More rewarding speaker directivity pattern for in-room stereo listening

The LX521 upper mid baffle width is 4.5 inches, 114mm. Can't speak for the Nao note as I don't own them.

Keith

Cool, thanks! :)

My humble guess is that 10-17cm wide trapezoid (or other variable width shape like sandglass) baffle should give optimal results to cover necessary midrange elevation + less dispersion in 2-3kHz region. Or in case of oval waveguide these could be optimal outer dimensions. Have to prove it yet experimentally.
 
That's right. And I'm not happy about that labeling either.
But I was referring to the Klippel experiment explained on pages 457 - 461 in Toole's book. I don't have the original Klippel source, and Toole isn't very precise about how he defines "Ldiffuse". He calls it the "total sound power". Since in the experiment (Ldiffuse - Ldirect) is always >0, Ldiffuse must be the total response including the direct response.

Rudolf

I don't have the book; but what you describe in making assumption seems reasonable to me as well.

Using REW and your IR: Hoerp_links_waag.wav I attempt reconstructing your result. Granted measurement and EQ of your setup may be different now for your above results. Perhaps you might post IR.wav used in post #32.

Two copies of Hoerp_links_waag.wav are imported to REW. One copy has long window applied, and the other a 5ms window. Overlays of the two results are presented:

hoerp_links_waag 1000ms v 5ms window.png

For comparison I apply same treatments to Pluto Clone measurement from 42":

pluto clone 42 inch 1000ms v 5ms.png

From previous dialogs, I understand that you listen to your OB setup from fairly close too.

Perhaps we need more enlightenment into (Ldiffuse - Ldirect)?
 
Hello Barleywater,

Thanks for following. My apologies for bad English, I'm really struggling as it's not my native language, and this sure makes understanding of idea less clear. What do you mean by 'sensitivity' plot? If you meant an equal loudness curve it actually deals with perception.

352198d1370166941-more-rewarding-speaker-directivity-pattern-room-stereo-listening-diffusfeld.png


The above is relationship of diffuse field, derived from unspecified reflective environment, referencing free-field condition which is free of reflections. Such free-field condition is completely uniform and independent of microphone and source locations, based on premise that both source and microphone are truly omnidirectional.

Speakers in normal listening rooms are poor representations of diffuse field, providing instead a primarily reverberant sound field composed of specular reflections below 3kHz, where both timing cues are strong, and where most fundamental sounds of interest to humans are generated.

Human hearing is not omnidirectional as demonstrated with HRTF data.

I'd like to continue using term 'correct perception' by meaning 'accurate' instead of 'desired' unless we agree that what we desire is accuracy :)

Accuracy of information transmission system is readily measured, and indeed extensive works of Toole to eliminate subjective bias show strong correlation of listener preference for speakers with wide uniform radiation pattern that have flat frequency response on listening axis.

'correct perception' = 'natural' = 'live' for similarly placed listener with speaker and real source interchanged.


This is true only for anechoic chamber, don't you agree? If done in the room results will be vary depending of variations of off-axis response of speakers even if the all have flat on-axis response.

No, it works in live spaces too. I've demonstrated this with recordings of voice and acoustic instruments in live room. Close microphone technique is used for live voice/instrument. For playback speaker is placed in same location as voice/instrument was. Playback is stunningly realistic using speaker with flat response over wide pattern.

Attached is statistically average typical Behringer 8000 curve. In my opinion it's good enough for the midrange region as it fits within Class 1 tolerance up to 5K. By using correction file I believe to have reduced the non-linearity to production variations limit. Mic has the same capsule as used by Linkwitz but having more limited dynamic range (noise floor, THD) as it is connected the standard way unlike Linkwitz's which uses non-standard connection.

What is the mic you are using?

Linkwitz has long since switched to Earthworks measurement microphone, that has spectacularly flat response. I use slightly older Earthworks OM-1 microphones that also have extremely flat response, to point where applying correction curve makes very small difference visually and no difference to me when listening.

