How to Choose a Loudspeaker -- What the Science Shows - Page 112 - AVS Forum | Home Theater Discussions And Reviews
Forum Jump: 
 6387Likes
Reply
 
Thread Tools
post #3331 of 5320 Old 07-06-2019, 12:04 PM
Member
 
Floyd Toole's Avatar
 
Join Date: Jun 2017
Location: California
Posts: 836
Mentioned: 89 Post(s)
Tagged: 0 Thread(s)
Quoted: 875 Post(s)
Liked: 3079
Quote:
Originally Posted by motrek View Post
I haven't seen the algorithm, but this is somewhat problematic. If you're creating an algorithm to see if two things are correlated then you can design the algorithm in such a way that you're basically manufacturing correlation. I'm not saying this would be done intentionally but there's a possibility that it happened unintentionally. This has been a problem for other branches of science in the past.

Related: There have been tremendous developments in machine learning just in the last few years. I wonder if the Harman algorithm couldn't be replaced with an artificial neural network that has even more predictive power re: loudspeaker preferences. Such a neural network could be made to analyze all the data collected in your anechoic chamber and not just the final spin-o-rama plots. Then again, the ideas of "flat" and "smooth" aren't that complicated, it might not be worth it to go to all that effort to end up at (probably) the same destination.
Indeed one can "fudge" algorithms. In our case it was an exercise prompted by Consumer Reports loudspeaker reviews which routinely trashed loudspeakers we knew sounded good (some of them Harman products) and praised products that routinely lost in our double-blind evaluations. CU did no listening tests. Their ratings and rankings were entirely based on a 1/3-octave measurement of sound power and calculations based on misinterpretations of the perceptual process. Some people in the industry knew the truth but could do nothing about it.

One day our CEO called me into his office and asked me to explain why our products did not always rank highly. Long story short, he and Sidney Harman authorized us to spend the time and money to force them to put up or shut up. Research was done. We let them preview the two definitive Olive papers before they were made public via the AES.
Olive, S.E. (2004a). “A multiple regression model for predicting loudspeaker preference using objective measurements: part 1 – listening test results”, 116th Convention, Audio Eng. Soc., Preprint 6113.
Olive, S.E. (2004b). “A multiple regression model for predicting loudspeaker preference using objective measurements: part 2 – development of the model”, 117th Convention, Audio Eng. Soc., Preprint 6190.

Consumer Reports stopped doing speaker reviews and started a process of improving their evaluations. It did not last long and the effort was abandoned. So, in a strange way, Harman did the entire loudspeaker industry a favor by removing a source of influential, but incorrect, consumer product evaluations.

We, of course, gained more confidence in the measurements we were doing. Could it have been taken farther? Probably. But, the present reality is that we can design loudspeakers with high confidence that they will rate highly in double-blind subjective evaluations - to the point where the highest ranked products end up in statistical ties. The weakest link is no longer the loudspeaker, it is the program - i.e. the Circle of Confusion. To this must be added the fact that everything we hear below about 400 Hz, which is determined by the room and arrangements within it, accounts for about 30% of the factor weighting in sound quality evaluations. Only in-situ measurements and solutions can work. This was the motivation for our work on multiple subs.

There seems to be little point in spending more time and money on predictive algorithms.
RichB, krabapple, Muza and 3 others like this.
Floyd Toole is offline  
Sponsored Links
Advertisement
 
post #3332 of 5320 Old 07-06-2019, 12:05 PM
Member
 
spkr_diy's Avatar
 
Join Date: Jul 2007
Location: Chantilly, VA
Posts: 57
Mentioned: 2 Post(s)
Tagged: 0 Thread(s)
Quoted: 44 Post(s)
Liked: 57
Quote:
Originally Posted by Floyd Toole View Post
The algorithm simply proved, with high statistical significance, that there was a very high correlation between what we were measuring and listener ratings of sound quality. That is all. It is not used as a metric to evaluate new products - we still rely on two ears and a brain (in case we missed something), and use the technical data to guide engineers during the design process. It turns out that the "best looking" spinoramas, informally assessed by experienced observers are sufficient to identify essentially neutral loudspeakers - which is the goal.
My point is that when you tout what a "very high correlation" you get between your assessment of your measurements and listener preferences, but you don't test your assessment technique in a predictive way, you exaggerate the level of correlation. If you want people to infer causation from your correlation (which it's obvious you do), then you really ought to test your assessment techniques on closed data and report the results. I'm glad to hear that you still use two ears and a brain, which supports my statement that there is still more to learn. I don't think anyone from Harman has ever claimed otherwise. When I say there is still more to learn I don't intend to contradict the Harman research in any way. Rather, it's a reminder for other participants in this thread, some of whom see those high correlation numbers and seem to get the impression that the entire issue of preferences has been figured out.
motrek and Elkerton like this.
spkr_diy is online now  
post #3333 of 5320 Old 07-06-2019, 12:16 PM
AVS Forum Special Member
 
