• WANTED: Happy members who like to discuss audio and other topics related to our interest. Desire to learn and share knowledge of science required. There are many reviews of audio hardware and expert members to help answer your questions. Click here to have your audio equipment measured for free!

Neumann KH 80 DSP Monitor Review

Status
Not open for further replies.

bobbooo

Major Contributor
Joined
Aug 30, 2019
Messages
1,479
Likes
2,079
Half-truths? Did you really say that? Good grief. :(

By half-truth I meant simply something "that conveys only part of the truth". To be clear, I in no way meant the more specific secondary definition, of something "used deliberately in order to mislead someone". It was a poor choice of words. I've edited it out.
 
Last edited:

bobbooo

Major Contributor
Joined
Aug 30, 2019
Messages
1,479
Likes
2,079
All is not lost. I have taken the level issue at heart and I am calibrating the loudness better per CEA-2034 recommendation.

Glad to hear this. Heated debate and discussion often bring progress in the end.
 
Last edited:

Blumlein 88

Grand Contributor
Forum Donor
Joined
Feb 23, 2016
Messages
20,523
Likes
37,056
I have to say Amir, if you did this one test here in the beginning again, on the axis Neumann suggests, and about 6 db lower in level, it might not be necessary, it might not contribute anything beyond excess precision, but is well worth doing. A couple or three hours of your time very well spent. Because it will shut down any critics, and give some people what they are asking for. So they'll be able to trust you again in the future without all this same turmoil. That seems very much worth your doing as much as it may rub you the wrong way to repeat a measurement you think is of little value. The value is in the minds of others, and not the data in this case.

I'd suggest you do the pointless, because it isn't pointless in regards to this project going forward and being highly convincing to anyone and everyone.
 

bobbooo

Major Contributor
Joined
Aug 30, 2019
Messages
1,479
Likes
2,079
"Science" says objective measurements of speakers approximate what we hear from a speaker as far as preference. No way does this require, or encourages the level of "accuracy" you are asking about.

Yes, the science says the correlation between measurement and preference is approximate (0.86 when using Sean Olive's formula to be exact), but the measurements the correlation formula was based on were highly precise. This means, if you measure speakers to the same standard and precision as Olive and his team did, statistically we will achieve the same correlation using his formula. However, if you measure speakers differently (e.g. with a different SPL or assuming different acoustic axes) or with lower precision (e.g. 10 data points per octave to Olive's 20), then an approximate correlation of 0.86 will be lowered even more. So in order to minimise any further approximation of the correlation (who wouldn't want that?), the same standards and precision of measurement used by Olive should be maintained.
 
Last edited:

dshreter

Addicted to Fun and Learning
Joined
Dec 31, 2019
Messages
794
Likes
1,226
I actually think with collective feedback and learning, it’s possible to make progress and have even more predictive measures than Olive’s formula. Who knows what else might be out there that could be an element of performance that they did not measure?

While it’s certainly an important reference work, it’s not a ceiling, and we have a lot of bright minds contributing here.

Regarding the measurement methodology, I’m sure things will be learned with each subsequent test for a little while until it’s humming along like a commercial laboratory. While this learning takes place, you can either recursively test each speaker so you have absolutely comparable results, or accept that things will get dialed in further over time.

Given this is a project born from pure enthusiasm, I’d suggest moving forward. Retesting the prior speakers seems more like drudgery, and I know there’s a strong desire to see more speakers on the bench. I really sympathize that Amir might not want to redo tedious work.

I’m personally interested in discovering which expensive speakers are crap, and as many “secret” bargains that might exist out there. Rectifying the last dB on the KH80 doesn’t seem necessary to accomplish that.
 

thewas

Master Contributor
Forum Donor
Joined
Jan 15, 2020
Messages
6,747
Likes
16,181
One thing I wonder is what about measuring speakers that the manufacturer recommends using with no toe-in? This is not un-common, especially for budget speakers where the designer knows the buyers won’t use toe-in, Andrew Jones for instance has stated that he does this.
Dali also does that on most of their speakers, on axis they are too bright, a way/trick to have more sound power in that region. Why not measure those with horizontal axis reference tilted at 30° as recommended by the manufacturer to get a more realistic direct sound, listening window and early reflections for their recommended use? Sound power will anyway stay the same and so what if the horizontal angle measurements are not symmetrical anymore, they also aren't on speakers with no vertical symmetry.
 

