One way to address these cavity resonances would be with one of the new ultra-low-profile woofer designs. This could minimize the separation between the two diaphragms. If they operate at low enough frequency and are closely spaced, say less than 1/4 wavelength, then there shouldn't be problems with resonance.
Of course that doesn't necessarily mean it's the best design solution.
Of course that doesn't necessarily mean it's the best design solution.
An externally hosted image should be here but it was not working when we last tested it.
If the woofers were facing each other, the cavity (although still unnecessary from the perspective of a conventional speaker) could be made quite small. But that of course would mean the speaker would be ugly, so not commercially acceptable.
Gary Dahl's having a lot of fun with this project, and he's OK with the fact I'll be building Version 2 while he does all the hard work with Version 1. Some of his hard work is related to his wanting grille frames (he has cats, and you know what they do to speakers), while I like to have the drivers exposed where I can admire them.
Here's a pix of the upper TD15M enclosure:
Attachments
Last edited:
Goldmund gets around the ugly part with a really big grille over their woofer clamshells in their woofer modules.
Lynn's point about resonance at the crossover point is still an issue though. If you have a big peak or null at the crossover point due to the design, it's really not worth using unless it's a sub which are peaky due to a number of reasons. Even as a woofer with a 200-350hz crossover, a big suckout or peak will be clearly audible.
Lynn's point about resonance at the crossover point is still an issue though. If you have a big peak or null at the crossover point due to the design, it's really not worth using unless it's a sub which are peaky due to a number of reasons. Even as a woofer with a 200-350hz crossover, a big suckout or peak will be clearly audible.
Tru dat. There's something about cavity resonances that make them uniquely audible. Our much-disliked cabinet colorations are nothing more than the sound of a larger cavity, and more audible than the small ripples on the frequency response would indicate.
What isn't shown in the picture are the methods Gary and I are discussing to break up the front-to-back mode and stiffen the enclosure at the same time (basically, large V-shaped corrugations made with sections of 1/2" plywood). The big reflection off the back wall is by far the most objectionable, and we are taking steps to attenuate and diffuse it (spread out in time). MLSSA and ARTA will come in handy for looking for these 1~2 mSec reflections and seeing how successful we are in attenuating them. They will not be swept under the rug.
Last edited:
Won't know 'til we get there. High-efficiency speakers might get a partial exemption from the cavity-resonance coloration, but that could be nothing more than expecting a certain sort of "big-speaker" sound from the big boys, while we don't extend the same allowance to little speakers with big pretensions.
The whole business of "expected coloration" is something I'm a bit wary of. We get used to the sound of dynamic direct-radiator drivers, until we hear an electrostat, and realize how much coloration we've been accepting all along. Listen to the electrostat long enough, and you start to hear its faults too. Same for CD, which fooled a lot of people for a long time, until we got wise to issues with dithering and jitter - but that took nearly ten years.
Although I didn't mention it earlier, my exposure to OB's made me more aware of the hanky-panky going on with heavy equalization, and my decreasing tolerance for it. Remember, I've been fiddling around with solid-state and vacuum tubes for a while, and don't care for the sound of most electronics - and the more signal bending the less I like it.
Looks familiarThose corners are great to work with.
Rob
looks like products offered from that place in Los Angeles
Only before the response is eq'ed.
Simply incorrect.
More or less the same behaviour will occure from a baffled OB speaker with a relatively wide baffle.
Omitting the baffle completely will - for the most outer parts of the membrane (and surround) leave almost no path length to travel before overlaying.
This actually could be seen to be a good thing as the time interval for "consecutively min phase behaviour" is practically no longer existent. Meaning for those nude dipole speakers we get "some improvement" here with respect to min phase behaviour.
Just some thoughts for the passionate thinkers around here
Michael
Michael,
I am afraid saying it is so doesn't make it so. Being passionate is fine but coming to the correct conclusion requires that the thinking be rational. The frequency response of the doublet system is the classic dipole response with peak and 1st order roll off. Each source is minimum phase but the sum is also minimum phase. When MP eq is applied to the summed response to make the match some minimum phase acoustic target over the useful band width the result is still minimum phase and the impulse with be the impulse that correspond to that minimum phase target. Shortening the path length only shifts the dipole peak to a higher frequency. We can normalize the frequency response by the peak frequency = C/(2d) where d is the effective separation, and normalize time by 2d/C and the results in the frequency and time domain then all collapse to a single curve with the dipole peak at 2dF/C = 1.0.
I will assume you would accept that IF the double impulse system is minimum phase then application of minimum phase eq would correct both the frequency and time domain, correct? After all, thsi is what happens. So the problem reduces to, Is the double impulse system minimum phase? Or, Under what conditions is it minimum phase?
