Neumann KH150 Review

[xaviescacs]

Great collection of mesuraments! What a beast....

 

[dananski]

Wow, thanks for all these extra insights! I always wondered what the padding was like inside to deaden the decay so immediately.

 

[Karmacoma]

View attachment 469179
After quite some time, Neumann finally released a new model in late 2022.
What makes this one especially exciting is that it neatly fills the gap between the KH80 (4") / KH120 (5") and the KH310 (8") — at a very versatile 6.5" size.




View attachment 469180
The back panel follows Neumann’s traditional style.
With the XLR connector mounted from the bottom up, you can place the speaker flush against a wall or bracket without interference — a thoughtful touch.




Frequency Response
View attachment 469181
As expected from a professional monitor, performance is excellent.
The overall balance is flat and smooth, with the -6 dB low-end extension measured at 39.4 Hz.


For reference, my graphs use 1/24-oct smoothing. (The CEA-2034 standard specifies a minimum of 1/12-oct; a larger denominator means higher resolution in the data.)




Cross check with ASR Data
View attachment 469182
I overlaid my results with the official measurements from Audio Science Review.
The two sets of data match extremely well, except for the low-end extension.
The Klippel NFS system used by @amirm is especially reliable in the low-frequency range.




Nearfield Measurements
View attachment 469184

As you’d expect from Neumann, all responses are clean and well-controlled.
The woofer shows no signs of breakup modes at higher frequencies, and the port is free from pipe resonance noise.


Let’s take a closer look at that last point.




View attachment 469183
In the rear-panel teardown photo, note the white section I’ve outlined.

This is an effort to control the inherent pipe resonance that can occur in any port design.
For a more detailed explanation of this topic, please see my article here:


https://audiore.kr/스피커-성능-측정-보고서/


Interestingly, there’s an opening at roughly one-quarter of the port’s length, backed with acoustic damping material to absorb resonances.
On closer inspection, there’s also a layer of black foam sandwiched between the white foam and the port opening — with each layer having noticeably different density.




This appears to be a multi-stage acoustic impedance transition.
By gradually increasing the acoustic impedance between the port air column and the damping material, the design likely improves absorption efficiency and reduces internal reflections.



In simple terms:
When sound travels through different media, the more similar their acoustic impedance, the more easily it passes through.
A sudden mismatch in impedance causes reflections instead of transmission.
Neumann’s layered approach likely minimizes these reflections and boosts damping efficiency.




CEA-2034
View attachment 469185
The on-axis response is smooth, but the Listening Window curve is even smoother and flatter.
This suggests Neumann may have designed and tuned the KH150 with the Listening Window in mind, rather than focusing solely on the exact 0° axis.




Directivity
View attachment 469186
View attachment 469187
Even with no smoothing applied, the response is smooth across nearly all angles — it almost looks smoothed.
Classic Neumann performance.




View attachment 469188
View attachment 469189
Vertical directivity is also excellent: narrow, consistent, and stable well into the top end.
(The apparent widening above ~17 kHz is due to a slight dip in the 0° response at that frequency.)




Beamwidth
View attachment 469190
The beamwidth narrows gradually and evenly, making it difficult to pinpoint the woofer–tweeter crossover just from the curve — a sign of a finely tuned waveguide and excellent driver matching.




View attachment 469191




Polar Plots
View attachment 469192
Among enthusiasts, there’s ongoing debate over whether narrowing or constant beamwidth is preferable.
Here, Neumann has gone for a gently narrowing pattern, achieving remarkably uniform attenuation at almost every radiation angle — even to the sides and rear — except for a small section in the extreme treble.




View attachment 469193
Vertical polar results are also tightly clustered within the intended radiation angles, with only a slight narrowing around 2 kHz due to the non-coaxial layout.





THD
View attachment 469194
As a reviewer, I try to avoid bias — but with brands like KEF, Neumann, and Genelec, it’s hard not to expect excellence even before measuring.
Fortunately, my microphone and test rig are free of such bias, so you can trust the objectivity of the data.