DJBonoBobo

Major Contributor
Joined
Jan 21, 2020
Messages
1,360
Likes
2,851
Location
any germ
You all guessing and putting trust in other measurements which you have had zero input to, or opportunity to criticise or analyse. Where is the proof that the acoustic center is where the company says it is? You have a fluid-dynamics simulation of it or other instrumentation that proves that? You don't. As it is, the most suspect thing is a frequency response that is dead flat from manufacturer.

Wasn't that exactly the reason why the KH80 was suggested and the whole discussion was held? From my perspective the review missed the chance to really clarify this. Personally i find it disappointing. For casual readers who only read the first post i even think it is misleading. What i get from the review in combination with 20 pages of discussion is: "If you don´t use the KH80 as intended, it´s still pretty good." People who only read post #1 will get: "It´s pretty good, but not as good as the manufacturer say. Sure thing, Neumann wants to make their product look pretty. Good to have an independent reviewer who shows us that it´s really not that perfect. Suspicions have been confirmed."

Anyway, if there is no chance of a repeat, i would suggest measuring another Neumann monitor soon to address this "trust issue". Thanks.
 

thewas

Master Contributor
Forum Donor
Joined
Jan 15, 2020
Messages
6,747
Likes
16,181
And it is not like we have a gold standard where we "know" it is the precise response or the truth as you call it. You all guessing and putting trust in other measurements which you have had zero input to, or opportunity to criticise or analyse.
As said before the measurements are from one of the most serious companies in the segment, also from one of Germanys best acoustics technical university professor and some hobbyist members of this forum. To pose it the opposite way around, how can you doubt them when yours have 2 obvious error sources?

Where is the proof that the acoustic center is where the company says it is?
Same here, why don't you give us the proof that it isn't, by re-doing the measurements as recommended. ;):D
By the way the vertical center is almost never directly at the exact middle of the tweeter at a vertical non-symmetric design (so unlike a coax or M-T-M), as said before from others its just that some speakers are more tolerant there due to their vertical directivity behaviour. If not stated from the manufacturer, I personally would rather place it in the mid distance between the tweeter and the midrange driver as this is more realistic and gives better results for most such speakers.

You have a fluid-dynamics simulation of it or other instrumentation that proves that? You don't.
Neumann monitor designs nowadays are completely done per numerical modelling and simulations, unlike 20+ years ago where the design of their first waveguides was done with several iterations making them with modelling clay and measuring them, I have been twice at their engineering centre which is only 150km from my home, last time this December. Their quality of measurements I think confirms that more than clearly.
 
Last edited:

Krunok

Major Contributor
Joined
Mar 25, 2018
Messages
4,600
Likes
3,065
Location
Zg, Cro
I’m personally interested in discovering which expensive speakers are crap, and as many “secret” bargains that might exist out there. Rectifying the last dB on the KH80 doesn’t seem necessary to accomplish that.

I'm affraid you got it all wrong: setting loudness according to standard and measuring at axis specified by manufacturer is about rectifying the measurement protocol, not "last dB on the KH80".
 

Krunok

Major Contributor
Joined
Mar 25, 2018
Messages
4,600
Likes
3,065
Location
Zg, Cro
Where is the proof that the acoustic center is where the company says it is? You have a fluid-dynamics simulation of it or other instrumentation that proves that?

Finding optimal listening axis is not really a rocket science, but nobody here expects you to do it for every speaker you'll be testting, so the only right thing to do is to use manufacturer spec.

And no, finding optimal listening axis doesn't involve fluid-dynamics simulation. As @NTK explained in his earler post, although that functionality may not be implemented, NFS has all the data needed to reconstruct the acoustic field to any set of coordinates from your first measurement so finding a true optimal listening axis is very doable by software .
 
Last edited:

Costas EAR

Active Member
Forum Donor
Joined
Jan 15, 2020
Messages
157
Likes
348
Location
Greece
The small Neumann is a really great little speaker, with by far the best measurements until now.
So_far.png


Amirm's way of measurement is like a magnifying glass of other types of measurements, nothing less.

Keep on walking, or measuring!

The only strange thing for me, is the listening results, where the Neumann should be by far the best speaker to listen. ;)
 

Juhazi

Major Contributor
Joined
Sep 15, 2018
Messages
1,717
Likes
2,897
Location
Finland
Amir seems to think that when he has the best measurement system, we must stop complaining. Sad. Technical measurement instruments can be accurate and reliable, but humans using them make mistakes and wrong conclusions, or are blind to systematic errors.

The measurements from NFS are precise, and because the system and we want to process that data to be able to show CEA, spinorama etc., it is terribly important that the reference 0 axis is set correctly. This gets emphasized when Amir wants to make goodness quantification from measurements (like Olive's)

Educated readers want to see much of the raw data to make their own conclusions and to find reasons to the behaviour of the speaker. Eg. optimal listening axis is easy to find then, and measurements reveal asymmetry of off-axis response directly (which is a bad thing and cannot be eq'd). Likewise distortion spectrum tells many things, so does nearfield measurement of low bass and port signal, step response, CSD and impedance spectrum for passives. These I want to see all, before making my own assessment!

During 50 years of reading as hobbyist, I haven't found a single magazine/website that can deliver all that I want to know, so I must read 3-5 different tests of same speaker! When designing and making diy speakers, we take hundreds of acoustic measurements in tens of different ways, of the driver units, whole system, port output, cabinet vibrations, air leaks, at various angles and distace, various spl etc. Modern computer aided measurements are cheap, easy and fun! And very difficult to analyze!

Remember the analogy to SINAD number of amps and dacs, do they tell the whole story of amps? Aren't L/R symmetry, quality of controls and other components etc. important too?
 
Last edited:

edechamps

Addicted to Fun and Learning
Forum Donor
Joined
Nov 21, 2018
Messages
910
Likes
3,620
Location
London, United Kingdom
Where is the proof that the acoustic center is where the company says it is?

But… that's neither here nor there. The reference axis, according to CTA-2034-A, is, by definition, where the manufacturer says it is. There's no need to "find" it or to "verify" it. This is about setting up the speaker the way it's intended to be used. To be perfectly precise and phrase this in IEC 60268-2 standards legalese, the reference axis is a rated value (by the manufacturer), not a measured value.

So what if "the acoustic center is not where the company says it is", as you say? Well, you would still measure according to the axis that they provide, and you might get subpar results. Which is perfectly normal, and this time it wouldn't be your fault - it would be the manufacturer's fault for putting the wrong axis in their instruction manual. Their loss.

As it is, the most suspect thing is a frequency response that is dead flat from manufacturer.

Why is it suspect? Given Neumann's philosophy, it seems quite likely to me that they use sophisticated DSP to make the on-axis response ruler flat. They might even do this automatically for every individual unit as it leaves the production line. In my opinion that's a perfectly reasonable thing to do. In fact, people in this forum are suggesting we use your data to generate our own EQs to do exactly this.

With all the measurements we have of the KH 80, including yours, there's every reason to believe the response of the KH 80 actually is ruler flat, if the speaker is used according to the manufacturer's instruction manual (i.e. at a reasonable output level and listening on the proper axis). Problem is, the casual reader will not get that by reading your review, unless they sift through 20 pages of discussion to find the caveats we raised.

Not one thing I have read here has anything to do with increasing the precision of instrumentation.

I agree. If your goal was to see what happens when you measure a speaker on the wrong axis while overloading its woofer, then yes, you achieved that goal to a very high degree of accuracy using very precise instruments. But what's the point?

I don't think anyone here ever claimed the instrumentation (i.e. NFS) was at fault. The discussion is about the measurement conditions/methodology/protocol. The NFS doesn't decide what output level to use or where the reference axis is. You do.

All is not lost. I have taken the level issue at heart and I am calibrating the loudness better per CEA-2034 recommendation.

Yay :D One problem solved, only one to go.
 
Last edited:

Costas EAR

Active Member
Forum Donor
Joined
Jan 15, 2020
Messages
157
Likes
348
Location
Greece
The reference axis, according to CTA-2034-A, is, by definition, where the manufacturer says it is. There's no need to "find" it or to "verify" it. This is about setting up the speaker the way it's intended to be used.
Exactly. ;)

I am using everywhere Neumann 310 in my immersive setup, and speaker placement was done with respect to the provided by the manufacturer acoustic center of the speaker.

It's not my jod to find it or verify it.

With this positioning, all monitors measure excellent at the sweet spot, as promised by the manufacturer. ;)

By the way, the acoustic center of the speakers, was verified by the software used for calibration (by Trinnov audio).
As stated by the manufacturer. ;)

Anyway, measurements provided by amirm, don't give any value to the acoustic center of the speaker, as they have another goal.
This is explained by amirm, no further explanation needed.
 

Pio2001

Senior Member
Joined
May 15, 2018
Messages
317
Likes
507
Location
Neuville-sur-Saône, France
As said before the measurements are from one of the most serious companies in the segment, also from one of Germanys best acoustics technical university professor and some hobbyist members of this forum.

Actually, we have two other measurements : the one from Sound and Recording, and the one from the member ctrl.
The curve from Neumann is not a measurement of an existing device, that they took as it is and put into their measurement jig, and found that it measured flat.

They created the device: they put a prototype into their measurement jig, got an average curve, and then tweaked the design, then got a new curve, then tweaked something else in the prototype etc, until their measurement says its flat. And the KH-80 was born !
The frequency response that they provide is obviously from the same measurement device that was used to make the prototype measure flat. It is the frequency response after they have tweaked the speaker so that it looks ruler flat according to their own measurements.
 

ctrl

Major Contributor
Forum Donor
Joined
Jan 24, 2020
Messages
1,616
Likes
6,086
Location
.de, DE, DEU
Actually, we have two other measurements : the one from Sound and Recording, and the one from the member ctrl.
Was not me, in Post#319 edechamps posted the measurement of his KH 80. I only put the Neuman, S&R, Amir measurements together in one diagram.

Maybe Amir, as customer of Klippel, could ask for their measurements of the KH 80 (or others) to compare the accurance of his system. It seems they used Neumann speaker for testing too. As posted by others before, the margin of error guaranteed by the manufacturer is very small.
The tolerances were provided earlierin this thread and awware on the Neumann site: 100 percent of speakers produced hit - +/-0.26dB from 100-10Khz, q50 percent 0.17dB. Each speaker is checked for calibration as one of the final steps.
index.php

The frequency response linearity is given as +/-0.7dB from 100 to 10KHz.
Klippel_neumann.jpg
 

thewas

Master Contributor
Forum Donor
Joined
Jan 15, 2020
Messages
6,747
Likes
16,181
They created the device: they put a prototype into their measurement jig, got an average curve, and then tweaked the design, then got a new curve, then tweaked something else in the prototype etc, until their measurement says its flat. And the KH-80 was born !
That's not exactly how design of louspeakers is done nowadays at high-tech companies like Neumann, Genelec, JBL etc., their enclosure and waveguide are designed, simulated and optimised on a PC, same as the chassis, which are also then built and measured after in a standardised environment like for "infinite baffle" and than they tweak the crossover also with simulations of both which can very well predict the acoustic response. Total measurements are done at a later stage with the first prototypes to confirm the simulations and to final optimise the crossover for the small deviations from simulation and measurements.
 

napilopez

Major Contributor
Forum Donor
Joined
Oct 17, 2018
Messages
2,109
Likes
8,424
Location
NYC
Studio monitors might be a slight exception since the direct sound is very important and I could see someone wanting to place them perfectly for that situation. I'm referring more to home setups where most of us place speakers somewhere close to tweeter height at ear level. Now as long as the reference axis is specified to be somewhere between the midrange and tweeter axis then sure go ahead and use the manufacturer's recommendation, my problem is specifying something that isn't a normal use case just to produce a better graph. My example of the M126 be is the only speaker I've seen that deviates from somewhere close to the tweeter axis and in that case I would measure along the tweeter's axis because that is more accurate to how people will be using them. Revel also specifies that they be placed with the tweeter near ear height in the manual, so it's odd that they want them measured 7" above the tweeter. So I say if a manufacturer specifies a reference axis then use it, otherwise just pick something like the tweeter axis to be consistent.

