KEF R3 Meta - Review & Measurements by Erin

[jackocleebrown]

Hi Guys,

Driver suspension break-in is a well known effect. For example, please see this paper from Klippel on the topic. https://www.aes.org/e-lib/browse.cfm?elib=16000

You can also find some slides from the paper here: https://www.klippel.de/fileadmin/kl...tigue_and_Aging_of_suspension_AES_NY_2011.pdf

Very typically our driver suspension stiffness directly from the production line is around 20-30% higher than the final target stiffness value. You can see from the Klippel paper that it typically requires a several hours of running at moderate levels to get to the final target value*. A good suspension will settle to a consistent final value and won't continue to fatigue. The frequency response variation depends on the exact loudspeaker but it can often be as much as 1dB, plus the pre-run-in response often tends to peak up around port tuning.

I'm travelling currently but here's some data from a lumped model using the Reference 1 for demonstration.

Kind regards,
Jack.

*as you can see from the Klippel paper, you can run speakers in very fast using high level signals but I'd never recommend that customers do it that way as there's significant potential to damage the drivers.

 

[Descartes]

Hi Guys,

Driver suspension break-in is a well known effect. For example, please see this paper from Klippel on the topic. https://www.aes.org/e-lib/browse.cfm?elib=16000

You can also find some slides from the paper here: https://www.klippel.de/fileadmin/kl...tigue_and_Aging_of_suspension_AES_NY_2011.pdf

Very typically our driver suspension stiffness directly from the production line is around 20-30% higher than the final target stiffness value. You can see from the Klippel paper that it typically requires a several hours of running at moderate levels to get to the final target value*. A good suspension will settle to a consistent final value and won't continue to fatigue. The frequency response variation depends on the exact loudspeaker but it can often be as much as 1dB, plus the pre-run-in response often tends to peak up around port tuning.

I'm travelling currently but here's some data from a lumped model using the Reference 1 for demonstration.

Kind regards,
Jack.

*as you can see from the Klippel paper, you can run speakers in very fast using high level signals but I'd never recommend that customers do it that way as there's significant potential to damage the drivers.

View attachment 333546

Thank you for all the information greatly appreciated!
Safe travels!

 

[Mnyb]

I am curious where have you read that? Have you gone scientific measurements to demonstrate that speakers need breaking time? If I remember correctly either Amir of Erin have shown that speakers measure the same out of the box and after hours of listening!

There is a difference between the complete systems behaviour and the drivers , even if the drivers break in somewhat ( not the hundreds of hours dreamed up by some ) the net effect on the complete system migth not be that audible in room ?
That’s my take away ?

 

[thewas]

Hi Guys,

Driver suspension break-in is a well known effect. For example, please see this paper from Klippel on the topic. https://www.aes.org/e-lib/browse.cfm?elib=16000

You can also find some slides from the paper here: https://www.klippel.de/fileadmin/kl...tigue_and_Aging_of_suspension_AES_NY_2011.pdf

Very typically our driver suspension stiffness directly from the production line is around 20-30% higher than the final target stiffness value. You can see from the Klippel paper that it typically requires a several hours of running at moderate levels to get to the final target value*. A good suspension will settle to a consistent final value and won't continue to fatigue. The frequency response variation depends on the exact loudspeaker but it can often be as much as 1dB, plus the pre-run-in response often tends to peak up around port tuning.

I'm travelling currently but here's some data from a lumped model using the Reference 1 for demonstration.

Kind regards,
Jack.

*as you can see from the Klippel paper, you can run speakers in very fast using high level signals but I'd never recommend that customers do it that way as there's significant potential to damage the drivers.

View attachment 333546

Thank you very much, I was sure there would be some well-founded objective facts before someone like you would make such a statement.