Dave in Green's Avatar
 
Join Date: Jan 2014
Location: USA
Posts: 8,355
Mentioned: 152 Post(s)
Tagged: 0 Thread(s)
Quoted: 3787 Post(s)
Liked: 3002
There is no perfect method of comparing speaker performance. We all just do the best we can with how much individual effort we are willing to put into getting the most out of what the best available science, the most credible reviewers and our own ears tell us. We can nitpick the imperfections in the best available science as well as what any random individual may tell us their ears tell them. But in the end we can never be 100% sure that we have selected the best possible speaker for our tastes at the price we're willing to pay. A reasonable goal is to end up with something close enough to the best to satisfy our craving for good sound reproduction. Those who chase after the elusive best imaginable speaker by all conceivable measures might as well be chasing after unicorns.
spkr_diy and bearr48 like this.
Dave in Green is offline  
Sponsored Links
Advertisement
 
post #3334 of 5320 Old 07-06-2019, 12:21 PM
AVS Forum Special Member
 
Soulburner's Avatar
 
Join Date: Oct 2007
Location: Lincoln
Posts: 4,892
Mentioned: 46 Post(s)
Tagged: 0 Thread(s)
Quoted: 194 Post(s)
Liked: 1630
Quote:
Originally Posted by motrek View Post
Thinking about this after a night's sleep, the solution seems fairly obvious.

Imagine a speaker B that produces more bass than speaker A. (B for bass.) In other words, the frequency where speaker B starts to roll off the bass is lower.

If you want to even out the bass reproduction between two speakers (so they're on an even footing before you apply e.g. a typical 80Hz high-pass filter) then you apply a filter that rolls off speaker B at the same frequency where A starts rolling off (at 24 dB/octave) and then stops decreasing the bass at the frequency where B starts rolling off. So that way, both speakers will roll off at a constant 24 dB/octave starting at the same frequency. Job done.
Pardon my primitive MS Paint skillz... but isn't this the result of the filter you are describing? (attached)

Actually, it might be even worse if designed exactly as you describe: the orange line would appear to jump back up to meet the green line when you "resume natural output".

I really think filters need to apply to both speakers unless you are prepared to exactly model one after the other, which would take a bit more work. Obviously, real speakers would have a bumpier contour that would need to be modeled than this MS Paint masterpiece you are feasting your eyes upon. That would truly level the playing field, which I understand is the goal.

However, applying a filter to one speaker, even if it were perfect, would violate another goal: setting up both speakers in the same way that they will be used in the complete system. That is, with a crossover and sub(s) enabled, no extra filters.

Click image for larger version

Name:	motrek filter.png
Views:	24
Size:	20.8 KB
ID:	2588128

HT: Samsung PN64H5000 (recommended settings) | NAD T758 V3 | Buchardt S400 (2) | Emotiva E2 (2) | Rythmik Audio F12 (2)
Soulburner is offline  
post #3335 of 5320 Old 07-06-2019, 12:28 PM
Member
 
Floyd Toole's Avatar
 
Join Date: Jun 2017
Location: California
Posts: 836
Mentioned: 89 Post(s)
Tagged: 0 Thread(s)
Quoted: 875 Post(s)
Liked: 3079
Quote:
Originally Posted by motrek View Post
Good point. If I could throw out an idea, also try listening to some familiar conversational spoken-word content like a podcast. (Make sure it's professionally recoded, i.e., with good microphones.) Most people don't do this but I don't know why not. You spend a lot of your life listening to other people talking--you will instinctively know how it should sound.

I remember watching No Country For Old Men with some speakers that I generally liked, but the opening scene with the Tommy Lee Jones voiceover sounded completely wrong in a way that I wouldn't have been able to pick out by just listening to music.
Sometimes voice - dialog replacement is common - is recorded in smallish rooms with a large script table in front of the talker, and a microphone a few inches from the lips. I think I have heard the table reflection in voices uttered by lips attached to heads on sailboats in open ocean . Vocal mics are not always neutral - another issue.
Floyd Toole is offline  
post #3336 of 5320 Old 07-06-2019, 12:36 PM
Advanced Member
 
motrek's Avatar
 
Join Date: Dec 2013
Posts: 653
Mentioned: 7 Post(s)
Tagged: 0 Thread(s)
Quoted: 518 Post(s)
Liked: 191
Quote:
Originally Posted by Soulburner View Post
Pardon my primitive MS Paint skillz... but isn't this the result of the filter you are describing? (attached)

Actually, it might be even worse if designed exactly as you describe: the orange line would appear to jump back up to meet the green line when you "resume natural output".

I really think filters need to apply to both speakers unless you are prepared to exactly model one after the other, which would take a bit more work. Obviously, real speakers would have a bumpier contour that would need to be modeled than this MS Paint masterpiece you are feasting your eyes upon. That would truly level the playing field, which I understand is the goal.

However, applying a filter to one speaker, even if it were perfect, would violate another goal: setting up both speakers in the same way that they will be used in the complete system. That is, with a crossover and sub(s) enabled, no extra filters.