And no we should be measuring all around the speaker but people who don't listen on-axis can safely disregard the On-axis curve and focus on the listening window, which is much more useful for most of us.

Do not disagree with any of this. The reason the reference axis was in contention with the Neumann was that it is very explicitly specified in the manufacturer instructions, and it is not at all an unusual location =]

You're right though. General procedure is to measure at the tweeter axis unless otherwise specified. Where things have sometimes gotten complicated for me is when the tweeter is placed asymmetrically on the baffle.

Actually, we have two other measurements : the one from Sound and Recording, and the one from the member ctrl.
The curve from Neumann is not a measurement of an existing device, that they took as it is and put into their measurement jig, and found that it measured flat.

They created the device: they put a prototype into their measurement jig, got an average curve, and then tweaked the design, then got a new curve, then tweaked something else in the prototype etc, until their measurement says its flat. And the KH-80 was born !
The frequency response that they provide is obviously from the same measurement device that was used to make the prototype measure flat. It is the frequency response after they have tweaked the speaker so that it looks ruler flat according to their own measurements.

I still think this is a very bizarre argument =] Aside from what @thewas_ said, it doesn't really matter if their measurements are flat because Neumann were aiming for flat.

A myriad of good speaker designers aims for ruler flat, especially pro products. I mean, that's what JBL is doing too in their monitors. So why aren't JBL's monitors as flat in their own measurements until you get to the M2? After all, they created the speakers. Shouldn't the 305P and 705P be flat too?

You could argue that that's because they wanted to save their best for their most expensive, but if so, that's another thing I appreciate about the neumann - the KH80 is actually their flattest speaker, despite being their cheapest. So that reasoning wouldn't hold.

(Edit: to be clear, it's not like I think being 100 percent flat even matters that much! It's about the discrepancies).

Point is, it's not that hard to make a speaker with a perfectly flat on-axis graph. All you need to do is some basic EQ in DSP. Make yourself some filters in REW. If manufacturers wanted to show a perfectly flat line on-axis, they could all do this. They don't(usually), because major deviations would show up in independent verification.

Moreover, the Neumann measurements are independently verified by sound and recording, and arguably my own. Though my measurements don't have quite the same tolerances because they were measured in a living room with three animals sometimes running around (lol) and the gate and splice method, they show the same low-Q tendencies.

Which as @edechamps pointed out, is crucial. I want to be clear that this is what those of us with concerns about these measurements were worried about before we learned about the SPL levels and reference axis. I doubt anyone cares about tiny differences like whether this line is more squiggly than that. That doesn't affect sound quality and can easily be explained by equipment, noise, procedure, tiny reflections in your setup, etc (within reason). It's the low-Q deviations that stood out to those of us familiar with this speaker because those would be the most audible issues.

Just seems weird to imply that just because a manufacturer provided a frequency response graph, it can't be accurate or useful or reliable . Obviously sometimes it's not, but we have no reason to believe that's not the case here - it's been independently verified.

Anyway... I'm glad we finally have another speaker to move on to :p. And I really appreciate Amir is taking the level issue to heart, as was noted in the new JBL 104 review.
 
Last edited:

ctrl

Major Contributor
Forum Donor
Joined
Jan 24, 2020
Messages
1,616
Likes
6,086
Location
.de, DE, DEU
And importantly mislead people into thinking there is some need for high precision in these electro-acoustic measurements.
For some (clearly not all) audio engineers it is important. They use studio monitors as a tool and want best precision (and linearity) in direct and diffuse sound field. For them +-0.5dB makes a difference.

But I have to admit, if the measured sample of the KH 80 is really, really bad, Amir's measurements would fit in the manufactures tolerance boundaries, if "strong" frequency smoothing is used.

Frequency response linearity is given as +/-0.7dB from 100 to 10KHz (by Neumann). The NFS has +-0.1dB tolerance (direction of of max SPL).

Again measurements by Amir (ver -10deg, to fit manufacturer recommendation), Sound&Recording and Neumann (most inaccurate curve from their website).

First measurements without further smoothing:
KH80_diff_tolerance_noSmooth.jpg

Now with additional 1/3oct. smoothing:
KH80_diff_tolerance_1-3oct.jpg
 
Status
Not open for further replies.
Top Bottom