Infinity R162 Bookshelf Speaker Review

[QMuse]

It's hard to believe that these EQs are really this good. I kinda doubt that you can EQ a speaker like this to sound as good as the Genelec. Maybe a limitation of the formula, or maybe it's true, but my instinct is to doubt at the moment.

Genelec is also EQ-ed to be linear. When you compare speakers EQ-ed to similar FR what you would notice first is difference in bass. I could easilly push LF response of the Infinity for few dB at 50Hz to be equal to Genelec 8341A, and it may easilly turn out that Infinity, being larger, would have better bass which can go louder too.

What you would also hear is a difference in directivity, but Infinity is very good with that too. I doubt you would hear the difference in the distortion. In other words, it may be a tough job to hear the difference between those 2 speakers in a proper blind test.

You may want to check article @mitchco wrote when comparing 2 different speakers EQ-ed to the same target.

 

[Jon AA]

We also have to keep in mind, actual measured results of the speaker after EQ won't necessarily match the predicted results. Some flaws can't be EQ's effectively.

 

[QMuse]

We also have to keep in mind, actual measured results of the speaker after EQ won't necessarily match the predicted results. Some flaws can't be EQ's effectively.

This speaker showed no resonances (except some at 100Hz) and response of the drivers is minimum phase so I would expect measurement to be very close to the predicted response.

 

[Jon AA]

Oh, I think this particular speaker should do quite well, but you never know for sure until you measure the results.

As for the comparison with the Genelec, while this speaker has really well controlled directivity, it's still significantly worse than the Genelec's (as are most speakers), so I don't think there's much danger of an EQ'd version of this speaker making Genelec owners feel they've wasted their money.

 

[QMuse]

Oh, I think this particular speaker should do quite well, but you never know for sure until you measure the results.

As for the comparison with the Genelec, while this speaker has really well controlled directivity, it's still significantly worse than the Genelec's (as are most speakers), so I don't think there's much danger of an EQ'd version of this speaker making Genelec owners feel they've wasted their money.

It is the room that introduces non-minimum phase artifacts so in-room response may sometimes differ from prediction, but when EQ-ing anechioc response it is pretty much never the case.

Difference in directivity is pretty small IMO and Infinity could easilly turn out to have better bass at higher SPL. If both would be paired with subs I wouldn't put my money that difference would be easilly spotted in a blind test.

 

[Cahudson42]

I remember earlier discussions about the Score being reported to two decimals - being decided that was 'meaningless' and one decimal suffices.

In reality, a single rounded digit and no decimal at all is probably all the Score deserves..there seem to be enough 'deficiencies' that a single digit is all it's worth - if even that. Just IMHO..

Personally, I think distortion, measured at various SPL levels, may tell us as much.. or more..

 

[richard12511]

Genelec is also EQ-ed to be linear. When you compare speakers EQ-ed to similar FR what you would notice first is difference in bass. I could easilly push LF response of the Infinity for few dB at 50Hz to be equal to Genelec 8341A, and it may easilly turn out that Infinity, being larger, would have better bass which can go louder too.

What you would also hear is a difference in directivity, but Infinity is very good with that too. I doubt you would hear the difference in the distortion. In other words, it may be a tough job to hear the difference between those 2 speakers in a proper blind test.

You may want to check article @mitchco wrote when comparing 2 different speakers EQ-ed to the same target.

I've read @mitchco 's article before. Great read!

You're right that the Genelec uses EQ to achieve the score it does. A better example would probably be the Revel F208, especially since it's in the same Harman family as the Infinity. I think it would be a really interesting concept to test. How much can speaker EQ really fix a speaker? Can you EQ this speaker to sound better than a Revel F208(at least with a sub and to a certain volume)? Going by the Olive score, it seems to be the case, but I still have my doubts. They're not confident doubts, though.

As for the comparison with the Genelec, while this speaker has really well controlled directivity, it's still significantly worse than the Genelec's (as are most speakers), so I don't think there's much danger of an EQ'd version of this speaker making Genelec owners feel they've wasted their money.

That difference in directivity control should be accounted for by the Olive score, though, right? Or maybe you're saying it might not be weighted heavily enough?

My question was more about what could the Olive score be missing. Or maybe it's not missing, and these really would have a greater than 50% probability of being preferred over the Genelecs. Right now I'm leaning more towards your side, but I think it would be interesting to test.

I remember earlier discussions about the Score being reported to two decimals - being decided that was 'meaningless' and one decimal suffices.

In reality, a single rounded digit and no decimal at all is probably all the Score deserves..there seem to be enough 'deficiencies' that a single digit is all it's worth - if even that. Just IMHO..

