Research Project: Infinity IL10 Speaker Review & Measurements

[QMuse]

My listening impressions may be wrong still but don't be in such a hurry to write them off this way.

I don't really tink your listening impression is "wrong" - I trust you when you said this speaker sounded "grunginess and lack of clarity to everything it played" to you. In other words, your listening impression may be correct for you but it may not apply to anybody else but you. If that same impression would be done with non-sighted listening and be backed by several more listeners than it would start to present value to the others. Of course, you don't have resources to do such testing, but the fact remains it is a single sighted listening impression and your training doesn't help much in that context.

The other thing is I can't relate those 2 distoriton peaks from THD measurement to that description as I think it can't explain nor justify that impression.

 

[QMuse]

This is a joke, right? Do you realize that you're commenting on the exact speaker model that was deemed to be superior to 12 others in a rigorous controlled double blind test, and is part of the initial set of speakers that was used as a starting point for the Olive preference model? This is actually one of the worst possible examples you could base this argument on!

Speaking of that double blind test, in which kind of listening environment was the test conducted? Was it anechoic, treated for room modes or similar to "normal" rooms? If it was non-anechoic, were the speakers response EQ-ed or not?

 

[amirm]

I don't really tink your listening impression is "wrong" - I trust you when you said this speaker sounded "grunginess and lack of clarity to everything it played" to you. In other words, your listening impression may be correct for you but it may not apply to anybody else but you. If that same impression would be done with non-sighted listening and be backed by several more listeners than it would start to present value to the others. Of course, you don't have resources to do such testing, but the fact remains it is a single sighted listening impression and your training doesn't help much in that context.

99% of audio research in the industry is performed sighted. Double blind tests are final confirmation. I research we are talking about took months. No way it is the method to develop products. By your notion no one can do any work that doesn't take decades.

Indeed the main reason trained listeners exist is because they can give quick answers to make progress.

We do want to be dismissive of average youtuber sitting there listening to two speakers with no formal training in anything, motivated by free loaners to say this or that. That is not what we are dealing with here.

So no, you don't have any evidence that my impressions don't apply to others. That is a supposition.

 

[amirm]

Speaking of that double blind test, in which kind of listening environment was the test conducted? Was it anechoic, treated for room modes or similar to "normal" rooms?

You wait to ask now???

No, the test was not anechoic. What kind of question is that?

Harman has an "IEC compliant" listening room they use for such tests. It is supposed to be standardized version of what a typical living room would be. Sean Olive has written a whole paper on it. Here is me in there taking a picture of Sean Olive while he is running his Harman training Quiz:

 

[amirm]

The other thing is I can't relate those 2 distoriton peaks from THD measurement to that description as I think it can't explain nor justify that impression.

What you can't relate to is not something that we care about. It is not like you are a trained listener or expert in psychoacoustics.

As an aside, a world where distortions in whole percentage values doesn't matter to audio is a very strange one. I suggest thinking through this one before saying stuff like this.

 

[QMuse]

So no, you don't have any evidence that my impressions don't apply to others. That is a supposition.

Do you have any evidence that sighted test done by a trained listener applies to others? Btw, don't take it personal, I wouldn't trust Olive's sighted listening test done in your room after peeking at measurements. I simply don't believe in sighted test done by a single person. I treat it as personal opinion, not measured fact.

 

[QMuse]

What you can't relate to is not something that we care about. It is not like you are a trained listener or expert in psychoacoustics.

As an aside, a world where distortions in whole percentage values doesn't matter to audio is a very strange one. I suggest thinking through this one before saying stuff like this.

I don't know who are "we", but those 2 peaks simply cover to narrow frequency range to have impact you described in your listening test. Are you saying those 2 distorion speaks are indeed responsible for "grunginess and lack of clarity to everything it played"? Blaming THD measurement like this one for such effect is quite a bold claim..

 

[QMuse]

You wait to ask now???

No, the test was not anechoic. What kind of question is that?

Harman has an "IEC compliant" listening room they use for such tests. It is supposed to be standardized version of what a typical living room would be. Sean Olive has written a whole paper on it. Here is me in there taking a picture of Sean Olive while he is running his Harman training Quiz:

View attachment 70663

So that room on the photo, without any furniture to speak about and with all those panels on the sides and behind the speakers is supposed to resemble typical living room? Mine is nothing like that. Is yours like that? This one looks similar to a home theater room, not a living room.