Performance is impressive across the board, especially the low distortion even below 100 Hz — notable for a speaker capable of solid output in the 30 Hz range.




View attachment 469195
View attachment 469196
3rd harmonic distortion is practically at the noise floor.




View attachment 469197
When increasing the measurement output, I noticed something odd in the 150–400 Hz range — not only visible on the graph, but also audible as a noise during playback.
First, the sharp spikes in the THD curve at those points are not part of the speaker’s intrinsic response; rather, they are the result of an interaction between my measurement stand and the speaker itself.

As with the KH120’s upgrade to the “II” version, the KH150’s enclosure has also been switched to plastic.
Because of this, the speaker’s total weight is relatively low compared to the strength of its woofer, meaning that if it’s placed directly on a perfectly smooth, flat surface, a certain resonance can cause the cabinet to “walk” at specific frequencies.

This is not unique to the KH150 — I’ve occasionally encountered other models with similar issues.
These days, when such problems arise, I apply additional damping measures during testing.
However, when I measured this unit back in 2023, my view at the time was that “the interaction between the product and the test environment is inevitable, and maintaining consistency is more important.”
So I decided to record the data as-is.

That said, the issue can be mitigated by placing a heavy book on top of the speaker (to add mass) or by putting a layer of felt or EVA foam sheet under the base.




View attachment 469198
View attachment 469199
Aside from this artifact, performance is excellent.




Woofer Measurement
View attachment 469200
With the owner’s permission, I opened the unit to examine the woofer in more detail.




View attachment 469201
Close-mic measurements of 3rd harmonic distortion showed similar trends to the far-field results (blue trace).





LSI(Large Signal Identification)
View attachment 469202
This might be the first time I’ve shown this type of graph in my speaker review series.
The Klippel measurement system includes a module called LSI (Large Signal Identification), which is a useful tool for observing the driver’s state in real time, by displacement, when the driver is operating at large excursions.
For this review, I’ve brought in only two parameters: Bl(x), which represents the force from the voice coil and permanent magnet, and Kms(x), which represents the restoring force of the suspension.

Since these graphs may look unfamiliar, let me explain briefly:
The x-axis indicates the distance the speaker cone moves forward or backward from the center position of 0 mm (measured in millimeters).
The y-axis shows the magnitude of each force (Bl and Kms) when the speaker is displaced by that amount.




View attachment 469212

If a speaker could maintain its best performance whether stationary or in motion, that would be ideal — but reality is different.
Because the voice coil moves and the permanent magnet remains fixed, Bl(x) will vary in real time depending on the coil’s position.




View attachment 469213
The same is true for the restoring force of the suspension.
In a loudspeaker, this role is handled by the surround and the spider.
The further these parts move away from rest, the greater their restoring force becomes by nature.
(Think of pulling on a piece of fabric or a rubber band!)




View attachment 469203

For this reason, even in well-made speaker drivers, the Bl(x) curve typically takes on a smooth inverted-U shape, and the Kms(x) curve takes on a smooth U shape.
(The example above shows the LSI measurement results for a KEF R-series 6.5-inch driver.)


The narrower and sharper these curves become, the more the 3rd harmonic distortion component will increase relative to displacement.
The more asymmetric the curves become, the more the 2nd harmonic distortion component will increase relative to displacement.

It’s a slightly difficult topic, but thank you for following along.
Now, let’s look at the KH150’s case.




View attachment 469202
Even over a very large displacement range, the KH150 shows almost no change in the values of Bl(x) and Kms(x), maintaining a consistent shape.
In the small displacement range, the symmetry is also excellent.

This, I believe, is the secret to how this speaker can reproduce frequencies in the 30 Hz range while maintaining clean quality.







Multitone Test
View attachment 469204
View attachment 469205
Multitone distortion performance is generally very strong.
It is especially impressive below 500 Hz.
Above that, the gradual rise up to around 1.7 kHz is likely dominated by Doppler-related IMD (Intermodulation Distortion), caused by the woofer handling upper midrange content in a 2-way design.




80Hz~
View attachment 469206




View attachment 469207
View attachment 469208

That’s pure strength, through and through.
In the mid-to-low frequencies in particular, I don’t think there’s anything in its class that can match it.




Compression Test
View attachment 469209
No words needed — it’s a beast, delivering well beyond what its size might suggest.




Deviation between 2 samples
View attachment 469210
“Neumann being Neumann.”
From the KH80 through the KH120 II, the integration of DSP has pushed their QC standards to an exceptional level.




Final Thoughts
In closing, here are my personal thoughts.
Once again, Neumann has lived up to expectations, delivering top-tier engineering.
That said, with all this added muscle paired to a lighter body, it might need just a little helping hand from the user.

That’s all.


----

Hi, if you have measurements for a purifi driver, I'd love to see a comparaison between both drivers.

 

[dfuller]

Hi, if you have measurements for a purifi driver, I'd love to see a comparaison between both drivers.

You're in luck.

Purifi Audio PTT6.5X04-NAA-08 6.5 Inch Woofer Review

Purifi Audio PTT6.5X04-NAA-08 6.5 Inch Woofer Review

www.erinsaudiocorner.com

 

[kyuu]

Also let me take this opportunity to express my personal grief with the Klippel colors:
-3dB: Dark red
-6dB and -9dB: almost the same kind of.. also red... in a rainbow map... cmon!

Amir's contour plots are especially challenging to read, since (to me at least, with a good color calibrated display) everything from 0dB to -10dB looks like a big red mess.

That might be intentional. Having hard color differentiation between neighboring levels might not be desirable if the intention is prevent a naive viewer from assuming that difference is meaningful. Amir generally draws in his windowing lines like so:

Window seems to be gated to the narrowest point as outlined by the -12dB outline in the highest frequencies. Alternately, it's possible he's basing it on the -6dB outline where it's narrowest between 1k-2k (perceptually probably the most important frequencies so this may be more correct). Either way, having hard color differentiation between each -1db or -3dB mark would be implying that those are important points to differentiate when that may not be the case.

Also, given the very different ways that the contour maps are presented between Nuyes, Amir, and Erin, I think it's safe to say that it's possible to change the presentation

"Friedrich" applied to one of my horn designs:
View attachment 469977

This is terrible for one simple reason: there is no legend so I have no idea what the different colors are even indicating.

 

[Newman]

I overlaid my results with the official measurements from Audio Science Review.
The two sets of data match extremely well, except for the low-end extension.

I don't think they match extremely well for resolution. Your curves are smoothed in a way that could potentially gloss over certian concerns with speakers (not this one).

cheers

 

[Nuyes]

I don't think they match extremely well for resolution. Your curves are smoothed in a way that could potentially gloss over certian concerns with speakers (not this one).

cheers

The FR curves I provide are shown with 1/24-octave smoothing. Interestingly, the smooth appearance is not due to processing—it’s because the Neumann speakers actually measure that way.


In particular, the data I provide is highly accurate in the mid to high frequencies above 1 kHz.


For reference, here are two additional measurements: one with no smoothing, and one with 1/12-octave smoothing.




No smoothing

1/12 oct smoothing

 

[Newman]

The FR curves I provide are shown with 1/24-octave smoothing. Interestingly, the smooth appearance is not due to processing—it’s because the Neumann speakers actually measure that way.

Yes, thanks for replying, I think I worded poorly. I did not mean that you are deliberately smoothing your plots. I was more interested in the possibility that your method could be less resolving in some sense that I don't understand.

Is there a better explanation for why your plot looks so much smoother than Amir's?

 

[mj30250]

Yes, thanks for replying, I think I worded poorly. I did not mean that you are deliberately smoothing your plots. I was more interested in the possibility that your method could be less resolving in some sense that I don't understand.

Is there a better explanation for why your plot looks so much smoother than Amir's?

View attachment 471475

I believe Nuyes uses a Klippel analyzer, but it doesn't provide the same level of resolution as the NFS that generates Amir's measurements.

 

[Matias]

Hi, if you have measurements for a purifi driver, I'd love to see a comparaison between both drivers.

Remeber that Purifi has a long range of drivers, many woofer options in all sizes, mids and a tweeter too. So it really depends to which Purifi driver you want to compare.

Products | PTT Shop

ptt.purifi-audio.com

 

[Nuyes]

Yes, thanks for replying, I think I worded poorly. I did not mean that you are deliberately smoothing your plots. I was more interested in the possibility that your method could be less resolving in some sense that I don't understand.

Is there a better explanation for why your plot looks so much smoother than Amir's?

View attachment 471475

I believe Nuyes uses a Klippel analyzer, but it doesn't provide the same level of resolution as the NFS that generates Amir's measurements.

Hey guys, I think there’s a small misunderstanding here.

I use the Merging Near & Far Field method for my measurements. While the accuracy depends on the gating window (which varies with the measurement environment), when performed correctly this method provides very reliable high-frequency data.


By contrast, nearfield bass measurements are less trustworthy, because they assume that a single measurement point can fully represent the low-frequency behavior of the speaker — which is rarely true in practice. That’s why I compared my data with NFS results in the first place.


Also, while NFS is an excellent tool, it does not directly provide the ‘actual’ far-field response. To obtain a truly accurate far-field result, you still need to measure in the far field. NFS works by scanning the speaker in the near field, then extrapolating its sound field within the room to predict responses at various angles and distances. Since this extrapolation is still based on nearfield scanning data taken in a real room, NFS is not completely free from room reflections — especially at high frequencies. That said, because the math is simpler and less error-prone at lower frequencies, NFS can be extremely accurate in the mid- and low-frequency range.


So rather than debating which method is ‘better’ or ‘more correct,’ I think it’s more productive to understand the strengths and limitations of each approach.

 

[boxerfan88]

Also, while NFS is an excellent tool, it does not directly provide the ‘actual’ far-field response. To obtain a truly accurate far-field result, you still need to measure in the far field.

For my understanding & learning, what is the "far field" distance that you do the measurement at?

/

 

[Nuyes]

For my understanding & learning, what is the "far field" distance that you do the measurement at?

/

I always measure at 1m.
Of course, depending on the speaker’s design and size, that distance can sometimes be far from ideal, but I stick to it for the sake of consistency and because of the limitations of my space (in my room, 1 m with a 5 ms window is the safe limit).

 

[peniku8]

I use the Merging Near & Far Field method for my measurements. While the accuracy depends on the gating window (which varies with the measurement environment), when performed correctly this method provides very reliable high-frequency data.

What gating times do you use? Your data looks like variable gate to me.

By contrast, nearfield bass measurements are less trustworthy, because they assume that a single measurement point can fully represent the low-frequency behavior of the speaker — which is rarely true in practice. That’s why I compared my data with NFS results in the first place.

This is true for large speakers, where the baffle introduces directionality in the bass region and also true for complex sources like anything with a vent or something like a cardioid speaker. It's also noteworthy that groundplane measurements virtually double the baffle size, so the "baffle step" moves lower, so that measurement also doesn't necessarily represent true free-field data, even if you substract the 6dB gain.

Also, while NFS is an excellent tool, it does not directly provide the ‘actual’ far-field response. To obtain a truly accurate far-field result, you still need to measure in the far field.

If you know the spherical wave expansion (which the NFS measures and computes), you can extrapolate the response at any point in space, which naturally also includes the true acoustic far field. This is what the NFS does and it's one of the main selling (and marketing) points of the system. By default, the NFS data is exported at a 7m reference distance. CEA-2034 is handicapped by mandating a 2m reference distance, so this data is inherently problematic for large complex sources (which the KH150 is definitely not).

Since this extrapolation is still based on nearfield scanning data taken in a real room, NFS is not completely free from room reflections — especially at high frequencies.

You set up a 'reflection free frequency' in the NFS software, which is the point where the software switches from its smart algorithms to traditional time-based gating (the same thing you do), to maintain high accuracy up to 20KHz (and probably some bit beyond that). The manual gives you recommended minimum room sizes for the NFS (slightly larger for the extended NFS), probably partly for this exact reason. Here is an example picture:

If set up correctly, the NFS can give you high resolution, true far field data completely devoid of any room interaction.
There is a reason why d&b put a nearfield scanner into their (large) anechoic chamber
Hope this helps

 

[Nuyes]

What gating times do you use? Your data looks like variable gate to me.


This is true for large speakers, where the baffle introduces directionality in the bass region and also true for complex sources like anything with a vent or something like a cardioid speaker. It's also noteworthy that groundplane measurements virtually double the baffle size, so the "baffle step" moves lower, so that measurement also doesn't necessarily represent true free-field data, even if you substract the 6dB gain.


If you know the spherical wave expansion (which the NFS measures and computes), you can extrapolate the response at any point in space, which naturally also includes the true acoustic far field. This is what the NFS does and it's one of the main selling (and marketing) points of the system. By default, the NFS data is exported at a 7m reference distance. CEA-2034 is handicapped by mandating a 2m reference distance, so this data is inherently problematic for large complex sources (which the KH150 is definitely not).


You set up a 'reflection free frequency' in the NFS software, which is the point where the software switches from its smart algorithms to traditional time-based gating (the same thing you do), to maintain high accuracy up to 20KHz (and probably some bit beyond that). The manual gives you recommended minimum room sizes for the NFS (slightly larger for the extended NFS), probably partly for this exact reason. Here is an example picture:

If set up correctly, the NFS can give you high resolution, true far field data completely devoid of any room interaction.
There is a reason why d&b put a nearfield scanner into their (large) anechoic chamber
Hope this helps

Thank you for your kind and detailed comment.
Of course, I do understand what the Klippel NFS is doing, and I know that, in practice, its high-frequency results may also rely on time-windowed data.

Over time I’ve answered many similar questions in my reviews, which is why I took the opportunity here to explain some of the fundamental strengths and limitations of different measurement methods. When I’m asked why my data sometimes looks ‘flatter’ compared to NFS measurements, my only honest answer is simply: ‘because that’s what was measured’ — at least in the high-frequency range.

It’s also true that with Klippel Holographic, the error rate increases as you move up in frequency. That’s exactly why, as in the example image you shared, reflection-free frequency responses become necessary.

For reference, I usually apply a 5ms time window, and I merge nearfield and farfield responses around 700–800 Hz — roughly two octaves above the 200 Hz lower limit of this method.

 

[peniku8]

For reference, I usually apply a 5ms time window, and I merge nearfield and farfield responses around 700–800 Hz — roughly two octaves above the 200 Hz lower limit of this method.

I personally find the resolution of a 5ms gate time to be unacceptably poor below 1KHz. I know that all too well because this is what you typically achieve with a 1m measurement distance in a residential room, which is the data I struggle with when I want to capture quick and easy spin at home. 5ms gating looks like 1/6th /octave smoothing at 1KHz and the resolution halves for each octave going down, which gets really rough really quickly. This can give you surprisingly good directivity data, but the absolute frequency response is not gonna be very accurate. And of course you can't capture (useful) directivity data with nearfield measurements.

Here is an example of measurement resolution for a measurement at 1 meter, indoors with 5ms gating and at 3m free field (with 25ms gating):

1m:

3m:

The overall measurement isn't too bad at 1m, even down to 300Hz, but the resolution of the data starts getting visibly lower around 2-3KHz already. Below 1K its very smoothed already and smoothing like this can obscure issues of lesser performing speakers as this.
Getting good data at low(-ish) frequencies is just a royal PITA. If it wasn't, there wouldn't be a market for anechoic chambers or the NFS

 

[PristineSound]

Proud owner of the KH-120 II DSP

Stunning speakers.

I wish, I had more budget to get the 150 instead.

 

[Nuyes]

I personally find the resolution of a 5ms gate time to be unacceptably poor below 1KHz. I know that all too well because this is what you typically achieve with a 1m measurement distance in a residential room, which is the data I struggle with when I want to capture quick and easy spin at home. 5ms gating looks like 1/6th /octave smoothing at 1KHz and the resolution halves for each octave going down, which gets really rough really quickly. This can give you surprisingly good directivity data, but the absolute frequency response is not gonna be very accurate. And of course you can't capture (useful) directivity data with nearfield measurements.

Here is an example of measurement resolution for a measurement at 1 meter, indoors with 5ms gating and at 3m free field (with 25ms gating):

1m:

3m:

The overall measurement isn't too bad at 1m, even down to 300Hz, but the resolution of the data starts getting visibly lower around 2-3KHz already. Below 1K its very smoothed already and smoothing like this can obscure issues of lesser performing speakers as this.
Getting good data at low(-ish) frequencies is just a royal PITA. If it wasn't, there wouldn't be a market for anechoic chambers or the NFS

That’s why, when I publish CEA-2034 plots, I only provide the data above 1 kHz for everything except the On-Axis and DI curves. It’s a compromise I make with full awareness of the limitations of my measurement method.

Likewise, I share both normalized and raw contour plots, because I don’t want either myself or readers to draw conclusions based solely on normalized data, which is inherently dependent on the on-axis response.

The same reasoning applies to why I measure THD and include both CHD and EHID results, even though my setup is not an anechoic chamber or an NFS.

While no single measurement is perfect, I believe that by looking at multiple plots from different perspectives, we can still obtain meaningful insights. More importantly, all of my data is produced under consistent conditions by the same reviewer — and I consider that consistency the top priority.

Please also see the link below.

'AsciLab A6B Review'

Jan 17, 2025

I sincerely apologize for the significant delay in posting this review. Thank you to everyone who patiently waited.

This review marks a milestone, as it is the first time I’ve measured a total of 11 samples simultaneously, thanks to the collaboration of the manufacturer and individual owners.
It was a monumental project for me personally.


Let’s dive in.

Impedance​

Frequency Response​

The curve is generally flat, with the bass roll-off extending to...

• Nuyes
• a6b ascilab
• Replies: 104
• Forum: Speaker Reviews, Measurements and Discussion

• 

'AsciLab C5B Review.'

Nov 24, 2024

Today, I’ll be reviewing the AsciLab C5B, which features a 5.25-inch woofer along with two passive radiators.

For this review, I measured one official sample provided by the manufacturer and another pair owned by an individual who had purchased the product personally.






Impedance

The impedance drops as low as approximately **2.2 ohms** at its minimum.

---




Frequency Response

Before delving into the frequency response graph, there’s an important matter to address.

As I reside in the same country (South Korea) as...

• Nuyes
• ascilab c5b
• Replies: 197
• Forum: Speaker Reviews, Measurements and Discussion

• 

 

[peniku8]

View attachment 472100

That’s why, when I publish CEA-2034 plots, I only provide the data above 1 kHz for everything except the On-Axis and DI curves. It’s a compromise I make with full awareness of the limitations of my measurement method.

I overlooked that - good idea! Maybe worth considering to make the contour plots black and white below 1KHz to indicate the limitations of the accuracy in the data aquisition. Or add some sort of dividing line or something.
And I agree with you - if the setup is constantly changing, you can't compare anything, so each test becomes less and less valuable. I have observed this with CEA-2010 data, where multiple reviewers compared their data with the manufacturer's and the total spread was like 4-5dB iirc. I do not remember who posted the write-up, but it was quite interesting. Despite everybody using the same testing methodology and similar testing environments (outdoor GP), the results were very different.

 

[pma]

So rather than debating which method is ‘better’ or ‘more correct,’ I think it’s more productive to understand the strengths and limitations of each approach

Well said.