ECM8000, and truly any microphone used for quality measurements must have individual calibration data. Class 1 limits although fairly good are becoming quite dated and are locked into very old, but well defined standardization techniques/practices of reference laboratories.

Example of 'typical' is of example calibration sheet being produced, and shows only single sample of ECM8000. Is it typical ECM8000 response? Can it be applied to any typical ECM8000 microphone? My answer is what a professional would say: no. Here is overlay of many ECM8000 response plots from a microphone calibration service:

An externally hosted image should be here but it was not working when we last tested it.


I also own several ECM8000 microphones. For recording and playback on speaker as described above, the results are less natural than when using OM-1 microphone.

When I reference ECM8000 microphones against OM-1, I get response curves that fit nicely into above posted ECM8000 overlays. When I use referenced response as calibration for microphone, results become virtually identical to results obtained using OM-1.

Subtle changes of tenths dB in EQ >3kHz can have significant effect on perceived sound.

I'm not in doubt about side-flat hard-toed-in loudspeaker setup sound being natural. But in my opinion their direct response must following specific curve to define right spectrum for reflections, and not be totally flat.

This comes back to how wide and uniform speaker pattern is, and how lively the room is, and the symmetry of listening setup.

My listening is primarily in my own living room that is fairly average in terms of size and furnishings. It does have complete carpet. RT60 times are typical to rooms makeup. Many similar sized rooms without carpet, and sparse furnishings are difficult to listen in over a range of SPL. I find that with Pluto Clone with flat on-axis EQ, changes in HF roll-off are useful, sometimes due to recordings, or when trying to listen at lower than realistic levels.

Lately I've been working with highly omnidirectional speaker with 16 tweeters. EQ is same in all horizontal directions. Sound character >500Hz has very little change regardless of speaker and listening positions; especially true when speakers and listener remain >1m from walls. This strongly supports all observations made of typical forward firing speakers, and of OB speakers.
 
Overlays of the two results are presented:
View attachment 352425
My published Hoerp_links_waag.wav was with the mic directed horizontally (waagerecht in German ;)) towards the speaker. It is an ECM 40 with a calibration file for on-axis measurements. In order to get most of the 360° diffusive reverb right, I pointed the mic up for the direct/diffuse measurement. This should explain the difference (and the steep loss to high frequencies). Listening situation was identical. Measurement distance = 51"
For comparison I apply same treatments to Pluto Clone measurement from 42":
View attachment 352426
I appreciate the comparative measurement of yours.:). Could you explain the orientation of your mic (forward, upward, sideward)?

Perhaps we need more enlightenment into (Ldiffuse - Ldirect)?
Yes, it's still guesswork on my behalf. Would cost me 30 $ to read the original Klippel publishing to know for sure. :(
 
Ldif is the sound power response of the speaker, free field ("anechoic" in Toole)
This is the Abstract to Klippel's document:

"The research reported here is based on listening tests and objective measurements of loudspeakers. The uncertainty of the subjective data and the dimensions of the listening impression were investigated by statistical analysis. By measuring relevant physical properties of the signal at the listening place and by modeling essential aspects of psychoacoustical processings, objective measures were determined which allow the results of subjective tests to be explained. These measures are the basis for an objective quality assessment of loudspeakers which refers to the sensations of the listener and considers the influence of the listening conditions"

There must have been listening tests which involved a real diffuse field, not an abstract "sound power response" a la Toole only.

So I'm not yet prepared to pay my dues. :p
 
Save the money. The paper is...a bit difficult to read and I think Toole has done a good job in translating it.

From what I can read, the expression "diffuse" has been used a bit loosely and it should have read "reverberant" in some places.

Anyway, it seems as if he has computed

Ldif = free field power response

from inroom measurements at the listening place - room response.

That is visible in a picture worth only 19.99 :D

Klippel.png

"Approximation of the sound pressure amplitude response of the direct and reverberation sound parts at the listening position on the basis of loudspeaker and room parameters".
 
