Neumann KH 80 DSP Speaker Measurements: Take Two

[NTK]

The balloon plots can be easily misinterpreted. The shape of the plot is derived from the SPL. It is usually not indicative of the direction of the lobes. The balloon can only show the lobes correctly when its coincide with the acoustic center. The r value of the balloon is scaled directly to the SPL. You can tell that from the color shading. The color shading and the r value both represent the same values.

Here are a few pictures of my simulations of a crude approximation of a 1" diameter tweeter. The 2 pictures below show the balloon center coincides with the acoustic center.

The 2 picture show a vertical offset of 10 cm of the acoustic center relative to the balloon center. You can see that the balloon tilted upwards because of the offset but that does not reflect the direction of the lobe (which is unchanged).

The 2 picture show a vertical offset of 10 cm and a lateral offset of 2 cm of the acoustic center relative to the balloon center. Now, there are other features showing up in the balloon too.

You can see that relatively "innocent" offsets of the balloon origin can cause pretty large changes in the look of the balloon which can be misleading.

 

[Sancus]

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

You're right that's more correct, I was lazy about putting the speaker on a flat surface in a well-lit area.

 

[restorer-john]

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.

So, what about multiple tweeter arrays or super tweeter >4 way loudspeakers? Back to a position in between the tweeters or follow the manufacturer's advice?

Which tweeter to pick...mmm

 

[briskly]

Addressing some earlier posts, I don't see how the field itself has changed. The only change I can see is the flux of the sound field intersecting the integration surface. Misalignment breaks axial/lateral symmetry, which should force higher orders for reasonable convergence.
This does seem to confirm that this is all computed in spherical coordinates, which goes back to why Klippel chose cylindrical sampling. I would guess that they really wanted to strip out the extra DOF needed for positioning.

 

[Francis Vaughan]

There are a number of independent axies being talked about. The manufacturer may specify the axis that their measurements are made on, this may even be different to the axis they intended the speaker be listened to. Then there is the axis that Klippel would prefer its measurements be relative to , which is chosen for reasons of computational stability. Finally there is the axis which we define as the listening axis for the purpose of deriving metrics of the speaker’s performance. If the manufacturer defines a preferred axis for listening it is reasonable that the last of this 4 be chosen to match that. So far there has been no such explicit manufacturer advice. Which is a pity, but one suspects that case by case toe in of speakers remains a free tuning parameter designers are unwilling to give up, and deliberately remain vague about.

 

[Blumlein 88]

So, what about multiple tweeter arrays or super tweeter >4 way loudspeakers? Back to a position in between the tweeters or follow the manufacturer's advice?

Which tweeter to pick...mmm

View attachment 49266

It may be that such a speaker with so many drivers and not symmetrical would require high order solutions in the math and more measurements to get a good accurate result. I'd think Amir knows the Klippel well enough now to choose correctly on that.

A similar question would apply to panel speakers like the Maggies with a long ribbon on one side or large ESL panels. Where is the right spot to ease the calculation complexity? Or do you need to do some of those 7 hour measurement routines. Maybe Klippel could give guidance on those.

 

[BYRTT]

Okay here we go again thanks to @Pio2001 pointed out the simple problem into post 146 so even myself got it, then we could use the huge detailed 360 deg vertical polars in post 149 to spot ridge errors where a new vertical offset need to be set in software, to correct that error needed distance well the difference in distance between original reviews tweeter position and the right one and thanks to @Sancus @thewas_ @LeftCoastTim to support that 53mm number. In long run need to check up on how offset in details relate all the 76 polars in softwares graphs and also need some sleep, but so far can share some good news for KH 80 owners when original reviews polars are offset -53mm on Y axis, have a look below ...

Actual think probably amirm and MZKM could aprove the 53mm offset down the road and then preference rating for KH 80 should be updated because on axis is pretty smoother now and line close up to listening window plus DI improve a mili squizze but lets see what they say about it, also because there is some small questions so as to live up to CTA2034 standards, my used software is set to 10 deg steps, listening window hor +/-30 deg ver +/- 10deg and a listening distance at 2 meter and those 2 meters are probably important was same in amirm's files from 21. of January.

 

[restorer-john]

Or do you need to do some of those 7 hour measurement routines. Maybe Klippel could give guidance on those.

I think it will end up being the physical placement (and safety) on the device that limits testing such speakers.

 

[Francis Vaughan]

So, what about multiple tweeter arrays or super tweeter >4 way loudspeakers? Back to a position in between the tweeters or follow the manufacturer's advice?

Which tweeter to pick...mmm

Well, this is a tweeter, super tweeter pair. So the best answer for computational stability will be on the super-tweeter axis. Simply a matter of giving the software the most symmetry at the hardest to manage wavelengths. Not that this mess is going to be easy at the best of times. Given how close the tweeters are to the frame edge it is clear the designer was directing most of their effort to the cosmetic appearance of the design, with little left for the performance. Nice bit of furniture however.

 

[Blumlein 88]

I think it will end up being the physical placement (and safety) on the device that limits testing such speakers.

Reading more about the way the Klippel works, using the tweeter reduces the calculations to a minimum. If something is more complex it may require additional calculation and even more measurement positions. A speaker like the one you posted a picture of would fit into that category. I don't see any problem that would crop up in regards to physical safety or placement.

 

[pma]

Someone might like to send JBL 4355 to Amir for measurements

 

[restorer-john]

I don't see any problem that would crop up in regards to physical safety or placement.

I would have thought Klippel would need multiple large platforms and a considerably greater vertical height (might mash into the garage roof) for such speakers. Also, there must be a load limit. And a safety aspect- he can't screw expensive and massive speakers down- surely?

And a way of manhandling the things up there- some type of lift stand. A bit risky once you go over 30kg/65lbs.

Even the Yamaha NS-1000M is 31kg and it's classified as a "bookshelf" speaker...

 

[Francis Vaughan]

Another way of talking about the axis is perhaps this.

The Klippel system provides the sphere of response at a quasi-far-field. So far there is no notion of a listening axis. The sphere is referenced to the measurement axis, which was defined for no other purpose than to ease the computational complexity and to help any issues of computational stability.

Once the measurement and computational processing are done the question of listening and metrics of sonic quality can be addressed. It is clear that some speakers are explicitly not designed to be listened to on-axis. There is no justification in insisting that the metrics of quality must be determined on-axis. Indeed it is scientifically indefensible to do so*.
Given we have the full sphere, we are in the position to be able to determine what the best listening angle is. Minimally it would seem that moving across the horizontal axis from on-axis to about 30 degrees off axis, choosing that as our listening axis, evaluating the various measures (on axis response and directivity being the ones to vary) and calculating the metric, and choosing the axis that provides the best result would be a defensible approach.

*There is the argument that standards require on-axis measurement. This misses the point. Standards have two purposes. One is to provide a standardised way of comparing things, in which case it doesn't matter if the technique is sub-optimal, so long as it is repeatable. The second is to capture best practice. Current standards for speaker measurement are the former. We are concerned with the latter.

 

[JIW]

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.

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

Okay here we go again thanks to @Pio2001 pointed out the simple problem into post 146 so even myself got it, then we could use the huge detailed 360 deg vertical polars in post 149 to spot ridge errors where a new vertical offset need to be set in software, to correct that error needed distance well the difference in distance between original reviews tweeter position and the right one and thanks to @Sancus @thewas_ @LeftCoastTim to support that 53mm number. In long run need to check up on how offset in details relate all the 76 polars in softwares graphs and also need some sleep, but so far can share some good news for KH 80 owners when original reviews polars are offset -53mm on Y axis, have a look below ...

Actual think probably amirm and MZKM could aprove the 53mm offset down the road and then preference rating for KH 80 should be updated because on axis is pretty smoother now and line close up to listening window plus DI improve a mili squizze but lets see what they say about it, also because there is some small questions so as to live up to CTA2034 standards, my used software is set to 10 deg steps, listening window hor +/-30 deg ver +/- 10deg and a listening distance at 2 meter and those 2 meters are probably important was same in amirm's files from 21. of January.

View attachment 49268

I have measured this a bit differently. Given that the tweeter is somewhat deep in the waveguide (several cm), the angle at which it is observed relative to a ruler changes the measurement (parallax). I have therefore used a picture where the recorded surface is parallel to the front of the speaker (orthogonal to the acoustic axis) (see below).

Then, I zoomed into the picture such that it almost filled my screen (2560x1440) and used macOS's screenshot tool to measure the number of pixels from the bottom to the top (1297 +/- 1) and from the bottom to the centre of the tweeter (1000 +/- 2). Knowing the height as 233 mm, gives a tweeter height of 179.6 +/- 0.4 mm, which since the height of the acoustic axis is known to be 129 mm gives a difference in height of 50.6 +/- 0.4 mm.

This 2.4 mm upward difference comes out to 2.7 degrees at the specified minimum listening distance of 500 mm, 1.4 degrees at 1 m and 0.7 degrees at 2 m.

https://static-neumann.s3.amazonaws.com/img/1145/x1_KH-80-DSP-Front_Neumann-Studio-Monitor_G.jpg

 

[Juhazi]

Good morning, darlings! Glad to see this measurement axis issue seems to be in order now!

I summarize it like this:
- Klippel NFS must be given the physical tweeter position as primary landmark for raw data
- measured raw data from two layers will be calculated/processed to compensate for reflections and to make graphical presentations of spl response etc.
- Amir said that it is difficult to give NFS the listening axis for calculations. Does it have a default, what is it? Obviously it is the tweeter axis? BYRTT uses freeware VituixCAD simulation where it is very easy to set 0-axis coordinates x,y and z on the baffle.

Those of us who are familiar with taking acoustic measurements, know painfully well how deviations in mic position change response of multiway speakers, which are basically a bunch of imperfections.

I've been surprised of how well (in my mind) NFS calculates bass response of BR speakers with front vs. rear port. In this manufacturers can play games (to get a response that looks good), and speaker positioning is very important for the end user IRL.

Cheers, Juha

 

[thewas]

Well, this is a tweeter, super tweeter pair. So the best answer for computational stability will be on the super-tweeter axis. Simply a matter of giving the software the most symmetry at the hardest to manage wavelengths. Not that this mess is going to be easy at the best of times. Given how close the tweeters are to the frame edge it is clear the designer was directing most of their effort to the cosmetic appearance of the design, with little left for the performance. Nice bit of furniture however.

The problem remains though for loudspeakers that use real tweeter arrays like for example some Geithain monitors https://www.hifi-selbstbau.de/mobile/testberichte/fertiglautsprecher/536-me-geithain-804k , M&K home cinema systems https://mksound.com/ or Hifi and PA line arrays.

 

[tuga]

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.

What do you mean by "correct at the reference point"? If the cabinet produces diffraction artifacts and they're audible at the listening spot?

Will the convolution produce an accurate representation of lobbing?

 

[Francis Vaughan]

The problem remains though for loudspeakers that use real tweeter arrays like for example some Geithain monitors https://www.hifi-selbstbau.de/mobile/testberichte/fertiglautsprecher/536-me-geithain-804k , M&K home cinema systems https://mksound.com/ or Hifi and PA line arrays.

It does. But you can get at least some useful symmetry. Choose on axis of one of the tweeters, and if the others are in a line, you have some significant symmetries, just not full rotational symmetry about the chosen axis for that driver. That is going to help a great deal, and no doubt is what is done with arrays. The Klippel system seems to be very much intended for use with modern PA line arrays, but it is here that you will need the optional multi-input management.

 

[Francis Vaughan]

What do you mean by "correct at the reference point"? If the cabinet produces diffraction artifacts and they're audible at the listening spot?

Will the convolution produce an accurate representation of lobbing?

One would say, correct at whatever point you choose to use as a reference. And yes, you get the entire mess, lobing and diffraction included. About the only thing that the mathematics seems to be short in is in modelling the effect of the emitted sound bouncing off other drivers. It assumes a solid surface in deconvolving the reflection, and thus can't correct for either a compliant surface or a moving surface. So it assumes a solid immobile reflector. Other than that, you get a very comprehensive prediction of what is heard at any point you choose.

 

[tomtoo]

So, what about multiple tweeter arrays or super tweeter >4 way loudspeakers? Back to a position in between the tweeters or follow the manufacturer's advice?

Which tweeter to pick...mmm

View attachment 49266

What speaker is this?

A Interference 2020?