 

[HarmonicTHD]

Hi Guys,

Driver suspension break-in is a well known effect. For example, please see this paper from Klippel on the topic. https://www.aes.org/e-lib/browse.cfm?elib=16000

You can also find some slides from the paper here: https://www.klippel.de/fileadmin/kl...tigue_and_Aging_of_suspension_AES_NY_2011.pdf

Very typically our driver suspension stiffness directly from the production line is around 20-30% higher than the final target stiffness value. You can see from the Klippel paper that it typically requires a several hours of running at moderate levels to get to the final target value*. A good suspension will settle to a consistent final value and won't continue to fatigue. The frequency response variation depends on the exact loudspeaker but it can often be as much as 1dB, plus the pre-run-in response often tends to peak up around port tuning.

I'm travelling currently but here's some data from a lumped model using the Reference 1 for demonstration.

Kind regards,
Jack.

*as you can see from the Klippel paper, you can run speakers in very fast using high level signals but I'd never recommend that customers do it that way as there's significant potential to damage the drivers.

View attachment 333546

Thanks for responding and providing some data. (Maybe update the previously quoted web article with appropriate links as well to avoid confusion).

Unfortunately I am not an AES member and can’t access the paper so I might be missing essential information. The slides however don’t provide information to how material fatigue affects audible frequency response change. The slides show that yes fatigue (stiffness over time change) exists, but that we know since ages (Woehler) and the slides even show that environmental factors seem to have higher effects than fatigue (page 20) (btw one can also see that in Amir’s initial testing of the Neumann KH80). I am also missing in the slides if the “load” (work) leading to the stiffness changes corresponds to “normal” speaker operation or if the load is artificially high to prove the concept.

Therefore it would be great if you could expand and elaborate a bit more on the graph of the Reference 1 you posted once back from your travels. Is the graph an actual measurement or a model assuming 30% in stiffness reduction? Where does the 30% come from? Is that a realistic value assuming “normal” operation of a speaker? Or does the 30% only occur with artificially high loads as shown in the Klippel slides?

And even if all that is the case, I am still a bit skeptical if the change shown in the graph (Reference 1) lead to such audible effects which your company describes in the web article Quote: “After a period of time, you'll notice the speakers open up and begin to sound more natural.” And if those differences are recognized in an ABX test?

Thanks very much appreciate it. Safe travels. Again I would hate my nice KEF speaker to deteriorate ;-)

 

[sweetchaos]

My thought is…

Not related to burn in, per se, but to speaker longevity.

How to ensure speaker longevity for Kef speakers?

For example, should you turn on speakers and let them play music, every couple of weeks, months, etc?
Temperature control, moisture control?

I am considering Kef R meta passive speakers, so would like to know the secret to ensuring a long lifespan.

Any guidance would help.

Thanks.

 

[HarmonicTHD]

My thought is…

Not related to burn in, per se, but to speaker longevity.

How to ensure speaker longevity for Kef speakers?

For example, should you turn on speakers and let them play music, every couple of weeks, months, etc?
Temperature control, moisture control?

I am considering Kef R meta passive speakers, so would like to know the secret to ensuring a long lifespan.

Any guidance would help.

Thanks.

My hypothesis is that for example temp changes in your room effects the stiffness much more yet nobody notices it. Ergo it is irrelevant with regard to audibility (as is material fatigue )Of course it would need proper ABX testing to prove or disprove that hypothesis.

Edit. Typo.

 

[MAB]

You can also find some slides from the paper here: https://www.klippel.de/fileadmin/kl...tigue_and_Aging_of_suspension_AES_NY_2011.pdf

This paper describes 'wear-out', not 'break-in'. The paper's focus is long term reliability, and accelerated life testing of components to fail. Weibull's analysis. Not HiFi break-in where the speakers start to sound dramatically better after the first few hours or days. I couldn't read the other one. I'm eager to see data where a driver changes parameters appreciably where the sound

My hypothesis is that for example temp changes in your room effects the stiffness much more yet nobody notices it. Ergo it is irrelevant with regard to audibility (as is material fatigue )Of course it would need proper ABC test to prove or disprove that hypothesis.

