Neumann KH 80 DSP Speaker Measurements: Take Two

[aarons915]

So it is clear, the time and effort that it took to make these two measurements, cost us measuring two new speakers. The work did confirm that setting all measurements to the same axis, i.e. tweeter axis as the system asks for, is the right approach. Randomly moving that reference based on some line in a manual is not wise. It radically changes the response and generates meaningful differences in how one would think the speaker measures.

I was one that wanted to move on to new speakers but hopefully this re-measure shows the measurements can be trusted. I also said all the measurements should be on the tweeter axis as it makes the most sense and it makes all measurements apples to apples comparisons.

 

[Juhazi]

Yes, sorry that I've been unpolite sometimes. It comes from frustration of how difficult it is to get specific answers for the details of measurement protocol. And we know that the devil is in details! Louspeakers are not easy at all to measure and evaluate!

One open question is why 2nd measurement's balloon shows best beam even more downwards from the already too down axis. one reason could be that at 20kHz tweeter's and specially waveguide-tweeter's response is unstable, haphazard directivity because of interferences. But when we look at response plots, they show flatter response from 8kHz up in negative angles too. Very strange, is +/- upside down?

The primal importance of directivity and DI I agree, but bass extension and spl capacity without extensive distortion ( in bass) is very important too! Some say it is 30% of overall preference, but it can easily be more when we discuss small speakers. Distortion profile is important too, despite Amir has many times disagreed.

However Amir himself said about Pioneer:
Informal Listening Tests
My first impression was positive, listening to what was mostly a balanced response. In a few minutes though, I found the speaker bright a bit due to lack of deep bass. I toed the speaker away from my ears and that made a drastic difference, balancing the tonality considerably. Compared to my JBL LSR305P Mark II, the JBL had much better bass but otherwise had similar tonality which is to say the Pioneer was doing well. Let me backtrack that a bit: there is a bit too much mid-range in the pioneer.

On loudness test, deep lows play a bit but then the speaker falls apart drastically, getting super distorted. Techno music with higher frequency bass was handled much better. If you have an option of cutting out the very lows in your playback software, I would do that.

Overall, we are going to good direction gradually, still learning to do and interprete measurements better with Klippel NFS!

 

[BYRTT]

Hi Juhazi, allow me suggest try work on ignore whatever those informal listening tests tell or take as is, and never document them upon each other as long they not performed in same more or less controlled enviroment in a good mood as with most HP amp listening tests, maybe a thing down the road, also think about it amirm seems love clean ear blowing lows in head phone domain and 1000 watts or more in private audio system so why should he love or sound positive for these toy woofers and amps. Sweep in original review and 2nd sweep today is two different samples from two different owners, you can have technical qustions yes it doesn't change its the most smooth and impressive curves so far and probably many behind facade is working fast to get that low end knee exposed and solved one or the other way, sleep well up north

 

[ctrl]

I also said all the measurements should be on the tweeter axis as it makes the most sense and it makes all measurements apples to apples comparisons.

This statement is unfortunately, as has often been said here, completely wrong.

It is decisive on which reference point the loudspeaker was developed. The further apart the acoustic centres of the drivers are from each other and the higher the crossover frequency, the more decisive is the reference point towards which the loudspeaker was developed.

Perhaps a simple example will help to understand this better.
Suppose we have a two-way loudspeaker with an 8'' woofer and a 1'' tweeter. The cabinet is 0.25m wide and 0.40m high - a "big" bookshelf speaker.
The acoustic centres of the drivers are 0.19m apart vertically.

The loudspeaker was developed to a reference point between tweeter and woofer.

First of all we want to have a look at the half-space simulation of a crossover with LR4@2000Hz, "measured" with the reference point between tweeter and woofer as intended by the developer.
Then "measured" at the height of the tweeter as reference point.

In the second case we change the crossover a bit and separate higher, but with flatter filters - LR2@3000Hz. Again in the same order as above.

I don't think anyone would think of calling these differences irrelevant.

Anyone who had to choose between two speakers would always take the one with the "better" measurements.

Depending on the loudspeaker (distance between acoustic centers and crossover frequency) it sometimes makes a huge difference.

UPDATE: Replaced the simulations with better, more accurate ones.

 

[aarons915]

I don't think anyone would think of calling these differences irrelevant.

Anyone who had to choose between two speakers would always take the one with the "better" measurements.

Depending on the loudspeaker (distance between acoustic centers and crossover frequency) it sometimes makes a huge difference.

I agree the differences in your simulation are significant but these 2 measurements of the KH80 show the "reference axis" actually being worse, so it doesn't always matter what a manufacturer specifies as their reference axis. Also, how many people are this anal about making sure their speakers are within inches of the "reference axis"? I would wager not very many, the point is we should be measuring speakers to be able to make comparisons between them, not just what creates the best graph. Most of the time that is going to be on the tweeter axis just due to the finicky nature of high frequencies as was shown in the re-measure.

