KEF R3 Speaker Review

[echopraxia]

I don't disagree with you about industry standards. My crusade has always been against the validity of the preference score as a way to determine listener preference.

The speaker the R3 lost the blind test to has good spins and overall, a neutral response. If you read the details of the blind test, the differences weren't in coloration of the response, but clarity, realism, and the harshness in the R3's.

I’m the one who performed that blind test, and I agree — I couldn’t find any obvious major coloration differences once bass was normalized. They sounded remarkably similar actually, in terms of frequency response. But they did not sound similar in terms of “fidelity”, for lack of a better word.

This objective and subjective review from Amir is really fascinating to me, because of how closely it aligns with my KEF R3 experience: I initially bought the R3 because the published specs seemed to indicate a speaker that should outperform just about anything else out there — and now thanks to Amir, we’ve confirmed that KEF was not sugarcoating the measurements. The KEF R3 really does measure fantastically well.

And yet in my blind test, they didn’t really dominate, despite their phenomenal measurements suggesting that they should. They were clearly fantastic speakers, but the truth is when compared side by side, they lost pretty severely to the Ascend Sierra 2EX in a bass-normalized blind test. And this result is mirrored in Amir’s subjective listening where the R3 loses subjectively to another speaker (the Revel), which purely according to measurements shouldn’t be happening. And there are many other accounts you will find online mirroring this.

As for what could explain this? I do not know. Many of us suspect dispersion breadth to be the likely culprit, but we need more data to better establish exactly how much more important wide dispersion may be than how current preference scores weight it. And thanks to Amir’s equipment and hard work, many of us are happy to look forward to much more valuable data

FWIW, ultimately I ended up returning the KEF R3’s, but I still have my Revel F206 and all my Ascend flagship speakers. I still have not done a blind test between the Revel F206 and Ascend RAAL Towers, but I do still plan to get around to it eventually.

 

[gags11]

I'm sure your setup is phenomenal. =)

I think you misunderstood though. I was commenting on the dynamic graphs @BYRTT generated using the speaker data Amir provided.

Totally misunderstood, my bad. I did not want to comment on Amir’s measurements. I just wanted to state that I would like to see such measurements.
Cheers

 

[BYRTT]

I really do. So much of what you all want can be trusted to an assistant. Heck, with speaker measurements, all the work can be done, sans analysis by someone else. I spend so much time also boxing, labeling, and shipping stuff.

Importantly, I could also use some foot massage at times. And back rub.....

Thanks the hard work and ideas of site will wish a trusted assistant appeared out of the blue down the road for other future hard stuff as magic acoustics and mods call maty.

 

[spacevector]

I really do. So much of what you all want can be trusted to an assistant. Heck, with speaker measurements, all the work can be done, sans analysis by someone else. I spend so much time also boxing, labeling, and shipping stuff.

Importantly, I could also use some foot massage at times. And back rub.....

This may be a good spot to land an intern this summer Amir: https://www.apl.washington.edu/departments/acoustics/home.php

I imagine you sitting in your hi-fi lounge smoking a cigar when said intern brings you data printout hot off the NFS and sets up the speaker. You scribble a few notes and the intern cues up your tunes. When done listening, you flick your hand dismissively and the intern takes the speaker away and prepares for the next one. On their way out, they ask if you care for a foot rub.

The good life.

 

[BYRTT]

How are you generating these stunning visualizations man? Very impressive!

Thanks you like it, used program is free VituixCad feeded amirm's nice spin data normal hanged on into first few post of each acoustic review, one can then right click and capture whatever those nice plots VC can present and i use amateur version Photoshop Elements 11 to output the animated gif-file. So if interest is there you or anybody to use or study how to navigate VC then its nice freeware, and i can share via PM MS Excell spreadsheet with a homebrew macro automatic that extract the 72x colonnes of directivity from amirm's spin data into 72x directivity named txt-files that VC use to output its magic as Spinorama and directivity vitualizations in various forms.

 

[TomJ]

Is it possible you can ... Eq them and then compare them?

Yes, nice suggestion to adjust the "house sound" of one to the other and then reconsider the findings. It would be helpful to understand more about why speakers differ as to how loud they need to play for optimal tonal balance.

 

[miike888]

Great review. More evidence that better objective measurements don't necessarily equate to subjective preference.

Its what the science is all about isnt it?

you have to listen blind then the objective messurements will follow the subjective preference!

//Mike

 

[stevenswall]

Remember guys the R3 "does not quite integrate the additional woofer with the coaxial", don't be fooled by these measurements

What does this mean? The dispersion doesn't match? Or they messed up on the crossover?

 

[Vuki]

I guess bass tuninig of the box also plays significant role in listening test. Despite lower reaching bass, r3 could sound 'drier' tnan m16.

 

[TimVG]

Two samples just isn't enough to draw any conclusions about unit-to-unit variation. Depending on the sigma of variation you're looking for at least 100s of units would need to be measured.

@amirm - please continue measuring new units instead of repeats.

I said some. As in better than nothing. You'd be surprised at how bad it can get even within a single pair of speakers.. Not that I'd expect something like this KEF to fall under that. Anyhow, doesn't matter.

 

[QMuse]

Here are the predicted in room responses of both, although at the (too) high listening distance of 12 feet for such speakers the sound power part should dominate even more, I matched them level wise at the ground tones region 200-500 Hz:

View attachment 54032

We see 2 main differences, one is the bass boost of the M16 which is often used for small speakers to compensate their early bass drop and make them sound also "larger" then they are and possibly also done for the US market where rooms are usually bigger and often have not solid walls like in Europe were the R3 was voiced. The other is the mid region were a higher level of the Revel makes something sound more "lively".

That said, bass behaviour should imho be equalized to a room or optimised by placement as otherwise just a random combination of loudspeaker, room acoustics and geometric placement will "win" although it shouldn't be objectively generalised as superior and we know from Toole that bass plays the biggest role in the subjective ranking of a loudspeaker. Also we shouldn't forget that when just one loudspeaker is used vs. a stereo pair, then its tonality is perceived as more bass shy as at the lower frequencies the addition of 2 loudspeakes is closer to 6dB (coherent signals) vs. to higher frequencies where its closer to 3dB (non-coherent) so a single loudspeaker with a bass boost will sound more neutral than a neutral one.

I have one suggestion: when matched this way R3 response is over the entire range either equal or it is lower than M16, which means their average SPLs are not really matched. I suggest ou try to match them int he region 1-2kHz as that may result in a more fair matching.

 

[thewas]

Great review. More evidence that better objective measurements don't necessarily equate to subjective preference.

Is it? It was a short sighted (and thus not free from bias unless you are Chuck Norris) listening test in mono (where the relative bass is a bit lower than in stereo) in a too large listening distance for such loudspeakers in a room with unknown acoustics and "random" placements regarding to the bass. I don't want to say that the existing Harman metric is perfect which it isn't and I have criticised it for that, but on the other hand we cannot seriously take such a quick test as a reason for doubt.

 

[thewas]

I have one suggestion: when matched this way R3 response is over the entire range either equal or it is lower than M16, which means their average SPLs are not really matched. I suggest ou try to match them int he region 1-2kHz as that may result in a more fair matching.

My level matching was on purpose done like that to better show the main tonal differences.

 

[thewas]

Let's remember to put more weight on the measurements than listening tests. The fact that Amir preferred the Revel is his truth, one listener's truth. Informed by measurements sure, but that doesn't eliminate his preferences, biases, and the individual testing conditions that led to his preferences.

I restate that the preference formula has an 86 percent accuracy and was based on multiple listeners. The M16 and R3 are simply each too good a speaker with close enough measurement performance to make the impressions of one listener carry so much weight. Someone calculated that a speaker needs to be 1.6 points ahead of another for a 95 percent confidence it will be preferred these are within that window.

I'd also suggest not being too concerned about personal biases. I've been writing reviews professionally for a good 6 years now, and well before that on forums. One of the first things I learned is that bias does not go away, and that being accurate, fair, and objective is not the same as being unbiased.

I can make my major biases known and be fair in testing, but I can't presume to eliminate bias in anything written. If I try to counteract my bias too much, then I'm just coloring my results another way; over the years I've been quite often accused of being a fanboy of companies I don't like and of being a hater of companies I do like. It's part of the reason I began to do measurements with speakers; regardless of my preferences, the data is objective and revealing enough for anyone to make their own interpretations.

That's not to say I don't think the subjective impressions are useful though; they inform you of a particular listener's preferences and listening conditions and can help you make decisions if you see particular similarities or differences with Amir's setup and preferences. They can also help reveal qualities that may not be obvious in the measurements, which we can then leave for further investigation. But we are primarily here for the measurements, not the panthers The subjective reviews are all out there and you can choose to tie those with the data where you see fit.

Fully agree with you, although for me for subjective impressions to be also useful I would like more care in the listening tests, thus sensible placement and listening distance and placement according to the bass propertiies or EQin the bass/modal region, stereo listening and measurements at the listening room position. For example if I would just plug and play loudspeakers in my current room a loudspeaker with a huge bass boost would win tonally due to my unfortunate room behaviour, but this wouldn't really help anyone unless they has a room with very similar acoustic properties which is very unlikely.

 

[QMuse]

My level matching was on purpose done like that to better show the main tonal differences.

But level matching is not about showing differences, it is about, well, matching levels.

IMO you either match the levels or you shift one curve down for say 5dB so you can observe the differences when they run in parallel.

 

[thewas]

But level matching is not about showing differences, it is about, well, matching levels.

IMO you either match the levels or you shift one curve down for say 5dB so you can observe the differences when they run in parallel.

No, in this case it is made as said to show the tonal differences which is optically much more easy if they have large regions where they coincide. Also about tonality it is said that our hearing is perceiving it in relation to the root tones region, that's another reason I did it the way I did. Our brain doesn't base that it on a computed average level which could much different due to for example a large peak in the lower bass.

 

[BYRTT]

What does this mean? The dispersion doesn't match? Or they messed up on the crossover?

Probaly they wired a single poweramp to coxial terminals and set switches there on the back to biwiring, means exactly it does not quite integrate the additional woofer with the coaxial

 

[maty]

I don't disagree with you about industry standards. My crusade has always been against the validity of the preference score as a way to determine listener preference.

The speaker the R3 lost the blind test to has good spins and overall, a neutral response. If you read the details of the blind test, the differences weren't in coloration of the response, but clarity, realism, and the harshness in the R3's.

I also had that problem with my Q100s for years until I started modifying them. The 5.25" coaxial driver is very good but KEF has made some mistakes in its implementation, most likely due to the policy of saving costs on everything that is not seen.

 

[QMuse]

No, in this case it is made as said to show the tonal differences which is optically much more easy if they have large regions where they coincide. Also about tonality it is said that our hearing is perceiving it in relation to the root tones region, that's another reason I did it the way I did. Our brain doesn't base that it on a computed average level which could much different due to for example a large peak in the lower bass.

It all depends what you want to show. Take a look at this comparison. Obviously both speakers aimed for flat PIR in the 1kHz-10khz so I aligned them that way (let's ignore M16 hump in the 1200-1800Hz range).

From that comaprison it turned out that M16 makes slightly less nergy in the entire sub 1kHz region and that is also rolls-off more steeply after 10kHz.

 

[Johan Liebert]

I wonder how much directivity and dispersion account for individual preference? The horizontal plot for the M16 is quite a bit wider than the R3 along with the warmer FR. And I think in large rooms directivity can have a pretty big influence.

And in turn I'm also curious how well Kef R300 measures against the R3.

I would also like an indication of how well the speakers reviewed so far work near field like on a desk?

Has anyone offered to send you one of their Buchardt S400's for you to review? As of now I mainly have to rely on napilopez review which I found very interesting.

What other speakers do you have lined up to review in the near future?