• 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

amirm

Founder/Admin
Staff Member
CFO (Chief Fun Officer)
Joined
Feb 13, 2016
Messages
26,894
Likes
63,388
Location
Seattle Area
#1
A few weeks ago I reviewed the Neumann KH80 DSP powered monitor. The measurements while very good, did not match manufacturers published graph which showed dead flat on-axis response. Third-party measurements had shown the same so objection was raised that my measurements must be in error. I had to send that sample back to its owner but as it happens, another member had offered his at the same time. Having another sample allows us to both test sample to sample variations and test the hypothesis put forward for difference in measurements.

First, let's review what we had in the review:



And here is the manufacturer "interpolated" response:

Neumann KH 80 DSP Monitor Active Studio Pro Speaker Anechoic Measurements.png


The small graph compresses differences due to lack of (display) resolution. That aside, we have two differences to investigate:

1. There is no low bass dip a I show.

2. Their response is flatter than what I measured.

Dealing with the second factor first, observation was made that the manufacturer states that the "acoustic center" of the speaker is not the usual tweeter axis but half-way between tweeter and woofer:

1581125481430.png


The hypothesis around bass dip was that I had pushed the speaker too hard in level and therefore it must have limited the low frequency response. This, despite the fact that there was no visible indication from the speaker that it was doing so.

New Measurements
To address both of these factors, I set the levels much lower then before and for the first pass of measurements, I set the reference to half-way between the two drivers which landed it around the edge of the woofer as they show.

Before I show you that, let me show the in-room response:
Neumann KH 80 DSP Monitor #2 Fundamental + Harmonic distortion components.png


This graph shows the proper measured SPL which is around 85 dB. For some reason when the CEA graphs are calculated for powered monitors, the levels are incorrect. I know what is causing it but don't know why. I will be contacting Klippel on that. For now, accept that the levels were very reasonable and not loud at all. Indeed, the levels were so low that some noise got into the measurements around 20 to 30 Hz which is not material.

Note that the above has a vertical dimension of almost 120 dB so that flattens the response.

For completeness, there is the distortion data as a percentage:

Neumann KH 80 DSP Monitor #2 Fundamental + Harmonic distortion percentage Audio Measurements.png


Now let's look at the spin data:

Neumann KH 80 DSP Monitor Active Studio Pro Speaker Spinorama CEA CTA-2034 Audio Measurements.png


Clearly this is not the acoustic center as one would define has having the flattest response. The droop in high frequencies clearly indicates we are not on-axis. We can confirm that by looking at the vertical dispersion:

Neumann KH 80 DSP Monitor Active Studio Pro Speaker Spinorama CEA CTA-2034 Vertical Directivit...png


The highlighted red graph is showing the "on-axis" as measured. The graphs above it are at -10 and -20 degree so they are more "correct." Translating, following the Neumann recommendation has clearly messed up the measurements.

As to bass dip, we still have that. Note that the new graphs are the more proper 50 dB vertical response which the original review did not follow. So there is more roughness still in there than the review data. This is the CEA-2034 recommendation so what we see, is what is required.

As a sanity check, I select the highest measured frequency of 20 kHz and plot that. Usually this is a clean output with just the tweeter "beaming" (narrowing dispersion). That is NOT what I see in this setup:

Neumann KH 80 DSP Monitor Active Studio Pro Speaker Balloon Measurements.png


There are tons of sound sources mixing up to generate this complex soundfield. I can't explain it all. Maybe there are multiple edge differections we are seeing from around the tweeter and woofer rings.

Re-measurements Using Tweeter Axis
Wondering that there may be sample to sample variations, I remeasured the same unit without changing anything but the reference axis to be at the center of the tweeter (as reasonably as I could eyeball it).

Here is our spin data now:

Neumann KH 80 DSP Monitor Active Studio Pro Speaker Spinorama CEA CTA-2034 Audio Measurements.png


This is more reasonable and nearly mirrors the measurement shown in the original review.

Here are the two new measurements side by side:
Neumann KH 80 DSP Monitor Active Studio Pro Speaker Compared to Acoustic Center Measurements.png


It is pretty clear that if Neumann says the speaker has flat response, its acoustic center can't at the same time be where they say it is. One or the other is wrong. I may have an explanation for this though. See later.

Here is our 20 kHz balloon soundfield again:

Neumann KH 80 DSP Monitor Active Studio Pro Speaker Balloon Measurements.png


This is much more correct. We have a single cone now. I am not clear why it is pointing down. Maybe the speaker base is not level? Or drivers are not 100% perpendicular to it? Regardless, the overall picture looks correct so I tend to trust this new measurement with reference axis being the tweeter than below it.

Discussion
Addressing the low frequency measurements, I think most measurements outside of what I am using with Klippel NFS, have some kind of correction for bass region. They do this either because the anechoic room is not anechoic that low, or they are splicing two measurements together. Here is Sound and Recording note on how they measured low frequency response:

1581126890206.png

Neumann have zero detail so likely have similar restriction. A calibration curve is applied that may be causing it to push up the response a bit versus reality.

Klippel NFS' claim to fame is the ability to be even more accurate than anechoic chambers so if I were to split the tie, I would split it in favor of our measurements being correct. Then again it is possible that my measurement mic has a slight error there. Either way, it is best to not scrutinize these graphs at such close levels.

As to the rest of the response, unfortunately S&R's graph is also of poor resolution:

1581127075306.png


Lots of sins can be buried in low resolution graphs like that.

On the other side of the coin, there is a potential source of variation in Klippel NFS due to reflections from the fixture at higher frequencies.

With respect to acoustic center, I think this is one of those instances where near-field measurement may cause some issues. Specifically, the Klippel system measures the speaker at 1 foot/0.3 meter or so. This heavily accentuates angular differentials. In anechoic chambers attempt is made to measure in the far field by using 2 meter measurement distance or so. This lowers the accuracy of the acoustic center since small differences in location of the microphone there, doesn't make as much angular difference.

Summary
The new data shows that the original measurement data was correct. The bass dip is not level sensitive. And changing the reference point makes the KH80 speaker look a lot worse, not better.

Ultimately I think both anechoic and NFS measurements have their own slight slants. The two measurements may never 100% sync. The "good news" is that there is hardly any anechoic data out there for speakers whereas I hope to build a large database of measurements using NFS system. So as long as we are consistent, our relative scores should hold.

Above all, let's focus on overall picture of these measurements and not 1 or 2 dB here and there.

------------
As always, questions, comments, recommendations, etc. are welcome.

The pink panthers sat this review out because of all the arguing over the last one. One of them may actually need therapy sessions as a result of that ordeal. Please donate a bit of money so I don't have to pay for that out of my own pocket using: https://www.audiosciencereview.com/forum/index.php?threads/how-to-support-audio-science-review.8150/
 

napilopez

Major Contributor
Joined
Oct 17, 2018
Messages
1,130
Likes
3,331
Location
NYC
#7
First of all: thank you for remeasuring. I think it means a lot and I appreciate it after things took such a bad turn in the original thread! This brings some fascinating insight.

On the other side of the coin, there is a potential source of variation in Klippel NFS due to reflections from the fixture at higher frequencies.

With respect to acoustic center, I think this is one of those instances where near-field measurement may cause some issues. Specifically, the Klippel system measures the speaker at 1 foot/0.3 meter or so. This heavily accentuates angular differentials. In anechoic chambers attempt is made to measure in the far field by using 2 meter measurement distance or so. This lowers the accuracy of the acoustic center since small differences in location of the microphone there, doesn't make as much angular difference.

Summary
The new data shows that the original measurement data was correct. The bass dip is not level sensitive. And changing the reference point makes the KH80 speaker look a lot worse, not better.

Ultimately I think both anechoic and NFS measurements have their own slight slants. The two measurements may never 100% sync. The "good news" is that there is hardly any anechoic data out there for speakers whereas I hope to build a large database of measurements using NFS system. So as long as we are consistent, our relative scores should hold.

Above all, let's focus on overall picture of these measurements and not 1 or 2 dB here and there.
This is really interesting. It somewhat lowers my trust in the klippel system as being an ultimate, do it all solution, because I believe sound captured at a distance *is* the real sound, whether that's anechoic or gated(edit: within their reasonable operating ranges at higher frequencies). That high frequency dip in the last two octaves is really quite puzzling.

But this isn't necessarily a bad thing - it's good to know the limitations of any system. You're right that in the grand scheme of things, focusing on a dB here or there doesn't matter that much. Given the dramatically different results from measuring at the manufacturer-stated reference axis (far worse than simply measuring from the tweeter). Seems to me like it might be best to continue to measure from the tweeter axis for more accurate results on the klippel going forward. Currently I find myself more inclined to trust the Klippel with regards to bass, an anechoic chamber on the treble. But of course we have very little anechoic data in the first place.

One question @amirm, I remember some klippel document mentioning the response can be extrapolated to different listening distances. Does changing that affect these measurements at all? Just curious.
 
Last edited:

Blumlein 88

Major Contributor
Forum Donor
Joined
Feb 23, 2016
Messages
8,993
Likes
11,656
#8
1000 thank you's for doing a 2nd sample of this speaker. As I hoped if this occurred, while it raises a few questions, it answers many more, and bolsters your results using the Klippel NFS system. While also expanding our understanding of both this speaker and the Klippel results.
 