Personally, I think distortion, measured at various SPL levels, may tell us as much.. or more..

I disagree that a single rounded number with no decimal is good enough, but I can agree with the 1 decimal point.

I also think it depends on what you're trying to get out of the score. If you're trying to use it to make purchasing decisions, then I probably wouldn't be comfortable making such decisions completely on score differences of less than 1.0(more like 1.5). Even then, I think there's evidence to support it doesn't do well with really weird designs(like omnis), and you still have to take into account max output and personal directivity preferences. In that sense, I agree with you that differences smaller than say 1.0 are probably not so meaningful.

If, however, you're just trying to make general statements of preference with a large sample size of say 1,000 listeners, then I think small differences in score can be relevant. For example, lets put this Infinity R162 against it's bigger brother Revel F208:

The EQed Infinity has a score of 6.6, while the Revel has a score of 6.1, so a 0.5 difference. Personally, that's no where near a big enough difference for me to make a purchasing decision off of, but for predicting preference in the general population over a large enough sample size, I see no reason why it shouldn't be relevant. Im assuming an equal score to roughly mean that one should be preferred 50% of the time over another. How does that % chance change with a 0.5 difference? 55%, 60%, 65%? I honestly have no idea. I know someone calculated that a 1.6 difference would lead to a 95% chance, but I don't know how they calculated that; I asked in another thread, but got no response.



PS: I'm listening to this speaker's bigger brother(R263) this morning, and I think it sounds great for the money.
PSS: I know @QMuse is an advocate for room EQ being *mostly equal for speakers below a certain point. I'm not sure if I agree or disagree, but in this case I found that the R263 measured very similarly to my mains below 300Hz, and so I was able to just reuse the sub 300Hz EQ that I use for my mains, and it worked well.

 

[QMuse]

I've read @mitchco 's article before. Great read!

You're right that the Genelec uses EQ to achieve the score it does. A better example would probably be the Revel F208, especially since it's in the same Harman family as the Infinity. I think it would be a really interesting concept to test. How much can speaker EQ really fix a speaker? Can you EQ a speaker to sound better than a Revel F208(at least with a sub and to a certain volume)? Going by the Olive score, it seems to be the case, but I still have my doubts. They're not confident doubts, though.




That difference in directivity control should be accounted for by the Olive score, though, right? Or maybe you're saying it might not be weighted heavily enough?

My question was more about what could the Olive score be missing. Or maybe it's not missing, and these really would have a greater than 50% probability of being preferred over the Genelecs. Right now I'm leaning more towards your side, but I think it would be interesting to test.

You should check the preference rating score, but IIRC score evaluates smoothness of directivity curves, not particular angle. IMHO wider horizontal directivity may be soemthing that can improve scooring.

EQ doesn't affect directivity so assuming 2 speakers with similarly smooth directivity they should sound pretty much the same once you EQ their FR.

So yes, you can EQ some imaginary speaker to sound better than F208. In fact, you can even EQ F208 to sound better than without EQ.

 

[richard12511]

You should check the preference rating score, but IIRC score evaluates smoothness of directivity curves, not particular angle. IMHO wider horizontal directivity may be soemthing that can improve scooring.

EQ doesn't affect directivity so assuming 2 speakers with similarly smooth directivity they should sound pretty much the same once you EQ their FR.

So yes, you can EQ some imaginary speaker to sound better than F208. In fact, you can even EQ F208 to sound better than without EQ.

Oops, I actually mean to say "this" speaker, not "a" speaker.

Can you EQ *this speaker to sound better than the Revel F208(at least up to a certain volume)?

 

[QMuse]

Oops, I actually mean to say "this" speaker, not "a" speaker.

Can you EQ *this speaker to sound better than the Revel F208(at least up to a certain volume)?

You would need a good pair of subs and do the integration work properly. As directivity of R162 is on paar with F208 it will all be about that distortion spike around 2kHz. If that turns out to be a non issue you could as well end up with a better sound as F208 wouldn't be able to match LF response of a sub.

 

[edechamps]

The EQed Infinity has a score of 6.6, while the Revel has a score of 6.1, so a 0.5 difference. Personally, that's no where near a big enough difference for me to make a purchasing decision off of, but for predicting preference in the general population over a large enough sample size, I see no reason why it shouldn't be relevant. Im assuming an equal score to roughly mean that one should be preferred 50% of the time over another. How does that % chance change with a 0.5 difference? 55%, 60%, 65%? I honestly have no idea. I know someone calculated that a 1.6 difference would lead to a 95% chance, but I don't know how they calculated that; I asked in another thread, but got no response.