Btw, did Harman ever published measured actual in-room response of any speaker in that room?

 

[napilopez]

I'm guessing you meant "room EQ" when you said EQ as there's nothing much to EQ in terms of speakers response here. Sure, you can correct those small resonances at 700, 1100 and 4700 Hz and correct hot tweeter response north of 6kHz, but it's nothing major - this speaker is pretty good the way it is and Amir didn't complaint about tonality.



I really can't relate those 2 narrow distortion peaks with the description of " grunginess and lack of clarity to everything it played".View attachment 70662

I agree with you about distortion. Perhaps a different kind of distortion test would be more useful?

But I no, I didn't mean Room EQ, although that could play into things as well. Purely looking at the spin, I think the resonances are quite a bit more prominent on the Infinity IL10 compared to the revel. The 700Hz and 4700 Hz bumps seem like theyd be quite audible flaws to focus on - the resonances on the M16 by comparisons seem much less audible.

I don't have a good an explanation for 'grunginess' but as mentioned in my earlier post, Amir commented that the revel and infinity had overall similar tonality but the IL10 suffered from a 'lack of clarity'. Immediately I thought of the fact that after 'great bass' the second most common comment about the IL10 in the Olive Paper was 'Dull,' followed by 'Mid depression.'

Here are the listening comments most often attributed to the IL10 in the paper:

Great bass (22), Dull (20), Mid Depression/hole (16), Mellow (11), Smooth (10), Neutral (8), Balanced (7), Tubby (6), Colored (6), Muffled (5), Lacks Bass (5), Raspy (4), Warm (4), Bass Extended (4), Flat (4).

Regardless of the fact the speaker was ranked number 1, these are qualities listeners did hear and described in their assessments.

Notably, the speaker ranked 3/13 in 'Dull,' 1/13 in 'mid depression,' and 1/13 in 'raspy.' I don't know about you, but that seems to correlate pretty well with what Amir described. These qualities are likely highlighted in the IL10 because the others had much more audible flaws, (Rank 4 had 74 complaints of 'bright,' for instance), but I think most all of these can be explained by the spin. Except for maybe raspy, which might be the 'grunginess' amir heard?

Preference score putting the M16 ahead aside, at least some of the IL10's flaws are clearly evident from the spins. That Amir did or didn't like the speaker is of less interest to me than the actual tonal qualities he heard.

Concordance between the Klippel and the Harmann data is very good. Below 100Hz, Klippel shows more output (or Harmann anechoic chamber is not perfect below 100Hz).

View attachment 70648

P.S.: I agree with others that this thread is deviating from a data driven approach ...

@pierre, what's going on in that graph, did you use the wrong speaker? I personall do not consider that very good correlation . But when I imported the data into REW I get much closer correlation:

Note, I recalculated the Early Reflections/ERDI curves with the proper formula, which shows the scoop in the upper mids region.

Notably, the ERDI curve in particular is essentially a perfect correlation above 60-70Hz (presumably when Harman's chamber stops being fully anechoic).

P.S. @amirm FYI, just a closer look at how your measurements compare to Harmans and where room effects kick in. Don't think you could ask for much better than this, especially the ERDI curve.

 

[GelbeMusik]

Of course, I am referring to classical instruments/recordings only.

Well, there it is again, the superstition of natural, true instruments. A double bass, for example, practically only emits harmonics, the basic tone is usually not present. How should one evaluate this situation? On an electric bass, the level of the harmonics is around 100% - for each one separately! THD 500% or so?
Synthetic babbling down below (e/g Yello, Kraftwerk) has to add harmonics to fit into the musical context at all. Otherwise it doesn't mix ... I don't believe that musicians weigh every half percent on the gold scale.

I don't know who are "we", but those 2 peaks simply cover to narrow frequency range to have impact you described in your listening test. Are you saying those 2 distorion speaks are indeed responsible for "grunginess and lack of clarity to everything it played"? Blaming THD measurement like this one for such effect is quite a bold claim..