 

[Pio2001]

I set the reference to half-way between the two drivers which landed it around the edge of the woofer as they show.

I also said all the measurements should be on the tweeter axis as it makes the most sense and it makes all measurements apples to apples comparisons.

This statement is unfortunately, as has often been said here, completely wrong.

It is decisive on which reference point the loudspeaker was developed.

There seem to be an awful misunderstanding from the very beginning of this affair.

Amirm and Aaron915 are talking about Klippel's tweeter point, which should be the origin of the holographic sound field.
Ctrl is talking about the reference axis of the speaker.

These two concepts have absolutely nothing in common.
What people were asking for was the frequency response on the reference axis given by the manufacturer, which was present from the beginning in the original review, among the grey curves of the diagram "SPL Vertical" on the right hand side (the one that represents the frequency response on-axis mid-way between the tweeter and the woofer) :

View attachment 46814

Klippel's tweeter point has nothing to do with it and must remain in front of the tweeter.

 

[ctrl]

... but these 2 measurements of the KH80 show the "reference axis" actually being worse, so it doesn't always matter what a manufacturer specifies as their reference axis.

Because something went wrong with the second measurement.

As Amir explained, it is advantageous to start the Klippel NFS measurements at tweeter level. Afterwards, however, it should be possible to move the coordinate system to the manufacturer's reference point for evaluation.

Also, how many people are this anal about making sure their speakers are within inches of the "reference axis"?

It is about fairness in the evaluation of the measurement results.
Manufacturers who develop their loudspeakers on the tweeter axis would always have an advantage over those who do not.

 

[napilopez]

So it is clear, the time and effort that it took to make these two measurements, cost us measuring two new speakers. The work did confirm that setting all measurements to the same axis, i.e. tweeter axis as the system asks for, is the right approach. Randomly moving that reference based on some line in a manual is not wise. It radically changes the response and generates meaningful differences in how one would think the speaker measures.

I think the remeasurement was very much worth it. It:

-Established the consistency of both the klippel and the KH80. The tweeter axis measurements are very similar.
-Showed what appears to be an important requirement to the klippel. Now we know that measurements need to be made at the tweeter axis for proper results (although it makes me wonder about how it will handle odd designs sans proper tweeters?). So I won't complain if I see another speaker measured at the tweeter that otherwise wouldn't be

I don't think that's a trivial distinction; from the perspective of someone measuring speaker the old school way, it seems an odd limitation to need to need capture the sound field from the tweeter axis if that's not how the speakers were designed. I think that's why you see many of who've had questions are people who measure or design speakers regularly. Klippel's public documentation is limited so we ask(although the tone needs work!)

I'm totally willing to accept that compromise, but it's important context to know going forward. Thanks for taking the time to perform these additional meaaurements.

 

[BYRTT]

.....Klippel's tweeter point has nothing to do with it and must remain in front of the tweeter.

Thanks thinks sounds excelent simple finding, below visual extract view for all 36 verticals including minus steps at left side and positives on right side.

 

[gr-e]

I'm totally willing to accept that compromise, but it's important context to know going forward. Thanks for taking the time to perform these additional meaaurements.

There's no compromise really. The tweeter point is just the data that klippel needs to work properly. The measured data can then be extrapolated to some farfield measurement of choice, including the reference axis @ 2m as CEA2034 suggests.

 

[Pio2001]

(although it makes me wonder about how it will handle odd designs sans proper tweeters?).

In the Klippel documentation posted above, we can see that if the tweeter point is not properly defined "This requires a very high order of expansion to identify the sound field".

I guess that in this case, Amir would have to ask the Klippel for a higher order holographic sound field expansion, and I suppose that it will require a lot more measurement points.

By the way, maybe this is the reason why the graphs with the reference point (tweeter's point ?) in the middle of the speaker look so strange. Maybe the sound field expansion was not "high order" enough for such an offseted center point.

Thanks thinks sounds excelent simple finding, below visual extract view for all 36 verticals including minus steps at left side and positives on right side.

View attachment 49243

Thanks,
So the curve that we are looking for (frequency response on the manufacturer's reference axis) is in the top left diagram, between the red and the first orange one. Closer to the red one.

 

[BYRTT]

Thanks,
So the curve that we are looking for (frequency response on the manufacturer's reference axis) is in the top left diagram, between the red and the first orange one. Closer to the red one.

Yes its rather fantastic you pointed it out because i can offset it precise into milimeters in VituixCad and seems will make any KH80 owner real happy, now have we any KH 80 owners @Pio2001 or @napilopez or @thewas that can tell me precise offset in milimeter from tweeter to Neumann's recommandation based data used used for original review

 

[Pio2001]

Sorry, I've got the KH-120, not the KH-80...

