• 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!

Understanding How the Klippel NFS Works

Three quality sources Neumann (full anechoic probably), ASR (NFS) and S&R (hemi anechoic plus nearfield for bass) showing KH80 within 1 or 2dB on axis.

Empirical data-gathering is of course at the heart of the scientific method and is the only way to prove a theory. It does require proper documentation for meaningful interpretation however.

What we need to properly interpret these graphs and draw conclusions is documentation on measurement hardware (frequency response, signal to noise), measurement environment (reflections, ambient noise, ..), measurement method (full space, groundplane, far field / near field, ...?), software (especially frequency resolution), and possibly more.

The appointment of well known names as "quality source" is irrelevant. No source of data is always automatically above doubt, and suggesting that is questionable and goes contrary to the scientific method.


All in all, I'd really like to see some more properly documented comparisons of NFS vs. free field measurements. I've seen these, I've seen Erin's.. well, they mostly match, which itself suggests accuracy - if multiple sources of data yield the same result, statistics would imply it's probably not because they all record the same mistake, but that it's more likely they all record correctly. But I'd still really like more documentation. No offense, but for example, most of these measurements seem to be smoothed / have pretty low resolution, which by itself complicates interpretation / can hide differences.
 
Last edited:
Empirical data-gathering is of course at the heart of the scientific method and is the only way to prove a theory. It does require proper documentation for meaningful interpretation however.

What we need to properly interpret these graphs and draw conclusions is documentation on measurement hardware (frequency response, signal to noise), measurement environment (reflections, ambient noise, ..), measurement method (full space, groundplane, far field / near field, ...?), software (especially frequency resolution), and possibly more.

The appointment of well known names as "quality source" is irrelevant. No source of data is always automatically above doubt, and suggesting that is questionable and goes contrary to the scientific method.


All in all, I'd really like to see some more properly documented comparisons of NFS vs. free field measurements. I've seen these, I've seen Erin's.. well, they mostly match, which itself suggests accuracy - if multiple sources of data yield the same result, statistics would imply it's probably not because they all record the same mistake, but that it's more likely they all record correctly. But I'd still really like more documentation. No offense, but for example, most of these measurements seem to be smoothed / have pretty low resolution, which by itself complicates interpretation / can hide differences.
You're making a lot of general comments.

The only source here which is in partial question is the Neumann data, where the conditions are not known. However S&R describes the approach and so does ASR. It's been demonstrated many times over that the agreement is good.

The onus is on you to do that research if you haven't yet.
 
You're making a lot of general comments.

The only source here which is in partial question is the Neumann data, where the conditions are not known. However S&R describes the approach and so does ASR. It's been demonstrated many times over that the agreement is good.

The onus is on you to do that research if you haven't yet.

It seems reciting the very basics of scientific research was / is necessary in this case. You can't just throw data into a room and claim "that is so", or even just vaguely name the source and expect the other party to do the rest.
There are some basic principles for the gathering of scientific facts that have been established for centuries, that every one of us is taught in basic mandatory education. The corresponding Wikipedia article covers the whole topic nicely imo: https://en.wikipedia.org/wiki/Scientific_method

Documentation is necessary for proper interpretation, as well as the possibility of replication / reproduction. Providing documentation is always the responsibility of the party that presents data. When presenting measurements, one has to list the conditions of the measurement, or clearly and obviously name / reference the source where full documentation can be found.

"Research it yourself" or "everyone knows this" may be common and convenient ways / attempts to get rid of that responsibility and shift the burden of proof to the other party in general society, but have no validity in scientific circles. Same goes for publicizing data / texts from commercial works, like books, or in this case the magazine "Sound & Recording", without the proper context, requiring the other party to purchase the work if they'd like to know more. Publication of data for scientific discussion is allowed under Intellectual Property rights within the Fair Use clause, and this includes documentation / context of data - so again, it is your responsibility to provide it.
 