Perhaps the assumption is not entirely wrong if resonances are to shape the picture. Resonances, which are reflected in distortion measurements, could form a sound that is constantly running in the background and is ultimately excited by every acoustic event. Quite notorious in this context are the first metal membranes and their use at that time. Although the excitation by the clean input signal was suppressed, the excitation by harmonics from the motor or suspension non-linearities was not.

Intermodulation thickens the sound image, that is well known.

… the second most common comment about the IL10 in the Olive Paper was 'Dull,' followed by 'Mid depression.'

If you look at it in relation to the peak at 4.5kHz? And that's from the bass/midrange driver. It's resonance. People probably (unconsciously) took this peak as a reference to judge the midrange?

I know the JBL2026 very well as a low/midrange driver. Its great quality might have something to do with the very unusual resonance-free midrange. It has its own signature of a kind of clear silence in the spaces in between, without any shine.

 

[patate91]

99% of audio research in the industry is performed sighted. .

That's not a valid argument. It's not scientific too

 

[edechamps]

Speaking of that double blind test, in which kind of listening environment was the test conducted? Was it anechoic, treated for room modes or similar to "normal" rooms? If it was non-anechoic, were the speakers response EQ-ed or not?

From the paper in which that specific speaker was assessed:

[35] leads to this paper which seems to be freely available here.

 

[QMuse]

I agree with you about distortion. Perhaps a different kind of distortion test would be more useful?

But I no, I didn't mean Room EQ, although that could play into things as well. Purely looking at the spin, I think the resonances are quite a bit more prominent on the Infinity IL10 compared to the revel. The 700Hz and 4700 Hz bumps seem like theyd be quite audible flaws to focus on - the resonances on the M16 by comparisons seem much less audible.

I don't have a good an explanation for 'grunginess' but as mentioned in my earlier post, Amir commented that the revel and infinity had overall similar tonality but the IL10 suffered from a 'lack of clarity'. Immediately I thought of the fact that after 'great bass' the second most common comment about the IL10 in the Olive Paper was 'Dull,' followed by 'Mid depression.'

Here are the listening comments most often attributed to the IL10 in the paper:

Great bass (22), Dull (20), Mid Depression/hole (16), Mellow (11), Smooth (10), Neutral (8), Balanced (7), Tubby (6), Colored (6), Muffled (5), Lacks Bass (5), Raspy (4), Warm (4), Bass Extended (4), Flat (4).

Regardless of the fact the speaker was ranked number 1, these are qualities listeners did hear and described in their assessments.

Notably, the speaker ranked 3/13 in 'Dull,' 1/13 in 'mid depression,' and 1/13 in 'raspy.' I don't know about you, but that seems to correlate pretty well with what Amir described. These qualities are likely highlighted in the IL10 because the others had much more audible flaws, (Rank 4 had 74 complaints of 'bright,' for instance), but I think most all of these can be explained by the spin. Except for maybe raspy, which might be the 'grunginess' amir heard?

Preference score putting the M16 ahead aside, at least some of the IL10's flaws are clearly evident from the spins. That Amir did or didn't like the speaker is of less interest to me than the actual tonal qualities he heard.

Looking at CSD graph I would also stick to what I said about the resonance at 1200Hz as it is visible at LW and PIR as well. Anyway, I have no trouble to believe that once EQ-ed this speaker would sound better but it is hard to say if it woul start to sound like M16 even if the preference score would be equal after EQ.

But that is not the point here - the thing seems to be Amir's opinion to blame THD measurement for the flaws he was hearing althoug Toole claimed THD is rarely an issue. I certainly find it hard to believe THD would be an issuse with the speaker ranked #1 in Olive's test.

 

[QMuse]

From the paper in which that specific speaker was assessed:

View attachment 70668

[35] leads to this paper which seems to be freely available here.

I don't really think typical listening room has average reverberation time of 0.3 sec, at least not in EU. If I have to guess it would be closer to 0.5 sec.

Btw, it would be nice to see comaprison between PIR and actual measured in-room response for at least several speakers that were tested there.

 

[ROOSKIE]

But that is not the point here - the thing seems to be Amir's opinion to blame THD measurement for the flaws he was hearing althoug Toole claimed THD is rarely an issue. I certainly find it hard to believe THD would be an issuse with the speaker ranked #1 in Olive's test.