Last edited:
My published Hoerp_links_waag.wav was with the mic directed horizontally (waagerecht in German ;)) towards the speaker. It is an ECM 40 with a calibration file for on-axis measurements. In order to get most of the 360° diffusive reverb right, I pointed the mic up for the direct/diffuse measurement. This should explain the difference (and the steep loss to high frequencies). Listening situation was identical. Measurement distance = 51"
I appreciate the comparative measurement of yours.:). Could you explain the orientation of your mic (forward, upward, sideward)?

Yes, it's still guesswork on my behalf. Would cost me 30 $ to read the original Klippel publishing to know for sure. :(

For Pluto Clone, original referencing was done with microphone pointed at tweeter from 23cm. Since manufacture calibration chart is sufficiently flat I use OM-1 as uncorrected reference microphone. With raw tweeter measurement inverse transfer function is generated for flat linear phase response, and band limited to my tastes with FIR filters.

Here I have just setup microphone and tweeter for demonstration.

A 23cm measurement is made, and inverse transfer function generated.

A second measurement is made, without touching microphone. Inverse transfer function is applied as calibration correction and the result is displayed with no smoothing:

ref.png

To demonstrate extreme sensitivity of spatial characteristic of sound field, a measurement is made with the microphone tip moved 5mm closer to tweeter. The calibration correction is applied to the result, and ripples appear revealing ring radiator properties of circular piston driver:

ref 5mm shift.png

This is the sort of clutter that prior to modern computational power could not be handled/demonstrated well, leading to multiple measurement techniques using spatial averaging.

Next the microphone is rotated 90 degrees in horizontal plane, and placed back at 23cm distance and at same height as for referencing measurements. Measurement is made and previous correction function is applied. Resultant response shows primary effects of microphone body, and change from annular tip reflection/diffraction being replaced with circular tip aperture being principally perpendicular to sound from driver. The dominant feature of this is HF roll off, and is mathematically consistent with the microphone tip having diameter of 7mm:

90 degree.png

Final picture is overlays with 1/3 octave smoothing:

overlays.png

The universe is made of mighty fine fabric.

Information theory with DSP and readily available processing power make for concise measurements, and allow fine control of sound. This is apparent with my Pluto Clone; windowing/gating of results to reveal direct sound compares closely with EQ derived at 23cm.
 
Save the money. The paper is...a bit difficult to read and I think Toole has done a good job in translating it.

From what I can read, the expression "diffuse" has been used a bit loosely and it should have read "reverberant" in some places.

Anyway, it seems as if he has computed

Ldif = free field power response

from inroom measurements at the listening place - room response.

That is visible in a picture worth only 19.99 :D

View attachment 352474

"Approximation of the sound pressure amplitude response of the direct and reverberation sound parts at the listening position on the basis of loudspeaker and room parameters".

Measurement system of choice for seminal works of Toole, and others of similar vintage was dual channel FFT, typically 1024 bins. Just not that much frequency resolution compared to what is readily possible with swept sine, and FFT sizes available for analysis today. All the same, Toole's efforts in experimental design, execution and data reduction are amazing.
 
Barleywater,

the OM1 has (as apparently many other Earthworks mics) an impressively constant spherical directivity. Just 2 dB rolloff at 90° and 10 kHz :eek:
I pretend to know how FIR filters work, but can't see why a 5 mm change in distance should have such an impact at a wavelength of 10 cm.

Regarding Toole and Klippel: There seems too much computation and weighting involved in their experiment to make it repeatable with mere gating. But nevertheless: Creative gating can give us a better understanding of the relationship between pure source sound and room response. Maybe PRTG could show us some more of his?
 
5mm for wavelength of 10cm is 18° of phase.

The entirety of microphone body, and boom make impact. In this case mic is on camera tripod, and reflections from this come into play.

Entirety of room comes into play as well. Previous results are with 100ms window. Inverse transform for producing null reference is same basic beast for DRC; which is very tricky territory. The D/R ration at 23cm from speaker is very high, and inversion techniques two edged sword of strength/weakness is readily observed with spectrogram of measurements using 1.5ms window:

spectro ref 200ms.png

Horizontal striations indicate solid reflections. In above picture 180ms would be round trip from single surface about 30m away. In this setup it is speaker to back-wall to front-wall to back-wall to microphone. At -63dB to direct sound it dispels notion of diffuse fields in domestic listening space.

