Kef R3 Bookshelf Speaker Review (Erin's Audio Corner)

[thewas]

Yeah, fair enough. I know the LS50 really don't do well when asked to play loud, I have heard them lol. To an extent that's why I think much of this is moot when regarding that particular speaker.

Was it the original or Meta? I have both and use the original ones (with EQ) nowadays for my desktop system and the Meta (plus sub) for my main listening system (2 meters listening distance) as they have lower distortions (and thus play cleaner at higher levels) due to the several improvements of their drivers.

Also would like to remind of the nice test and binaural recording comparison of @mitchco of the original LS50 vs the huge JBL 4722 which was done also at 83 dB reference level

KEF LS50 (David) Versus JBL 4722 Cinema (Goliath) Speaker Comparison with Binaural Recordings

Here are two extreme opposite loudspeakers when it comes to sensitivity, size, directivity, topology and looks. Does the big speaker sound perceptually larger than the smaller speaker? Do the speakers sound the same if eq’d the same and level matched? Does the LS50 sound perceptually bigger when ...

audiophilestyle.com

I do find it interesting that 1-2khz showed 40dB of IMD in this test. That is significantly more distortion than the THD in Amir's test. If that correlation holds at higher levels, it does imply that the IMD may become audible before the THD. Especially since the THD in that region is primarily 2nd harmonic.

Possibly, but as you can see similar distortion in that region also is measured at the non-coaxial B&W 707 S2 and B&Ws are rather known for low distortion drivers.

 

[Sancus]

Was it the original or Meta? I have both and use the original ones (with EQ) nowadays for my desktop system and the Meta (plus sub) for my main listening system (2 meters listening distance) as they have lower distortions (and thus play cleaner at higher levels) due to the several improvements of their drivers.

It was the original, I haven't heard the Meta yet.

Possibly, but as you can see similar distortion in that region also is measured at the non-coaxial B&W 707 S2 and B&Ws are rather known for low distortion drivers.

It would be a better comparison if both were actually at 86dB. Given nonlinearity, even 4dB could make a substantial difference to these small 5" drivers. But yeah, you might be right. This was discussed in the LS50 Meta thread and the simulation here implies that this should be a much bigger issue at 6+ khz, but I don't see anything there, so I dunno. I don't know that these tests would capture the psychoacoustic impact of the directivity change either. But who even knows how much IMD is audible? That's not a question we have an answer to either.

I really have to wonder why reviewers do this, surely if you wanted the IMD to actually be comparable from speaker to speaker, and you KNOW distortion is nonlinear, you would use at least use a defined set of levels, like 86dB for really small speakers, 86/96 for mid-size, and 96/106 for very large. Or just anything that is actually comparable...

 

[thewas]

It would be a better comparison if both were actually at 86dB. Given nonlinearity, even 4dB could make a substantial difference to these small 5" drivers. But yeah, you might be right. This was discussed in the LS50 Meta thread and the simulation here implies that this should be a much bigger issue at 6+ khz, but I don't see anything there, so I dunno. I don't know that these tests would capture the psychoacoustic impact of the directivity change either. But who even knows how much IMD is audible? That's not a question we have an answer to either.

Yes, unfortunately so many unanswered questions about audibility and preference and it seems Harman research nowadays only cares about headphone targets...

I really have to wonder why reviewers do this, surely if you wanted the IMD to actually be comparable from speaker to speaker, and you KNOW distortion is nonlinear, you would use at least use a defined set of levels, like 86dB for really small speakers, 86/96 for mid-size, and 96/106 for very large. Or just anything that is actually comparable...

I agree, if I find time and motivation in the next weeks I might do some comparable multitone measurements of my Meta vs my dusty JBL LSR 305 at various levels.

 

[nc535]

I appreciate the posting of the IMD test. That kind of test gives a binary answer; its not diagnostic. You look at the results and get a qualitative feel for whether or not there is a problem.

IMD produces sum and difference frequencies. When you have tone spaced evenly across the spectrum you don't have any idea which pair of tone produced any particular cross product. A 1 khz IMD product could be produced by a 100 Hz tone beating with a 1100 Hz tone or by a 10 khz tone beating with 11 khz.