This is the part that gets me as well.
I just don't yet see a clear part to Harmonic distortion specifically being the reason for amirs subjective experiences with the speakers thus far reviewed.
There may be other factors and yet somehow these levels of harmonic distortion get blamed because we have been trained to fear 1%&2%(even narrow band) & because harmonic distortion happens to be easy to measure and put in a chart.

It seems to keep getting lost here as to what harmonic distortion actually is.

Maybe it would put everyone at a good point if @amirm could talk a bit about that type of distortion, what it is, how it plays out in the playback and what one could expect to hear. Let's say a speaker has 2% 3rd order harmonic distortion at 2khz. What is happening and how does the sound change vs .5%?

This would be helpful IMHO.

 

[napilopez]

I don't really think typical listening room has average reverberation time of 0.3 sec, at least not in EU. If I have to guess it would be closer to 0.5 sec.

Btw, it would be nice to see comaprison between PIR and actual measured in-room response for at least several speakers that were tested there.

I agree it would be nice, but so far I think there's enough anecdotal data to show PIR ends up being quite close to the real thing even in very different rooms. I know you saw the following, but for others following along for example, a while back @tecnogadget posted MMM measurements of the R3 in a very irregular room:

Yet it still matched my calculated PIR for the R3 almost perfectly:

 

[QMuse]

99% of audio research in the industry is performed sighted.

Last time you mentioned it it was 90%, nice to see the percentage is increased as neeed.

And than there's this quote. Once you started to provide speaker listening tests they obviously don't need to be bias controlled and blind:

"As to listening tests, one hopes those are bias controlled and blind. Otherwise, such tests are not reliable at all. And certainly not indicative of "voicing" or whatever fix you are applying."

One more here:

"Those are all subjective terms created out of sighted listening. I doubt there is truth to much of that. "

I especially like this one, I like the :

"Here is a convincing example. Even though speakers are thought to sound different enough to not need blind tests, they absolutely do. See: http://seanolive.blogspot.com/2009/04/dishonesty-of-sighted-audio-product.html "
In fact, I like this one so much I'll quote what Olive said i the blog:

So much about consistency regarding bias and sighted listening tests - now you are saying your sighted listening tests are to be trusted by everyone here. No offense, but I think not.

 

[amirm]

That's not a valid argument. It's not scientific too

It is the reality. I know, that is how we developed audio technology including world standards.

Trained listeners are hired because they can provide reliable answers that you can use to design products. They can occasionally misfire as I have but most of the time they get it right.

 

[QMuse]

I agree it would be nice, but so far I think there's enough anecdotal data to show PIR ends up being quite close to the real thing even in very different rooms.

Oh, for sure, I don't doubt PIR at all in that context, I would just like to see how room modes are looking in that room they were doing the testing in.

 

[amirm]

Last time you mentioned it it was 90%, nice to see the percentage is increased as neeed.

And than there's this quote. Once you started to provide speaker listening tests they obviously don't need to be bias controlled and blind:

"As to listening tests, one hopes those are bias controlled and blind. Otherwise, such tests are not reliable at all. And certainly not indicative of "voicing" or whatever fix you are applying."

One more here:

"Those are all subjective terms created out of sighted listening. I doubt there is truth to much of that. "

I especially like this one, I like the :

"Here is a convincing example. Even though speakers are thought to sound different enough to not need blind tests, they absolutely do. See: http://seanolive.blogspot.com/2009/04/dishonesty-of-sighted-audio-product.html "
In fact, I like this one so much I'll quote what Olive said i the blog:

You continue to be confused due to lack of real hands on experience with this topic. We absolutely want to rail against inexperienced people passing judgments on audio products in sighted tests. Speakers are included in that. I already explained this to you.

You can't mix that with trained listeners who get paid to do a job, not to provide online reviews for people. As I noted, they can make mistakes as I have explained about my own experiences. But in general, they provide reliable information. We had a handful of them at Microsoft and this is what they did all day, every day. Every few months we would conduct blind tests with larger scale but day to day development of products relied on trained opinion.

Look at the discrimination ability of Harman trained listeners:

They don't just lose that skill when you ask them to evaluate a speaker sighted. They continue to have powerful tools that you don't have. So you best listen to them.

Sure, you can have allowance of mistakes but to say that their opinion is worthless makes that statement worthless and no more.