Zooming in with 25ms view is reflection at about 4m from a corner at -38dB to direct sound:

spectro ref.png

Vertical line at about 12kHz is artifact/defect resonance of Peerless 2" driver design. Inverse transfer function effectively cloaks what is really fairly strong sound field. Cloaking effect is reduced by 5mm shift:

spectro mic shift 5mm.png

In above picture many more broadband reflections are seen, of particular interest is horizontal line at about 1ms; this is microphone setup on tripod. Only microphone was tweaked 5mm, leaving tripod alone. For 90 degree measurement, tripod was moved, and speaker was moved (maybe 50mm) to get 23cm mic distance. This changes picture quite a bit:

spectro mic 90 deg.png

13.4ms reflection is backwall. For this series, referencing inverse was applied to results. Same results are obtained when referencing inverse is applied to the measurement sweep. When this is done, sound of sweep in room is very uniform through changing frequencies.
 
Last edited:
I am new to this thread and have only had a glance at it.
A lot of text to be digested.
Actually i am preparing a text about my work on free field and diffuse field equalization.
In a nutshell i am using open baffle speakers, first dipole and now cardioide.
I equalize then free field on the front and diffuse field on the back.
I do this simply with a back firing, reversed tweeter that also destroys the impulse response to the back.
This text will be published by Brieden Verlag in a volume called :
The Loudspeaker Bible.
 
Joachim,

Diffuse field: Please define this in context of loudspeakers.

Speaker only has one response to a given stimulus. It can not reproduce simultaneously two separate equalization settings.

'The Loudspeaker Bible' is a pretentious title for a book; but a good name in reference to a well regarded book as such becomes widely accepted.
 
Joachim,

Diffuse field: Please define this in context of loudspeakers.

I'd like to know that too :)

Speaker only has one response to a given stimulus. It can not reproduce simultaneously two separate equalization settings.

With an aditional tweeter on the back one can do that to some extend.
My question is "what for?"
 
Tweeter on back of OB dipole extends range over which good approximation of dipole radiation is maintained. Dipole tweeter, such as some ribbon types may be incorporated for achieving this result.

Cardioid pattern is achievable with enclosure employing resistive attributes.

Cardioid pattern may be also achieved by integrating dipole with monopole of matching bandwidth.

Not withstanding, observation of speaker radiation is from single perspective that is summation of all radiation from speaker that reaches observation point. No dual equalization can apply if radiation system is only resolvable as single source.

Diffuse field as developed by Sabine is lumped element analysis to simplify a working model. From work of Toole and his numerous references this model is of little or no use, especially in small rooms.

My previously posted sonograms of room reflections makes this very clear. Sound field remains highly coherent at all points in room. It remains so well below threshold of hearing within room. As energy decays, it is effectively lost and doesn't create a secondary independent field of noise with any relevant impact on sound qualities perceivable or measurable.
 
Not withstanding, observation of speaker radiation is from single perspective that is summation of all radiation from speaker that reaches observation point. No dual equalization can apply if radiation system is only resolvable as single source.

Think of it as an equalization of reflections. If they are more similar to the direct sound, this should help the precedence effect.

Furthermore localization, loudness and timbre are different processes in the brain. A lumped model is not helpful.
 
It appears that throughout this thread the term "diffuse" field is being used when in fact it should be "reverberant" field. There is no guarantee that any place in a real room is diffuse, nor is there any reason to believe that it matters. But we should use the proper terms and "diffuse" is not the correct one. Steady state or reverberant - either suites me.

Dual EQ will change the sound radiation pattern and hence change all of the "fields". The question is: Is this a good thing?

What radiation pattern do we want? Do we want sound radiated from the rear of the source to the back wall? Toole and I would say "No, we don't want that." I go out of the way to eliminate all rearward sound radiation (from the source as well as the wall reflection.) I cannot imagine trying to enhance it.
 
Last edited: