Hello,
Please take into account that for Ambiophonics there is cross talk cancelling involved so the situation is different than in a free sound field.
- Elias
Please take into account that for Ambiophonics there is cross talk cancelling involved so the situation is different than in a free sound field.
- Elias
Hello,
Thx for the link Skorpion, but I have to emit a warning here, if not so any discussion with these data in mind will be useless (immediately you can try it on the (big) post just before, it gives an other lightning).
There is a typo mistake :
ILD cues work well only for signals with energy between 90 Hz and 1,000 Hz.
and :
Like ILD cues, ITD cues work really well only for signals with energy between 90 Hz 1,000 Hz .
Hello,
Let me act as an opponent for a moment
Yes, I want to hear the acoustics of the recording room. And No, I will not use headphones!
If we are talking about the midrange (that's where the music lives) then omnidirectivity is propably the worst choice in normal domestic rooms ! More reflections in this range will only introduce forward and backward masking by the room which leads to less articulation and to loss of small detail and actually prevents the recording acoustics to come through.
Also, concert hall is a poor comparison for the small room problem, first of all the temporal spread of reflection does not have similarity of a small rooms and halls are designed to have reflection patterns that match the temporal bandwidth of the music played therein, that a small room can never achieve.
Griesinger has some nice presentation on the subject of halls and acoustics.
- Elias
Let me act as an opponent for a moment
Yes, I want to hear the acoustics of the recording room. And No, I will not use headphones!
If we are talking about the midrange (that's where the music lives) then omnidirectivity is propably the worst choice in normal domestic rooms ! More reflections in this range will only introduce forward and backward masking by the room which leads to less articulation and to loss of small detail and actually prevents the recording acoustics to come through.
Also, concert hall is a poor comparison for the small room problem, first of all the temporal spread of reflection does not have similarity of a small rooms and halls are designed to have reflection patterns that match the temporal bandwidth of the music played therein, that a small room can never achieve.
Griesinger has some nice presentation on the subject of halls and acoustics.
- Elias
With respect to you Elias - and I do respect you - buuuuuuuut - I'm an Omni devote as are many others. You got your perspective 'N I have mine. We have both been at this little hobby for many years - and I have listened to some of the very best equipments and they all have their usefulness to meet a listeners wants and needs. Which takes me back to - I like Omni!
I will not attempt to prove my perspective by listing the many makers of highly regarded speakers that are Omni in design and use as I feel it would be argumentative and perhaps insulting - and I don't want to go there. Just making the point that I am not alone in my enjoyment of Omni (as well as OB).
You have experienced these results with the "fingerprint" reflections Aleksandar was talking about? Could you give some more information/details of your layout/design?
I believe that Aleksandar wasn't comparing, but differentiating between concert hall and small room. Notably because concert halls have vastly different acoustics, it is easy for our brain to distinguish the recorded hall from the small listening room acoustic.
Rudolf
Both camps have strong point in this discussion, I think. My goal is to combine both worlds, corner points being the open, "in-the-air" 3D sound of omnis in properly treated large rooms and the uncolored directness and big/close image of freefield listening, especially with very dynamic speakers...
... which I happen to have now (15"-PA-coax, OB, DSP-corrected for identical and flat amplitude and phase at listening pos), as well as heavily damped room -- too much for many, but I like it "dry" in the living room. RT60 levels are within EBU studio limits. Speaker polars are sure not perfectly independent of freq, but at least much better than conventional, average 3-way system.
I completely agree with Linkwitz and Aleksander that this kind of domestic sound is neither adequate nor pleasing with the majority of recordings, it just sounds flat (imaging) and lifeless (tonally), pretty much unengaging and sterile... unless the recording includes excellent acoustics and engineering to provide enough and working spatial clues that make up a comfortable auditory scene. Not many albums in my collection qualify, say 5 or so...
Now comes in what some may call blasphemy, I worked long and found good ways to dial in virtual room acoustics that allow me to adjust representation to my likings within a range from low ambience "in-your-face" to moderate room ambiences and finally to concert-room size, all with increased vertical size/position imaging and ageneral "3D"-quality (a speciality topic for mine that drives me all the time). This sure is far from perfect but personally I arrive at strong moments of emotion when listening, when I open my eyes after I while more often than not I'm shocked that the speakers really are that close (2m) and I'm listening in that (rather small) room, while the action is located in all the space behind and around them which I can modify to get the most realistic space illusion (again that drives me, I can't help it).
Before my current setup I had conventional, rather big 3-ways in the standard simple setup, CD->Amp->Spkr. I tried live room acoustics but I didn't work with them, neither did damping the room, the speakers never really disappeared as sources and always sounded colorered...
My 2 ct's, Klaus
... which I happen to have now (15"-PA-coax, OB, DSP-corrected for identical and flat amplitude and phase at listening pos), as well as heavily damped room -- too much for many, but I like it "dry" in the living room. RT60 levels are within EBU studio limits. Speaker polars are sure not perfectly independent of freq, but at least much better than conventional, average 3-way system.
I completely agree with Linkwitz and Aleksander that this kind of domestic sound is neither adequate nor pleasing with the majority of recordings, it just sounds flat (imaging) and lifeless (tonally), pretty much unengaging and sterile... unless the recording includes excellent acoustics and engineering to provide enough and working spatial clues that make up a comfortable auditory scene. Not many albums in my collection qualify, say 5 or so...
Now comes in what some may call blasphemy, I worked long and found good ways to dial in virtual room acoustics that allow me to adjust representation to my likings within a range from low ambience "in-your-face" to moderate room ambiences and finally to concert-room size, all with increased vertical size/position imaging and ageneral "3D"-quality (a speciality topic for mine that drives me all the time). This sure is far from perfect but personally I arrive at strong moments of emotion when listening, when I open my eyes after I while more often than not I'm shocked that the speakers really are that close (2m) and I'm listening in that (rather small) room, while the action is located in all the space behind and around them which I can modify to get the most realistic space illusion (again that drives me, I can't help it).
Before my current setup I had conventional, rather big 3-ways in the standard simple setup, CD->Amp->Spkr. I tried live room acoustics but I didn't work with them, neither did damping the room, the speakers never really disappeared as sources and always sounded colorered...
My 2 ct's, Klaus
Basically it's just adding simulated room reflexions and reverb. One point is that these "reflexions" can be easily made to have zero problems with heavily disturbed frequency and time responses, in fact that is the easiest way. I'm using various reverb plugins (my system is PC-based), either real-time or with extracted and further manipulated impulse responses.
The second component is a bit more complex to explain, in short terms, it is using specialized crossfeeds inspired by HRTF-relationships which produce different, richer localisation cues also by subtle comb filtering effects steered by level and time differences of the encoded stereo content, this makes the individual L/R-signals become a bit of phantom-sources themselves which helps the speakers to disappear better. The added reverberation passes through a different "projection cells" (as I call them, actually they are cross-convolution kernels) than the main signal.
Needless to say that all this is very difficult to dial-in, the signal flow is complicated and you can quickly ruin it or get lost, side-effects (there are some, of course) must be at tolerable levels and there must be a true net benefit. To judge that, I use long periods of testing to find out if the experienced "improvement" holds.
As you can see, this might be worth its own thread, I might run one when I have gained more solid experience and insight, hopefully.
The second component is a bit more complex to explain, in short terms, it is using specialized crossfeeds inspired by HRTF-relationships which produce different, richer localisation cues also by subtle comb filtering effects steered by level and time differences of the encoded stereo content, this makes the individual L/R-signals become a bit of phantom-sources themselves which helps the speakers to disappear better. The added reverberation passes through a different "projection cells" (as I call them, actually they are cross-convolution kernels) than the main signal.
Needless to say that all this is very difficult to dial-in, the signal flow is complicated and you can quickly ruin it or get lost, side-effects (there are some, of course) must be at tolerable levels and there must be a true net benefit. To judge that, I use long periods of testing to find out if the experienced "improvement" holds.
As you can see, this might be worth its own thread, I might run one when I have gained more solid experience and insight, hopefully.
Last edited:
Hello,
Neither of them is more important ! Let me explain.
For me number one important thing is for stereophonic effect is that the loudspeakers must not be localised, even when turning your head.
Because if a speaker is localised stereo is gone! No phantoming anymore.
For me the minimum turning of head is about +/-30 degrees. Take into account that head rotation happends subconsciously you want it or not. Only head fixture can prevent it. Also if you are jamming with the music your head will shake
In my experience speaker localisation in this circumstance is due to energy in the treble range.
Now there are two methods that I've succesfully implemented to achieve non localised speakers even when turning the head.
1) Ambiophonics. Speakers at front + cross talk cancelling. No explanation needed here, check out their home page.
2) Normal speaker placement but with very wide dispersion treble, almost close to omni above 3kHz. When placed in a normal domestic room, this will spatially homogenise high freq cues so no direction is preferred over another. Stereo works with the midrange cues ILD + ITD.
How does the frequency response helps in this? Nothing ! Phantoming is possible regardless of frequency response.
How does the power response helps in this? Nothing ! It's mainly to define tonal balance.
Achieving this can be somehow critical (but I'm not yet sure how critical) and individual differences can be big. And of course this can be very much room dependent, because treble absorbtion can vary from room to room!
What I do is I find the optimum dispersion for the treble by measuring the room impulse response and visualise it in the time-freq domain to see the temporal spread and frequency content of reflections. By modifying the relation of direct sound to the reflections I can find an optimum where speakers are not localised.
There is an optimum! Namely, if I remove the direct sound speakers are not localised for sure, but small detail is also lost. And if there are no reflections detail is great but coming from the speakers. The optimum is fine, it allows non localised speakers and small details of the recording.
Nice
- Elias
Neither of them is more important ! Let me explain.
For me number one important thing is for stereophonic effect is that the loudspeakers must not be localised, even when turning your head.
Because if a speaker is localised stereo is gone! No phantoming anymore.
For me the minimum turning of head is about +/-30 degrees. Take into account that head rotation happends subconsciously you want it or not. Only head fixture can prevent it. Also if you are jamming with the music your head will shake
In my experience speaker localisation in this circumstance is due to energy in the treble range.
Now there are two methods that I've succesfully implemented to achieve non localised speakers even when turning the head.
1) Ambiophonics. Speakers at front + cross talk cancelling. No explanation needed here, check out their home page.
2) Normal speaker placement but with very wide dispersion treble, almost close to omni above 3kHz. When placed in a normal domestic room, this will spatially homogenise high freq cues so no direction is preferred over another. Stereo works with the midrange cues ILD + ITD.
How does the frequency response helps in this? Nothing ! Phantoming is possible regardless of frequency response.
How does the power response helps in this? Nothing ! It's mainly to define tonal balance.
Achieving this can be somehow critical (but I'm not yet sure how critical) and individual differences can be big. And of course this can be very much room dependent, because treble absorbtion can vary from room to room!
What I do is I find the optimum dispersion for the treble by measuring the room impulse response and visualise it in the time-freq domain to see the temporal spread and frequency content of reflections. By modifying the relation of direct sound to the reflections I can find an optimum where speakers are not localised.
There is an optimum! Namely, if I remove the direct sound speakers are not localised for sure, but small detail is also lost. And if there are no reflections detail is great but coming from the speakers. The optimum is fine, it allows non localised speakers and small details of the recording.
Nice
- Elias
Elias,
maybe "linear" was misleading
Rudolf
Thank you, Elias!
I'm happy again to draw some information from you, that really makes sense to me. Of course this intermittent lack of understanding must be a deficit on my side, but I'm working on it ...
When I insist on finding something, that you want to be "linear" to the frequency response, it would be the decay of the reverberant field. Could you agree to that?
I can easily follow your explanation how the ratio of direct to reverberant field controls the difference between best detail and best homogenity of the stereo illusion in your system.
In post #162 you oppose the usability of an omnidirectional midrange. I have seen the Bark wavelets of your system at the S15 Econowave thread, but can't identify any special attenuation of the midrange in the diagrams.
It would help me to compare, where I stand related to your effort, if I could see the same wavelet for my loudspeaker.
I have attached an ARTA zip of my ungated impulse response at the listening distance of 2 m. Would be very nice indeed if you could translate it to the time/frequency domain too.
Rudolf
Hi,
Sharing his room impulse response is like inviting for a visit lol Thank you.
Here comes:
20ms:
There is a cross over at 200Hz?
20ms normalised:
150ms:
150ms normalised:
300ms normalised:
You seem to have a great room!
How you keep below 300Hz so clean? You have bass traps? Wooden house?
Look at my room with a monopole speaker:
- Elias
Sharing his room impulse response is like inviting for a visit lol
Thank you.Here comes:
20ms:
An externally hosted image should be here but it was not working when we last tested it.
There is a cross over at 200Hz?
20ms normalised:
An externally hosted image should be here but it was not working when we last tested it.
150ms:
An externally hosted image should be here but it was not working when we last tested it.
150ms normalised:
An externally hosted image should be here but it was not working when we last tested it.
300ms normalised:
An externally hosted image should be here but it was not working when we last tested it.
You seem to have a great room!
How you keep below 300Hz so clean? You have bass traps? Wooden house?
Look at my room with a monopole speaker:
An externally hosted image should be here but it was not working when we last tested it.
- Elias
I have attached an ARTA zip of my ungated impulse response at the listening distance of 2 m. Would be very nice indeed if you could translate it to the time/frequency domain too.
Rudolf
Hi,
See post 279:
http://www.diyaudio.com/forums/multi-way/164029-wtf-wavelet-transform-audio-measurements-what-how-6.html#post2319713
At the moment it may be the best summary of Bark wavelet. Cannot say it's very user friendly
- Elias
See post 279:
http://www.diyaudio.com/forums/multi-way/164029-wtf-wavelet-transform-audio-measurements-what-how-6.html#post2319713
At the moment it may be the best summary of Bark wavelet. Cannot say it's very user friendly
- Elias
While on that topic (and while we're off-topic from this thread) - is there a summary in the WTF thread that describes how you're generating those wavelet plots? I stopped reading that thread once it got too deep into theory that I could not follow
Hi Elias,
that room is somewhat my office. I would not dare to let your sharp eye look into one of my more private rooms.
Indeed you are extracting (too?) much detail from the dataThere is a cross over at 200Hz?
. The cross over from the H frame woofers to the OB mids is ~300 Hz. I have published some more information about that room and system at audio circle. The front wall (behind the speakers) is a dry wall in the lower part and wood in the pitched roof area. There is a brick wall to the right and back. Left wall is a combination of glass, brick and dry wall. Floor and ceiling are concrete.You may be right. But notice that there are no special acoustic treatments in the room. No absorbers, diffusors or bass traps. Just room high shelf units at the back wall, which might work as a sort of diffusor. And a dense short pile wall-to-wall carpet.
Elias, your wavelets are really great. Thank you very much for taking the trouble to extract them. Somewhere in the WTF thread JohnK remarked that wavelets are nothing else than burst decays. I have tried the burst decay function in ARTA with the same data - no competition at all! It is a pity that there is no easy way to extract wavelets from the IR. What I have seen in the WTF thread is too complicated to get me involved.
Nevertheless I'm still not exactly sure what in these diagrams is real data, and what is artefacts or visual noise. The old "What you see is NOT what you hear" syndrome. It looks like I have a high density of broadband (not much attenuation at high frequencies) reflections, which decay quite fast and evenly. I very much attribute that to the controlled dipole pattern, which illuminates the room in almost all horizontal directions - excluding only the points of first wall reflections.
I am trying to get the best possible dipole 8 pattern from my system.
As a last detail I include the (not normalized) RT30 diagram:
Attachments
This reminds me in the experiments, Wolfgang Klippel has conducted 1990, which are also mentioned in Toole's book. I am not sure you have it or know the content so I'll summarize briefly:
It was not directly about hiding a speaker by finding the optimum direct to reflected sound ratio but it was about the optimum direct to reflected sound ratio for a "feeling of space", which contributed to attributes "pleasantness" and "naturalness". Klippel found that the optimum is reached when the reflections of the room exceed the direct sound by 5dB for music (3dB for speech). The values do not consider frequency dependent absorption in the room.
For me hidden speakers belong to the attributes "pleasantness" and "naturalness". That's why I see a relation to what you describe and maybe this info is helpful for you.
I agree, read on.
I believe that you have a misconception about sound power in mind. So first of all, I am talking about dipole tweeters now to match the topic of the thread:
With dipole tweeters you cannot reach constant directivity all the way up, which actually equals constant sound power, due to the ratio of size (also dipole length) and wavelength radiated. Rudolf has shown that perfectly in this thread.
Every dipole tweeter will start to beam, be it domes or planars. The highs will sooner or later have a forward/backward bias => directivity index goes up as the sound power falls and there are less reflections => there is more direct sound => the speakers unhide. I strongly believe that this is "your" above mentioned energy problem. You have more direct energy than reflected energy; there is an imbalance. Yes, wide dispersion will help. With that you can reach above 3KHz. Way above 5K for 1" dome tweeters...nope.
All you would need to perfectly hide a speaker is one with constant directivity (constant power), wide dispersion (e.g. omni or dipole) and a room that is able to reflect the sound in a similar shape as the direct sound.
So sound power is much more than tonal balance. The more uniform it is the better it supports e.g. the precedence effect.
I would also be interested in knowing how to create and read your diagrams !
What about a special dipole-tweeter that uses one voicecoil for a backside and a frontside tweeter. That should be small and thin enough to work perfectly up to highest frequency. The problem is you can't use pole pieces in this configuration, but two rings of NdFeB should to pretty good as well
Here's a sketch (voicecoil is 25mm):
Sorry to revive a 6 month old thread, and specifically this post from more than a year ago. But I was bored this afternoon (on call... waiting for the pager to start blowing up...) and have been thinking about dipole tweeters for the past couple of weeks. I remembered reading this thread but had to do some digging to find it and this post.
I didn't want to start a new thread when so much good information was already contained here... though I was tempted.
Any more thoughts on the above proposal? One obvious problem is that the inner magnet or steel circuit has to be supported somehow... I think a sort of double voice coil sleeve (inner and outer portion) with the outer being split might work. Sort of two tweeters joined back to back by an inner sleeve. Like the quick sketch below shows.
More poking around and I realize this sketch is pretty close to how some tweeters of today look, particularly the hexadym motor geometry. Well, they are half of this sketch at least. I actually think if I had the money and some time I could take two hexadym tweeters and construct a single good dipole tweeter that was effectively a dual motored tweeter... sort of like JBL's differential drive geometry on a small scale. But this is unnecessarily doubling up on some geometry.
I actually think constructing a good dipole tweeter that had minimum thickness and also minimal motor diameter is entirely possible, such that the dipole length was quite small and true dipolar behavior was maintained well into the upper midrange.
What say ye o' wise ones?
Attachments
Well, if by "o'", you mean old, then I'm at least 50% qualified.
I find the idea clever, but I'm at a loss as to why you'd want to go to the trouble for a tweeter. I haven't read this thread since it was active, so I may be off-base, but here goes.
For much of its range the front and back waves of this tweeter will be decorrelated, i.e., the wavelengths are so short that it doesn't matter if you run them +/-, or +/+ (dipole vs bipole). For there to be a correlation, you have to go low enough in frequency to accomplish it (1 - 2 kHz), and I'm prejudiced against running a tweeter that low (even though millions do it
). From my perspective, your idea becomes really clever when you figure out how to do the same thing with a midbass driver. This way, you get around all of the asymmetry caused by the magnet's "shadow".OK, now for a minor rant: about the same time that this thread was running, another thread was aflame over Linkwitz's stunning discovery that adding a rear firing tweeter helped his Orion ... of course it did. He had painted himself into a corner by using a metal mid-bass, which, in turn, forced him to use an uber-expensive tweeter crossed at a too low frequency. At that x-over frequency, he was still in a range where a dipole has a positive effect on sound.
According to my personal prejudices, L:ionkwitz made the wrong design decision from the start in going with the excel driver (but the distortion curves, they're beautiful!!!
). He could have chosen one of the scanspeaks, which have a much more extended range, and saved people real money - since there would have been no need for the silly priced tweeter, and he would have never needed the second tweeter in the first place...at least, not as much. ;rant over.
I find the idea clever, but I'm at a loss as to why you'd want to go to the trouble for a tweeter. I haven't read this thread since it was active, so I may be off-base, but here goes.
For much of its range the front and back waves of this tweeter will be decorrelated, i.e., the wavelengths are so short that it doesn't matter if you run them +/-, or +/+ (dipole vs bipole). For there to be a correlation, you have to go low enough in frequency to accomplish it (1 - 2 kHz), and I'm prejudiced against running a tweeter that low (even though millions do it
For this to be a useful idea the tweeter would certainly have to be physically small. The sketch above is of a 3/4" VC diameter. I'm not sure just how compact the motor structure could be made and maintain thermal performance, but perhaps a dipole length of 4cm or so is possible, corresponding to useful dipole behavior up to 8-9kHz. With such a small tweeter, crossing over low isn't an option, so this leads one on a trail of a significant mulitway undertaking. However, the numbers just work out. That is, if the tweeter is small, and xo point is somewhere around 3-4kHz, the next driver can be sized appropriately to maintain good dipole behavior and still be small enough to maintain close CC distance to approximate point source behavior as well. If you follow the rabbit hole, you wind up with xo points dictated by the two constraints of dipole behavior and CC spacing something like 60Hz, 250Hz, 1kHz, 4kHz, with each driver covering a reasonable 2 octave bandwidth.
OK, so constructing a 5 way dipole system wasn't really the point of this thread or my post, just a progression of answering your question.
Short answer... I think there may still be some use if one is ready to tackle a statement system design that maintains dipole behavior up to within an octave of our hearing limit. I think it can be done. I just don't have the financial resources at the moment to sacrifice a few drivers to find out.
But in that line... I know the hexadym drivers are spendy. Any thoughts on a suitable cheap sacrificial alternative that has a natural compact motor structure? I know there are a lot of small profile and small diameter car audio drivers but I know nothing of their actual motor structure.
My thoughts exactly. In fact, it probably gets easier as the driver gets larger. I guess cavity resonance between diaphragms/cones is a unique problem that would have to be addressed but I think this could actually be leveraged to make very stiff radiating surfaces (think really thick sandwich type construction). But there is no reason the motor topology once figured out can't be scaled in size. Suspension would take a little experimenting... my first thought for larger units is to replace the traditional spider with a sort of integrated spider and surround, with one on each face as is the case with the tweeter sketch. Big subs using crazy stuff like seven spiders... yeah, that might get a little tricky!