OP
amirm

amirm

Founder/Admin
Staff Member
CFO (Chief Fun Officer)
Joined
Feb 13, 2016
Messages
26,894
Likes
63,388
Location
Seattle Area
Thread Starter #9
It somewhat lowers my trust in the klippel system as being an ultimate, do it all solution, because I believe sound captured at a distance *is* the real sound, whether that's anechoic or gated.
Far field is good. But that is NOT what anechoic measurements typically show. Due to size limitation and modes in the anechoic chamber, measurements are usually NOT in far field but somewhere between it, and near field. In contrast, the Klippel NFS measurements properly show the true far field sound radiation. The larger the speaker, the worse the typical anechoic measurement in this regard.

If you are performing far field measurements outdoor using gating, then you have signal to noise ratio, temperature and humidity variations.
 

LTig

Major Contributor
Forum Donor
Joined
Feb 27, 2019
Messages
1,930
Likes
2,789
Location
Europe
#12
I agree with @StevenEleven's sentiment. Thanks, Amir, for your coolness and professionalism in returning to this speaker. And the results are indeed very interesting.
Me too. Many thanks. Seems the Klippel NFS is in relation to measurements in anechoic rooms similar to the AP555 in relation to the RME ADI-2 PRO fs. It just digs deeper into the details. Of course you pay for it.
 
Joined
Aug 30, 2017
Messages
20
Likes
16
#13
Great test Amir

I was thinking that the JBL 305pmkii has also the acoustic axis in the center distance of tweeter and woofer

why the 120 dB scale that’s cheating
 
OP
amirm

amirm

Founder/Admin
Staff Member
CFO (Chief Fun Officer)
Joined
Feb 13, 2016
Messages
26,894
Likes
63,388
Location
Seattle Area
Thread Starter #14
why the 120 dB scale that’s cheating
It has to simultaneously show the output of the speaker and distortion products which are at very low level.
 

LeftCoastTim

Active Member
Forum Donor
Joined
Apr 15, 2019
Messages
208
Likes
417
#16
It's very impressive to me that all the small deviations / ripples in frequency response are the same between the two examples up to 10Khz. So Neumann's unit-to-unit variance control is as good as they promise. The slightly sagging bass suggests to me that there is a systemic error there, likely Neumann not having very good data in that region.

If I'm to make a bold prediction, I would say that these Neumann's (and likely their other speakers too) are one of the flattest measuring, and most consistent speakers ASR will see at any price. Kudos to Neumann.

I just wish Sennheiser would be as transparent as Neumann (their own subsidiary) about their specs and target frequencies.
 

JohnYang1997

Major Contributor
Joined
Dec 28, 2018
Messages
3,247
Likes
4,071
Location
China
#17
Ok, no work for me; the measurements are so very close. Odd about the stated reference axis not being correct. I wonder if Neumann now knows the bass is not actually flat, they can adjust the internal DSP of further models.
Well. On the other hand, if anechoic chamber is not correct then there is no way in the world any room is going to reflect the pure dry response. While it would be good to know, it's not that meaningful.
Also a suspecting thing would be chirp signal will measure differently than using noise+average. I had been measure earphones and headphones so in that regard it's ideal for both methods and there's difference between the two. I think @solderdude is probably have something to comment on this too.
 

LeftCoastTim

Active Member
Forum Donor
Joined
Apr 15, 2019
Messages
208
Likes
417
#18
Well. On the other hand, if anechoic chamber is not correct then there is no way in the world any room is going to reflect the pure dry response.
This is not quite true. Anechoic chambers also have to be "calibrated" at <100Hz, unless it is very large (and maybe even then for really low frequencies). So I would trust the math here (used by Klippel) over the "calibrated measurement" in an anechoic chamber.

My 2 cents, of course.

Edit: PS I'm speaking as someone who has experience with difficult physics measurements and how a mathematical model can predict better than an actual measurement under certain conditions. Yes, that's lot of uncertainty there, but at least in a model, you can quantify it and model it out. In the real world, noise of your instruments and calibration could lead to misleading conclusions.
 

JohnYang1997

Major Contributor
Joined
Dec 28, 2018
Messages
3,247
Likes
4,071
Location
China
#19
This is not quite true. Anechoic chambers also have to be "calibrated" at <100Hz, unless it is very large (and maybe even then for really low frequencies). So I would trust the math here (used by Klippel) over the "calibrated measurement" in an anechoic chamber.

My 2 cents, of course.
I didn't disagree that using klippel is more accurate. But how much does it mean in designing a pair of speakers that's not going to be used in any room remotely close to it, and not even anechoic chamber. And we should agree that most rooms have issue in the bass, so I don't see it being very useful to correct the bass reference to Klippel. And again I agree Klippel is more accurate.
 

Wombat

Major Contributor
Joined
Nov 5, 2017
Messages
5,791
Likes
4,493
Location
Australia
#20
How does the Klippel System with mic. at the baffle(position not critical?) correlate with mfrs. measurements with mic. at one metre, 4ft. or further, on mfrs. preferred baffle axis?
 
Top Bottom