• WANTED: Happy members who like to discuss audio and other topics related to our interest. Desire to learn and share knowledge of science required as is 20 years of participation in forums (not all true). Come here to have fun, be ready to be teased and not take online life too seriously. We now measure and review equipment for free! Click here for details.

Neumann KH 80 DSP Speaker Measurements: Take Two

tuga

Addicted to Fun and Learning
Joined
Feb 5, 2020
Messages
558
Likes
377
Location
Oxfordshire
#62
This should not be the case, as this is the most unusual thing in the whole system - even if I don't understand it :)
So the Klippel NFS website says:

But the statements all seem to refer to a single-source. Because Klippel continues and says to measure multi-way speakers:

That again I can easily understand. Only by measuring the individual drivers individually can their "behaviour" be extrapolated for any distance and by combining the individual driver measurements the frequency response can be determined for any point in the room.

But if this is the case the measurements of a multi-way loudspeaker in 0.3m are not transferable to 1m distance by the NFS.

Can anyone with more insight into the functioning of the NFS comment on this?




At low frequencies the limiter intervenes much more. Here is a comparison of the first and second measurements, normalized to the second measurement:
View attachment 49142
Does it make sense to "avoid diffraction problems"?

Does it make sense to measure each driver individually?
What about driver interference, does that not affect dispersion?
 

pozz

Machine
Forum Donor
Editor
Joined
May 21, 2019
Messages
1,472
Likes
2,106
#63
But it is conflicting the known Neumann design philosophy (at least Markus Wolff keeps emphasizing it) which is: All speakers are designed to be flat on-axis anechoic, which means baffle step fully compensated, diffraction fully compensated. Which means off-axis it cannot and does not measure flat, nor does power response (as measured eg in a reverberation room).
In-room response has to be established by user EQ/DRC, of course with the intrinsic directivity pattern overlaid, the goal being to find a best compromise for the given situation.
Where's the contradiction? The off-axis measurements have shown characteristic smooth downwards slopes, which is equally true for power response. With that, if you take the in-room response as representative, and being generous for whatever's going on below the transition band, then whatever debate about accuracy and process we're having here is less important for those who actually want to buy these speakers. A sub is necessary in any case.
 

ctrl

Active Member
Joined
Jan 24, 2020
Messages
128
Likes
465
Location
DE
#64
Looks like you've nailed the problem.
I hope not, because in this case all the measurements made would be completely useless because of great inaccuracies due to the phase shift/measuring angle of the distant chassis.
Only the low frequency range in which a single driver plays would be measured accurately.
 
Last edited:

HammerSandwich

Addicted to Fun and Learning
Forum Donor
Joined
Nov 22, 2018
Messages
653
Likes
696
#65
Only the low frequency range in which a single driver plays would be measured accurately.
Okay - for the holographic separation. How are the (gated) top-octave measurements affected here? That's the biggest disparity with Take Two...
 

pozz

Machine
Forum Donor
Editor
Joined
May 21, 2019
Messages
1,472
Likes
2,106
#66
I hope not, because in this case all the measurements made would be completely useless because of great inaccuracies due to the phase shift of the distant chassis.
Only the low frequency range in which a single driver plays would be measured accurately.
I very much doubt that the system fails to account for the differences in driver position and orientation in bookshelf/tower/point-source/array-like speakers. The Revel center channel speaker being the best representation of that.
 
Joined
Jul 10, 2019
Messages
61
Likes
65
#67
That again I can easily understand. Only by measuring the individual drivers individually can their "behaviour" be extrapolated for any distance and by combining the individual driver measurements the frequency response can be determined for any point in the room.
Solutions to the Kirchhoff-Helmholtz integral do not need to concern themselves with isolating individual emitters. The requirement of isolating the drive units is for the concern of adjusting the summed response of the array of drivers, such as a line array.
 
Last edited:

sejarzo

Active Member
Forum Donor
Joined
Feb 27, 2018
Messages
275
Likes
193
#68
As to the rest of the response, unfortunately S&R's graph is also of poor resolution:

View attachment 49072

Lots of sins can be buried in low resolution graphs like that.
I'm a bit confused by what you mean by "resolution". Yes, the graph is small and of low visual resolution, but given the y-axis scale, all the wiggles are well within a +/- 1 dB range. Even if there are many of them that are not visible, their magnitude is very small compared to most speakers.

Enlarged and sharpened:

1581176314216.png
 

ctrl

Active Member
Joined
Jan 24, 2020
Messages
128
Likes
465
Location
DE
#69
Does it make sense to "avoid diffraction problems"?

Does it make sense to measure each driver individually?
What about driver interference, does that not affect dispersion?
The convolution of the individual measurements does not avoid diffraction problems, but rather ensures that these are actually correct at the reference point.

