Cause and effect have to be separated !
Reflections in a small room vitually always cause "comb filtering". But there are wanted reflections, e.g. lateral >10ms. As a result you will have comb filtering. But where is the damage then ? All the combs are, is something that can be measured with a microphone.
"Damage" only comes from unwanted reflections, which also causes comb filtering.
So it is not the comb filtering that is bad in general but the specific reflections causing it, e.g. floor reflections for some people here.
Ref. Toole, chapter 9, The Audibility of Acoustical Interfrence - Comb filtering, page 151. The four bullets on the page together with the paragraph above and below.
The reason I write all this is: In forums there are certain buzz words that are being used loosely, uncontrolled and even incorrectly. The rookie readers will likely catch these words and more confusion and BS will emerge from that such as "comb filtering must always be bad. It looks ugly so it must sound ugly". And it takes long discussions to sort it out again. Sorry for making you a victim
Cause and effect have to be separated !
Reflections in a small room vitually always cause "comb filtering". But there are wanted reflections, e.g. lateral >10ms. As a result you will have comb filtering. But where is the damage then ? All the combs are, is something that can be measured with a microphone.
"Damage" only comes from unwanted reflections, which also causes comb filtering.
So it is not the comb filtering that is bad in general but the specific reflections causing it, e.g. floor reflections for some people here.
Ref. Toole, chapter 9, The Audibility of Acoustical Interfrence - Comb filtering, page 151. The four bullets on the page together with the paragraph above and below.
The reason I write all this is: In forums there are certain buzz words that are being used loosely, uncontrolled and even incorrectly. The rookie readers will likely catch these words and more confusion and BS will emerge from that such as "comb filtering must always be bad. It looks ugly so it must sound ugly". And it takes long discussions to sort it out again. Sorry for making you a victim
exactly!
I happened to listen e.g. in/near dipole "nulls" with different arrangements
and also experienced some systems based on high reverberant ratio in the
Khz range.
A friend of mine had an experimental (fun) setup using a small fullranger
aiming into the two upper front corners of his room.
Some of those arrangements sounded "better than expected" to me, if
you get tonal balance right as often the power response of the system
gets more dominant, which is usually falling with frequency.
Such arrangements may have phantom sources sound (very)
wide (ASW), also deep, maybe even somewhat "involving" in a
wide range of the room, but resolution and neutrality is always missing
to an extent, that makes it unacceptable to me personally even if
appropriate EQ is applied.
For easy listening, say in a bar or restaurant, why not.
and also experienced some systems based on high reverberant ratio in the
Khz range.
A friend of mine had an experimental (fun) setup using a small fullranger
aiming into the two upper front corners of his room.
Some of those arrangements sounded "better than expected" to me, if
you get tonal balance right as often the power response of the system
gets more dominant, which is usually falling with frequency.
Such arrangements may have phantom sources sound (very)
wide (ASW), also deep, maybe even somewhat "involving" in a
wide range of the room, but resolution and neutrality is always missing
to an extent, that makes it unacceptable to me personally even if
appropriate EQ is applied.
For easy listening, say in a bar or restaurant, why not.
ok! so You have listened to "something", found it even "involving" and so... suitable for "easy listening"... ???
and You think that You can criticize things You haven't heard but which You think are somewhat similar, yes?
Great topic guys. This is core-business.
I think that in discussing this topic there are some effects that have to be considered:
- Stereo is a compromise in a spatial sense. You can only hope to listen into the recording venue.
- I think most people agree that early reflections are unwanted. However, when listening to very directional speakers, some people almost experience in-head or very near localisation of the auditory scene. Lateral reflections, which in a natural setting aid localisation, are unfortunately absent with stereo (or arrive only very early from the side walls).
- People differ greatly in their sensitivity and preference for early reflections, as described by Toole in 'Sound Reproduction'. There is no doubt that at least some preferences are learned or differ depending on the situation. Toole writes for example that most recording engineers prefer an absence of early reflections during mixing, but many prefer them for casual listening. Some find stereo to be spatially unsatisfying otherwise, which is not really surprising (see previous point). It may be considered desirable by some to 'enhance' the spatial impression of a stereo system with early reflections.
--
- On the requirement of a flat on-axis response: I'm not convinced that this requirement is correct because of the large difference in HRTFs between 0 and 30 degrees. Most sources (or the most important ones) in stereo reproduction are in the center. Personally I find a slight on-axis HF down-shelving to be desirable with virtually every speaker.
- I'm also not sure about the requirement for either rising directivity. It might be 'necessary' to preserve the balance intended by the recording engineer because all recordings are mixed on speakers with such directivity. It might also be 'required' to compensate for the HRTF coloration of the phantom center. There are proponents for both flat and rising directivity; I'm not sure about either however.
I think that in discussing this topic there are some effects that have to be considered:
- Stereo is a compromise in a spatial sense. You can only hope to listen into the recording venue.
- I think most people agree that early reflections are unwanted. However, when listening to very directional speakers, some people almost experience in-head or very near localisation of the auditory scene. Lateral reflections, which in a natural setting aid localisation, are unfortunately absent with stereo (or arrive only very early from the side walls).
- People differ greatly in their sensitivity and preference for early reflections, as described by Toole in 'Sound Reproduction'. There is no doubt that at least some preferences are learned or differ depending on the situation. Toole writes for example that most recording engineers prefer an absence of early reflections during mixing, but many prefer them for casual listening. Some find stereo to be spatially unsatisfying otherwise, which is not really surprising (see previous point). It may be considered desirable by some to 'enhance' the spatial impression of a stereo system with early reflections.
--
- On the requirement of a flat on-axis response: I'm not convinced that this requirement is correct because of the large difference in HRTFs between 0 and 30 degrees. Most sources (or the most important ones) in stereo reproduction are in the center. Personally I find a slight on-axis HF down-shelving to be desirable with virtually every speaker.
- I'm also not sure about the requirement for either rising directivity. It might be 'necessary' to preserve the balance intended by the recording engineer because all recordings are mixed on speakers with such directivity. It might also be 'required' to compensate for the HRTF coloration of the phantom center. There are proponents for both flat and rising directivity; I'm not sure about either however.
Last edited:
Member
Joined 2009
Guys! You're moving so fast, I have trouble keeping up just reading! But don't let me hold you back.
You are actually reading all of this?
On this note let me respond to something from over 70 posts a go. I mean *yesterday*
I have a similar experience with controlling the imaging width and clarity by varying the toe in-out. I'm convinced the ratio of the direct to reflected signal from the sidewalls plays a great role in this. I also suspect that longer time gaps between the direct and reflected sound allow for different tolerances of the intensity of the reflection. When I have the spare time I'd like to try different toe in angles and take some listening tests. I think plots of the Initial Pulse and the first side wall reflection accompanied with the subjective image perception of each speaker position will be enough to make some initial assumptions.
Maybe if more of us are interested in something like this we could organize an official format by which people can conduct such listening tests at home and upload the data to a central repository. If indeed a relation between the sidewall reflections and imaging exists, patterns in the data should be easy to discover.
Transients, those bästards ! 
I used to be in that camp also. I like dipole line arrays too, remember! I'm not sure if I still am with that since gained new knowledege, but it certainly feels wierd to think otherwise. Old habits are too deep in me.
See again the document I posted earlier:
Transient is an AM modulation. How fast can it be perceived? How fast is the human auditory system?
From the document we can read -10dB cut off points of the modulation transfer function of the auditory nerve fibers, and they are something like this:
at 500Hz: 100Hz
at 1kHz: 200Hz
at 2kHz: 400Hz
etc
To convert this into the time response in the modulation domain it is something like:
at 500Hz: 10ms
at 1kHz: 5ms
at 2kHz: 2.5ms
etc
It seems to follow constant Q characteristics.
Is this fast? Or rather slow?
At 1kHz the 'speed' is 5ms. How many reflections can fit in this time frame already
What is your "transient" ?
- Elias
I used to be in that camp also. I like dipole line arrays too, remember! I'm not sure if I still am with that since gained new knowledege, but it certainly feels wierd to think otherwise. Old habits are too deep in me.
See again the document I posted earlier:
Transient is an AM modulation. How fast can it be perceived? How fast is the human auditory system?
From the document we can read -10dB cut off points of the modulation transfer function of the auditory nerve fibers, and they are something like this:
at 500Hz: 100Hz
at 1kHz: 200Hz
at 2kHz: 400Hz
etc
To convert this into the time response in the modulation domain it is something like:
at 500Hz: 10ms
at 1kHz: 5ms
at 2kHz: 2.5ms
etc
It seems to follow constant Q characteristics.
Is this fast? Or rather slow?
At 1kHz the 'speed' is 5ms. How many reflections can fit in this time frame already
What is your "transient" ?
- Elias
Last edited:
What's wrong with "tough nut" soffa pillow?
First you test, by any means necessary to get a result, then only try to make it practical. That's the way to do research.
- Elias
..piffle on your sofa pillow.
-FAR more advanced than a sofa pillow. Multi-miniature curtains! This way you can vary the pattern.
I'm thinking to do it with a radial will require 2 identical drivers (side to side) with delay and possible phase rotation. Not exactly my area of expertise.
Last edited:
Erjee asks “Besides the 'broadband CD point source' characteristic there are two other aspects of the SH50 which are not found in normal indoor hifi speakers:
- very narrow directivity (50 deg beamwidth)
- the lack of any baffle which probably reduces the amount of diffraction
Do you consider the second aspect as an important property for a loudspeaker?”
The constant directivity is a useful thing in commercial sound where you have a listening area as opposed to one location, the benefit is that everyone hears essentially the same spectrum.
Two ways these are different though is that they radiate as a single source, as if it only had one small broad band driver and as a side benefit, they can reproduce a square wave over a decade wide band. That is not a key but an indicator that they system radiates the same place and time or phase over a wide band.
A fun thing to demonstrate is that you can remove the grill and even when your head is inside the horn mouth, the sound always sounds like it is somewhere floating in front of you, you can’t ever hear woofers mids or the hf driver. A friend used a pair on his computer desk as his monitors.
The other thing is that at any level one would ever what in a living room, they are loafing along and the mid section especially has very low distortion. Consider using shaped pink noise, raising the level every 5 min, the point the response shaped changed (relative to 1W) at just one frequency by 3dB was at 56 V rms (1w 1m sensitivity about 100dB)
So far as diffraction, actually that is one thing a horn can deal with, it is the deliberate control of the radiation angle and in a conical horn the wall angle doesn’t change but as the mouth gets larger, the pattern control loss frequency goes down in frequency. Above that frequency, the mouth is large enough to define the radiation angle and so assuming one drives it with an acoustically small source or one that was curved already, then what one radiates is a segment of a sphere.
What is very different from most speakers is how much energy is radiated to the sides and rear compared to normal. In commercial sound one wants as much of the energy as possible where the ears are and as little as possible on the walls floor and ceiling especially in front of the audience. I don’t know that there are many polar plots for hifi speakers handy but this part one can see by examining the polars for an SH-50.
If you down load the .CLF data file for it and the file viewer (free) you can examine the polars and a number of other things.
Anyway, remembering we are talking about the energy in vs out of the pattern, here is what an SH-50 looks like at say 90 degrees off axis.
At 160Hz it is about -5dB at 90 degrees off axis
At 250Hz it is about -8dB at 90 degrees off axis
At 500Hz it is about -20dB at 90 degrees off axis
At 1KHz it is about -20dB at 90 degrees off axis
At 2KHz it is about -23dB at 90 degrees off axis
At 5KHz it is about -28dB at 90 degrees off axis
At 10KHz it is about -35dB at 90 degrees off axis
At 16KHz it is about -26dB down 90 degrees off axis.
Consider that when the beam is -20dB, it is 1% of the intensity on axis and so the spectrum both on and off axis (reverberant field) is primarily the on axis response.
The result of that directivity in my living room is that the measured response at the listening position is very much like the response at one meter.
A high resolution measurement of small cone/dome 2 way measures about + - 15 dB over the 1 meter response at the LP, the SH-50 measured about + -3dB over the 1 meter response at the LP.
It would be my guess that the large reduction of the reflected sound that allows the L.P. response to be much better combined with the time response and lack of depth cues radiated by the speaker is also why the mono phantom can be so strong.
The near field is very large and the reflected energy very low.
A funny thing too, if you place two normal speakers side by side and play music, there is obvious comb filtering if you move around.
You can place two SH-50s side by side tight together and there is no audible seam at all. That acoustic array-ability is useful in commercial sound but this arrangement also lets you place a speaker against a wall or boundary without reflections or an audible seam. . If you scroll down to page 7 here; you can get the idea how this is possible.
http://www.danleysoundlabs.com/pdf/danley_tapped.pdf
I have used this directivity / boundary trick in a narrow room which makes a stereo image the entire width of the room and in commercial sound, people often place them on a ceiling aiming down and the floor monitors work that way but pointed up. as long as the array-able side is used, there is no audible seam or reflection
Hope that helps explain.
Fwiw, a fellow on another forum just posted about his impressions of one of the smaller horns.
RE: That is exactly what SoundLab uses... - johny - Speaker Asylum
Best,
Tom Danley
Danley Sound Labs, Inc. | Facebook
- very narrow directivity (50 deg beamwidth)
- the lack of any baffle which probably reduces the amount of diffraction
Do you consider the second aspect as an important property for a loudspeaker?”
The constant directivity is a useful thing in commercial sound where you have a listening area as opposed to one location, the benefit is that everyone hears essentially the same spectrum.
Two ways these are different though is that they radiate as a single source, as if it only had one small broad band driver and as a side benefit, they can reproduce a square wave over a decade wide band. That is not a key but an indicator that they system radiates the same place and time or phase over a wide band.
A fun thing to demonstrate is that you can remove the grill and even when your head is inside the horn mouth, the sound always sounds like it is somewhere floating in front of you, you can’t ever hear woofers mids or the hf driver. A friend used a pair on his computer desk as his monitors.
The other thing is that at any level one would ever what in a living room, they are loafing along and the mid section especially has very low distortion. Consider using shaped pink noise, raising the level every 5 min, the point the response shaped changed (relative to 1W) at just one frequency by 3dB was at 56 V rms (1w 1m sensitivity about 100dB)
So far as diffraction, actually that is one thing a horn can deal with, it is the deliberate control of the radiation angle and in a conical horn the wall angle doesn’t change but as the mouth gets larger, the pattern control loss frequency goes down in frequency. Above that frequency, the mouth is large enough to define the radiation angle and so assuming one drives it with an acoustically small source or one that was curved already, then what one radiates is a segment of a sphere.
What is very different from most speakers is how much energy is radiated to the sides and rear compared to normal. In commercial sound one wants as much of the energy as possible where the ears are and as little as possible on the walls floor and ceiling especially in front of the audience. I don’t know that there are many polar plots for hifi speakers handy but this part one can see by examining the polars for an SH-50.
If you down load the .CLF data file for it and the file viewer (free) you can examine the polars and a number of other things.
Anyway, remembering we are talking about the energy in vs out of the pattern, here is what an SH-50 looks like at say 90 degrees off axis.
At 160Hz it is about -5dB at 90 degrees off axis
At 250Hz it is about -8dB at 90 degrees off axis
At 500Hz it is about -20dB at 90 degrees off axis
At 1KHz it is about -20dB at 90 degrees off axis
At 2KHz it is about -23dB at 90 degrees off axis
At 5KHz it is about -28dB at 90 degrees off axis
At 10KHz it is about -35dB at 90 degrees off axis
At 16KHz it is about -26dB down 90 degrees off axis.
Consider that when the beam is -20dB, it is 1% of the intensity on axis and so the spectrum both on and off axis (reverberant field) is primarily the on axis response.
The result of that directivity in my living room is that the measured response at the listening position is very much like the response at one meter.
A high resolution measurement of small cone/dome 2 way measures about + - 15 dB over the 1 meter response at the LP, the SH-50 measured about + -3dB over the 1 meter response at the LP.
It would be my guess that the large reduction of the reflected sound that allows the L.P. response to be much better combined with the time response and lack of depth cues radiated by the speaker is also why the mono phantom can be so strong.
The near field is very large and the reflected energy very low.
A funny thing too, if you place two normal speakers side by side and play music, there is obvious comb filtering if you move around.
You can place two SH-50s side by side tight together and there is no audible seam at all. That acoustic array-ability is useful in commercial sound but this arrangement also lets you place a speaker against a wall or boundary without reflections or an audible seam. . If you scroll down to page 7 here; you can get the idea how this is possible.
http://www.danleysoundlabs.com/pdf/danley_tapped.pdf
I have used this directivity / boundary trick in a narrow room which makes a stereo image the entire width of the room and in commercial sound, people often place them on a ceiling aiming down and the floor monitors work that way but pointed up. as long as the array-able side is used, there is no audible seam or reflection
Hope that helps explain.
Fwiw, a fellow on another forum just posted about his impressions of one of the smaller horns.
RE: That is exactly what SoundLab uses... - johny - Speaker Asylum
Best,
Tom Danley
Danley Sound Labs, Inc. | Facebook
I am not sure what qualifies as a messed up transient for LineArray. What is a satisfyingly clean-transient example speaker?
Most decent dome tweeters are jet-fast and don't have a problem with audible transients. Some famed transient speakers like ribbon tweeters are actually producing audible distortion compared to a good dome, and this gives them a little extra (extra wrong) liveliness in the ultra highs.
Last edited:
I concur with that. Speaker/room interaction has to be in the focus
of discussion IMO.
_________________
This was taken in an untreated room, except for floor bounce (see below)
Floor: Approx. 70% wood tiles, ca. 30% glass, nearly plane
At the reflection point floor covered with a 2cm thickness wool carpet, approx. 70cm x 70cm
Ipsilateral side wall: Approx. 50% stone and 50% glass, nearly plane, stone non porous.
Speaker position
----------------
Floor standing.
Distance to side wall approx. 65cm
Distance to front wall approx. 150cm
Distance to back wall >500cm (non uniform)
Ceiling height approx. 240cm
Microphone (omni) position
--------------------------
Distance approx. 150cm from speaker.
Height approx. 80cm in direction of listening position
Approx 25 degrees off speaker's axis and speaker toed in slightly
__________________
This means floor bounce and ipsilateral sidewall bounce are included ...
Surely frequency response of this experimental proto was not a beauty,
but i posted it due to "early reflections" we were discussing.
That early proto resembles a planar bending wave transducer in open baffle
configuration. Side wall reflection at LF is suppressed by the dipole notch.
Ringing around 5Khz is due to that proto version itself.
Attachments
Last edited:
Transients, those bästards !
...
Transient is an AM modulation. How fast can it be perceived? How fast is the human auditory system?
From the document we can read -10dB cut off points of the modulation transfer function of the auditory nerve fibers, and they are something like this:
at 500Hz: 100Hz
at 1kHz: 200Hz
at 2kHz: 400Hz
etc
To convert this into the time response in the modulation domain it is something like:
at 500Hz: 10ms
at 1kHz: 5ms
at 2kHz: 2.5ms
etc
It seems to follow constant Q characteristics.
Is this fast? Or rather slow?
At 1kHz the 'speed' is 5ms. How many reflections can fit in this time frame already
What is your "transient" ?
- Elias
____________________
OK, my first objection/question would be, whether it is admissible to conclude from neural
data to perception thresholds directly.
To my knowledge this is not admissible, even if neural data may give valuable insights into
the ecoding mechanism.
My second objection:
Your interpretation of the -10dB cutoff of the modulation transfer function as well as the
resulting "time response" is arbitrarily when applied to temporal resolution (of perception).
A modulation frequency of 100Hz causes a full cycle within 10ms, which means the amplitude of
a 100Hz AM modulated carrier rises from
- neutral level to maximum amplitude and back to neutral level
and then
- to minimum level and back to neutral.
during that time.
A step or a single impulse may be encoded in a quarter
of that cycle or a half cycle respectively ...
So even 1/4 or 1/2 of your "time domain resolution" could
be claimed the right value.
If a certain frequency fu is the highest to be encoded,
then time resolution (time between two samples) according
to Shannon's therorem is t <= 1/(2 * fu).
While explicitly not following your conclusions, i suggest your own notion
of "time response" to halve (at least), which would then be:
at 500Hz: 5ms
at 1kHz: 2.5ms
at 2kHz: 1.25ms
In the range of maximum sensitivity between 2Khz and 5Khz
(if i may deliberately supplement an approx. value at 4Khz according to cited data)
that would be
4Khz: 0.61ms
Which corresponds to a sound travelling distance of approx. 0.2 meters.
______
That 0.2 meters, as a sound travelling pathlength would reach or exceed the
detection threshold for group delay at 4Khz as e.g. found by Blauert 1978.
Of course discussion of perception thresholds for aspects of group delay
and reflection seems rather pointless, unless you manage to design a speaker/room
system which provides at least a few milliseconds of clearly dominant direct sound
to the listener.
Last edited:
That cited sentence i wrote was just a supplement. Overread it
if you like. It does in no way change the situation, that your own
standpoint looks rather fragile due to data cited by yourself.
______________
Excuse me Elias, weren't you citing even neural data to
support your point of view ?
Naturally in any study concerning perception idealized conditions
have to be setup.
And surely such results will be applicable more closely to non-reflective
environment than to "live" rooms.
But, like it has been shown more than once now, a sufficient ITD is achievable
even in small rooms, given speaker directivity, position and room
treatment (if necessary) work together.
if you like. It does in no way change the situation, that your own
standpoint looks rather fragile due to data cited by yourself.
______________
Excuse me Elias, weren't you citing even neural data to
support your point of view ?
Naturally in any study concerning perception idealized conditions
have to be setup.
And surely such results will be applicable more closely to non-reflective
environment than to "live" rooms.
But, like it has been shown more than once now, a sufficient ITD is achievable
even in small rooms, given speaker directivity, position and room
treatment (if necessary) work together.
Last edited:
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.