Klippel Near-field Scanner - room reflections and the estimated in-room response

[Yorkshire Mouth]

Can anyone tell me how room reflections are calculated with the Klippel?

To be specific, the combined effect of reflections will be greatly determined by the distance of the speaker from the listener in relation to the distance of reflecting surfaces.

This is going to greater then just a binary nearfield/farfield demarcation.

Let’s take a room, and compare it to others, same speakers, same distance from front wall, same seating distance, same distance between speakers.

Now the first room is 3m wide, the second 3.5m wide, and let’s increase that in 0.5m increments up to 6m wide. With every increase in room size the distance from side walls (and side wall reflections) increase, which (all else being equal) will make these reflections quieter, and will therefore they’ll have a different level of net effect on the sound.

I think I’ve heard Erin say in one of his videos that the Klippel’s measurements represent the reflections in a ‘typical room’.

So I suppose I’m asking, does anyone know the dimensions of this ‘typical room’? And wouldn’t it be great if there were a calculator into which you could enter the results from a scan, and change the measurements to see what different results different rooms would create.

In addition, I wonder if it’s possible for either @amirm or Erin to do this - I suspect they’ll have results files saved to an archive somewhere - so they could see what would happen in different rooms.

I just feel ‘nearfield’ and ‘farfield’ is just too binary.

Many thanks.

 

[amirm]

Can anyone tell me how room reflections are calculated with the Klippel?

Klippel itself measures the speaker without the influence of the room (it uses phase differential to detect direct vs reflected sound). It then computes the full 3D soundfield radiated from the speaker alone.

What you are asking about is a different matter. A special module in Klippel NFS computes specific set of reflection angles as specified by CEA/CTA-2034 standard. Since Klippel NFS has full radiation patter of the speaker, this is just a matter of summing specific angles.

The standard relies on research performed by Allan Devantier when he was working at Harman (he moved to Samsung before they bought Harman). Here is the paper and abstract:

Characterizing the Amplitude Response of Loudspeaker Systems
Allan Devantier
Harman International Industries Inc., Northridge, CA, 91329, USA

ABSTRACT
The amplitude response of a loudspeaker system is characterized by a series of spatially averaged
measurements. The proposed approach recognizes that the listener hears three acoustical events in a typical
domestic environment: the direct sound, the early arrivals and the reverberant sound field. A survey of 15
domestic multi-channel installations was used to determine the typical angle of the direct sound and the early
arrivals. The reflected sound that arrives at the listener after encountering only one room boundary is used to
approximate the early arrivals, and the total sound power is used to approximate the reverberant sound field.
Two unique directivity indices are also defined and the in-room response of the loudspeaker is predicted from
anechoic data.

Put simply, it is a standardized set of reflection angles. It would be different than any room you may be using. Its purpose is to show how uniform the reflections are in a standardized way. Klippel has full set of angles so other rooms can be simulated. I export all the relevant angles in my reviews so you could build your own model if you wanted.

 

[Yorkshire Mouth]

I believe that answer can be described as ‘comprehensive’.

Many thanks.

 

[Thomas_A]

@amirm,

Not sure but is the simulated response based on the model room position of speaker in stereo setup? The reference mentions multichannel installations so there would be one for each speaker position? Diid you ever compare Klippel results with your own room?

 

[napilopez]

@amirm,

Not sure but is the simulated response based on the model room position of speaker in stereo setup? The reference mentions multichannel installations so there would be one for each speaker position? Diid you ever compare Klippel results with your own room?

Although the paper shows and mentions multi-channel installatinos, only the LCR speakers were used for evaluation.

Full paper is here: https://www.aes.org/e-lib/browse.cfm?elib=11234

"Figures 2 and 3 show the geometry of the 15 rooms including the primary listening location and the location of the left, center, and right loudspeakers. Due to gross variations in the location of the surround speakers these were not included in the analysis. All of the rooms were located in Southern California, nine were in homes and six were in apartments."