The interference of the individual chassis is not lost. In this way, the Klippel NFS works like a conventional simulation program for multi-way loudspeakers for a given reference point.
 

BYRTT

Senior Member
Forum Donor
Joined
Nov 2, 2018
Messages
379
Likes
712
Location
Denmark (Jutland)
#70
The 20kHz balloons (or any other frequency, for that matter) *must* have the minutely exact same shape no matter where you put the acoustical axis, the only difference is that the whole thing is tiltet a bit. If they don't, something is wrong... at least if the assumption is that the balloons shall represent far field conditions.
My educated guess is that the measurement distance is too close, and more importantly, dynamically varying depending on choosen reference axis and that produces systematic errors despite of the technique using two distances and calculating the far field from that.

It is not for nothing that there is the recommondation, for traditional style measurements, using a distance of at least ten times the largest dimension of the DUT, and preferably larger than 20 times. Otherwise you're not in the far field.

I kindly suggest @amirm to do some experimenting to establish a baseline of repeatiblity and error bounds. Measure the same speaker (this one or any other) using a set of somewhat arbitrary reference positions. Say, the upper left corner and the lower left corner, and also a different set of the two distances the far field extraction is made from. Unless the ballons (at high frequencies) perfecty match in shape (ignoring the tilt), it's not working as intended.

EDIT: This assumes a mic that has close to no directivity (way less than 1dB) at angles up to least 30degree. A condition not met by the mic currently in use, it seems.

EDIT2: I also suggest to really think about double checking the whole measurement rig in the way I've explained here.
Can this help microphone polar patterns copyed from MK 255 datasheet, notice its not numbered 1/4" but 1/2", also take one spinorama with and without the microphone correction offset, at low reach mic is a 3,5Hz 1st order rolloff before whatever used preamp will rolloff.

Can be wrong here but have bit of a hard time see why not the high resolution is possible, signal to noise ratios are state of art at one feet so transducer plus difraction should be captured maybe showing more wriggles than we used to see and that low end is absoute possible tunning for ported systems, if Neumann know about that low end knee in that resolution is a quistion.

MIC.png
 
Last edited:

Juhazi

Addicted to Fun and Learning
Joined
Sep 15, 2018
Messages
931
Likes
1,083
Location
Finland
#71
These 2nd round measurements make us ask more and new questions... I've been tinkering about that 20kHz 3D disfference, but I can't get hold of it.
An elliptic waveguide+tweeter should measure pretty well until interference with midrange (or woofer in this case) happens. The KH80 obiously has LR4 symmmetric slopes (correct is LR8!) and measurements represent only the tweeter above say 5khz.

Here is measurement of KH 420 vertical dispersion, tweeter looks identical. Notice symmetric up/down response at 20kHz
https://www.fidelity-online.de/neumann-kh-420-messungen/

 
Last edited:

napilopez

Addicted to Fun and Learning
Joined
Oct 17, 2018
Messages
685
Likes
1,918
#72
I have always wondered how Klippel manages to convert a loudspeaker measurement in 0.3m into a correct 1.0m measurement.

As others have said, there is a problem with the measurement angles to the tweeter and woofer when measuring at a distance of 0.3m, compared to measuring at 1m.
In addition, the influence of the baffle depends on the measuring distance (has already been mentioned by others).

Therefore, measurements (in anchoic chambers) for large loudspeakers are carried out at distances of up to 4m (in exceptional cases even further) to keep the problems described above as small as possible.

The Klippel NFS must therefore convert all these influences to a distance of 1m (when measuring at a distance of 0.3m). Otherwise the measurements would be useless.

Even if measurements were not taken on the acoustic axis specified by the developer, the measurement result can later be converted to any point in space as a measuring point - if I have understood the theoretical approach to this correctly.

This can only mean that something is wrong with the second measurement - wasn't the result converted to 1m distance?

In the first measurement, the vertical measurement with -10deg has approximately the acoustic axis required by Neumann - in the diagram in red.
In cyan is the second KH 80 measurement of Amir with now correctly adjusted measuring point in 0.3m distance.
Something went completely wrong!

View attachment 49131
Thank you for elaborating my concern clearly. I'll try again: The Klippel is supposed to use fancy math to calculate the frequency response at a realistic listening distance. Nobody cares about how a pair of speakers sounds at 1 foot.

My presumption is that even if measurements are made at 0.3m, the plotted frequency response is the mathematical prediction of the measurement at some greater distance. I presume this is 2m, as that is what is suggested for the spinorama. (The SPL is then supposed referenced to 1m for standardization purposes, but the plot itself is supposed to represent measurements at 2m.

While at 0.3m the difference between the tweeter and reference axis might be large, at 2m, it should only cause a minimal change in the treble response.

