salas said:
1. I always use a reference axis between the mid and tweeter, and design the crossover so the center of the lobe is aimed right at the listener. I want the nulls as far away as possible.
2. Yes, compression drivers can be used "barefoot" with the rear chamber removed. But that won't fix the inherent diaphragm breakup mode in the 14~21 kHz region, and throws away a tremendous amount of horn-gain/impedance-matching and resultant drop in IM distortion at lower frequencies. So it amounts to a tradeoff - a big, well-made dome tweeter (that needs a lot of EQ) versus the HOM's (or multipath distortion, as I think of it) of a horn or waveguide (and much less EQ).
It's significant that over in the Dr. Geddes thread, at least one person has tried a short foam plug in the throat of the horn and is reporting good results. Since the throat area has been identified as one of the main culprits in generating HOM/multipath distortion, it looks like a little bit of open-cell foam in the right area could do a lot of good. Since pressures are highest in this part of the horn or waveguide, the lossy foam in turn should be most effective in damping unwanted modes.
The old advice of "put a sock in it" might turn out to be true.
Re: Very interesting thread
Hi Dave,
You have to put what I said in context. That was a reply to Tom in reference to pro audio drivers, not the SS, Seas, etc.
I'm well aware that paper can be stiffened significantly, ala the new Seas Nextel cones, to the point where they have almost a metal like upper resonance peak.
I agree with the rest of what you said. There are rigid cones that do not exhibit the the same amplitude peak(s) as the W Seas magnesium. Toole's Ceramic skinned aluminum appear to push the peak further up and lower in amplitude. Even coated aluminum can have relatively benign peaking.
Now lets take a look at the type of driver I was referring to.8CX21
I suspect there may be some bending and flexing there.
My point was (again) there are bad and good examples of both.
I still prefer good metal over good paper or cloth. Others obviously don't.
cheers,
AJ
dlr said:
Hi Dave,
You have to put what I said in context. That was a reply to Tom in reference to pro audio drivers, not the SS, Seas, etc.
I'm well aware that paper can be stiffened significantly, ala the new Seas Nextel cones, to the point where they have almost a metal like upper resonance peak.
I agree with the rest of what you said. There are rigid cones that do not exhibit the the same amplitude peak(s) as the W Seas magnesium. Toole's Ceramic skinned aluminum appear to push the peak further up and lower in amplitude. Even coated aluminum can have relatively benign peaking.
Now lets take a look at the type of driver I was referring to.8CX21
I suspect there may be some bending and flexing there.
My point was (again) there are bad and good examples of both.
I still prefer good metal over good paper or cloth. Others obviously don't.
cheers,
AJ
Lynn Olson said:
Lynn,
I believe that time-fidelity is at least as important as frequency-response to maintain the perceptual parsability of complex orchestral and choral recordings. As I continue on my personal 'audio-trek' - which shares a lot in common with your current efforts - I have recently been working on time-fidelity... with very encouraging results.
Here is a recent impulse response measurement for the Gecko system:
This measurement has ZERO editorializing.
For me, this accomplishment has been a personally and aurally rewarding step in the right direction.
Edward
Re: Re: Very interesting thread
Though it uses neodymium, the impedance curve would indicate that it may still not be a low distortion design. I'm curious about that aspect of the driver. At that price, I'd expect to see better than standard distortion figures. Too bad they don't show any. I also always question a manufacturer that does not post 30 and 60 degree off-axis FR graphs.
Dave
I hadn't see the context on that, my mistake then. There are still a lot of folks around who make the same reference but in the context in which I took it, guess I was reacting to that in seeing your post.AJinFLA said:
I'm not sure what I think of the Nextel. I haven't heard any, so I'd only be conjecturing as to their sound.
That's an interesting driver, I've not seen it before. I see that the description is "wide range driver", more apt, but it seems that it's being touted as a full range for usage in the examples.
Though it uses neodymium, the impedance curve would indicate that it may still not be a low distortion design. I'm curious about that aspect of the driver. At that price, I'd expect to see better than standard distortion figures. Too bad they don't show any. I also always question a manufacturer that does not post 30 and 60 degree off-axis FR graphs.
With that driver it's hard to say what is related to cone material, cone geometry (resonance modes) or possible diffraction due to the coincident driver as well as diffraction due to the cone profile. I wouldn't even hazard a guess without knowing more details.
Dave
soongsc said:
I’ not going to get all upset if I can not convince you but I wish you only to consider this. Forget about nonlinear effects and concentrate only on the direct sound. Now, if I have two perfectly linear sources which have different frequency response, one with a big resonance peak at 4k Hz, and I equalize the resonance so both drivers have identical on axis response, then the impulse response associated with the direct sound will be identical for both systems and the impulse response is the time response so how can the direct sound that reaches my ear be different? Yes, it takes time for the resonance to build, but if the resonance grows as exp(At) and I apply equalization that goes as exp(-At) the system output is exp(0) = 1.0. You are correct in that what happens first reaches your ear first, but the equalization corrects the input time response so that what happens first at the output is controlled in such manor that it is what we want to happen first. I also agree that this eq. may not produce the perfect correction off axis. That was one of the other factors I grouped into “There are numerous other potential sources of problems too” and would possibly account for why two different but perfectly linear sources with difference frequency response may still sound different when equalized to have perfectly identical on axis response. This is part of the what make the problem so complex and another possible reason to stay away from driver with nasty breakup peaks.
This is not a feedback system. This is a linear correction to a linear system. In a linear system if you know before hand that the system produces an error you can subtract the error from the input to eliminate it in the output. This is different than a nonlinear system because you can not predict the error in a nonlinear system. It will vary with the input. In a nonlinear system we look at the output to see what the instantaneous error is and then we subtract the instantaneous error from the input. This in turn changes the input which may change the error created in the output, so we have to do this continuously because the error is constantly changing. I’m not an expert in feedback systems like amplifiers but I do know that in a nonlinear system the best approach is to design the system to be as stable and as close to linear as possible and then apply the least amount of feedback required to clean up the system.
gedlee said:
And is this unedited 3-way system, on axis impulse response impossible too? Sorry that it is inverted. My mic inverts. (~60k/sec sampling rate, first 3 msec).
An externally hosted image should be here but it was not working when we last tested it.
I understand what you are saying, and I agree it covers most part of the signal regeneration. Preshaping a signal is a very common technique used in control, and I understand that we are not talking about feedback. However, we must realize that presure is generated from different parts of the surface and different time throughout the signal variation.john k... said:
Well thinking about direct sound, at the first cone mode, I'm visualizing in the VC moving first, part of surface close to where VC and Cone connect move a bit later, then finally the outer edge of the cone follows. Throughout this first 1/4 cycle is the onset, if compensation prevents this part from being faithfully reproduced, then the sound will not be as good as a more well behaved cone. I won't be upset if I cannot let you understand this part.
Now if we wanted to do somthing really exotic using real=time software technology to it's extreme, in might be possible to model the energy storage charateristics of a cone, and depending on what the previous state and new signal state is, more adequately compensate for cone defficiencies. But I don't think anyone is going to spend millions on a project like that.
john k... said:
John
An impulse response from an acoustic source must average to zero as there can be no DC response to any acoustic signal. What you have shown is indeed in error in that clearly the LF data is missing, thus somehow the missing data must be "assumed" and clearly here it was "assumed" to be the same as the passband - which is impossible. I am not saying that it is not possible to make a bad measurement that does not average to zero,but I am saying that in all of my work I test the DC to make sure that it is well below the passband or I don't trust the data. You should do likewise.
You will never see me show an acoustic impulse response that does not average to zero.
Hi
Soongsc, don't make a common mistake.
EVRY TIME the air is moved by the speaker there is this same time delay between the current in the voice coil and the air getting pressurised. This is simply due to the fact that every material has its maximal speed of information transport ( sound speed here in the voice coil and the membrane not to mention that the membrane has to build up speed before pressure is created).
This is independent of resonance or not.
In case of equalisation you compensate phase to force the membrane to be exactly in time.
Its just like compensating a certain electrical filter ( high Q peak ) with its mirror function ( high Q notch ).
The main question that is left is how linear the driver really acts in its resonant frequency - as JohnK said.
---------------------------
Again – any suggestions for EQ units that can compensate more accurate than a DCX ?
EQ the peak of the driver like shown makes a big difference. I would like to investigate that further.
As it seems even a minor notch of – 3 dB makes tweaking necessary to keep a good balance.
Greetings
Michael
By the way – is it only me that gets all the " the " in this thread displayed in beautiful red ?
soongsc said:
Well thinking about direct sound, at the first cone mode, I'm visualizing in the VC moving first, part of surface close to where VC and Cone connect move a bit later, then finally the outer edge of the cone follows. Throughout this first 1/4 cycle is the onset, if compensation prevents this part from being faithfully reproduced, then the sound will not be as good as a more well behaved cone. I won't be upset if I cannot let you understand this part.
Soongsc, don't make a common mistake.
EVRY TIME the air is moved by the speaker there is this same time delay between the current in the voice coil and the air getting pressurised. This is simply due to the fact that every material has its maximal speed of information transport ( sound speed here in the voice coil and the membrane not to mention that the membrane has to build up speed before pressure is created).
This is independent of resonance or not.
In case of equalisation you compensate phase to force the membrane to be exactly in time.
Its just like compensating a certain electrical filter ( high Q peak ) with its mirror function ( high Q notch ).
The main question that is left is how linear the driver really acts in its resonant frequency - as JohnK said.
---------------------------
Again – any suggestions for EQ units that can compensate more accurate than a DCX ?
EQ the peak of the driver like shown makes a big difference. I would like to investigate that further.
As it seems even a minor notch of – 3 dB makes tweaking necessary to keep a good balance.
Greetings
Michael
By the way – is it only me that gets all the " the " in this thread displayed in beautiful red ?
I'm checking on the model
we use for Theater installations on Symetrix. In the meantime...there is some interesting software being used over here.
http://www.diyaudio.com/forums/showthread.php?s=&postid=1301530#post1301530
C
we use for Theater installations on Symetrix. In the meantime...there is some interesting software being used over here.
http://www.diyaudio.com/forums/showthread.php?s=&postid=1301530#post1301530
C
gedlee said:
Earl,
What leads you to the conclusion that the impulse responses which John and I posted do not average to zero?
As you know a normalized impulse response for a perfect system - flat from DC to daylight - is a single spike of magnitude 1. If the system is low-pass, then the spike is lower magnitude, dull and has a small finite width. If the system is high-pass it will have a sharp spike with a magnitude of nearly 1, and then a small undershoot with a slow decay back to zero. If the low-frequency cut off is at 20 Hz, then the undershoot and tail will be very low magnitude and very long - on the order of 10's of milliseconds. Given the very small magnitude and extended length of the tail due to the low-frequency cut off, one does not 'notice' it in the presence of the hash of the noise floor.
Of couse, this whole description is predicated on the assumption that the pass-band phase response is minimum phase. If you have a conventional 3-way speaker system with two crossover transitions between 20 Hz and 20 KHz, then the acoustic sum is not minimum phase.
The response will be like this:
But if the same system has a minimum phase acoustic sum, then one gets this:
Perhaps, there is a misunderstanding due to the common lack of familiarity with minimum phase acoustic sum systems.
Edward
gedlee said:
Earl,
We when through this before. This system has a 25 Hz Q=0.5 high pass cut off (woofer). The statement is correct that the signal must average to zero, but how? Relatively speaking the impulse has a very short, high amplitude positive spike. The area is approximately H x dt where H is the amplitude and dt is the width of the pulse (one sample here, ~0.0167msec) The negative part of the pulse, by comparison, has a very low amplitude, very long duration decay to zero from below for this system. While there is a little ringing in the impulse following the initial spike due to the nature of the system's low pass characteristic (tweeter cut off at about 25k Hz), the low frequency, negative swing is so low in amplitude that it's not visible on the vertical scale and if the time axis were extended room reflections would contaminate the result further. In fact, at the level the system was tested, the negative part of the impulse might even be lost in the noise floor of the measurement environment. That there is no visible negative swing or oscillation around zero is because there is not time/transient distortion introduced by the woofer/midrange or midrange/tweeter crossovers. These, since they occur at much higher frequencies relative the high pass corner of the present system, are typical much more visible in the impulse. May I suggest you look at the impulse response for a system with a band width defined by a 2nd order, 25Hz Q=0.5 high pass filter cascaded with a 25k Hz Q= 0.7, 2nd order low pass filter. You will see the initial part of the impulse looks very similar to what I have posted with the magnitude of the negative, after the initial tweeter ringing, on the order of 0.5% of the magnitude of the initial positive pulse. Average to zero? Yes. Necessarilly visible in the impulse response,? No.
DSP EQ with high Q and fine F stepping
Hi
Edward, At least for me its hard to decipher that " minimum phase " thing – not necessarily a problem of translation only -. would be great if you give a short survey .
-----------------------
chrismercurio, thanks for the link which lead me to
http://www.acourate.com/
something I will check out in more detail. If it were not a computer solution with all its drawbacks ( acoustic and electric noise and handling ) it seemed to be what I was looking for.
Sure worth a try anyway !
Greetings
Michael
Hi
EdwardWest said:
Edward, At least for me its hard to decipher that " minimum phase " thing – not necessarily a problem of translation only -. would be great if you give a short survey .
-----------------------
chrismercurio, thanks for the link which lead me to
http://www.acourate.com/
something I will check out in more detail. If it were not a computer solution with all its drawbacks ( acoustic and electric noise and handling ) it seemed to be what I was looking for.
Sure worth a try anyway !
Greetings
Michael