Could you grasp the relation between the loudness curve and the Blauert directional bands ? Those are the German links.
The 500Hz and 3KHz peak are related to presence (added focus in the phantom image) while the 8...9KHz sag is related to diffuseness (removed focus from tweeters).
Anyway, good to hear that I am not the only one, who prefers such a FR (for wide dispersion speakers) !
The higher the directivity of the speakers, the less of this required. But at the same time the listening distance has to become greater in order to stay within the preferred D/R ratio.
Its from the Toole's book. Experiment was made within variable D/R ratio (from an-echoic to echoic) room involving many people with different background including experienced listeners like sound professionals, but also people totally unrelated to audio industry. To cover group with special needs some persons having diagnosis of partial hearing loss were also participating.
Resulting matrix had shown that majority of preference went for D/R ratio of 1. Some of sound engineers gave preference for reflected sound to be attenuated by 3dB (attributed to professional habit to pay more attention to direct sound) while people with impaired hearing wanted reflected signal attenuated by at least 6dB (obviously when you have trouble with resolution then spatial envelopment works more like masking blur and becomes undesirable). Only few preferred reverberant signal to exceed direct signal, but nobody liked reflections to be significantly reduced or eliminated as it reduced pleasantness of spatial envelopment of reflected sound that we humans seem to possess naturally.
I must look into the book again to see if I remember the data correctly, unfortunately not possible at the moment as I'm fighting fever in my father-in-law house to be quarantined from kids.
Go and have another look. I am sure you will find it
You're trying to place the burden of proof on the wrong side. You made the claim, not me.
Totally agree with that! Now to find whats the historical reason to prefer equalization in smaller cavities compared to open-air is the question for athropologists, I guess it should be part of evolution - our ancestors had to hear differently in caves compared to open air. What also came into my mind is that nature provides at least three great sources to normalize our hearing: heavy rainfall, big waterfalls and ocean waves all providing spectrum close to pink noise. No wonder that many people use these recordings as a therapy - we probably get tired for being in rooms too long partially because of our hearing is in constantly room-adapted state.
My feelings exactly
A huge relief, needless to say.You nailed it!
Now I'm going to summarize what methods we can use to implement it on different speaker types.
Correction for in post #6
Initial text:
"My conclusion from this picture is that for in-room listening for frontally lineary responding speaker the best dispersion pattern curve would follow this suggested equalization line, or directivity index curve following just inverted version of it."
Corrected version:
"One that designs correct loudspeaker system for in-room reproduction must aim for most linear frontal response measured in anechoic chamber combined with DI (directivity index) curve that follows attached psycho-acoustic correction EQ curve suggested by Zwicker/Fastl and that was suggested by Joachim Gerhard in his thread. In other words narrower dispersion at each given frequency would will excite less reflections which equals higher DI and makes direct signal less masked by reflections while wider dispersion (lower DI) will do the opposite therefore creating desired D/R EQ acoustically rather than electronically. Sum of direct/reflected signal will therefore be correct for in-room equall loudness perception. As this will most probably give correct finally perceived sound in particular room type achieving this DI for particular reference speaker will just set an average baseline for usage in different room sizes. Confirmed yet by few the curve shows with some probability how majority of people would like the sound to be reproduced in rooms to be perceived as linear in terms of equal loudness in particular room used in their experiments. Necessity of correction is highly disputable as it seems to be not widely adopted by industry yet.
For most precise reproduction experience in any particular room that has feature set (size, reflectivity, diffusivity) different from room used for EQ curve creation using appropriate Q (to be determined easily) for each parametric filter. This final EQ must be applied to direct signal to achieve resulting sound power curve that complements with the psychoacoustic EQ curve and satisfies desire for most linear perceived signal (yet to be determined). This may shift direct signal off linear on-axis response for very small levels, but it is the best compromise I can think of unless speakers are designed for the one particular room. Experiments with equalization of symmetrical omnidirectional speaker has shown that only correct sum is what matters, regardless of what we change, in this case EQ of both - direct and reflected signal. This part is the most controversial as it goes against to some historical canons I guess.