Without posting all the images from the paper, the rooms do have some radically different layouts, with about half in corner setups and half in more "rectangular" setups. Just a few examples.

One of these people appears to be a pickier audiophile than the rest

 

[Duke]

[quoting from the Devantier paper's abstract] The proposed approach recognizes that the listener hears three acoustical events in a typical domestic environment: the direct sound, the early arrivals and the reverberant sound field... The reflected sound that arrives at the listener after encountering only one room boundary is used to approximate the early arrivals, and the total sound power is used to approximate the reverberant sound field.

I figured it was something like this but didn't know for sure. Thank you.

I assume these three acoustical events are weighted, perhaps by SPL, in calculating the predicted in-room preference score... ?

This is just curiosity on my part; I don't take the in-room preference scores as gospel. Floyd Toole on the subject in a recent thread:

"... please stop putting any reliance on the calculated "scores". Learn to interpret the spinorama curves. That will have to do until we have an "educated" AI version of sound quality prediction. The ratings that were calculated by the Harman research group were done to prove a scientific point, and that done, they ceased to be used even by the people who created them. We rely on visual interpretations of the family of curves."

 

[amirm]

I assume these three acoustical events are weighted, perhaps by SPL, in calculating the predicted in-room preference score... ?

Yes, the formulas are in the spec although membership found a bug in there which we conveyed to Klippel and they fixed their implementation:

"Early Reflections
The early reflections curve is an estimate of all single-bounce, first-reflections, in a typical listening room. • Floor Bounce: 20º, 30º, 40º down • Ceiling Bounce: 40º, 50º, 60º up • Front Wall Bounce: 0º, ± 10º, ± 20º, ± 30º horizontal • Side Wall Bounces: ± 40º, ± 50º, ± 60º, ± 70º, ± 80º horizontal • Rear Wall Bounces: 180º, ± 90º horizontal"

"13 Estimating In-Room Response from Anechoic Data
The Estimated In-Room Response shall be calculated using the directivity data acquired in Section 5 or Section 6. It shall be comprised of a weighted average of 12 % Listening Window, 44 %, Early Reflections, and 44 % Sound Power. The sound pressure levels shall be converted to squared pressure values prior to the weighting and summation. After the weightings have been applied and the squared pressure values summed they shall be converted back to sound pressure levels."

 

[napilopez]

I figured it was something like this but didn't know for sure. Thank you.

I assume these three acoustical events are weighted, perhaps by SPL, in calculating the predicted in-room preference score... ?

This is just curiosity on my part; I don't take the in-room preference scores as gospel. Floyd Toole on the subject in a recent thread:

"... please stop putting any reliance on the calculated "scores". Learn to interpret the spinorama curves. That will have to do until we have an "educated" AI version of sound quality prediction. The ratings that were calculated by the Harman research group were done to prove a scientific point, and that done, they ceased to be used even by the people who created them. We rely on visual interpretations of the family of curves."

Yeah, and totally agree about not putting too much weight into the score, but to be clear the PIR is just part of the score.

I forget where exactly the PIR calculation comes from, but I recall it is:

• 12% Direct: Listening Window,
• 44% Early Reflections,
• 44% Late: Sound Power.

Once you have that, the preference score is then computed from:

• PIR Narrow Band Deviation: 20.5%
• PIR Smoothness: 17.5%
• Low Frequency Extension: 30.5%
• Narrow Band Deviation on-axis: 31.5%

So the PIR counts for 38% of the score.

To me the score is most useful as a calibration check. Sometimes I see a speaker that doesn't seem that great to me, but it has a high preference score, so that compels me to take another look.

On the other hand, even in the paper it comes from, Olive points out a potential flaw is that speakers with smooth directivity but flattish sound powers may be negatively affected. Cardioids, some horns, very wide directivity speakers, and other speakers with unconventional but good directivity are all potentially hurt by the preference score having an "ideal" directivity.