setup to filter (Sheets has built in filtering but it is not user friendly, I set it up so you type in the text box the values to set the min/max).

I have not added driver count nor driver material as I don’t think it has any importance. I have not added size (bookshelf or tower) because I am lazy, I can get that done easily and quickly though if enough people care.

My presentation is alphabetical as it’s easier to navigate, and if you narrow down your search enough there shouldn’t be too many to sort through. I don’t have any updating graph view either, as it’s a pain.