It seems reciting the very basics of scientific research was / is necessary in this case. You can't just throw data into a room and claim "that is so", or even just vaguely name the source and expect the other party to do the rest.
There are some basic principles for the gathering of scientific facts that have been established for centuries, that every one of us is taught in basic mandatory education. The corresponding Wikipedia article covers the whole topic nicely imo: https://en.wikipedia.org/wiki/Scientific_method

Documentation is necessary for proper interpretation, as well as the possibility of replication / reproduction. Providing documentation is always the responsibility of the party that presents data. When presenting measurements, one has to list the conditions of the measurement, or clearly and obviously name / reference the source where full documentation can be found.

"Research it yourself" or "everyone knows this" may be common and convenient ways / attempts to get rid of that responsibility and shift the burden of proof to the other party in general society, but have no validity in scientific circles. Same goes for publicizing data / texts from commercial works, like books, or in this case the magazine "Sound & Recording", without the proper context, requiring the other party to purchase the work if they'd like to know more. Publication of data for scientific discussion is allowed under Intellectual Property rights within the Fair Use clause, and this includes documentation / context of data - so again, it is your responsibility to provide it.
It's your lack of familiarity that's apparent, a lack which you position as everyone else's (or my) problem. This entire thread is about the operation of the NFS. For a single quick example of validation of ASR results I provided extracted on axis data from the hemianechoic measurmements of S&R, and Neumann's own. The results are very close. There is no smoothing applied. If you want to understand measurement conditions of S&R you have explanatory articles published by Dr. Anselm Goertz, who is responsible for them. For Neumann's I haven't found a good source explaining conditions, but they are among the most transparent manufacturers when it comes to sharing performance data. For ASR, this is one of the best threads on the subject, and if you want more for the KH80 specifically you can read the not one but three threads dedicated to it alone.
 
All in all, I'd really like to see some more properly documented comparisons of NFS vs. free field measurements.
My Neumann and Genelec speaker measurements are compared to both company's anechoic data. I don't publish the reviews without this sign off. Both companies have per-sample measurements which makes this comparison very feasible. There is tiny amount of disagreement here and there which neither one of us can explain. It could be their error, mine, or both. Either way, it is very small deviation of no material impact. We are talking about speakers after all.

Satisfying some member's skepticism is beyond the scope of what I do. You are welcome to have your doubts and ignore the measurements. But please don't make it sound like you have data that indicates there is an issue here. There is none.
 
Documentation is necessary for proper interpretation, as well as the possibility of replication / reproduction. Providing documentation is always the responsibility of the party that presents data. When presenting measurements, one has to list the conditions of the measurement, or clearly and obviously name / reference the source where full documentation can be found.
Level of documentation depends on application. A defense contractor may have to produce hundreds of pages of documentation on some system. They get paid to produce such documentations that may have to comply with many laws and standard. Likewise someone researching into a new field, would need to have extensive documentation and database of results.

What we do here is none of that. Heck, companies can't even be bothered to product a simple frequency response. My reviews on the other hand, are very comprehensive and look at the speaker performance from multiple vantage points. Critical test conditions such as measurement axis is documented. And when weather gets cold, even measurement temperature is supplied. If I sat here and wrote a book to go with these, I would get nothing else done. Engineering is about optimization of resources. My time and equipment time is not free to fill the needs of someone who just wants to create FUD. I rather have 200 speakers measured per year this way, than 10 done your way. You disagree? You start a venture doing it your way.
 
All in all, I'd really like to see some more properly documented comparisons of NFS vs. free field measurements.
To get this information you only have to browse the KH80 Thread No.3.

If you want to get the comparison of Amir's NFS measurement compared to different sources right away, you can find a summary here.
There I compare the measurement of the KH80 from Amir (newest and older measurement via NFS) with the measurements from Neumann and S&R (both use anechoic chambers to my knowledge).

The deviations in the measurements depend strongly on the smoothing used. Amir's unsmoothed measurements of the KH80 are in the range +-1dB compared to Neumanns own smoothed measurement of the KH80 (see link second image).

Here's the comparison (Amir's measurement with 1/6oct smoothing to make it more fair, both other measurements use smoothing - I don't use smoothing in the link).
It was normalized to the manufacturer's measurement of the KH80 and then compared to the S&R measurement in an anechoic chamber and to Amir's NFS measurement.

1664580786009.png
 
It's your lack of familiarity that's apparent, a lack which you position as everyone else's (or my) problem. This entire thread is about the operation of the NFS. For a single quick example of validation of ASR results I provided extracted on axis data from the hemianechoic measurmements of S&R, and Neumann's own. The results are very close. There is no smoothing applied. If you want to understand measurement conditions of S&R you have explanatory articles published by Dr. Anselm Goertz, who is responsible for them. For Neumann's I haven't found a good source explaining conditions, but they are among the most transparent manufacturers when it comes to sharing performance data. For ASR, this is one of the best threads on the subject, and if you want more for the KH80 specifically you can read the not one but three threads dedicated to it alone.

Again, claiming general knowledge or deflecting the burden of proof are not valid methods in scientific circles (which this website / forum implies by its very name).

If a person cannot provide clear concise well documented evidence, it is almost always because they have none. I strongly assume you cannot provide said documentation, like Anselm's S&R measurement methodics. If so, you would have simply quickly linked to them, instead of taking forever to try to explain the topic away.

As I've stated myself first, close correlation of multiple sets of data itself implies accuracy. But this is a simplification, and it's not actually even necessarly about accuracy. Different measurement methodics and environments, even hardware and software, do by definition produce different results. If we would know them, we could associate them - "this measures differently because this and this".

As for smoothing, you don't even have to apply dedicated smoothing (which actually seems to have been done, as I suspected, as crtl just indicated) for the measurement to have a different resolution. Varying FFT sizes alone make a huge difference. For example, load up a measurement in ARTA, switch between the default 4K and let's say 32K FFT, and compare the difference.


Level of documentation depends on application. A defense contractor may have to produce hundreds of pages of documentation on some system. They get paid to produce such documentations that may have to comply with many laws and standard. Likewise someone researching into a new field, would need to have extensive documentation and database of results.

What we do here is none of that. Heck, companies can't even be bothered to product a simple frequency response.

Agreed. I actually explicitly specified the level of documentation that would be needed, and it wouldn't be something that fills a few hundred pages; probably not even one. Problem is, Curvature provided none, except vaguely naming the source.

Also agreed on the problem of manufacturer data. I have probably seen hundreds, but never seen one single properly documented manufacturer measurement. The most you're getting in some cases is information like "2,83V, 1m half space", which is better than nothing, but tells us nothing about the measurement environment and its influence etc. - for example, simply stating "anechoic" tells you next to nothing, since every anechoic chamber is different in size, length, shape, material of its damping elements, etc., and has a different lower "corner frequency".

As for the rest of your statements that I haven't quoted, please don't put words into my mouth. Never at one point did I criticize or even refer to your measurements or their accuracy, or even the Klippel NFS' accuracy or a supposed lack thereof.

We could simply conclude that these measurements are within a reasonable margin of error and that it's probably all okay and simply let it rest at that. That's probably okay for most purposes. Again though, we are 11 pages into a thread full of intricate technical discussion in a forum that names itself AudioScienceReview - and suddenly we're getting too technical? Yeah, right. But, if you as the owner say so, I am happy to not get further into the discussion.

For the record, I / we, in a German DIY community, are currently organizing a measurement comparison with a fixed DUT with multiple participants, that includes an NFS, free field measurements, and many more . Here's a timelapse of our DUT in the NFS:


If all goes well we'll arrive at some 10 data sets. We have some preliminary results so far. Once the project is complete I think it would be interesting to compare and diagnose the results vs. those presented here, less in a sense of "right vs wrong", but "which factor / variable influences the measurement in which way". Even the size and shape of the DUT alone might play an important role. Like someone else stated in the course of this thread, also, there hasn't been a single KH80 tested here, but multiple devices - series deviation / production spread is a reality. Etc. The more (well documented) data is available, the more accurate the conclusions that can be drawn.

BUT, again, I am happy to stop at this very point, if this kind of technical discussion and sharing of data is not wanted. I am posting in various other audio forums and am happy there. I do not have to post here.


To get this information you only have to browse the KH80 Thread No.3.

If you want to get the comparison of Amir's NFS measurement compared to different sources right away, you can find a summary here.
There I compare the measurement of the KH80 from Amir (newest and older measurement via NFS) with the measurements from Neumann and S&R (both use anechoic chambers to my knowledge).

Thank you! As usual, Anselm is very thorough in his documentation:

"Der Klasse-1-Messraum erlaubt Messentfernung bis zu 8 m und bietet Freifeldbedingungen ab 100 Hz aufwärts. Alle Messungen erfolgen mit einem B&K 1/4″-4939-Messmikrofon bei 96 kHz Abtastrate und 24 Bit Auflösung mit dem Monkey-Forest Audio-Messsystem. Messungen unterhalb von 100 Hz erfolgen als kombinierte Nahfeld-Fernfeldmessungen."

It would be relevant to name the method for SPL matching of the near field measurements of the low frequency sources (membrane and port). In the original documentation of this measurement method, Low-Frequency Loudspeaker Assessment by Nearfield Sound-Pressure Measurement, D.B. Keele, AES, 1973, the author names a formula for doing so - square root of sd/sv. Another known method is SPL matching by hand at a frequency of appx. 1/2 fb. I believe I remember though, from another of his tests, or maybe from a personal conversation with him, that Anselm uses the latter - so, question answered.
 
The project detailed in the previous post is still ongoing. An Interim Report however has just been finished. It features acoustic measurements of, as stated, a fixed / unchangeable DUT, by the NFS, as well as other participants employing various other measurement methods - including a free field one.

So, here's another, in this case I'd say pretty well documented (we could have went even more into detail, but the report is already 18 pages long; if anything is not clear, please ask), empirical test / comparison of the NFS:

Interim Report 11/2022 ARTA Round Robin #3 (English, machine translated)
Zwischenbericht 11/2022 ARTA Ringversuch #3 (German / Deutsch) - if you're a native speaker, or for your own translation with the software of your choice
 
The project detailed in the previous post is still ongoing. An Interim Report however has just been finished. It features acoustic measurements of, as stated, a fixed / unchangeable DUT, by the NFS, as well as other participants employing various other measurement methods - including a free field one.

So, here's another, in this case I'd say pretty well documented (we could have went even more into detail, but the report is already 18 pages long; if anything is not clear, please ask), empirical test / comparison of the NFS:

Interim Report 11/2022 ARTA Round Robin #3 (English, machine translated)
Zwischenbericht 11/2022 ARTA Ringversuch #3 (German / Deutsch) - if you're a native speaker, or for your own translation with the software of your choice
For each measurement, where possible, can you please export to csv?
 
The project detailed in the previous post is still ongoing. An Interim Report however has just been finished. It features acoustic measurements of, as stated, a fixed / unchangeable DUT, by the NFS, as well as other participants employing various other measurement methods - including a free field one.

So, here's another, in this case I'd say pretty well documented (we could have went even more into detail, but the report is already 18 pages long; if anything is not clear, please ask), empirical test / comparison of the NFS:

Interim Report 11/2022 ARTA Round Robin #3 (English, machine translated)
Zwischenbericht 11/2022 ARTA Ringversuch #3 (German / Deutsch) - if you're a native speaker, or for your own translation with the software of your choice

Nice job, indeed. A lot of efforts. Thank you for sharing.
 
I think it is best to first define what far-field (and near-field) means. In internet audio fora such as this one, people often use the term "nearfield" to mean something that is completely unrelated to its meaning in acoustics. (See this article by E. Sengpiel)

Briefly, in what most internet audio forum posts, "nearfield listening" means the listening distance is close and the majority of the sound heard is direct sound from the speakers (i.e. direct field listening). The opposite case is "far-field listening", and in this case the majority of the sound heard is from reflections (i.e. reverberant field listening). Here I took from one of Genelec's monitor selection guides and use the model 8350A for illustration.
View attachment 193610

The red zone is the acoustical "near-field". It is where the listening distance to the speaker is too short for its 2 drivers to integrate. In the acoustical near-field, frequency response change with listening distance because of lack of driver integration. Once beyond the red zone, we are in the acoustical far-field. The green zone is where direct sound dominates. The black vertical separator is the critical distance. This is the point the direct sound energy (which decreases with listening distance) and the reflected sound energy becomes the same. The critical distance is room dependent (mostly volume and reflectivity).

Beyond the critical distance, as in reverberant field listening, the sound heard is dominated by reflections and therefore room effects are much more prominent.

In acoustics, far-field means the listener is far enough away from the sound source that the source appears to be compact, and thus behaves like a point source. For a speaker with multiple drivers, that means the sound radiated from the drivers are well integrated at that distance. If you look at the Genelec chart, the far-field for the 8350A begins at ~2.5 m ft. For the 8351B, which has the same SPL output rating as the 8350A, because of its coaxial design, far-field begins at a much shorted distance of ~1.5 m ft. Where farfield begins is determined by the speaker design, and is not room dependent.

In the acoustical far-field, the sound radiating properties of the speaker only depends on distance. If you have measurements at one distance (in the far-field), you can easily scale it to a different distance by using the 1/r relationship (i.e. SPL decreases by 6 dB every doubling of distance). Therefore, there is little difference between the reporting numbers at 3 m and 10 m. They are just a scaling factor offset of each other. The ANSI/CTA2034 standard for anechoic chamber measurement is to measure at 2 m, and report the scaled measurements at 1 m.

[Edit] Corrected the distance unit error. Distances quoted are in ft instead of m.
Doesn’t the red area belong to the near field of the speaker?
1. Speaker far field
2. Room far field
3. Acoustic far field
4. far field of loudspeaker unit
May I consider that there are four concepts in total?
 
Last edited:
Why is the default setting for NFS 10 meters and not 3 or 5 meters?
 
Doesn’t the red area belong to the near field of the speaker?
1. Speaker far field
2. Room far field
3. Acoustic far field
4. far field of loudspeaker unit
May I consider that there are four concepts in total?
Red area in the Genelec chart is the speaker's nearfield.

1 through 4 are misunderstandings.

The only concepts are nearfield and farfield, and they apply to sound sources, like speakers or other objects. Reflections in a room are secondary sound sources, They also have a near and farfield.

Nearfield means chaotic. Physically, this means that the object producing the sound does not radiate coherently, and acts instead like multiple smaller sources.

Farfield means predictable, stable. Once you are far enough away, all of the sound radiated by the object integrates and it appears to be one source.

The distance for nearfield and farfield depend on the wavelength of the frequency being radiated.
 
I have a question that has been bugging me for some time. How do we differentiate between an "acoustically small" speaker (e.g. KEF LS50) from an acoustically large one (e.g. Duntech Sovereign, Wilson Chronosonic, Avantgarde Trio, etc) with measurements? An Avantgarde Trio has multiple cones of radiation spaced widely apart, it is very far from a "point source". How can we tell if something is point source or not?
 
I have a question that has been bugging me for some time. How do we differentiate between an "acoustically small" speaker (e.g. KEF LS50) from an acoustically large one (e.g. Duntech Sovereign, Wilson Chronosonic, Avantgarde Trio, etc) with measurements? An Avantgarde Trio has multiple cones of radiation spaced widely apart, it is very far from a "point source". How can we tell if something is point source or not?
My understanding is that whether a sound source can be considered as a point source all depends on the distance from it, and that is basically the definition of (the acoustic) far field. Here is a post on the rules of thumb by Klippel on when far field begins.

The following figures are for line arrays to show the difference between near field and far field. The left figure show the on-axis sound pressure from a 4 meter long line uniform source at 8 kHz in anechoic (free field) condition. In the near field, the sound pressure attenuation with distance is not monotonic, and at a slope of less than -6 dB per double distance. In this case the near field to far field transition point is a little over 100 m. When in the far field, the sound pressure with distance attenuation is constantly at -6 dB per double distance.

The right figure is to show that the near field to far field transition is frequency dependent.
Source: https://engineering.purdue.edu/ece40020/Homework/SomeRefs/Line_Array_Theory.pdf

line array theory.png
 
Thank you for your response. I read the paper you linked as well as the Klippel paper in that other post: https://www.klippel.de/fileadmin/kl...ld Measurement of Loudspeaker Directivity.pdf

You also said:

  1. The listening distance must be larger than the largest dimension of the speaker -- pretty easy.
  2. The listening distance must be larger than the wavelength -- difficult/impossible to achieve at low frequencies.
  3. The ratio of the listening distance to largest speaker dimension must be larger than the ratio of the largest speaker dimension to wavelength -- may be difficult to achieve at high frequencies.

I am not sure about the "wavelength" criteria. The KEF LS50 is capable of going down to 50Hz, that has a wavelength of 6.8m. According to Klippel's criteria, the farfield begins at that distance? I don't think you could even get a meaningful measurement at that distance, the speaker would have to play so loud that it's distorting, not to mention the freq response would be severely tilted downwards. As you said, "impossible/difficult" to achieve.

If the last two criteria are not fulfilled, and we only take "listening distance > largest dimension of the speaker" as the sole criteria, does this mean that there is a lower freq limit where the "farfield" measurement will be meaningless? Also, I presume that "largest dimension" includes the baffle. What about soffit mounted speakers? The "largest dimension" might include the entire width and height of the wall. Does this mean that a small bookshelf speaker mounted in a soffit becomes an acoustically large speaker?

Your other post also mentions comparing the real sound power vs. the apparent sound power. I had a look at the sections of the paper that you quoted, and the maths went a bit over my head. I take it that "real sound power" is what the Klippel actually measures, but I am not clear what "apparent sound power" means.

Sorry if the questions seem silly. It just seems counterintuitive to me.
 
I have a question that has been bugging me for some time. How do we differentiate between an "acoustically small" speaker (e.g. KEF LS50) from an acoustically large one (e.g. Duntech Sovereign, Wilson Chronosonic, Avantgarde Trio, etc) with measurements? An Avantgarde Trio has multiple cones of radiation spaced widely apart, it is very far from a "point source". How can we tell if something is point source or not?
Every speaker is acoustically small if the listening distance is far enough.

Hence I am not sure what you are actually asking, or to what end?

Perhaps you want to know what is the minimum listening distance for a given speaker to be perceptually undisadvantaged by the separation of the drive units?

Near field/far field does not even require separation of the drive units. A single drive unit alone behaves as near-field up to a certain distance and far-field beyond that distance, and the 'distance' in question varies with frequency.

If you want to ensure that you are listening in the far field at all frequencies, then your minimum listening distance will be dominated by the bass driver (and port if used). Hence, a speaker comprising a full-range coaxial 300mm driver will have the same listening distance for far-field listening (about 3m) as a 3-way unported speaker with a 300mm bass driver and mid and treble drivers strung out in a line above it. I doubt that this is the measure that you are looking for?

That's why I ask, what do you actually want to be able to do, other than just assign labels to speakers? Labels to what end?

cheers

PS my understanding is that acoustics guys define far field as the distance at which the source is small enough compared to the measurement distance for the SPL to transition to a nice predictable -6 dB per doubling of distance. In the near field that nice predictable gradient does not apply. Once could say the far field represents the point where the wavefront stabilises. A rough guide suggested by Beranek in 1986 is a distance 3 to 10 times the largest dimension of the sound source, depending on which frequency of the driver's bandwidth you are listening to. This is unrelated to the term 'near-field speaker' used in studio control rooms, where it simply means whatever speakers are used close to the recording console operator. They might not even be achieving near-field functioning, they will still be called 'the near-field monitors'.
 
That's why I ask, what do you actually want to be able to do, other than just assign labels to speakers? Labels to what end?

I am trying to understand what Linkwitz meant when he said that an ideal speaker is an acoustically small point source speaker. I can hear obvious differences between small speakers and large, and I am trying to understand that difference. I do not see any measurements that correlate with the difference that I can hear, hence the question.
 
Back
Top Bottom