For loudspeakers having high average DI and/or used in larger rooms (latter to be yet determined) less of this type of correction will be needed. DI=0 at particular band sets sets maximum of the correction level needed in some 'standard' listening room used by Zwicker/Fastl to equal the correction level shown on the curve. Minimum correction level is yet to be determined or must be found emphirically. 'Critical midbass' or band of 400-500Hz may be a good starting point as speakers directivity at this band is usually is very low already (close to Shroedders frequency) but not yet influenced by room modes".
This essentially answers what in my opinion defines more rewarding loudspeaker directivity pattern for in-room stereo listening. May apply to mono loudspeakers as well. For multi-channel systems where frontal speakers are required to have much narrower dispersion compared to side speakers appropriately corrected DI curves must be achieved for each type.
If moderator could move this to the top before prologue it would be easier for more busy people to get to the point quickly without reading all through.
Any criticism is welcome -as always. Thanks!
Initial text:
"My conclusion from this picture is that for in-room listening for frontally lineary responding speaker the best dispersion pattern curve would follow this suggested equalization line, or directivity index curve following just inverted version of it."
Corrected version:
"One that designs correct loudspeaker system for in-room reproduction must aim for most linear frontal response measured in anechoic chamber combined with DI (directivity index) curve that follows attached psycho-acoustic correction EQ curve suggested by Zwicker/Fastl and that was suggested by Joachim Gerhard in his thread. In other words narrower dispersion at each given frequency would will excite less reflections which equals higher DI and makes direct signal less masked by reflections while wider dispersion (lower DI) will do the opposite therefore creating desired D/R EQ acoustically rather than electronically. Sum of direct/reflected signal will therefore be correct for in-room equall loudness perception. As this will most probably give correct finally perceived sound in particular room type achieving this DI for particular reference speaker will just set an average baseline for usage in different room sizes. Confirmed yet by few the curve shows with some probability how majority of people would like the sound to be reproduced in rooms to be perceived as linear in terms of equal loudness in particular room used in their experiments. Necessity of correction is highly disputable as it seems to be not widely adopted by industry yet.
For most precise reproduction experience in any particular room that has feature set (size, reflectivity, diffusivity) different from room used for EQ curve creation using appropriate Q (to be determined easily) for each parametric filter. This final EQ must be applied to direct signal to achieve resulting sound power curve that complements with the psychoacoustic EQ curve and satisfies desire for most linear perceived signal (yet to be determined). This may shift direct signal off linear on-axis response for very small levels, but it is the best compromise I can think of unless speakers are designed for the one particular room. Experiments with equalization of symmetrical omnidirectional speaker has shown that only correct sum is what matters, regardless of what we change, in this case EQ of both - direct and reflected signal. This part is the most controversial as it goes against to some historical canons I guess.
For loudspeakers having high average DI and/or used in larger rooms (latter to be yet determined) less of this type of correction will be needed. DI=0 at particular band sets sets maximum of the correction level needed in some 'standard' listening room used by Zwicker/Fastl to equal the correction level shown on the curve. Minimum correction level is yet to be determined or must be found emphirically. 'Critical midbass' or band of 400-500Hz may be a good starting point as speakers directivity at this band is usually is very low already (close to Shroedders frequency) but not yet influenced by room modes".
This essentially answers what in my opinion defines more rewarding loudspeaker directivity pattern for in-room stereo listening. May apply to mono loudspeakers as well. For multi-channel systems where frontal speakers are required to have much narrower dispersion compared to side speakers appropriately corrected DI curves must be achieved for each type.
If moderator could move this to the top before prologue it would be easier for more busy people to get to the point quickly without reading all through.
Any criticism is welcome -as always. Thanks!
Attachments
I measured the direct/diffuse ratio at my listening point:
Don't look for the response curve as is - the microphone was directed to the ceiling to get the full horizontal response.
Red is diffuse field without gate (total power response), green is direct response with 5 ms gate. It looks like I prefer only 2-3 dB difference compared to the 5 dB Klippel is talking about.
The graph you've posted is only true for pure tones. We don't listen to pure tones.
Is that graph also true for phantom sources?
How much is perceived loudness determined by indirect sound and how much by direct sound?
The indirect sound field differs considerably between rooms - how is this incorporated in the proposed directivity-based equalization attempt?
Spectral balance is already corrected for in the production process, so why apply it a second time in reproduction?
Last edited:
PRTG,
Posted sensitivity plot and thesis is totally nonsensical approach for achieving desired perceptions.
Use a measurement microphone with flat response to record a sound and then play it back with speaker referenced to same level using same microphone placed at same distance used in recording. Speaker with flattest response sounds most natural.
Typical crap measurement microphone such as ECM8000 has rising response. Equalization of speaker with such requires calibration curve applied as correction, and likewise if microphone is used for above simple demonstration, correction equalization needs to be applied to recording as well.
If speaker is equalized flat from off axis, i.e. 30 degrees, and listened to at same angle for above demonstration, result is natural. If speaker has typical characteristic of direct radiator, and is listened to on axis using such EQ, it will sound too bright.
With flat on axis response, the further back the listening position is in a live room, the more apparent the perception becomes that the sound is being beamed.
With flat off axis EQ, listening further back, the bright on axis energy is illuminating the room, accentuating its liveliness. This leads to all manner of perceptions that are all highly dependent of the room and listener position.
Walking the polar pattern of speaker set up outside makes this very clear. EQ on axis may lead to region 0°±10° with very natural sound. EQ at 30° off axis leads to two zones in horizontal plane for possible coverage of 40° with natural sound.
With off axis EQ of conical radiator, much greater volume is controlled. For Pluto type speaker with 2" full range as tweeter, I tilt tweeter up about 5°, and EQ at about 10° off axis. At listening position tweeter shoots over head, and sound changes very little with toe-in changes of ±20°. Swapping to 1" dome tweeter extends this to usable toe-in range of about ±30°.
Better overall balance is achieved with speaker with wide uniform pattern, or treatment of the room, or combination of the two. Treatment also includes particular placement of listener and speakers, including orientation of speaker axis (unless it is truly omnidirectional), and choice of equalization axis.
Posted sensitivity plot and thesis is totally nonsensical approach for achieving desired perceptions.
Use a measurement microphone with flat response to record a sound and then play it back with speaker referenced to same level using same microphone placed at same distance used in recording. Speaker with flattest response sounds most natural.
Typical crap measurement microphone such as ECM8000 has rising response. Equalization of speaker with such requires calibration curve applied as correction, and likewise if microphone is used for above simple demonstration, correction equalization needs to be applied to recording as well.
If speaker is equalized flat from off axis, i.e. 30 degrees, and listened to at same angle for above demonstration, result is natural. If speaker has typical characteristic of direct radiator, and is listened to on axis using such EQ, it will sound too bright.
With flat on axis response, the further back the listening position is in a live room, the more apparent the perception becomes that the sound is being beamed.
With flat off axis EQ, listening further back, the bright on axis energy is illuminating the room, accentuating its liveliness. This leads to all manner of perceptions that are all highly dependent of the room and listener position.
Walking the polar pattern of speaker set up outside makes this very clear. EQ on axis may lead to region 0°±10° with very natural sound. EQ at 30° off axis leads to two zones in horizontal plane for possible coverage of 40° with natural sound.
With off axis EQ of conical radiator, much greater volume is controlled. For Pluto type speaker with 2" full range as tweeter, I tilt tweeter up about 5°, and EQ at about 10° off axis. At listening position tweeter shoots over head, and sound changes very little with toe-in changes of ±20°. Swapping to 1" dome tweeter extends this to usable toe-in range of about ±30°.
Better overall balance is achieved with speaker with wide uniform pattern, or treatment of the room, or combination of the two. Treatment also includes particular placement of listener and speakers, including orientation of speaker axis (unless it is truly omnidirectional), and choice of equalization axis.
Markus, thanks for the questions!
It could be the same regarding complex signals due highly resolutive nature of human perception of the sound. At least that's what I'm hearing which isn't satisfactory proof alone, of course. I'll try to find the ground for this in the Toole's book once I get to it.
I also can't answer about phanthom sources, but if I remember correctly their perception are mostly delivered through specific frequency band. I must check this, too.
Spectral balance of recordings can vary depending on recording techniques, skill, hearing condition and personal taste of recording, mixing, mastering and finalizing engineers. If engineers have used controlled directivity monitors in moderately diffusive studio control room, or mastering grade headphones or both, result will have good chances to be close to pleasant. I personally always have used Sennheisser HD25-C II for ocassional finalizing tasks of already recorded and mixed material and usually was happy with the result. If D/R EQing gives better percieved balance of the loudness of particular audio bands it should just more precisely reveal the result of studio work as it is.
Perhaps you are asking what will happen to reverberant cues that are put into the final mix in some proportion with direct recording - will they be multiplied or perceived 'as is'?
And you are certainly not alone with this. This probably goes into the same direction like musicians and mixing people like less reflected sound.
Meanwhile I see the 5dB as an upper limit and I would probably find myself in the 3...4dB range.
Red is not diffuse response. It is direct + diffuse. So plots compare windowing of complete response of perhaps 100ms v direct response with windowing of 5ms. Without appropriate normalization and referencing methods for different window lengths, the results do not show a D/R relationship as the area between two plots.
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
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.
I'd like to continue using term 'correct perception' by meaning 'accurate' instead of 'desired' unless we agree that what we desire is accuracy
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.
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?
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.
Side-flat hard-toed-in loudspeaker setup surely is another method how to create correct spectral balance between direct and reflected signal and create stronger engagement of late reflections as 'on-axis' signal hits the opposite wall first. Plus it allows for wider coverage in bigger rooms. While travelling to opposite wall its on-axis (now rather assymetric of-axis) spectrum changes as decay rate is higher for higher frequencies and walls and audience absorb some part of it. It may quite well complement with last part of Zvicker/Fastl curve which could be even measureable by placing mic near the wall and taking measurements against the wall with poper gating applied. By the way don't you ever do additional EQing to better fit such loudspeakers to particular room size?
So on-axis becomes non-flat actually. Doesn't it really matter what spectrum you're sending to opposite walls to illuminate the room? If it matters, what is the correlation of optimal direct response curve with different room sizes? I bet you don't know the answer but you're working on it.
Sounds reasonable.
I agree. But that means very wide variations of actions without good starting point as a baseline. Also big problem being our adaptation rate being too fast. We just start liking every the new setup too quickly.
Your post gave more food for thought, thanks!
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.
I'd like to continue using term 'correct perception' by meaning 'accurate' instead of 'desired' unless we agree that what we desire is accuracy
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.
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?
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.
Side-flat hard-toed-in loudspeaker setup surely is another method how to create correct spectral balance between direct and reflected signal and create stronger engagement of late reflections as 'on-axis' signal hits the opposite wall first. Plus it allows for wider coverage in bigger rooms. While travelling to opposite wall its on-axis (now rather assymetric of-axis) spectrum changes as decay rate is higher for higher frequencies and walls and audience absorb some part of it. It may quite well complement with last part of Zvicker/Fastl curve which could be even measureable by placing mic near the wall and taking measurements against the wall with poper gating applied. By the way don't you ever do additional EQing to better fit such loudspeakers to particular room size?
So on-axis becomes non-flat actually. Doesn't it really matter what spectrum you're sending to opposite walls to illuminate the room? If it matters, what is the correlation of optimal direct response curve with different room sizes? I bet you don't know the answer but you're working on it.
Sounds reasonable.
I agree. But that means very wide variations of actions without good starting point as a baseline. Also big problem being our adaptation rate being too fast. We just start liking every the new setup too quickly.
Your post gave more food for thought, thanks!