I haven't watched this thread for a longer time, but the in phase/antiphase measurements inspired by David starting here
http://www.diyaudio.com/forums/mult...-pattern-stereo-speakers-156.html#post2737030
made me thinking, whether a comparison of results between
the "optimum cancellation point" near the listening position
with alternative microphone positions
(say +/- 0.5m to the left/right and front/rear)
could be useful to estimate some kind of measure at least correlated
to "image stability" due to variation of the listening position.
I had that idea a few days before and then intuitively was re-reading
this thread, finding people conducted measurements using correlated pink
noise with their stereo setups ... with very interesting results.
http://www.diyaudio.com/forums/mult...-pattern-stereo-speakers-156.html#post2737030
made me thinking, whether a comparison of results between
the "optimum cancellation point" near the listening position
with alternative microphone positions
(say +/- 0.5m to the left/right and front/rear)
could be useful to estimate some kind of measure at least correlated
to "image stability" due to variation of the listening position.
I had that idea a few days before and then intuitively was re-reading
this thread, finding people conducted measurements using correlated pink
noise with their stereo setups ... with very interesting results.
Omni shock.
Hi,
Recently attended John Watkinson's lecture at the London AES section. Attendance was about 75 persons. He reminded us about the human hearing system, and the many things that he sees wrong with transducer and cabinet design. Then followed some recordings and live music presentations through his speakers. These were 2 floor standing full range, and a compact instrument cabinet. He emphasised the importance of the first 1ms of any reproduced signal, and how it determines the quality of what we hear. He also linked this to position location detection. Further he dimissed some of the long held held views on reverberant fields, phase, distortion and resonant systems(ported, TL etc). So, after some listening, in what was a large conference room, with hard surfaces. He dropped the bombshell, that these speakers were omni. Also that they were able to reproduce square waves over a wide bandwith. So how did they sound? Pretty amazing! I was sitting 3/4 of the way down the room, and didn't expect very much. But it felt as if I was only a few feet away. Jazz bass was really impressive. Closed box with two custom 12inch drivers with 6 inch voice coil and custom carbon composite cones.
So back to the drawing board for me, when it comes to directivity! In my pro audio world, directivity was always the holy grail, according to some. But I never really believed that wholeheartedly.
Iain.
Recently attended John Watkinson's lecture at the London AES section. Attendance was about 75 persons. He reminded us about the human hearing system, and the many things that he sees wrong with transducer and cabinet design. Then followed some recordings and live music presentations through his speakers. These were 2 floor standing full range, and a compact instrument cabinet. He emphasised the importance of the first 1ms of any reproduced signal, and how it determines the quality of what we hear. He also linked this to position location detection. Further he dimissed some of the long held held views on reverberant fields, phase, distortion and resonant systems(ported, TL etc). So, after some listening, in what was a large conference room, with hard surfaces. He dropped the bombshell, that these speakers were omni. Also that they were able to reproduce square waves over a wide bandwith. So how did they sound? Pretty amazing! I was sitting 3/4 of the way down the room, and didn't expect very much. But it felt as if I was only a few feet away. Jazz bass was really impressive. Closed box with two custom 12inch drivers with 6 inch voice coil and custom carbon composite cones.
So back to the drawing board for me, when it comes to directivity! In my pro audio world, directivity was always the holy grail, according to some. But I never really believed that wholeheartedly.
Iain.
Resurrecting a REALLY old thread here, and I can't quite find the correct section of the thread that I had in mind, but this post will do.
As I recall there was some debate earlier in the thread as to the significance of traditional "floor bounce" cancellation versus the 1/4 wavelength cancellation from a boundary, in relation to choosing woofer height from the floor.
On the one hand there were those who were of the opinion that the floor bounce effect was the primary issue of concern - with the floor bounce being the notch formed by path length differential between the direct sound from the woofer and the longer path length of the ray reflecting off the floor arriving at the same listening point.
This particular cancellation frequency is very spatial - as you get further away the angle becomes shallower and the frequency of cancellation goes up. It's not just one fixed frequency but depends on the listeners location, so it stands to reason that whilst it might cause localised cancellation it won't have much if any effect on the overall power response within the room.
On the other hand there were those who were more concerned with the 1/4 wavelength cancellation frequency - occurring at the frequency where the woofer is a 1/4 wavelength from the nearest boundary, often the floor. This frequency is always significantly lower than the "floor bounce" cancellation frequency, for example for a woofer 75cm from the floor its around 115Hz, while floor bounce from the same woofer at a 2.5 metre listening distance may be more like 250Hz. (Rough numbers plucked from memory for the sake of discussion without being checked for accuracy)
The question is whether this cancellation which occurs at 1/4 wavelength is audible at far points of the room, given that the woofer to floor axis is roughly at right angles to the listener.
Months later I was reading an early 90's interview with Roy Allison which contained an interesting little nugget of information which I'm not sure I've ever seen anywhere else. The interview can be found here, starting on page 46 of the PDF:
http://theaudiocritic.com/back_issues/The_Audio_Critic_18_r.pdf
I'll quote the interesting bits:
So in other words he's saying that the boundary cancellation caused by 1/4 wavelength cancellation from each of the near boundaries doesn't just cause cancellation at one (or a few) point(s) in space along that axis (for example if you were measuring from above the woofer) but that it actually reduces the overall power response within the entire room at that frequency.
EG you could take a 100 different measurements in a 100 different randomly chosen locations in the room, average them, see most of the room / modal variations average out spatially, but see persistent dips in the response at frequencies corresponding to 1/4 wavelength from the nearest boundaries including the floor.
This is something I've observed in my own measurements anecdotally, but have never done enough spatially averaged measurements to confirm, nor have I seen it talked about. (The usual explanation is "everything is due to room modes")
He goes on to say (interviewer also quoted):
So again, if multiple boundaries are roughly the same distance from the woofer their interference is cumulative. Not a big surprise, but given that each boundary is along a different axis in 3d space, perhaps a bit of a surprise that all will contribute to the loss in power response, thus a loss in response in every part of the room, once modal effects are smoothed out.
Having a high mounted woofer in the 75-90cm height range is very likely to be roughly the same distance from both the floor and the front wall in many typical listening setups, causing the cancellation in the power response to be cumulative at the same frequency. Not good. Having a low mounted woofer at least eliminates one source of 1/4 wavelength cancellation.
He goes on to say:
Now this is where I start to have a bit of trouble. I've actually suggested myself before perhaps even in this thread, that the nearby floor boundary out of phase reflection might alter the radiation resistance seen by the woofer enough to alter its total (eg power) output, but I was shot down in flames, and afterwards more or less accepted the fact that, no, boundary cancellation in rooms does not significantly alter the movement of the cone of a woofer or its radiation resistance, thus can't affect its total power output to any significant degree.
So is he right that it does, or is he just accepting it as his explanation for a real effect that he's measured without rigorously proving it as the cause ?
I'm wondering if the overall loss in power response is simply because the bass is more or less cancelled entirely at certain frequencies in certain directions (in line with the woofer/boundary axis for a given reflection) which after reflecting all around the room results in a net loss in power response without any alteration of the excursion of the driver itself or significant change in radiation resistance.
Thoughts anyone ?
One thing is for sure, I think it definitely shows the typical upper bass 100-200Hz hole that Wayne persistently keeps reminding us all about is very real and IS a result of 1/4 wavelength boundary cancellations NOT a floor bounce differential cancellation. (Which is also present but much less troublesome)
This also has implications when choosing the correct crossover frequency for a low woofer high midrange design, since the 1/4 wavelength cancellation occurs much lower in frequency than the floor bounce cancellation. Ideally you'd want to keep both effects on the right side of the crossover frequency for each driver...
Last edited:
Having done a plethora of work in LFs in small rooms, I don't buy it. First, averaging the sound throughout the room WILl NOT average out the modes (been there, done that). The modes create a very real change in the acoustic load presented to the source and as such the power response from this source will retain this modal character. To be able to see a chage that could be correlated to 1/4 wavelength amongst the multitude of very real modal effects is like trying to find a needle in a haystack.
In a large room this might certainly be possible (high modal density at the relavent frequencies), but not in a room where these frequencies are below the Schroeder Frequency.
I am not saying that this phenomina does not exist only that it will pale before the modal situation in this frequency region and it is in fact already incorprated into the modes. ALL LF effects are accounted for in the modal solution. One does not need to add an effect related to a 1/4 wavelength its already there to the extent that it exists. Solutions based on single planes are misleading when there are other planes within a few wavelengths. One must consider all planes that exist within a few wavelengths, not just one.
Last edited:
Doesn't that reinforce my point though ?
The uniformity of a given speakers woofer height presents a persistent hole in the response that is independent of the size of the room or its modal characteristics. (When I say independent I mean in addition to all these other effects)
As you point out the listeners point being a fairly uniform distance from the floor and (often) the ceiling is an issue too. We can't place the listeners head at the floor or the ceiling but we can place the woofer at the floor to at least eliminate one issue.
I think the behavior is modal. We're talking about small rooms with fewer LF modes than large rooms - rooms with a somewhat uniform ceiling height and a somewhat uniform placement of speakers and listener. This results in dips in more or less the same frequency region.
Ok I'll give you that point. I'm not suggesting that averaging would give a flat response, clearly not.
Really ?
Room modes occur at frequencies dependant on the room dimensions such as length height and width. Boundary cancellations occur at frequencies dependant on the distance from the boundary to the woofer or listener.
Move a speaker along one axis of a room and the amplitude of excitation of that mode changes but the frequency does not. Move a speaker closer and further from a boundary and the frequency of cancellation changes but amplitude only varies gradually due to spreading losses.
By simply taking measurements with the speaker in a number of different locations it should be possible to identify which features in the response are due to modes and which are boundary cancellation effects based on which ones move in frequency and which only change in amplitude.
If this were true, show me how your modal solution copes with a room with only three boundaries - a floor, a front wall and one side wall. No roof or other walls. Where are your modes now ?
There will still be boundary cancellation and reinforcement occurring from each boundary in such a "room". This is why I don't buy the "modes explain everything" explanations no matter how tempting the maths might look.
Reflections explain both boundary cancellation and modes. Modes don't explain everything by themselves.
Last edited:
I had a debate with a fellow over at the Classic Speaker website where Allison is much revered. He interpreted Allison as meaning that the woofer saw its reflection and modified its cone motion to create the power dip.
I don't believe that is the case. What we usually find with loudspeakers is that the radiation impedance is typically pretty small compared to reactances of mass and compliance and moving a speaker from an anechoic chamber to a room seems to have no visible effect on cone velocity.
The Allison papers show 1, 2 and 3 boundary power responses. Remember that power response can be measured by sampling in space around the surface of a sphere. You simply unlog to get back to pressure, square the curve to get to intensity, add all the curves, divide by the number and then log again to get back to "power". I'll put it in quotes because its reall an intensity average (as Toole often shows), but if you use the right calibration shift for the distance you can turn it into true Watts.
Now, you can calculate this for a pair of woofers where one woofer is the real woofer and the second is the reflected woofer behind the boundary. Equally, you can calculate the same for 4 woofers (back wall, floor simulation) or 8 woofers (corner simulation). If you calculate for all the points in your space (i.e. 1/8th space for the corner case) then you would be calculating the radiated "power" assuming the 8 woofers had absolutely no impact on each other. In other words, the calculation is of the 2, 4 or 8 units and their pressure sums from the geometry of the situation. All are given a starting impulse and the real woofer is closest, the others add with delayed response.
My belief is that this is all that is needed to create Allison's curves and so he is modeling the response of woofers that sum independently, they do not interact in a way that one impacts the volume velocity of the next. Allison talks as if that is the case but his curves don't support it.
The next question is whether the power response is the right way to look at this when considering a speaker in a room.
If you were going into a real room and averaging all over the room, then the power response of the woofer and its near boundaries may be the way to look at it. The dip in the power curve might be seen in the room average, with the addition of all of the room modes as well.
Still, it would seem to be more useful to calculate the real response at a particular listener position. From that you would use the image model and the virtual woofers would play a more specific role since their exact geometry can be considered. (Note that interaction between woofers isn't considered in the frequently used image model.)
Returnng to the Allison curves you will see that the speaker in a corner has a much more severe null than the 2 or 1 boundary cases. I would explain that as this: when you have a single boundary you get to move around the virtual pair over a wide range of angles (180 degrees each way). At some angles to the side, the sound from the two units arrives in phase and no cancellation occurs. This softens the look of the null as well as shifting it up in frequency. For the 3 boundary case you are much more constrained in your sampling sphere section and over most of the angles the cancelation is strong and virtually at a fixed frequency. I take this as further proof that the nulls are due to vector summation effects rather than any changes in cone volume velocity.
As a practical observation, I see the Allison dip very strongly when I measure a woofer in a room from a fairly close distance. As I move the microphone away, the first dip frequency rises and the room effects become relatively stronger. It becomes hard to find the Allison dip at most typical listening positions, unless the speaker is pretty close to a back wall.
Now when Allison was analyzing rooms in the Boston area, he averaged response around each room. That would tend to reveal the Allison dip, while tending to downplay the room modal response. I'm just saying...
Regards,
David S.
I don't believe that is the case. What we usually find with loudspeakers is that the radiation impedance is typically pretty small compared to reactances of mass and compliance and moving a speaker from an anechoic chamber to a room seems to have no visible effect on cone velocity.
The Allison papers show 1, 2 and 3 boundary power responses. Remember that power response can be measured by sampling in space around the surface of a sphere. You simply unlog to get back to pressure, square the curve to get to intensity, add all the curves, divide by the number and then log again to get back to "power". I'll put it in quotes because its reall an intensity average (as Toole often shows), but if you use the right calibration shift for the distance you can turn it into true Watts.
Now, you can calculate this for a pair of woofers where one woofer is the real woofer and the second is the reflected woofer behind the boundary. Equally, you can calculate the same for 4 woofers (back wall, floor simulation) or 8 woofers (corner simulation). If you calculate for all the points in your space (i.e. 1/8th space for the corner case) then you would be calculating the radiated "power" assuming the 8 woofers had absolutely no impact on each other. In other words, the calculation is of the 2, 4 or 8 units and their pressure sums from the geometry of the situation. All are given a starting impulse and the real woofer is closest, the others add with delayed response.
My belief is that this is all that is needed to create Allison's curves and so he is modeling the response of woofers that sum independently, they do not interact in a way that one impacts the volume velocity of the next. Allison talks as if that is the case but his curves don't support it.
The next question is whether the power response is the right way to look at this when considering a speaker in a room.
If you were going into a real room and averaging all over the room, then the power response of the woofer and its near boundaries may be the way to look at it. The dip in the power curve might be seen in the room average, with the addition of all of the room modes as well.
Still, it would seem to be more useful to calculate the real response at a particular listener position. From that you would use the image model and the virtual woofers would play a more specific role since their exact geometry can be considered. (Note that interaction between woofers isn't considered in the frequently used image model.)
Returnng to the Allison curves you will see that the speaker in a corner has a much more severe null than the 2 or 1 boundary cases. I would explain that as this: when you have a single boundary you get to move around the virtual pair over a wide range of angles (180 degrees each way). At some angles to the side, the sound from the two units arrives in phase and no cancellation occurs. This softens the look of the null as well as shifting it up in frequency. For the 3 boundary case you are much more constrained in your sampling sphere section and over most of the angles the cancelation is strong and virtually at a fixed frequency. I take this as further proof that the nulls are due to vector summation effects rather than any changes in cone volume velocity.
As a practical observation, I see the Allison dip very strongly when I measure a woofer in a room from a fairly close distance. As I move the microphone away, the first dip frequency rises and the room effects become relatively stronger. It becomes hard to find the Allison dip at most typical listening positions, unless the speaker is pretty close to a back wall.
Now when Allison was analyzing rooms in the Boston area, he averaged response around each room. That would tend to reveal the Allison dip, while tending to downplay the room modal response. I'm just saying...
Regards,
David S.
There is no such thing as boundary cancellation effects that are independent of modal effects. They are the same thing.
If this were true, show me how your modal solution copes with a room with only three boundaries - a floor, a front wall and one side wall. No roof or other walls. Where are your modes now ?
It can't of course, but then a three walled space IS NOT A ROOM!! It's not what we are talking about - or certainly not what I am talkking about. That space would be called a "large room" where the Schroeder Frequency has gone to zero - free space. Nothing modal applies.
Yes they do. The modal solution is complete, nothing is lost and nothing can or needs to be added for it to be correct. This is fundamental physics.
Last edited:
Which means there is an unlimited number of modes which is beyond my imagination if we think of modes as the superposition of two or more standing waves.
I have to agree.
One of the reasons why I posted a link to the interview is that whilst I believe that the effects of 1/4 wavelength cancellation that he observed are real enough, he seems to have taken a punt on the cause of it, without really taking any steps to prove that his theory is correct.
It does indeed seem to be easily explainable through simple reflection based modelling without any interaction between the real source and image sources. In fact the room reflection calculator I use which is based on the image model does show the effects he discusses and almost certainly doesn't model driver interaction.
I think where the power response becomes useful in the discussion rather than the response at a particular location is that if you have a hole in the power response that is almost independent of the size and shape of the room or speaker and listener locations, but dependant of the height of the woofer from the floor (a known constant in a floor standing speaker) then it gives us the ability to "fix" at least one room problem in the speaker design in a way that will be "correct" in virtually any room.
In other words almost anything else that you might do to a speaker to "compensate" for room behaviour in the bass region is almost certainly going to make some rooms better but a lot more worse, however a floor mounted woofer (provided that the floor gain is adjusted for in the network) can be an improvement which is "correct" or at least not detrimental in any particular arbitrary room.
Even though there will still be some cancellation from front and side walls, you now only have 2 distances that need to be unequal to minimize "stacking" of notches, instead of trying to evenly distribute 3 notches, which means less total cancellation and more freedom to vary front and side wall distances without large combined notches forming.
I think that's valuable, not only because it measures better in room, but also because subjectively it sounds better. To me there is just something very "hollow" about mid/upper bass that is produced too far from any room boundaries. (I have to wonder if its related to the huge spike in group delay that often forms at those cancellation frequencies due to delayed reflections exceeding the amplitude of the direct signal)
Sure, its easy to say "but the room will make a total mess of the bass response anyway, why bother", to which I would argue that systematic one by one elimination of flaws is the way you get from an average result to an excellent result, especially if you can eliminate a problem more or less "for free". (Design constraints permitting)
What you're talking about is the floor bounce reflection though - a discrete reflection at one location which shifts up in frequency with distance. When you're at the woofer the 1/4 wavelength and floor bounce cancellation occur at the same frequency.
As you've noted once you get a certain distance away it becomes hard if not impossible to measure or hear this notch in a room, most likely because the multitude of unequal path length reflections from elsewhere in the room "fill in" the notch.
However if the 1/4 wavelength effect really does put a dent in the total power response in the room, on average there will be a general depression in this area at most listening positions which doesn't shift in frequency with listener position.
Last edited:
It makes no sense. If there are no parallel walls to allow a standing wave to build up there can't be any modes.
Infinite mode density is not equal to no modes.
Last edited:
Actually Simon, it is. Think of a room that gets bigger and bigger, What happens to the modal density? It gets denser and denser until it forms a continuum of modes - no more discrete modes, i.e. no modes. This too is a typical aspect of physics. It's true of optics, quantumm mechanics, virtually all fields of physics will exhibite this kind of behavior in one form or another.
Then don't think of it as a superposition of standing waves. Modes are a mathematical contruct and need not conform to our intuitive sense of what is right. Think of quantuum mechnaics where things are not at all what they seem. There is no "empty space", it's all filled with particles and anti-particles that exactly cancel to make it "seem" empty.
One thing that I have learned studying physics is not to let your intuition get the better of you. The math is always right but intuition need not be so.
Then don't think of it as a superposition of standing waves. Modes are a mathematical contruct and need not conform to our intuitive sense of what is right. Think of quantuum mechnaics where things are not at all what they seem. There is no "empty space", it's all filled with particles and anti-particles that exactly cancel to make it "seem" empty.
One thing that I have learned studying physics is not to let your intuition get the better of you. The math is always right but intuition need not be so.
Good thing speakers aren't like some aspects of quantum mechanics where simply observing or measuring something can change it's behavior.
Or does it....
*boggle*
This kind of gets us back to the radiation resistance subject that we debated some time back. Put a pair of sources in space and measure them from a position equidistant from both. Pressure doubles, or the response curve goes up 6dB.
6dB would be 4 times the power, so where did the power increase come from (an efficiency doubling if we doubled the input by having 2 speakers)? Answer: mutual coupling or an increase in radiation resistance. Is it doubled? Depends on whether the sources are near to each other or not. Consider them as two independent sources and find the spherical power average around them. If it is less than 6dB greater than one source (or perhaps it is 6dB at low frequencies and less at higher frequencies) then the difference is a directivity gain.
In all cases the axial gain from two sources is 6dB. The 6dB must be the sum of the radiation impedance rise and the directivity rise (usually primarily radiation impedance at low frequencies and directivity at high).
So we talk about this mystical factor of radiation impedance gain or mutual coupling, but we calculate it simply by an integration of 2 independent sources.
Allison says the two woofers see each other and, due to their spacing, cause a power response hole in radiated output. He calculates it via math assuming two independent sources.
Finally, if the sources saw each other and this impacted the power response, every MTM we made would have odd abberations in the forward response due to the spacing between the units. This doesn't happen. There is no impact in the response at 90 degrees to the line of the two units.
David S.
6dB would be 4 times the power, so where did the power increase come from (an efficiency doubling if we doubled the input by having 2 speakers)? Answer: mutual coupling or an increase in radiation resistance. Is it doubled? Depends on whether the sources are near to each other or not. Consider them as two independent sources and find the spherical power average around them. If it is less than 6dB greater than one source (or perhaps it is 6dB at low frequencies and less at higher frequencies) then the difference is a directivity gain.
In all cases the axial gain from two sources is 6dB. The 6dB must be the sum of the radiation impedance rise and the directivity rise (usually primarily radiation impedance at low frequencies and directivity at high).
So we talk about this mystical factor of radiation impedance gain or mutual coupling, but we calculate it simply by an integration of 2 independent sources.
Allison says the two woofers see each other and, due to their spacing, cause a power response hole in radiated output. He calculates it via math assuming two independent sources.
Finally, if the sources saw each other and this impacted the power response, every MTM we made would have odd abberations in the forward response due to the spacing between the units. This doesn't happen. There is no impact in the response at 90 degrees to the line of the two units.
David S.
Although a reflection may technically be a mode, I still suspect they have unique qualities in a real situation. Although you say that they are not modes without the other three walls but are with them, it seems only the first three that count. These reflections typically above 100Hz with smaller wavelengths have clear cut cancellations that become more diffuse by the time they come back from across the room. They may not even be reproduced long enough to make the journey, and may be measured in a way that reduces the later reflections anyway. Seeing them as reflections, they converge over a large space compared to their wavelength as opposed to normal modes.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.