Attachment 2588128
Hmm, I think you're right. Well, there has to be some way to get those curves to look the same. Seems like it shouldn't be too hard, I just can't think of how to do it at the minute. It's been about 20 years since I took a class on signal filtering.

When evaluating two speakers for use with a subwoofer, I would argue that it is fairly important to get them on a level playing field re: bass, because presumably each one can be integrated correctly with the subwoofer. It might require different settings but you should be able to get there. So you really want to know how the speakers compare outside of that frequency range. (Of course I'm talking about nice speakers that can be crossed over at ~100Hz or lower, not Bose Acoustimass satellites.)
motrek is online now  
post #3337 of 5320 Old 07-06-2019, 12:44 PM
AVS Forum Special Member
 
Soulburner's Avatar
 
Join Date: Oct 2007
Location: Lincoln
Posts: 4,892
Mentioned: 46 Post(s)
Tagged: 0 Thread(s)
Quoted: 194 Post(s)
Liked: 1630
Quote:
Originally Posted by motrek View Post
Hmm, I think you're right. Well, there has to be some way to get those curves to look the same. Seems like it shouldn't be too hard, I just can't think of how to do it at the minute. It's been about 20 years since I took a class on signal filtering.
Perhaps if there were a way to input two speaker measurements, and tell the software to "make filters to EQ B to equal A". It's sort of a difference engine. It shouldn't be hard for the designer of REW - John - for example, to do. Maybe there already is such a function - I haven't used REW to make filters so someone with more experience may be able to provide a better answer.

HT: Samsung PN64H5000 (recommended settings) | NAD T758 V3 | Buchardt S400 (2) | Emotiva E2 (2) | Rythmik Audio F12 (2)
Soulburner is offline  
post #3338 of 5320 Old 07-06-2019, 12:54 PM
Advanced Member
 
motrek's Avatar
 
Join Date: Dec 2013
Posts: 653
Mentioned: 7 Post(s)
Tagged: 0 Thread(s)
Quoted: 518 Post(s)
Liked: 191
Quote:
Originally Posted by Floyd Toole View Post
Sometimes voice - dialog replacement is common - is recorded in smallish rooms with a large script table in front of the talker, and a microphone a few inches from the lips. I think I have heard the table reflection in voices uttered by lips attached to heads on sailboats in open ocean . Vocal mics are not always neutral - another issue.
Indeed... Also, I have a pair of Ascend Acoustic CBM-170 bookshelf speakers that resolve enough detail that I can often tell how/where an actor is mic'ed based on how the sound changes when the character moves their head, or I can tell where a boom mic is located based on how the sound changes as the mic moves around the set (e.g., when characters are walking down a hallway).

So far I have never heard another set of speakers (other than headphones) that have allowed me to hear such subtle auditory cues. I was sad that my new Infinity floorstanders don't. Maybe you can speak to why this might be. I think maybe the floorstanders are putting out so much more bass that it overwhelms the higher frequencies. Or maybe the Ascends are a bit louder in the frequency range where such cues are present.

Re: my suggestion to listen to a podcast: maybe this wouldn't allow a listener to choose between two higher-end speakers but I've heard many poor audio setups where human voice is obviously very wrong, so it might allow people to quickly filter out speakers with obvious issues (that might not be immediately obvious when listening to music).
motrek is online now  
post #3339 of 5320 Old 07-06-2019, 01:04 PM
AVS Forum Special Member
 
Dave in Green's Avatar
 
Join Date: Jan 2014
Location: USA
Posts: 8,355
Mentioned: 152 Post(s)
Tagged: 0 Thread(s)
Quoted: 3787 Post(s)
Liked: 3002
Variable miking is part of audio's Circle of Confusion.
Dave in Green is offline  
post #3340 of 5320 Old 07-06-2019, 01:28 PM
AVS Forum Special Member
 
aarons915's Avatar
 
Join Date: Oct 2012
Location: Cincinnati, OH
Posts: 1,324
Mentioned: 21 Post(s)
Tagged: 0 Thread(s)
Quoted: 885 Post(s)
Liked: 696
Quote:
Originally Posted by motrek View Post
Hmm, I think you're right. Well, there has to be some way to get those curves to look the same. Seems like it shouldn't be too hard, I just can't think of how to do it at the minute. It's been about 20 years since I took a class on signal filtering.
Quote:
Originally Posted by Soulburner View Post
Perhaps if there were a way to input two speaker measurements, and tell the software to "make filters to EQ B to equal A". It's sort of a difference engine. It shouldn't be hard for the designer of REW - John - for example, to do. Maybe there already is such a function - I haven't used REW to make filters so someone with more experience may be able to provide a better answer.
It's pretty easy to match the bass response of 2 speakers in REW, all you need is some way to EQ. A minidsp would be great for this because you can add in high pass filters along with PEQ. I always audition with around a 100Hz 2nd order high pass to minimize bass differences but most speakers I try use a 5" midwoofer so the bass response is pretty close anyway.
aarons915 is online now  
post #3341 of 5320 Old 07-06-2019, 02:02 PM
AVS Forum Special Member
 
Join Date: Mar 2017
Posts: 1,401
Mentioned: 15 Post(s)
Tagged: 0 Thread(s)
Quoted: 956 Post(s)
Liked: 288
Quote:
Originally Posted by motrek View Post
Thinking about this after a night's sleep, the solution seems fairly obvious.



Imagine a speaker B that produces more bass than speaker A. (B for bass.) In other words, the frequency where speaker B starts to roll off the bass is lower.



If you want to even out the bass reproduction between two speakers (so they're on an even footing before you apply e.g. a typical 80Hz high-pass filter) then you apply a filter that rolls off speaker B at the same frequency where A starts rolling off (at 24 dB/octave) and then stops decreasing the bass at the frequency where B starts rolling off. So that way, both speakers will roll off at a constant 24 dB/octave starting at the same frequency. Job done.



I know such filters are pretty trivial to construct in audio-editing software. I'm not sure how to implement such a filter in software myself though but I'm sure it can be sorted out with some reading.
In all these discussions, let's remember that we are trying a simple speaker audition system that can be used at any dealer and is better than listening to two speakers playing 30 min apart at two SPL levels.

Ease of use and ability to use it at any dealer is paramount. That means it MUST use components and features commonly available at any dealer.
SouthernCA is offline  
post #3342 of 5320 Old 07-06-2019, 02:06 PM
Advanced Member
 
motrek's Avatar
 
Join Date: Dec 2013
Posts: 653
Mentioned: 7 Post(s)
Tagged: 0 Thread(s)
Quoted: 518 Post(s)
Liked: 191
Quote:
Originally Posted by SouthernCA View Post
In all these discussions, let's remember that we are trying a simple speaker audition system that can be used at any dealer and is better than listening to two speakers playing 30 min apart at two SPL levels.

Ease of use and ability to use it at any dealer is paramount. That means it MUST use components and features commonly available at any dealer.
Absolutely. Well, what I was thinking is an app where you just plug your phone into a receiver, and the receiver's left channel is connected to one speaker and the right channel is connected to another.

Couldn't really be any easier.

From the app, you'd be able to instantly toggle between speakers, instantly rewind whatever you're listening to by N seconds, change the volume of each speaker individually, and probably some more features that I'm forgetting. But there's no reason why it wouldn't be extremely easy to use.

All this bass-matching stuff would be extra, optional stuff that people can ignore... but there's no reason why it couldn't be easy to use too.
motrek is online now  
post #3343 of 5320 Old 07-06-2019, 02:30 PM
Member
 
spkr_diy's Avatar
 
Join Date: Jul 2007
Location: Chantilly, VA
Posts: 57
Mentioned: 2 Post(s)
Tagged: 0 Thread(s)
Quoted: 44 Post(s)
Liked: 57
Quote:
Originally Posted by Dave in Green View Post
Variable miking is part of audio's Circle of Confusion.
Interesting point.

To my ears, Megan Davies (right) and Jaclyn (left) are recorded and/or mixed significantly differently in this video, and Jaclyn sounds more natural to me, on my speakers. In fact, Jaclyn sounds as natural as any recorded female vocals I've heard. I would be very interested to know if anyone finds Megan to sound more natural on their system. I know some audio reviewer used a Megan Davies song (without Jaclyn) to help evaluate speakers and mentioned it in their review, so there's evidence that someone who cares about audio ranks the presentation of Megan's voice very highly.



For male vocals, this recording of Harry Connick, Jr. is the most natural I've heard :


The very next song on that same album sounds less natural to me (too much reverb).


If I'm going off topic, I apologize. I think it would make more sense to test speakers with these videos than with a random podcast though. We may also get some insight into the circle of confusion if different people with different systems have different preferences. On my system, my daughter agrees that Jaclyn sounds more natural for whatever that's worth.

I'd also be interested to know what others consider the most natural vocal recordings they've heard. PM me if you want to keep that out of this thread.
spkr_diy is online now  
post #3344 of 5320 Old 07-06-2019, 02:39 PM
Advanced Member
 
motrek's Avatar
 
Join Date: Dec 2013
Posts: 653
Mentioned: 7 Post(s)
Tagged: 0 Thread(s)
Quoted: 518 Post(s)
Liked: 191
Quote:
Originally Posted by spkr_diy View Post
...
If I'm going off topic, I apologize. I think it would make more sense to test speakers with these videos than with a random podcast though. We may also get some insight into the circle of confusion if different people with different systems have different preferences. On my system, my daughter agrees that Jaclyn sounds more natural for whatever that's worth.

I'd also be interested to know what others consider the most natural vocal recordings they've heard. PM me if you want to keep that out of this thread.
My idea with podcasts is that it's normal spoken voice, which we all listen to all the time in real life, so we're going to have a good inherent understanding of what it should sound like.

I rarely listen to anybody sing in person.
spkr_diy and Soulburner like this.
motrek is online now  
post #3345 of 5320 Old 07-06-2019, 02:54 PM
Member
 
spkr_diy's Avatar
 
Join Date: Jul 2007
Location: Chantilly, VA
Posts: 57
Mentioned: 2 Post(s)
Tagged: 0 Thread(s)
Quoted: 44 Post(s)
Liked: 57
Quote:
Originally Posted by motrek View Post
My idea with podcasts is that it's normal spoken voice, which we all listen to all the time in real life, so we're going to have a good inherent understanding of what it should sound like.

I rarely listen to anybody sing in person.
I understand your idea, but the equipment and technique used on the vast majority of podcasts is pretty suspect. In contrast, for music, there's a concerted effort to get the right sound. Could you recommend a particular podcast or youtuber who talks and sounds very natural to you?
spkr_diy is online now  
post #3346 of 5320 Old 07-06-2019, 03:13 PM
Advanced Member
 
motrek's Avatar
 
Join Date: Dec 2013
Posts: 653
Mentioned: 7 Post(s)
Tagged: 0 Thread(s)
Quoted: 518 Post(s)
Liked: 191
Quote:
Originally Posted by spkr_diy View Post
I understand your idea, but the equipment and technique used on the vast majority of podcasts is pretty suspect. In contrast, for music, there's a concerted effort to get the right sound. Could you recommend a particular podcast or youtuber who talks and sounds very natural to you?
Not only do I rarely listen to people sing, but I also don't want anybody making a "concerted effort to get the right sound" when I'm trying to minimize the circle of confusion.

It's not hard to find podcasts that are recorded in recording studios by professionals using professional gear. Basically anything from NPR that is/was a radio show, for example. There are literally dozens of NPR podcasts.

Conan O'Brien started a podcast in the past few months that I like and I think it sounds good.

Come to think of it, I know the stereotype of a podcaster is some guy rambling into his cell phone's microphone in his basement but I don't think I've ever heard such a podcast. I think everything I've listened to has actually sounded pretty good. I even listen to a daily podcast put out by a couple of guys and they usually record in an extra room in one of their houses, and it sounds good. But they are former radio guys and they have nice equipment and they know what they're doing.
motrek is online now  
post #3347 of 5320 Old 07-06-2019, 03:19 PM - Thread Starter
Member
 
avkv's Avatar
 
Join Date: Jan 2003
Location: Los Angeles, CA
Posts: 153
Mentioned: 26 Post(s)
Tagged: 0 Thread(s)
Quoted: 133 Post(s)
Liked: 376
Simple Single Speaker Comparisons

Quote:
Originally Posted by SouthernCA View Post
I or someone here more adapt at this should find/create an app that can play a song in mono and be able to turn on and off either channels. One more - it needs to have a left and right balance.

With that app, one can do blind test at any dealer. Dr. Toole can tell if I miss anything big.

1. Place two different speakers (one each side by side)
2. Connect one to left and one to the right channel
3. Ask your friend to switch between left and right at random and you score them for sound quality.

If anyone knows of an app already available, please let us know.

Note: if planning to use with a sub, do high pass filter at 80 Hz and do not use a sub during testing.
Analog comes in handy for simple single-speaker A-B comparisons. (Please note that Harman X (our research scientists) use summed mono. I personally use a single channel, and have just standardized on the left channel. There are a few selections for which this doesn’t work, but not many.)

Note that the sadly typical 2nd-order (12dB/octave) high-pass filters are too shallow to minimize low frequency bandwidth differences. If possible, use a 4th-order high-pass filter to do the comparison.

To set up the comparison, simply take an analog output from a DAC or whatever source, dig up a “y-connector,” and feed one output to the left channel of preamp input “A,” and feed the other output to the right channel of preamp input “B,” and simply switch between A and B. Even if the preamp doesn’t have .1dB input level adjustments, it might have an input-specific balance control. Since there isn’t a signal in the unused channel of each input, the balance control(s) will let you match levels—at least to the precision offered by the preamp. Try to match the levels in the midrange—like on vocals—so that the bandwidth doesn’t throw you off.
Aja, Karl Maga, SouthernCA and 1 others like this.

Kevin Voecks
Acoustic Technologies
HARMAN Luxury Audio Group
Revel Founder
avkv is offline  
post #3348 of 5320 Old 07-06-2019, 03:21 PM
AVS Forum Special Member
 
torii's Avatar
 
Join Date: Mar 2015
Posts: 7,242
Mentioned: 54 Post(s)
Tagged: 0 Thread(s)
Quoted: 3592 Post(s)
Liked: 2027
Quote:
Originally Posted by OldMovieNut View Post
With the parameters you have set, it doesn't tell you a whole lot. You need to know what your trying to analyze. The spectrogram tool is useful for looking at reflections. To do that you need to change the parameters. I prefer the wavelet to the Fourier transform for this. For example, if you want to look at reflections above the Schroeder frequency, then I suggest this:

thanks. I found your response helpful. but still hard for me to equate graphs with sound quality
Attached Thumbnails
Click image for larger version

Name:	spectro.jpg
Views:	23
Size:	127.2 KB
ID:	2588314   Click image for larger version

Name:	spectro1.jpg
Views:	20
Size:	125.4 KB
ID:	2588316  

Power: Marantz sr7008, NAD C 275Bee x 2, Video: Oppo 103, Samsung 75un6300
Speakers: Focal aria 948, Focal cc900, Klipsch synergy KSF 10.5, Magnepan LRS
Subs: Rythmik FV25HP, Rythmik FV15HP
torii is offline  
post #3349 of 5320 Old 07-06-2019, 04:45 PM
AVS Forum Special Member
 
Scotth3886's Avatar
 
Join Date: Jan 2016
Location: New Albany, OH
Posts: 8,229
Mentioned: 35 Post(s)
Tagged: 0 Thread(s)
Quoted: 3997 Post(s)
Liked: 2464
Quote:
Originally Posted by spkr_diy View Post
I understand your idea, but the equipment and technique used on the vast majority of podcasts is pretty suspect. In contrast, for music, there's a concerted effort to get the right sound. Could you recommend a particular podcast or youtuber who talks and sounds very natural to you?
One of the better ones that I regularly listen to and have listened to on the two primary system, the LRS ribbons and the ESLs, ironically, is CR's Talking Cars. I've met a couple of these people at the NAIAS several times and know what their voice sounds like real life. I know they also spent money a few months ago trying to upgrade their system including mics to sound more real over you tube.


Last edited by Scotth3886; 07-07-2019 at 03:37 AM.
Scotth3886 is offline  
post #3350 of 5320 Old 07-06-2019, 04:51 PM
AVS Forum Special Member
 
Scotth3886's Avatar
 
Join Date: Jan 2016
Location: New Albany, OH
Posts: 8,229
Mentioned: 35 Post(s)
Tagged: 0 Thread(s)
Quoted: 3997 Post(s)
Liked: 2464
Quote:
Originally Posted by spkr_diy View Post
Interesting point.

To my ears, Megan Davies (right) and Jaclyn (left) are recorded and/or mixed significantly differently in this video, and Jaclyn sounds more natural to me, on my speakers. In fact, Jaclyn sounds as natural as any recorded female vocals I've heard. I would be very interested to know if anyone finds Megan to sound more natural on their system. I know some audio reviewer used a Megan Davies song (without Jaclyn) to help evaluate speakers and mentioned it in their review, so there's evidence that someone who cares about audio ranks the presentation of Megan's voice very highly.

https://youtu.be/ryx6ww2Nyqs?t=31


For male vocals, this recording of Harry Connick, Jr. is the most natural I've heard :

https://www.youtube.com/watch?v=pm24...=youtu.be&t=12

The very next song on that same album sounds less natural to me (too much reverb).

https://www.youtube.com/watch?v=U0v2zppdnR8

If I'm going off topic, I apologize. I think it would make more sense to test speakers with these videos than with a random podcast though. We may also get some insight into the circle of confusion if different people with different systems have different preferences. On my system, my daughter agrees that Jaclyn sounds more natural for whatever that's worth.

I'd also be interested to know what others consider the most natural vocal recordings they've heard. PM me if you want to keep that out of this thread.

Thanks for posting the vids. The Megan Davies sounds really nice and natural, but I can't find this exact cut in Tidal.

I've never heard of Harry Connick before so have no idea what he's supposed to sound like. Doesn't sound as good as the Megan Davies though. I generally avoid music with vocals unless its in a language that I do not understand so then it become just another musical instrument.
Scotth3886 is offline  
post #3351 of 5320 Old 07-06-2019, 08:00 PM
Advanced Member
 
James A. McGahee's Avatar
 
Join Date: Dec 2005
Posts: 869
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 81 Post(s)
Liked: 17
Are you accepting requests for subjective speaker comparisons

Quote:
Originally Posted by Gooddoc View Post
Thank-you! 😀 Great thread topic!!

Are you accepting requests for subjective speaker comparos and/or measurements? 😁
Absolutely!
James A. McGahee is offline  
post #3352 of 5320 Old 07-06-2019, 08:27 PM
AVS Forum Special Member
 
PrimeTime's Avatar
 
Join Date: Mar 2003
Location: Lower California
Posts: 3,258
Mentioned: 10 Post(s)
Tagged: 0 Thread(s)
Quoted: 802 Post(s)
Liked: 499
Quote:
Originally Posted by DonH50 View Post
Untold story there, Scott? Or are we back to talking about our old Flame Linears again?
A long while back I visited a lab that was using a Phase Linear 700 as a test amplifier.


Right next to it someone had placed a card that said "Thank You For Not Smoking"
DonH50, Scotth3886 and drh3b like this.
PrimeTime is offline  
post #3353 of 5320 Old 07-06-2019, 08:47 PM
AVS Forum Addicted Member
 
DonH50's Avatar
 
Join Date: Feb 2010
Location: Monument CO
Posts: 12,340
Mentioned: 69 Post(s)
Tagged: 0 Thread(s)
Quoted: 3275 Post(s)
Liked: 3362
Quote:
Originally Posted by PrimeTime View Post
A long while back I visited a lab that was using a Phase Linear 700 as a test amplifier.


Right next to it someone had placed a card that said "Thank You For Not Smoking"
I had to repair mine an original 700) several times. It also managed to take out a few speakers along the way. The 700B was more reliable, and the 400 never seemed as bad IME.
Scotth3886 likes this.

"After silence, that which best expresses the inexpressible, is music" - Aldous Huxley
DonH50 is offline  
post #3354 of 5320 Old 07-06-2019, 08:47 PM
AVS Forum Special Member
 
Join Date: Mar 2017
Posts: 1,401
Mentioned: 15 Post(s)
Tagged: 0 Thread(s)
Quoted: 956 Post(s)
Liked: 288
Quote:
Originally Posted by avkv View Post
Analog comes in handy for simple single-speaker A-B comparisons. (Please note that Harman X (our research scientists) use summed mono. I personally use a single channel, and have just standardized on the left channel. There are a few selections for which this doesn’t work, but not many.)



Note that the sadly typical 2nd-order (12dB/octave) high-pass filters are too shallow to minimize low frequency bandwidth differences. If possible, use a 4th-order high-pass filter to do the comparison.



To set up the comparison, simply take an analog output from a DAC or whatever source, dig up a “y-connector,” and feed one output to the left channel of preamp input “A,” and feed the other output to the right channel of preamp input “B,” and simply switch between A and B. Even if the preamp doesn’t have .1dB input level adjustments, it might have an input-specific balance control. Since there isn’t a signal in the unused channel of each input, the balance control(s) will let you match levels—at least to the precision offered by the preamp. Try to match the levels in the midrange—like on vocals—so that the bandwidth doesn’t throw you off.
Thanks, Kevin. Do you know of an app that has those features? Or maybe someone at Harman knows of something similar. I tried Foobar2000 for a few minutes but could not figure out if it can do this.

Last edited by SouthernCA; 07-06-2019 at 08:53 PM.
SouthernCA is offline  
post #3355 of 5320 Old 07-06-2019, 09:18 PM
AVS Forum Special Member
 
Soulburner's Avatar
 
Join Date: Oct 2007
Location: Lincoln
Posts: 4,892
Mentioned: 46 Post(s)
Tagged: 0 Thread(s)
Quoted: 194 Post(s)
Liked: 1630
Quote:
Originally Posted by SouthernCA View Post
Thanks, Kevin. Do you know of an app that has those features? Or maybe someone at Harman knows of something similar. I tried Foobar2000 for a few minutes but could not figure out if it can do this.
So basically we need a software version of this:
Click image for larger version

Name:	81QfK7zp+aL._SL1500_.jpg
Views:	24
Size:	157.7 KB
ID:	2588378

By the way, here is what you need in foobar to downmix to mono:
Click image for larger version

Name:	foobar.png
Views:	18
Size:	40.3 KB
ID:	2588380

In stereo:
Click image for larger version

Name:	stereo.png
Views:	17
Size:	119.1 KB
ID:	2588382

In mono:
Click image for larger version

Name:	mono.png
Views:	17
Size:	121.5 KB
ID:	2588384

It helps if you assign hotkeys. Ctrl+M = Mono, Ctrl+N = Not Mono

(now playing: The Dark Knight Rises OST)

HT: Samsung PN64H5000 (recommended settings) | NAD T758 V3 | Buchardt S400 (2) | Emotiva E2 (2) | Rythmik Audio F12 (2)
Soulburner is offline  
post #3356 of 5320 Old 07-07-2019, 12:10 AM
Senior Member
 
Join Date: Mar 2019
Location: Russia
Posts: 345
Mentioned: 1 Post(s)
Tagged: 0 Thread(s)
Quoted: 245 Post(s)
Liked: 70
JACK software (and plugins for it) has everything you need and more for that. You can use its advanced routing to do whatever you want.
aats is offline  
post #3357 of 5320 Old 07-07-2019, 03:13 AM
AVS Forum Special Member
 
torii's Avatar
 
Join Date: Mar 2015
Posts: 7,242
Mentioned: 54 Post(s)
Tagged: 0 Thread(s)
Quoted: 3592 Post(s)
Liked: 2027
I use apo and peace plugins for foobar. https://sourceforge.net/projects/pea...apo-extension/


very powerful and useful

Power: Marantz sr7008, NAD C 275Bee x 2, Video: Oppo 103, Samsung 75un6300
Speakers: Focal aria 948, Focal cc900, Klipsch synergy KSF 10.5, Magnepan LRS
Subs: Rythmik FV25HP, Rythmik FV15HP
torii is offline  
post #3358 of 5320 Old 07-07-2019, 04:49 AM
Senior Member
 
highmr's Avatar
 
Join Date: Mar 2010
Location: Scottsdale, Arizona
Posts: 241
Mentioned: 3 Post(s)
Tagged: 0 Thread(s)
Quoted: 107 Post(s)
Liked: 67
Quote:
Originally Posted by spkr_diy View Post
Here are some problems with the methods (as far as I understand them).

They basically trained a machine learning algorithm to predict speaker preferences as best as possible in a statistical sense using features derived from taking measurements of the speakers. They then report the effectiveness of this algorithm. However, since it sounds like they are reporting the performance of the algorithm on the training data used to tune the algorithm, the assessment is certainly suspect. To put it in perspective, I could take 30 speakers and measure their baffle dimensions, colors, weight, and anything else that has very little impact on the sound of the speaker, and I could generate an algorithm that would predict preferences on that same set of speakers with fairly high accuracy (somewhere between 70 and 100%).

To test whether or not these type of algorithms are really giving us useful information, you HAVE to test them on closed data (speakers that were not used in training the algorithm). I haven't seen any such results reported. I'm not saying the algorithms developed would perform poorly, but providing results on closed data would go a long way to eliminate doubt. It's certainly true that reporting the results on the training data has the potential to exaggerate the accuracy of the algorithm.

Additionally, since all they are doing is ranking speakers that exist, it does little to help us push the boundaries of the science of speaker design. We have the potential to weed out poor designs, but it doesn't really inform us what the ideal approach to speaker design is, since that approach isn't even represented by the speakers included in the training data.

Having said that, I'm grateful for research that is as good as they have created. Let's just keep it in perspective. There's still more to learn.
One problem that would fit in to your closed system argument would be over fitting the data. This would be reflected by the number of degrees of freedom remaining. Probably this is represented in the papers, but I don't know that. Otherwise, I don't think it is necessary to test untested speakers to verify the model. There won't be a future change in people's hearing. Of course, it is bounded by the types of variations in speaker response present in the initial data set. So I would propose that it will be predictive as long as the untested speaker is within those bounds. A case where previous prediction can be less predictive would be stock market trends, where new data coming in may actually be influenced by the previous analysis and knowledge of it by the investors.

One specific method question I would have is what about removing the bass component with a high pass filter and redoing the double blind data and the resulting fitting. Since deeper and well behaved bass is obviously important to the subjective listener, but yet people will generally have subwoofers, removing it will improve the precision of fitting the other parameters.

Of course, I would want more disclosure of what comparison speakers were tested (for Harman double blind tests) and ideally I would want to be able to hypothesize speaker improvement ideas (or different curves) and be able to test them, but Harman doesn't owe us this. This would be more of a job for an independent organization, like Floyd Toole's prior research job.

Living room: Samsung PN60F8500, Sony X800M2, Roku Premiere+, Revel W763, Revel M16, Hsu HB-1 MK2, RSL C34E, Rhythmik D15SE, Denon X4200W; Zone 2: Speaker selector, Yamaha RX-V495 (1999); Bedroom: Panasonic P50U54, Oppo BDP-80, Chromecast Audio, Advent Legacy III (1994), Denon 2310ci; Cabin: Roku 3, NHT Super Zero (1996), Hsu VTF-2 MK4, Denon 1712
highmr is offline  
post #3359 of 5320 Old 07-07-2019, 05:52 AM
Member
 
OldMovieNut's Avatar
 
Join Date: Aug 2006
Location: Fort Wayne, IN
Posts: 191
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 50 Post(s)
Liked: 54
Quote:
Originally Posted by torii View Post
thanks. I found your response helpful. but still hard for me to equate graphs with sound quality
It looks like you're measuring both speakers simultaneously. Measure them one at a time, otherwise the data will be confused. Post those results.

Cheers,
OldMovieNut
OldMovieNut is offline  
post #3360 of 5320 Old 07-07-2019, 05:58 AM
Member
 
OldMovieNut's Avatar
 
Join Date: Aug 2006
Location: Fort Wayne, IN
Posts: 191
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 50 Post(s)
Liked: 54
Quote:
Originally Posted by DonH50 View Post
I had to repair mine an original 700) several times. It also managed to take out a few speakers along the way. The 700B was more reliable, and the 400 never seemed as bad IME.
I had a Phase 400 back in the 70's and never had a problem with it. I had the thermal cutout shut it down several times playing Also Sprach Zarathrustra into Infinity 2000As, though. I don't know what would happen if you try that with one of today's receivers.

Cheers,
OldMovieNut
OldMovieNut is offline  
Sponsored Links
Advertisement
 
Reply Speakers

Tags
cea 2034 , double-blind , listening tests , loudspeaker measurements , spinorama

Thread Tools
Show Printable Version Show Printable Version
Email this Page Email this Page


Forum Jump: 

Posting Rules  
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off
Trackbacks are Off
Pingbacks are Off
Refbacks are Off