It's just like you say. Speakers are sensitive to temperature more sensitive to temperature than break-in.

Do Audio Speakers Break-in?

I have a brand new woofer that I got as a replacement, the original was defective. The replacement driver has been sitting around un-played. Seems like it is time to see how it changes with 'break-in'. I got a DATS recently, so I'll use that. I did check it against an HP LCR with a bias...

www.audiosciencereview.com

This is a pretty modest change in parameters, I am sure I can't hear, would be interesting to see if measurable in a box.
While I agree drivers change in the first few moments of operation, it is just so tiny.

 

[jackocleebrown]

Unfortunately I am not an AES member and can’t access the paper so I might be missing essential information. The slides however don’t provide information to how material fatigue affects audible frequency response change. The slides show that yes fatigue (stiffness over time change) exists, but that we know since ages (Woehler) and the slides even show that environmental factors seem to have higher effects than fatigue (page 20) (btw one can also see that in Amir’s initial testing of the Neumann KH80). I am also missing in the slides if the “load” (work) leading to the stiffness changes corresponds to “normal” speaker operation or if the load is artificially high to prove the concept.

It's unfortunate that I can't post the Klippel paper. Anyhow, in the slides you can clearly see the "break-in" period marked on the graphs and distinct from the fatigue behaviour. In the paper Klippel defines this non-reversible change as "Initial exposure to mechanical load opens some bonded joints in the impregnated fiber structure (break-in effect).". This drop in driver stiffness is very well known. Kippel even has metrics to define how much work needs to be applied to a particular suspension to pass through the break-in phase. Klippel suggests that a accumulated work model is valid for the break-in period so this indicates that break-in would occur at any playback level. The timescale is fast compared to fatigue but it's still a significant number of hours playback time at moderate normal levels. For instance, this graph from Kippel shows a 30-40% drop in stiffness occurring within the first day of continuous playback, depending on the playback level.

Therefore it would be great if you could expand and elaborate a bit more on the graph of the Reference 1 you posted once back from your travels. Is the graph an actual measurement or a model assuming 30% in stiffness reduction? Where does the 30% come from? Is that a realistic value assuming “normal” operation of a speaker? Or does the 30% only occur with artificially high loads as shown in the Klippel slides?

The data I provided is a simulation with the stiffness of the suspension changed by 30%. This is realistic. The +30% would represent a factory fresh suspension and eventually we would expect it to reach a steady value. The time taken depends on how loud the speaker is played back. Fatigue is not a desirable characteristic in a suspension because it eventually leads to failure, so we try and avoid using suspensions that show fatigue and hence we would expect our suspensions to eventually arrive at a steady target stiffness for all users.

And even if all that is the case, I am still a bit skeptical if the change shown in the graph (Reference 1) lead to such audible effects which your company describes in the web article Quote: “After a period of time, you'll notice the speakers open up and begin to sound more natural.” And if those differences are recognized in an ABX test?

ASR forum readers can happily understand the technical details, but I'm sure that you would appreciate that we have a lot of customers who will not and hence we have literature aimed at different technical understanding levels. Here's the difference curve from the two SPL plots I provided. Why not dial it into Roon/Mini-DSP and see if you hear a difference.

Thanks very much appreciate it. Safe travels. Again I would hate my nice KEF speaker to deteriorate ;-)

No fear, they will not deteriorate. We have happy customers still using speakers from the 1970s and 1980s.

 

[HarmonicTHD]

It's unfortunate that I can't post the Klippel paper. Anyhow, in the slides you can clearly see the "break-in" period marked on the graphs and distinct from the fatigue behaviour. In the paper Klippel defines this non-reversible change as "Initial exposure to mechanical load opens some bonded joints in the impregnated fiber structure (break-in effect).". This drop in driver stiffness is very well known. Kippel even has metrics to define how much work needs to be applied to a particular suspension to pass through the break-in phase. Klippel suggests that a accumulated work model is valid for the break-in period so this indicates that break-in would occur at any playback level. The timescale is fast compared to fatigue but it's still a significant number of hours playback time at moderate normal levels. For instance, this graph from Kippel shows a 30-40% drop in stiffness occurring within the first day of continuous playback, depending on the playback level.
View attachment 333574


The data I provided is a simulation with the stiffness of the suspension changed by 30%. This is realistic. The +30% would represent a factory fresh suspension and eventually we would expect it to reach a steady value. The time taken depends on how loud the speaker is played back. Fatigue is not a desirable characteristic in a suspension because it eventually leads to failure, so we try and avoid using suspensions that show fatigue and hence we would expect our suspensions to eventually arrive at a steady target stiffness for all users.


ASR forum readers can happily understand the technical details, but I'm sure that you would appreciate that we have a lot of customers who will not and hence we have literature aimed at different technical understanding levels. Here's the difference curve from the two SPL plots I provided. Why not dial it into Roon/Mini-DSP and see if you hear a difference.

View attachment 333579



No fear, they will not deteriorate. We have happy customers still using speakers from the 1970s and 1980s.

Hi.

Thank you so much for responding in detail.

I never doubted that change in stiffness of polymer materials over time under load is not a thing. It is well researched and understood as it is much more relevant in safety related applications (eg just think of polymer o-rings in braking systems or airfoil actuators etc). And luckily it converges to some kind of steady state.
What was not clear from the slides was, if the amount of load (work) which needs to be applied to observe the effect is realistic for the “normal” speaker operation. But you clarified this and I take your word for it.

As for the audibility, yes the suggested test using DSP and that ABX test software would provide some clarity. I still doubt that it is audible with music (yes for sine tones) especially as Fletcher-Munson “kicks in” at these low frequencies. But only a test would tell.
Plus the wording of the web article “… the speaker opens up” is greatly exaggerated.

Edit. Typo. Braking

 

[Music707]

Unfortunately I am not an AES member and can’t access the paper

Try this one:

https://www.klippel.de/fileadmin/klippel/Bilder/Know-How/Literature/Papers/Aging%20of%20loudspeaker%20suspension_Klippel.pdf

 

[HarmonicTHD]

Try this one:

https://www.klippel.de/fileadmin/klippel/Bilder/Know-How/Literature/Papers/Aging%20of%20loudspeaker%20suspension_Klippel.pdf

Thank you. Appreciate it. Let me have a look later tonight.

 

[thewas]

Plus the wording of the web article “… the speaker opens up” is greatly exaggerated.

Would be interesting to see such curves also for a brand new mid driver and tweeter where such (if at similar levels) would be more audible and sure wording would make more sense.

 

[HarmonicTHD]

Would be interesting to see such curves also for a brand new mid driver and tweeter where such (if at similar levels) would be more audible and sure wording would make more sense.

I agree. Further data would be nice.

However my interpretation of the Klippel Slides (I still haven’t read the paper) and my engineering gut feeling tells me that the relaxation effect (reduced stiffness) is most pronounced the more work (force x displacement) is being put into the polymer suspension.

The force is biggest at lower frequencies and the displacement is largest at the lower frequencies and additionally at the eigen-modes of the polymer suspension.
This would explain also the behavior of the KEF Reference series woofer linked above by @jackocleebrown . However I might be completely wrong, so maybe @jackocleebrown could comment or ideally provide some data of their famous mid/tweeter coaxial?

 

[MAB]

For instance, this graph from Kippel shows a 30-40% drop in stiffness occurring within the first day of continuous playback, depending on the playback level.

I am quite frankly surprised by this graph. I will admit that I don't regularly measure driver changes during initial operation. But I do measure it from time to time, and I considered it to be minimal based on measurements. For instance, a new Seas W18 midwoofer I have many samples of seems to change appreciably during initial use, more so than some paper cone drivers. So I did a fairly detailed measurement to try to determine temperature and other environmental conditions. My stress is way different than the Klippel stress conditions you reference, I drove the woofer at 90% of max excursion for 30 mins and then let the driver cool down overnight (one of the points of my rambling study is to see the impact of temperature!)

After stress, I observe on a new and unused Seas W18 woofer after application of stress:

• Driver fs drops 1.4% from 37.0 to 36.5 Hz
• Qms drops 3.8% from 2.66 to 2.56

therefore:

• Cms reduces 5.1%

I am just not seeing dramatic changes here. Agree, if a driver has compliance change by nearly 50% it is going to be not only measurable but likely audible!!! But I see about an order of magnitude lower Cms changes.
I measured this Seas driver because it always seems to show larger changes than other drivers I have. I don't do this type of study often, and I didn't use the same stresses as Klippel. But Klippel does show and comment that the effect has low dependency on power. So it seems my blasting the speaker for 30 mins should do the trick...

If Seas already stressed the driver, then I will miss the effect per Klippel's data. I don't think Seas do this, but would certainly stand corrected if they do.
Maybe I am not stressing the driver as per Klippel and I really need to do 100s of mW for days rather than hundreds of Watts for 30 mins! Or some other methodology issue.
Maybe I am not exposed to drivers that demonstrate significant changes. I just did a spot-check on two new paper cone woofers, the results were in the noise and not worth reporting. I had considered this speaker break-in phenomenon to be a tiny effect. As you can tell, I am struggling to put this Klippel paper and conclusions in context of drivers I buy!

 

[HarmonicTHD]

I am quite frankly surprised by this graph. I will admit that I don't regularly measure driver changes during initial operation. But I do measure it from time to time, and I considered it to be minimal based on measurements. For instance, a new Seas W18 midwoofer I have many samples of seems to change appreciably during initial use, more so than some paper cone drivers. So I did a fairly detailed measurement to try to determine temperature and other environmental conditions. My stress is way different than the Klippel stress conditions you reference, I drove the woofer at 90% of max excursion for 30 mins and then let the driver cool down overnight (one of the points of my rambling study is to see the impact of temperature!)

After stress, I observe on a new and unused Seas W18 woofer after application of stress:

• Driver fs drops 1.4% from 37.0 to 36.5 Hz
• Qms drops 3.8% from 2.66 to 2.56

therefore:

• Cms reduces 5.1%

I am just not seeing dramatic changes here. Agree, if a driver has compliance change by nearly 50% it is going to be not only measurable but likely audible!!! But I see about an order of magnitude lower Cms changes.
I measured this Seas driver because it always seems to show larger changes than other drivers I have. I don't do this type of study often, and I didn't use the same stresses as Klippel. But Klippel does show and comment that the effect has low dependency on power. So it seems my blasting the speaker for 30 mins should do the trick...

If Seas already stressed the driver, then I will miss the effect per Klippel's data. I don't think Seas do this, but would certainly stand corrected if they do.
Maybe I am not stressing the driver as per Klippel and I really need to do 100s of mW for days rather than hundreds of Watts for 30 mins! Or some other methodology issue.
Maybe I am not exposed to drivers that demonstrate significant changes. I just did a spot-check on two new paper cone woofers, the results were in the noise and not worth reporting. I had considered this speaker break-in phenomenon to be a tiny effect. As you can tell, I am struggling to put this Klippel paper and conclusions in context of drivers I buy!

The graph shows the change of the stiffness of the suspension polymer not the audibility.

This one does show the audibility and you see how small it really is

 

[MAB]

The graph shows the change of the stiffness of the suspension polymer not the audibility.

Yes, that is my interpretation too!

This one does show the audibility and you see how small it really is

View attachment 333702

This Ref 1 example is realistic expectation, consistent with my limited experience. I just don't know what to make of the Klippel example, with about 50% change in compliance! Is that just one component of the driver??? I think I am not following or misinterpreting the Klippel example that shows such a large change.

 

[HarmonicTHD]

Yes, that is my interpretation too!

This Ref 1 example is realistic expectation, consistent with my limited experience. I just don't know what to make of the Klippel example, with about 50% change in compliance! Is that just one component of the driver??? I think I am not following or misinterpreting the Klippel example that shows such a large change.

By now I have also read the Klippel paper. The paper introduces a mathematical model which correlates the stiffness change of the polymer suspension (the rubbery thing around the diaphragm eg) to the subjected work (load) and then fits the model to one measurement for a speaker type A and B. The paper does not say how that stiffness reduction manifest itself in actual speaker performance change, let alone audibility of such. KEF @jackocleebrown took the paper and considered the worst case of ca 30% stiffness reduction during „break-in“ and plugged that stiffness reduction into their KEF Ref 1 woofer model which then resulted in the graph above. So why does the reduction in suspension stiffness not translate one to one to speaker performance change? In simple words, the driver controls the movement of the diaphragm and is much stronger than the resistance (stiffness) offered by the suspension.

And here come my questions/comments
A) the paper does not tell anything about the test variation nor how many tests had been conducted nor does it present the raw data of the tests. So it is unclear how repeatable and reproducible the data is.

B) Kef plugged the assumed 30% stiffness reduction into their Ref 1 woofer model, but did not measure if it is indeed a valid assumption for the specific Kef Ref1. Ideally I would like to see ca 5 drivers taken out of storage then run sweeps for maybe a week, record the data and compare the change in FR change over time. Maybe Kef has that data, but I can only comment on what was made transparent until now. @jackocleebrown please correct me if I misinterpreted your previous post.

C) Audiabilty. One could assume some good will and assume that the Kef Ref1 graph is indeed correct. Then construct a filter in a PEQ, run some music through it and ABX the two resulting files with that ABX software. (Personally I think I would be hard pressed to hear a difference, but have to admit that I never tried it). Maybe @pkane or @pma could help out to setup such a test if they find the time and motivation.

 

[MAB]

By now I have also read the Klippel paper. The paper introduces a mathematical model which correlates the stiffness change of the polymer suspension (the rubbery thing around the diaphragm eg) to the subjected work (load) and then fits the model to one measurement for a speaker type A and B. The paper does not say how that stiffness reduction manifest itself in actual speaker performance change, let alone audibility of such. KEF @jackocleebrown took the paper and considered the worst case of ca 30% stiffness reduction during „break-in“ and plugged that stiffness reduction into their KEF Ref 1 woofer model which then resulted in the graph above. So why does the reduction in suspension stiffness not translate one to one to speaker performance change? In simple words, the driver controls the movement of the diaphragm and is much stronger than the resistance (stiffness) offered by the suspension.

And here come my questions/comments
A) the paper does not tell anything about the test variation nor how many tests had been conducted nor does it present the raw data of the tests. So it is unclear how repeatable and reproducible the data is.

B) Kef plugged the assumed 30% stiffness reduction into their Ref 1 woofer model, but did not measure if it is indeed a valid assumption for the specific Kef Ref1. Ideally I would like to see ca 5 drivers taken out of storage then run sweeps for maybe a week, record the data and compare the change in FR change over time. Maybe Kef has that data, but I can only comment on what was made transparent until now. @jackocleebrown please correct me if I misinterpreted your previous post.

Hey, Kef can send me lots of new drivers, I would be happy to test them in my basement. Just saying... OK, I'll stop.

C) Audiabilty. One could assume some good will and assume that the Kef Ref1 graph is indeed correct. Then construct a filter in a PEQ, run some music through it and ABX the two resulting files with that ABX software. (Personally I think I would be hard pressed to hear a difference, but have to admit that I never tried it). Maybe @pkane or @pma could help out to setup such a test if they find the time and motivation.

Thanks.
Agree with your summary!
Even the Ref 1 delta seems small, perhaps less than moving the speaker a bit. I have debated audibility of this in the past, would be happy to see data that changes my view.

 

[Mnyb]

Regardless it shows that KEF do think about and tests these things and don’t just assume stuff