The Olive study contains a chart (that @Sancus helpfully digitized) showing the actual preference scores plotted against the scores the model predicted. From there, one can observe that the error follows a gaussian distribution, and @bobbooo was able to deduce that you'd need a 1.6 score difference to have 95% confidence that a representative listener would prefer one speaker over the over.

 

[QMuse]

The Olive study contains a chart (that @Sancus helpfully digitized) showing the actual preference scores plotted against the scores the model predicted. From there, one can observe that the error follows a gaussian distribution, and @bobbooo was able to deduce that you'd need a 1.6 score difference to have 95% confidence that a representative listener would prefer one speaker over the over.

Here's what's buggin me with that: I can introduce a very audible wide peak in the region where our ears are very sensitive (say around 1.5khz) at that EQ-ed version of R162 response that would spoil the score for say 1.0, and pretty much everyone should be able to hear the difference.

 

[edechamps]

Here's what's buggin me with that: I can introduce a very audible wide peak in the region where our ears are very sensitive (say around 1.5khz) at that EQ-ed version of R162 response that would spoil the score for say 1.0, and pretty much everyone should be able to hear the difference.

Yes. That's why it's a probabilistic confidence interval, not an exact estimate. But even if it was, there are many ways you can game the score if you do it on purpose, because the score assumes a "typical" loudspeaker, where "typical" means "similar to the loudspeakers used in the study". If you add a worst-case peak on purpose, you are breaking that assumption, and you shouldn't be surprised the score doesn't work anymore. Again this comes back to correlation vs. causality.

 

[QMuse]

Yes. That's why it's a probabilistic confidence interval, not an exact estimate. But even if it was, there are many ways you can game the score if you do it on purpose, because the score assumes a "typical" loudspeaker, where "typical" means "similar to the loudspeakers used in the study". If you add a worst-case peak on purpose, you are breaking that assumption, and you shouldn't be surprised the score doesn't work anymore. Again this comes back to correlation vs. causality.

Sure, all clear there - it is not very typical to have a speaker that does everything almost perfectly except a single wide peak.

What bothers me more is the fact that scoring model has not been tested with a different set of speakers than the one that was used to develop a model. It seems nobody is doing that kind of research anymore, which is a pity as model looks promising..

 

[Jon AA]

It is the room that introduces non-minimum phase artifacts so in-room response may sometimes differ from prediction, but when EQ-ing anechioc response it is pretty much never the case.

I was speaking of anechoic only. Many horns/waveguides, concentric drivers, etc, will have flaws in the response caused by acoustic cancellations/diffraction, etc, that won't respond to EQ in the real world as well as it does for you on a screen--if at all. We've also seen numerous speakers measured that have odd bumps/dips in the lower mid-range due to port resonances, etc, that are accompanied by abrupt directivity changes which makes the prospect of EQ flattening them out to perfection rather low.

Difference in directivity is pretty small IMO

There's enough difference to expect that even with the Infinity LW EQ'd perfectly flat they would not sound the same. In-room measurements would show significant differences in the steady state curve as well. Relative to the Genelec the Infinity has a narrowing from 1K-2.5K and a widening from 2.5-10K while the Genelec remains more nearly constant. I'm not saying 100% of people would prefer the Genelec in a blind test as they'd both be very good, but they'd certainly sound different--people should be able to easily tell them apart.

 

[QMuse]

There's enough difference to expect that even with the Infinity LW EQ'd perfectly flat they would not sound the same. In-room measurements would show significant differences in the steady state curve as well.

This is Genelec's predicted in-room response vs corrected estimated in-room response of Infinity. As DI's are similar ER and SP came similar too once you flatten the LW, and thus PIR is practically the same. This indicates that steady state in-room response would also resemble PIR once you remove the room effects.

I could have easilly pushed the 35-90Hz range to match the Genelec curve. Actually I believe Infinity has better bass than that tiny Genelec.

I'm not saying 100% of people would prefer the Genelec in a blind test as they'd both be very good, but they'd certainly sound different--people should be able to easily tell them apart.

As I said, I'd put my money on the bet they would sound extremely similar as LW, ER and SP (and thus PIR) are practically the same.

 

[Jon AA]

This is Genelec's predicted in-room response vs corrected estimated in-room response of Infinity.

While it looks subtle on the graph, you will have a very broad dip in the overall response of the Infinity relative to the Genelec in the 2K region with equal LW EQ.

As I said, I'd put my money on the bet they would sound extremely similar as LW, ER and SP (and thus PIR) are practically the same.

"Practically the same" can mean anything. The original measurements of the Infinity show a very flat LW between 1-4KHz with maybe a 1 db dip. The ER and SP show 3-4 db dip. That dip is going to be there unless you put a bump in the LW.

But you're missing the broader point.

As DI's are similar.....

I and others have noted for quite some time, that a single DI can be deceiving--or at the very least, incomplete in describing the dispersion of a speaker. These two speakers will have significantly different horizontal and vertical DI's. All discussion I've run into in the various research, indicate sidewall reflections have a different impact on the perception of the sound than do ceiling and floor reflections. If you look at the sidewall reflections alone:

Genelec:

Focus in on the pink line. Now the Infinity:

Again, the pink line. As you can see, the sidewall reflections from the Infinity have a very deep and broad dip from 1-3KHz and are much brighter above that relative to the Genelec. With differences that great, a listener is going to have little problem telling that they're different speakers.

And that's just for a single listener at the sweetspot. For multiple listeners, where horizontal dispersion is even more important, the differences between the two will be more obvious.

Like I said, the Infinity EQ'd to have a flat listening window should sound very good. But let's not get carried away with the meaning of a single in-room measurement prediction--our ears won't hear it that way.

 

[QMuse]

While it looks subtle on the graph, you will have a very broad dip in the overall response of the Infinity relative to the Genelec in the 2K region with equal LW EQ.

"Practically the same" can mean anything. The original measurements of the Infinity show a very flat LW between 1-4KHz with maybe a 1 db dip. The ER and SP show 3-4 db dip. That dip is going to be there unless you put a bump in the LW.

But you're missing the broader point.

I and others have noted for quite some time, that a single DI can be deceiving--or at the very least, incomplete in describing the dispersion of a speaker. These two speakers will have significantly different horizontal and vertical DI's. All discussion I've run into in the various research, indicate sidewall reflections have a different impact on the perception of the sound than do ceiling and floor reflections. If you look at the sidewall reflections alone:

Genelec:

View attachment 68814

Focus in on the pink line. Now the Infinity:

View attachment 68815

Again, the pink line. As you can see, the sidewall reflections from the Infinity have a very deep and broad dip from 1-3KHz and are much brighter above that relative to the Genelec. With differences that great, a listener is going to have little problem telling that they're different speakers.

And that's just for a single listener at the sweetspot. For multiple listeners, where horizontal dispersion is even more important, the differences between the two will be more obvious.

Like I said, the Infinity EQ'd to have a flat listening window should sound very good. But let's not get carried away with the meaning of a single in-room measurement prediction--our ears won't hear it that way.

I thought it is clear that I'm comparing EQ-ed Infinity response to the measured Genelec response. That means that side wall bounce (your pink line) gets EQ-ed as well.

It also means that it doesn't make sense to speak about original Infinity response as we don't know Genelec's original (non EQ-ed) response - and even if we would knew it it again wouldn't matter, as we are comparing both EQ-ed versions.

Let's now see that "very broad dip in the overall response of the Infinity relative to the Genelec in the 2K region with equal LW EQ", whatever "overall" means in this context. I have already shown PIR comparison, so let's see how ERs compare:

No "broad dip" there, Infinity's ER is a bit smoother as Genelec's ER is showing 2 small peaks, but for any practical purpose those 2 curves are identical.

Let's try SP comparison:

Here Genelec's response is a little smoother, but differences are so small they are within margins of measurement error so nothing to write home about.

As all 4 curves (LW, ER, SP and PIR) are shown to be identical for any practical purpose I have no idea why do you think anybody could tell them apart in a blind listening test. As preference ratings are the only proven indicator so far related to what we hear and what we prefere maybe you're the one missing the broader point of spinorama charts and preference ratings.

 

[Jon AA]

I thought it is clear that I'm comparing EQ-ed Infinity response to the measured Genelec response. That means that side wall bounce (your pink line) gets EQ-ed as well.

So you give the side wall bounce a 4 db bump to fix the big dip, but you don't add that to the LW EQ. That's a fine trick, if you figure out a way to market those skills you'll be rich!

As all 4 curves (LW, ER, SP and PIR) are shown to be identical for any practical purpose

They just aren't. Guessing preference is one thing, the rather absurd claim nobody would be able to tell they're different speakers is quite another--and far, far beyond a "practical purpose." To make the claim people will be unable to tell they are different speakers, you might want to shoot for "more identical" if you want to keep your money. For example the ER DI:

Far from identical. And again you're glossing over the point that several of the curves that make up the ER curve are dramatically different as shown in the previous post.

I know you've had trouble accepting the idea that two ears and a brain will hear things differently than the steady state line drawn from an omnidirectional microphone but you've never shown any evidence to counter the mountains of evidence that shows this to be the case.

 

[edechamps]

For example the ER DI:
View attachment 68897
Far from identical.

It looks like you plotted the Infinity RC263 instead of R162? The RC263 is a center speaker, it's no wonder the DI will be very different.