All that being said, even disregarding the above fact, I'm quite confident that a single measurement made at 0.3m at the reference axis would not show this droop in the treble (I'm travelling, otherwise I'd test). So I'm puzzled.
 
Last edited:

ctrl

Active Member
Joined
Jan 24, 2020
Messages
128
Likes
465
Location
DE
#73
Solutions to the Kirchhoff-Helmholtz integral do not need to concern themselves with isolating individual emitters. The requirement of isolating the drive units is for the concern of adjusting the summed response of the array of drivers, such as a line array.
But why then does Klippel explicitly refer to the individual measurements for multi-way systems?
The sound pressure output of loudspeaker systems with multiple transducers (line sources, sound bars, 3-way systems, etc.) can be determined by measuring each transducer separately using a multiplexer.
 

KSTR

Senior Member
Joined
Sep 6, 2018
Messages
440
Likes
1,083
Location
Berlin, Germany
#75
I very much doubt that the system fails to account for the differences in driver position and orientation in bookshelf/tower/point-source/array-like speakers. The Revel center channel speaker being the best representation of that.
I would also doubt that the experts at Klippel failed. OTOH, aren't a woofer and a tweeter emitting the same frequency (at the same phase, assuming a 0deg phase offset crossover like any lin-phase XO as well as any proper min-phase Linkwitz-Riley) the same thing as two identical fullrangers doing that? Somehow the Klippel quotes that @ctrl gave seem to contradict themselves, unless I missed the proper context (I didn't check the sources).
 
Joined
Sep 18, 2019
Messages
46
Likes
19
#76
All Class 1 free field measurement mic's are calibrated as in a free field. When you use them close to a reflective object ( aka a loudpeaker) the calibration will be off....
 

pozz

Machine
Forum Donor
Editor
Joined
May 21, 2019
Messages
1,472
Likes
2,106
#77
I would also doubt that the experts at Klippel failed. OTOH, aren't a woofer and a tweeter emitting the same frequency (at the same phase, assuming a 0deg phase offset crossover like any lin-phase XO as well as any proper min-phase Linkwitz-Riley) the same thing as two identical fullrangers doing that? Somehow the Klippel quotes that @ctrl gave seem to contradict themselves, unless I missed the proper context (I didn't check the sources).
It's just front-page explanations, advertising aimed at the technical. https://www.klippel.de/products/rd-system/modules/nfs-near-field-scanner.html

So not a formal source.
 

Juhazi

Addicted to Fun and Learning
Joined
Sep 15, 2018
Messages
931
Likes
1,083
Location
Finland
#78
More details here https://www.klippel.de/fileadmin/klippel/Bilder/Our_Products/R-D_System/PDF/C8 Near Field Scanner System.pdf !

A special module is needed for large speakers!

Multi Source Superposition

The sound pressure output of loudspeaker systems with multiple transducers (line sources, sound bars, 3-way systems, etc.) can be determined by measuring each transducer separately using a multiplexer. After the holographic processing the sound pressure of the individual source can be superimposed to determine the total sound pressure output of the audio device.

Basic module without MSS must use measurement distance that represents "free field conditions" for the DUT.

This is the problem/question I have been asking all the time...
Next question is "How do you determine the distance needed for FFC?"

nfs free field.jpg
 
Last edited:

ctrl

Active Member
Joined
Jan 24, 2020
Messages
128
Likes
465
Location
DE
#79
All that being said, even disregarding the above fact, I'm quite confident that a single measurement made at 0.3m at the reference axis would not show this droop in the treble. So I'm puzzled.
I would agree with that.
The acoustic centres of the tweeter and woofer are about 0.1m apart vertically in the KH 80.
At a distance of 0.3m there is a vertical angle of about 10° from the first reference point/measurement point to the tweeter. This cannot be the sole cause.
At 2m distance the angle error is only 1.5°. Can therefore be completely neglected.

I very much doubt that the system fails to account for the differences in driver position and orientation in bookshelf/tower/point-source/array-like speakers. The Revel center channel speaker being the best representation of that.
But we have no anechoic chamber measurement for comparison to detect possible measurement errors.

...loudspeaker systems with multiple transducers...
This wording also includes 2-way speakers - does it?
 

jhaider

Addicted to Fun and Learning
Joined
Jun 5, 2016
Messages
641
Likes
621
#80
About the measurement distance, I thought there was a standard (1 m) but each one measures as he wishes. The logical thing, for me, is 1 meter away and at the listening point.
1m is convention for individual drive units. Multi driver speakers generally need more integration distance.

Also, I wouldn’t take Danny Ritchie too seriously. He’s trying to sell (often useless) tweaks, not advance the state of reproduction.
 
Top Bottom