This can be answered fairly simply. The impulse of the doublet is
h(t) = u(t) + a x u(t-td)
where "a" is the magnitude of the second source and td is the time delay. Since the second source is alway further from the listener that the first, with two equal strength sources "a" will necessarily within the limits -1< a< +1, + or - depending on phase
The frequency response is
H(f) = 1 + a exp(-j x 2 x Pi x f x td)
Now, if we look at the discrete time representation of h(t) we have,
h
= u + a x u(n-N) where N is the number of samples corresponding the td.
N = sampling rate x td
Now, the Z transformation of h
isH(Z) = 1 + a Z^-1
H(Z) has N zeros uniformly spaced on a circle of radius "|a|" and an Nth order pole at the origin. Thus, for "a" as defined above, -1<a<1, all the poles and zeros of H(Z) will be inside the unit circle and H(Z) is minimum phase. I.E. the doublet response is minimum phase.
QED....
John
All correct, but only at a single point in space. There is no single filter that will do this globally. This is what makes the concept of minimum phase less than ideal for transducers - the fact that what happens at one point in space does not translate to any other points in space. This makes arguments like you present pretty limited.
All correct, but only at a single point in space. There is no single filter that will do this globally. This is what makes the concept of minimum phase less than ideal for transducers - the fact that what happens at one point in space does not translate to any other points in space. This makes arguments like you present pretty limited.
To atone for the rather cynical tone of post #6641, it's time for my semiannual Good Deed.
A fellow called Derek Goldsworth wrote me - as you can see from the picture he sent, Derek has a problem. He's made no less than six stunningly beautiful Ariels and only has drivers for two of them.
He needs to find 8 (eight) Vifa P13WH-00-08 drivers to complete the set. They were discontinued some time ago, but amongst the worldwide readership of diyAudio, maybe someone here can find some that they would be willing to part with. His e-mail address is (ignore the separator marks):
<|> dgolds <|> at <|> pacbell <|> dot <|> net
If anyone here has a lead, send him an e-mail at the address shown above. He might need Scan-Speak 9000, 9300, or 9500 tweeters, but I understand they are still available from the usual sources.
I'm sympathetic because I've recently been shopping for a Center speaker and it was more difficult than I expected. Rather than go through all the tedium of designing my own - I've got plenty to do already, and don't need another project - this time I decided to take the easier path and just buy the thing.
The test protocol was simple: Play my favorite CD's, set the receiver or processor to "Dolby PL II Cinema" and turn off all equalization. The "Cinema" mode is a brutal-sounding mode that forces any sound close to the center-region of the soundstage firmly into the center speaker. What you hear with a stereo CD is 50~60% Center and a little splash of sound from the Left and Right speakers - mono with a splash of stereo, if you will. This is a very hard test for the Center speaker, since it then carries the majority of the music.
What I was looking for was simple: intelligibility (duh, it's a center speaker and carries all the dialog) and the ability to play music at least as well as Karna's $129 Sharp stereo (hey, don't laugh, the thing sounds a lot better than many systems at the shows). This seemed like a pretty basic set of requirements - to be intelligible and not sound offensive on music. How hard could it be?
Well, guess what, not many Center speakers met the minimum requirements. I was kind of surprised, really. Not the Bowers & Wilkins consumer models, and not the entry-level Sonus Fabers, either. The high-line Sonus Faber Cremona (at US$7000/pair, or $5500 for the Center) was pretty decent-sounding but still short of the mark (treble and mids were nice, bass not as good). The installer-special B&W CT700 7.4 ($1000) sounded more or less OK, but too coarse for my taste - I'm still not a fan of Kevlar drivers, and can hear what's going on in the breakup region. But at least the tonal balance was acceptable.
I was under the impression the SF Center was in the price range of the Toy and Liuto, but no, what I heard at first was the Cremona. Auditioning the Toy and Liuto led to immediate rejection, along with the B&W CM Center 2. Well, this wasn't going anywhere. The only Center that I thought was even acceptable - not ideal, but acceptable - was the SF Cremona. I thought it was OK, but Karna didn't like it at all - she likes speakers with confident bass, and it didn't have that. Dynamics and power-handling was well down from the Ariels, not something I was willing to accept for a pretty steep price of $5500 (for one speaker!).
Time to go down the street to another shop. This time Karna and I auditioned the Dynaudio Contour SC, and yes, that was the one that finally sounded like an Ariel - at $2,100. Ouch. I called Dynaudio USA and found the crossover for the SC was 1400 Hz, lower than I would like, and decided to go for the next model up, the Contour SCX - at $3,300. But the presence of the Esotar tweeter and the 2200 Hz crossover made it easier to justify. (In my own mind, at least. Karna felt it wasn't so easy to justify one little loudspeaker costing as much as a state-of-the-art 58" Panasonic plasma TV. Or a secondhand Toyota Corolla for the grandson. But the SCX is hand-made in Denmark. By Danes! You can see how the value vs price discussion went.)