For diagnosis, we'd need a 2-tone test. For doppler, one tone would be at the peak excursion frequency of the mid cone- just above box crossover - and the other swept (manually) higher to see where IMD peaked. If the IMD peaked at or near the mid low pass frequency, I would suspect conventional IMD. If it peaked above it, I would suspect waveguide doppler IMD, to coin a phrase.

Having done waveguide simulations using the ATH tool in the DIYaudio forum and considering dimensions relative to wavelength, I find it hard to believe that moving the waveguide a mm or so back and forth would have an audible effect; nothing like what we saw with the trim ring displaced.

My suspicion is that if the KEF coax has an IMD problem at volume its due to the size of the mid cone, not to its being coaxial. What does one do when faced with the need to play louder with a small mid? Go 3-way as KEF did with the R3 to keep the excursion of the mid cone to a minimum.

 

[Sancus]

For diagnosis, we'd need a 2-tone test. For doppler, one tone would be at the peak excursion frequency of the mid cone- just above box crossover - and the other swept (manually) higher to see where IMD peaked. If the IMD peaked at or near the mid low pass frequency, I would suspect conventional IMD. If it peaked above it, I would suspect waveguide doppler IMD, to coin a phrase.

Having done waveguide simulations using the ATH tool in the DIYaudio forum and considering dimensions relative to wavelength, I find it hard to believe that moving the waveguide a mm or so back and forth would have an audible effect; nothing like what we saw with the trim ring displaced.

My suspicion is that if the KEF coax has an IMD problem at volume its due to the size of the mid cone, not to its being coaxial. What does one do when faced with the need to play louder with a small mid? Go 3-way as KEF did with the R3 to keep the excursion of the mid cone to a minimum.

Yeah, so I believe the assumption is the mid on the LS50 does need to move more than 1mm, the simulation post I linked above suggested it needs at least 3mm. A pure midrange doesn't, of course, but it's not a pure midrange.

Something else of note is that someone pointed out that the LS50 wave guide is actually partially static, as the center piece of the guide is larger than the tweeter and extends slightly above it. So that would mitigate this effect for that particular speaker to some extent.

The trade-off of doing that of course is that you cut a larger section from your woofer which reduces its capability. And indeed, the LS50 Meta does have more harmonic distortion in low - upper bass range than the Genelec 8030C or even the Adam T5V especially 86dB. Of course, it does still beat speakers with pretty bad distortion problems like the JBl 306p.

It's possible that Kef has engineered this trade-off of reducing woofer capacity for wave guide stability, so that the doppler distortion and the general harmonic distortion become problematic around the same time. Would be a pretty smart thing to do, though still leaves you with a speaker that is overall trading off output/distortion for coaxial directivity. Of course, going 3-way indeed sidesteps this problem.

 

[nc535]

Yeah, so I believe the assumption is the mid on the LS50 does need to move more than 1mm, the simulation post I linked above suggested it needs at least 3mm. A pure midrange doesn't, of course, but it's not a pure midrange.

Something else of note is that someone pointed out that the LS50 wave guide is actually partially static, as the center piece of the guide is larger than the tweeter and extends slightly above it. So that would mitigate this effect for that particular speaker to some extent.

The trade-off of doing that of course is that you cut a larger section from your woofer which reduces its capability. And indeed, the LS50 Meta does have more harmonic distortion in low - upper bass range than the Genelec 8030C or even the Adam T5V especially 86dB. Of course, it does still beat speakers with pretty bad distortion problems like the JBl 306p.

It's possible that Kef has engineered this trade-off of reducing woofer capacity for wave guide stability, so that the doppler distortion and the general harmonic distortion become problematic around the same time. Would be a pretty smart thing to do, though still leaves you with a speaker that is overall trading off output/distortion for coaxial directivity. Of course, going 3-way indeed sidesteps this problem.

I did my own simulation to see how much the cone needs to move at 86 db SPL. First, I measured outer diameter as 4" and inner dia as 1.8" for an Sd of 65 cm2. Then I went to a sealed driver simulation and fudged the driver Sd to 65, set the output SPL to 86 db., did a little EQ to flatten and looked at the excursion vs frequency. This showed me .1 mm at 400 Hz, which is crossover in the R3 and 1 mm at 100 Hz. I then plugged those numbers into Paul Klipsch's formula for IM distortion due to doppler. D(%) = .033 * ExcursionInInches * HighestFrequency. If using the coax in a 2-way, the cabinet would be vented and I would expect the vent would start taking over below 100 Hz. Quite the difference between a 2-way and a 3-way!

D(%) = .033A1(")F2
Excursion (mm)	High F	D(%)	comment
0.1	2900	0.377	400 Hz high pass
1	2900	3.768	100 Hz high pass

 

[nc535]

sharpened my eye reading the sim and added excursion at 450 Hz lower limit as used on Blade and Reference series. That higher cutoff reduces the calculated IMD by about 20%.

D(%) = .033X(")F2
Excursion (mm)	F2	D(%)	Lowest freq
0.07	2900	0.264	400 Hz R3
0.057	2900	0.215	450 Hz Blade&Ref
1	2900	3.768	100 Hz
trust chart even less below 100 Hz

 

[sigbergaudio]

An additional point may be the audibility of the distortion levels we're talking about at high volumes. Our coaxial speakers are designed to play pretty loud. They're 2.5-way designs, which means the coax is contributing down to the f3 point at 90hz. Xmax for the coax is +/- 4mm. At that point we will obviously experience the doppler effect. But at that point you're at 110dB@1m from one speaker. At a more reasonable 105dB@1m (so ~110dB from two speakers), the excursion is less than 1mm from 200hz and up.

I'd be really interested to see someone detect the already pretty moderate IMD distortion at those sound levels, not to mention normal listening levels.

Point being, within the sound levels the speakers are designed for (goes for the R3, LS50 etc as well), I doubt this is practical issue, and the benefit of a coax will far outweigh the theoretical problem of the doppler effect.

 

[nc535]

Since you have the coax contributing all the way down, you apparently didn't find a distortion issue of any kind to motivate you to limit its range and that buys some extra output capability. That is good to know. I know from online THD listening tests, I can't hear harmonic distortion until its several percent. I don't know of similar tests for IMD. If one exists, it would certainly put this question to rest.

The physics of the doppler is undeniable. It may or may not be audible in a set of tones. Even if so, it could well be masked by the music. Its certainly not an issue at the moderate levels I enjoy.

 

[sigbergaudio]

Since you have the coax contributing all the way down, you apparently didn't find a distortion issue of any kind to motivate you to limit its range and that buys some extra output capability. That is good to know. I know from online THD listening tests, I can't hear harmonic distortion until its several percent. I don't know of similar tests for IMD. If one exists, it would certainly put this question to rest.

The physics of the doppler is undeniable. It may or may not be audible in a set of tones. Even if so, it could well be masked by the music. Its certainly not an issue at the moderate levels I enjoy.

There's two ways to see this. I could worry about IMD and do a three way. Or by having the coax contribute all the way down, the required excursion is cut in half on the second driver. So in practice this solution REDUCES overall distortion of the entire system compared to a 3-way solution, and improves capacity and dynamics in the midbass area. It's all about choosing the best compromises for overall performance.

 

[Sancus]

There's two ways to see this. I could worry about IMD and do a three way. Or by having the coax contribute all the way down, the required excursion is cut in half on the second driver. So in practice this solution REDUCES overall distortion of the entire system compared to a 3-way solution, and improves capacity and dynamics in the midbass area. It's all about choosing the best compromises for overall performance.

Considering your systems are explicitly designed to be used with a subwoofer and have a frequency response that reflects that, I'm sure that helps to prevent the issue as well.

 

[aarons915]

Since you have the coax contributing all the way down, you apparently didn't find a distortion issue of any kind to motivate you to limit its range and that buys some extra output capability. That is good to know. I know from online THD listening tests, I can't hear harmonic distortion until its several percent. I don't know of similar tests for IMD. If one exists, it would certainly put this question to rest.

The physics of the doppler is undeniable. It may or may not be audible in a set of tones. Even if so, it could well be masked by the music. Its certainly not an issue at the moderate levels I enjoy.

I think a 2.5 way is possibly the best design compromise in coaxial speakers as well. My problem with 3 ways is it takes a bit of that point source magic away compared to a 2-way. I have no problem with IMD or doppler distortion in the LS50's with a 100Hz 2nd order high pass on them and I've said that I prefer it to my previous setup of the R3. I would like to see KEF use the 2 way coaxials in their 3 ways and cross them over around 150Hz to retain the pure point source quality while being able to play loud cleanly but this is very close to a 2.5 way.

 

[Head_Unit]

Do coaxial manufacturers even use 3-way designs noticeably more than other people?

No but that's due to other factors I'd say. In automotive, due to space, 3-ways are pretty useless, even in a 6x9, because you can't really extend the low frequency the "midrange" is covering. They are really still just tweeters. And a 3-way home speaker will automatically be at a higher price than a 2-way, so any makers wanting to make more budget priced stuff can't start with 3-way be they coaxial or not.

Coaxial design is just not easy. For the bridge type like most automotive, the bridge interferes with sound from the woofer, and the tweeter out in front of the woofer causes cancellation. And whenever I was doing those designs there was almost always a phase rotation whereby a simple capacitor crossover would give an in-out-in phase characteristic, giving peaks and dips you could move around but not eliminate. Sometimes a 2nd order crossover could help, or something more complex which is generally out of budget for all but the most expensive models (which in automotive then tend to be separates).

 

[theyellowspecial]

Yeah, fair enough. I know the LS50 really don't do well when asked to play loud, I have heard them lol. To an extent that's why I think much of this is moot when regarding that particular speaker.

I had them for a couple weeks. In my 20x11 room they gave up a few times watching blockbuster action movies. Sounded like the drivers were being torn apart. Also, just generally not clear in louder scenes. This was with a 100Hz crossover, too. I don't recall any problems with music, but I didn't test with anything with high dynamic range like orchestra, etc. After their near death experience with movies I sold them on.

 

[thewas]

Coaxial design is just not easy. For the bridge type like most automotive, the bridge interferes with sound from the woofer, and the tweeter out in front of the woofer causes cancellation. And whenever I was doing those designs there was almost always a phase rotation whereby a simple capacitor crossover would give an in-out-in phase characteristic, giving peaks and dips you could move around but not eliminate. Sometimes a 2nd order crossover could help, or something more complex which is generally out of budget for all but the most expensive models (which in automotive then tend to be separates).

Yes, with some effort such "bridge type" can be done also quite well as ME Geithain shows with their really good studio monitors.

 

[nc535]

Yes, with some effort such "bridge type" can be done also quite well as ME Geithain shows with their really good studio monitors.

Yes, with some effort such "bridge type" can be done also quite well as ME Geithain shows with their really good studio monitors.

well. I won't say I'm not impressed with the ME Geithain line but their coax, while clever and apparently well done, is not quite a point source.

 

[thewas]

well. I won't say I'm not impressed with the ME Geithain line but their coax, while clever and apparently well done, is not quite a point source.

Most people who have heard them say the opposite though, also being a point source doesn't necessarily mean that all drivers are centred but that their acoustic output doesn't create lobes on larger angles, which is achieved usually by the output distances being smaller than the radiated wavelengths at the crossover frequencies, see exemplary also https://www.soundandrecording.de/eq...ialer-hochleistungs-midfield-monitor-im-test/ , Genelec Ones, KEF 3-ways, Danlye horns etc.

 

[nc535]

That is what I meant by "not quite". I've DIYed a synergy horn. From that background, I would impose a 1/4 wavelength center to center distance for point source summation. I've done the trig to know that you can get away with greater distance if you limit the listening window width.

 

[thewas]

That is what I meant by "not quite". I've DIYed a synergy horn. From that background, I would impose a 1/4 wavelength center to center distance for point source summation. I've done the trig to know that you can get away with greater distance if you limit the listening window width.

Yes, some people call those designs also "pseudo-coaxials", I also usually try to stay under a one fourth wavelength center to center.

 

[Streamc]

View attachment 138341
50Hz for Prodigy - Breathe.

Made analysis by Audacity. It have even sub 40Hz like any other music. I think it would be subwoofer work.