 

[amirm]

In the Klippel documentation posted above, we can see that if the tweeter point is not properly defined "This requires a very high order of expansion to identify the sound field".

I guess that in this case, Amir would have to ask the Klippel for a higher order holographic sound field expansion, and I suppose that it will require a lot more measurement points.

I verified this in the last few hours. The default 500+ point and order 10 gets accuracy to about 6 kHz. Above that error climbs rapidly due to complexity of the soundfield:

Given the same number of points, the highest order I can increase in software is 14 and this is the results:

The error is pushed down in higher frequencies. Results are still not very good above 9 kHz.

Don't panic! The data I show you is independent of this:

Gating is used above 2 kHz where we have plenty of error margin in the previous graph. Indeed, I compared the generated frequency response graphs and they are essentially identical regardless of expansion order (these are for the on-tweeter-axis):

If you were measuring in a tiny space and had to push the reflection-free zone way up, then the order would matter more and more. Given the large space I am using (3-car American garage with high ceilings), that is not an issue.

One area that *seems* to be impacted is the 3-D visualizations. I think these are generated using ungated response which would explain why moving away from the tweeter axis made such a mess of them.

The computational cost is low for increasing the order. It is a super pain to do the same to measurement points though. The current measurements take about 70 minutes. Doubling the measurement points would push that to nearly 3 hours. And would sharply increase the wear on the mechanical system. I will resort to that if fitting error drops too low in the ungated area on complex speakers but for now, I think we are good with the number of measurement points.

 

[Sancus]

Yes its rather fantastic you pointed it out because i can offset it precise into milimeters in VituixCad and seems will make any KH80 owner real happy, now have we any KH 80 owners @Pio2001 or @napilopez or @thewas that can tell me precise offset in milimeter from tweeter to Neumann's recommandation based data used used for original review

The manual says the (x,y) coordinate (from the bottom of the speaker) of the "acoustic center" is (7.7cm, 12.9cm). Eyeballing this with a ruler, I see 17.6cm as the center of the tweeter giving 47mm difference. 18.2cm & 53mm seems more accurate on a second look.

No way this is accurate to better than +/- 2.5mm though, definitely not +/- 1mm.

 

[napilopez]

In the Klippel documentation posted above, we can see that if the tweeter point is not properly defined "This requires a very high order of expansion to identify the sound field".

I guess that in this case, Amir would have to ask the Klippel for a higher order holographic sound field expansion, and I suppose that it will require a lot more measurement points.

By the way, maybe this is the reason why the graphs with the reference point (tweeter's point ?) in the middle of the speaker look so strange. Maybe the sound field expansion was not "high order" enough for such an offseted center point.

This makes a lot of sense to me, and my money is on this explanation.

There's no compromise really. The tweeter point is just the data that klippel needs to work properly. The measured data can then be extrapolated to some farfield measurement of choice, including the reference axis @ 2m as CEA2034 suggests.

Sure, by compromise I just mean that all measurements will need to be referenced to the tweeter axis. We still haven't seen evidence the plotted data can be referenced to a different axis, have we? I always understood the klippel prefers to perform expansions from the tweeter axis, but I wasn't clear on the ability to plot data as referenced to a different point.

 

[amirm]

Showed what appears to be an important requirement to the klippel. Now we know that measurements need to be made at the tweeter axis for proper results (although it makes me wonder about how it will handle odd designs sans proper tweeters?).

As I just explained, this is not an issue in practice due to order expansion only being needed up to 2 kHz. But if needed, I will adjust.

 

[amirm]

Sure, by compromise I just mean that all measurements will need to be referenced to the tweeter axis. We still haven't seen evidence the plotted data can be referenced to a different axis, have we?

You can set this mechanically or in the software. The latter is a pain as you have to compute the coordinates to put in there. I will make another measurement by setting this using the robotic system and see what difference it makes.

 

[LeftCoastTim]

Yes its rather fantastic you pointed it out because i can offset it precise into milimeters in VituixCad and seems will make any KH80 owner real happy, now have we any KH 80 owners @Pio2001 or @napilopez or @thewas that can tell me precise offset in milimeter from tweeter to Neumann's recommandation based data used used for original review

53mm from the acoustic center (x=7.7cm y=12.9cm) on mine (+/- 1mm)

 

[napilopez]

As I just explained, this is not an issue in practice due to order expansion only being needed up to 2 kHz. But if needed, I will adjust.

Good to know, thanks!

You can set this mechanically or in the software. The latter is a pain as you have to compute the coordinates to put in there. I will make another measurement by setting this using the robotic system and see what difference it makes.

Also thank you. It's good to know the capabilities of the system better. Cheers for all the hard work!