Why go so far out of original budget? Well, I have a hard time justifying a $2100 purchase and promptly modifying it, which is what I would have done with the Contour SC, voiding the warranty and ruining the resale value. I've done that with sports cars, and it was a losing proposition each time. The SCX - in stock condition - looks like it will do the job - sonics similar to the Ariel, and the dynamics for which Dynaudio is known.
This has been sort of a pricey education. No, I didn't want to waste the next several months designing a Center speaker. No no no. I really want to get the high-efficiency project done with, and do not want any further audio distractions. No buying a turntable and shopping for a preamp, no more fiddling around.
So when Derek Goldsworth wrote me and sent the pix of the cabinets (way better looking than mine), I was sympathetic. Finding speakers that sound like the Ariels isn't easy, and when you do, they will cost a lot of money. Are they for everyone? Hardly. Almost everything in the stores sounds really different. I must be designing for the few weirdos who still listen to acoustic music. Everyone else must be listening to music that is transistor-amplified with multi-kilowatt amps in PA systems and theaters, and using that as a reference.
A fellow called Derek Goldsworth wrote me - as you can see from the picture he sent, Derek has a problem. He's made no less than six stunningly beautiful Ariels and only has drivers for two of them.
He needs to find 8 (eight) Vifa P13WH-00-08 drivers to complete the set. They were discontinued some time ago, but amongst the worldwide readership of diyAudio, maybe someone here can find some that they would be willing to part with. His e-mail address is (ignore the separator marks):
<|> dgolds <|> at <|> pacbell <|> dot <|> net
If anyone here has a lead, send him an e-mail at the address shown above. He might need Scan-Speak 9000, 9300, or 9500 tweeters, but I understand they are still available from the usual sources.
I'm sympathetic because I've recently been shopping for a Center speaker and it was more difficult than I expected. Rather than go through all the tedium of designing my own - I've got plenty to do already, and don't need another project - this time I decided to take the easier path and just buy the thing.
The test protocol was simple: Play my favorite CD's, set the receiver or processor to "Dolby PL II Cinema" and turn off all equalization. The "Cinema" mode is a brutal-sounding mode that forces any sound close to the center-region of the soundstage firmly into the center speaker. What you hear with a stereo CD is 50~60% Center and a little splash of sound from the Left and Right speakers - mono with a splash of stereo, if you will. This is a very hard test for the Center speaker, since it then carries the majority of the music.
What I was looking for was simple: intelligibility (duh, it's a center speaker and carries all the dialog) and the ability to play music at least as well as Karna's $129 Sharp stereo (hey, don't laugh, the thing sounds a lot better than many systems at the shows). This seemed like a pretty basic set of requirements - to be intelligible and not sound offensive on music. How hard could it be?
Well, guess what, not many Center speakers met the minimum requirements. I was kind of surprised, really. Not the Bowers & Wilkins consumer models, and not the entry-level Sonus Fabers, either. The high-line Sonus Faber Cremona (at US$7000/pair, or $5500 for the Center) was pretty decent-sounding but still short of the mark (treble and mids were nice, bass not as good). The installer-special B&W CT700 7.4 ($1000) sounded more or less OK, but too coarse for my taste - I'm still not a fan of Kevlar drivers, and can hear what's going on in the breakup region. But at least the tonal balance was acceptable.
I was under the impression the SF Center was in the price range of the Toy and Liuto, but no, what I heard at first was the Cremona. Auditioning the Toy and Liuto led to immediate rejection, along with the B&W CM Center 2. Well, this wasn't going anywhere. The only Center that I thought was even acceptable - not ideal, but acceptable - was the SF Cremona. I thought it was OK, but Karna didn't like it at all - she likes speakers with confident bass, and it didn't have that. Dynamics and power-handling was well down from the Ariels, not something I was willing to accept for a pretty steep price of $5500 (for one speaker!).
Time to go down the street to another shop. This time Karna and I auditioned the Dynaudio Contour SC, and yes, that was the one that finally sounded like an Ariel - at $2,100. Ouch. I called Dynaudio USA and found the crossover for the SC was 1400 Hz, lower than I would like, and decided to go for the next model up, the Contour SCX - at $3,300. But the presence of the Esotar tweeter and the 2200 Hz crossover made it easier to justify. (In my own mind, at least. Karna felt it wasn't so easy to justify one little loudspeaker costing as much as a state-of-the-art 58" Panasonic plasma TV. Or a secondhand Toyota Corolla for the grandson. But the SCX is hand-made in Denmark. By Danes! You can see how the value vs price discussion went.)
Why go so far out of original budget? Well, I have a hard time justifying a $2100 purchase and promptly modifying it, which is what I would have done with the Contour SC, voiding the warranty and ruining the resale value. I've done that with sports cars, and it was a losing proposition each time. The SCX - in stock condition - looks like it will do the job - sonics similar to the Ariel, and the dynamics for which Dynaudio is known.
This has been sort of a pricey education. No, I didn't want to waste the next several months designing a Center speaker. No no no. I really want to get the high-efficiency project done with, and do not want any further audio distractions. No buying a turntable and shopping for a preamp, no more fiddling around.
So when Derek Goldsworth wrote me and sent the pix of the cabinets (way better looking than mine), I was sympathetic. Finding speakers that sound like the Ariels isn't easy, and when you do, they will cost a lot of money. Are they for everyone? Hardly. Almost everything in the stores sounds really different. I must be designing for the few weirdos who still listen to acoustic music. Everyone else must be listening to music that is transistor-amplified with multi-kilowatt amps in PA systems and theaters, and using that as a reference.
Attachments
Last edited:
Yes Earl. We have been through this many times before. However, crossover filters are generally nothing more than a minimum phase filters which equalize a driver's raw response to a specific acoustic target, at some point in space. At some other point in space the acoustic response may deviate from that target because the driver's response may vary while the filter remains constant. The driver's response at that new point, while different than at the design point, will still be minimum phase, though the phase will be different than at the design point. And the filtered response will remain minimum phase with the impulse of a minimum phase system with the observed amplitude at that point.
But the point of this argument is whether or not the doublet is minimum phase and thus can be equalized to a single impulse with minimum phase eq. With the doublet the only thing that changes with position is td. This has no effect on whether or not the doublet is minimum phase or not as there are no restrictions on td in the above mathematical argument. The doublet response is always MP. So the application of a single minimum phase eq function to the doublet response will result in a minimum phase response at any point in space.
Whether or not this response is a function of position is an issue of design.
You could insert something like a phase plug between the woofers: a piece of wood that would more or less follow the shape of the woofers (face to face or back to back, or even face to back) to minimize the volume of air. That would also reduce problems with air compression.
This would be much simpler to build than a real phase plug, as we do not really need exactly equal path everywhere.
some inspiration: Phase Plugs
What do you think?
Yes, we have been through this before, just trying to keep you honest as your post made it sound like the EQ worked everywhere and I think that people should understand that it doesn't.
You point, as I said, is quite correct.
I still do not agree - though its maybe kind of silly to argue against paramount math skills
Yes the point I'm focusing on is solely the min phase behaviour of a doublet in my concept of "consecutively min phase behaviour" - the 3-D issue mixed in by Earl is solved perfectly by perfect constant directivity - there is no questioning.
The point where I disagree is not exactly that the doublet isn't min phase - my point is that it is not the same identical min phase throughout the time line.
Putting one point source - say - 1 m from you and the other point source another meter behind that will let the sound of the latter arrive some ms later for example.
That such dipole config will have "true" dipole behaviour only at pretty low frequencies is clear but not the point here - its just to give a picture for imagination.
No matter how you set any correction - until the second part of the doublet arrives, the SPL pressure at listening position is different than after those few ms.
So - my point simply is: there is no way to correct correctly for the time after overlay and for the time until then - in other words - there is a kink in the pressure line corresponding to the time of delay.
The issue is independent of the distance of the two sources of course - the problem lyes in the discrete time interval that separates the two peaks of the doublet.
Michael
Yes the point I'm focusing on is solely the min phase behaviour of a doublet in my concept of "consecutively min phase behaviour" - the 3-D issue mixed in by Earl is solved perfectly by perfect constant directivity - there is no questioning.
The point where I disagree is not exactly that the doublet isn't min phase - my point is that it is not the same identical min phase throughout the time line.
Putting one point source - say - 1 m from you and the other point source another meter behind that will let the sound of the latter arrive some ms later for example.
That such dipole config will have "true" dipole behaviour only at pretty low frequencies is clear but not the point here - its just to give a picture for imagination.
No matter how you set any correction - until the second part of the doublet arrives, the SPL pressure at listening position is different than after those few ms.
So - my point simply is: there is no way to correct correctly for the time after overlay and for the time until then - in other words - there is a kink in the pressure line corresponding to the time of delay.
The issue is independent of the distance of the two sources of course - the problem lyes in the discrete time interval that separates the two peaks of the doublet.
Michael
Last edited: