This thread is really shaping up to be a treasure trove of information on tapped horns AND bandpass boxes! It's quite exciting to have so much good information in one spot.
I think I'll have to give the tapped horn a 2nd round, this time with a woofer that's more appropriate. With a lot of sawdust behind me, I've come to the conclusion that a high FS woofer with high power handling makes more sense than a low FS woofer with low power handling.
It's like taking a woofer designed for a sealed box and cramming it into a vented enclosure. Sure, it will work, but the results will be wonky ESPECIALLY if the foot print is the same.
As we speak I'm burning in a P-Audio SN-12MB. That will likely be the candidate for round II. I have a Audax ht240g0 and a HiVi d6.8 cooking away too...
I think I'll have to give the tapped horn a 2nd round, this time with a woofer that's more appropriate. With a lot of sawdust behind me, I've come to the conclusion that a high FS woofer with high power handling makes more sense than a low FS woofer with low power handling.
It's like taking a woofer designed for a sealed box and cramming it into a vented enclosure. Sure, it will work, but the results will be wonky ESPECIALLY if the foot print is the same.
As we speak I'm burning in a P-Audio SN-12MB. That will likely be the candidate for round II. I have a Audax ht240g0 and a HiVi d6.8 cooking away too...
Tom Danley said:
Hi Tom,
While that might be the easy way, I suspect that the method used by Bjørn Kolbrek will produce the more accurate result
.http://www.diyaudio.com/forums/showthread.php?postid=1591527#post1591527
Kind regards,
David
David -
I have good news and bad news. The good news is that horn response doesn't throw a divide by zero error when the horn segments are straight or narrowing.
The bad news is that it still throws an error about 25% of the time when you do the following:
1: Change from a 4pi to a 2pi simulation
or
2: Model an enclosure where the segments aren't expanding. For instance, you can use hornresp to model a transmission line or a bandpass with a flared port at both ends now, but it errors out about 25% of the time.
I have good news and bad news. The good news is that horn response doesn't throw a divide by zero error when the horn segments are straight or narrowing.
The bad news is that it still throws an error about 25% of the time when you do the following:
1: Change from a 4pi to a 2pi simulation
or
2: Model an enclosure where the segments aren't expanding. For instance, you can use hornresp to model a transmission line or a bandpass with a flared port at both ends now, but it errors out about 25% of the time.
Attachments
The schematic diagram demonstrates that the port is indeed flared at BOTH ends. Up until recently, you could only model the flare at the mouth. Of course this is very valuable, as a flared port can be used to reduce the overall length and size of the port in a bandpass enclosure.
Attachments
Patrick Bateman said:
Hi PB,
Many thanks for the feedback.
Unfortunately, as far as I am aware, there in nothing that can be done about the second error. The problem is due to an incompatibility between the computer you are using, and Hornresp being a 16-bit program. You will note that it is a general "Application Error", not a specific "Hornresp Error", which means that it is being generated by something other than a coding bug.
You wouldn’t have an AMD Athlon processor in your computer, or be running another 16-bit application at the same time as Hornresp, by any chance?
Kind regards,
David
yes and yes.
I was running hornresp on my girlfriend's laptop with a intel processor and a vista OS, and noticed it crashes quite a bit less.
By the way, you know you're ADDICTED to building speakers when you're asking your girlfriend if you can install hornresp on her laptop while you're out of the country on vacation
I was running hornresp on my girlfriend's laptop with a intel processor and a vista OS, and noticed it crashes quite a bit less.
By the way, you know you're ADDICTED to building speakers when you're asking your girlfriend if you can install hornresp on her laptop while you're out of the country on vacation
Tom Danley said:
Tom
Thats not what I said. I said that at LFs the pipe must look like a lumped parameter mass. It has to. I keep talking about what the systems and models have to look like as the frequency falls and you keep talking about what they look like as the frequency increases. They aren't going to be the same things - not even close.
I am not sure the antennae analogy is of much use either as its different enough from acoustics to be misleading. At any rate, I'm pretty sure that this level of detailed discussion cannot be effective in posts like this.
Hi Earl
In that case, I misunderstood what you were getting at, we are talking about the same thing.
The Tapped horns always have a horn path length that is greater than a quarter wl within its operating range.
I agree that when the port is made much shorter than the lowest mode ( its low cutoff in a TH), it does act like a lumped element like the mass in a port etc.
I thought you were saying that in its operating range it was best considered a lumped element.
Best,
Tom
In that case, I misunderstood what you were getting at, we are talking about the same thing.
The Tapped horns always have a horn path length that is greater than a quarter wl within its operating range.
I agree that when the port is made much shorter than the lowest mode ( its low cutoff in a TH), it does act like a lumped element like the mass in a port etc.
I thought you were saying that in its operating range it was best considered a lumped element.
Best,
Tom
Tom Danley said:
Hi Tom
When simulated in Hornresp the 15TBX100 tapped horn (TH-115) and dual 18PS76 reflex have almost identical sensitivity and maximum output, as I posted some time ago.
When you say the reflex is about 3dB down from the TH-115 I assume you're comparing your measurements with simulation? This efficiency (and cone travel) discrepancy has still never really been explained, and other people have reported recently that simulated and measured cone travel do agree.
So either what I thought was inside the TH-115 was wrong (do please send me the drawing so I can model it!), or Hornresp is still wrong (which now seems less likely), or your measurements are wrong (which also seems unlikely) -- or maybe a bit of all three
Cheers
Ian
Tom Danley said:
You spec the system to go down to 28 Hz, which has a wavelength of about 40 feet. So your horn paths are all greater than 10 feet? It is really difficult to see from the photos how you achieve that. And if they are not at least ten feet then at the lower edge of the passband they are lumped masses.
iand said:
So either what I thought was inside the TH-115 was wrong (do please send me the drawing so I can model it!), or Hornresp is still wrong (which now seems less likely), or your measurements are wrong (which also seems unlikely) -- or maybe a bit of all three
This is the point that I was making (lets hope you don't get chastised too for saying it) that until the theory and measurements agree there is something that we don't understand going on. And quite honestly, from what I have seen they are pretty far off.
Absolute levels are always difficult to correlate and arguing about a few dB in level is not going to be very fruitful, but one should be able to get shapes and passbands to come out right. The models show, and this seems logical to me, a highly resonant response above the first two resonances. Toms results don't show this, and they, in fact, show a very very smooth response that one would not expect from either the theory of a highly resonant system or the models. If the absolute passband level is off a few dB, well thats the nature of the beast, but the disparancy between what the models predict, which is what I would expect, and what is measured is a real mystery to me.
One thing that's ironic about your inventions is that one of them reduces high order modes (the foam that Geddes uses in his waveguides) while the other one depends on it to some extent (the tapped horn.)
Here's what I mean by this:
A Geddes waveguide is meticulously designed so that the wave originating from the diaphragm doesn't suffer from diffraction. To really wallop high order modes, foam is placed in the waveguide to absorb waves which are reflected at the mouth and back into the waveguide.
In a tapped horn, there are four mechanisms meticulously designed to generate high order modes. First, the horn is not driven at the apex, so reflections are created almost immediately, and in large numbers. Second, under sized mouths create more reflections than properly sized mouths IIRC (and a tapped horn has a horn mouth which is dramatically undersized.) Third, a tapped horn does not have any mouth treatment to reduce reflections. For example, a labhorn uses a significant flare at the mouth, which increases output and reduces reflections. Last but not least, a tapped horn has a number of 180 degree bends, which add more reflections than a series of 90degree bends. For example, the Danley DTS-20 has two or three 180 degree bends. In comparison, the labhorn has none IIRC.
In a recent post Dr Geddes wondered how the tapped horn can have such a long path length. Part of the answer is that a great deal of it's output is the result of internal reflections.
Here's what I mean by this:
A Geddes waveguide is meticulously designed so that the wave originating from the diaphragm doesn't suffer from diffraction. To really wallop high order modes, foam is placed in the waveguide to absorb waves which are reflected at the mouth and back into the waveguide.
In a tapped horn, there are four mechanisms meticulously designed to generate high order modes. First, the horn is not driven at the apex, so reflections are created almost immediately, and in large numbers. Second, under sized mouths create more reflections than properly sized mouths IIRC (and a tapped horn has a horn mouth which is dramatically undersized.) Third, a tapped horn does not have any mouth treatment to reduce reflections. For example, a labhorn uses a significant flare at the mouth, which increases output and reduces reflections. Last but not least, a tapped horn has a number of 180 degree bends, which add more reflections than a series of 90degree bends. For example, the Danley DTS-20 has two or three 180 degree bends. In comparison, the labhorn has none IIRC.
In a recent post Dr Geddes wondered how the tapped horn can have such a long path length. Part of the answer is that a great deal of it's output is the result of internal reflections.
gedlee said:
Earl,
The 28Hz box is the much larger TH-215, where the TH-115 extends to ~38Hz.
I haven't seen inside of any of Tom's more recent designs, but most all of the tapped horns are folded significantly to allow the path length, and the longer path length past the tap is as long as you describe.
I expect the folding is a big source of the unknowns or approximations in the HF resonant Qs as the programs don't really allow a good approximation of this as frequency rises or the width of the bend becomes significant.
Hi Earl
The quarter wave resonance is what defines the low corner of the response for a TH.
Essentially, one finds that wherever the knee in the response is, is where the quarter wave resonance is.
As with a conventional horn, having a compliance volume at the throat combines with the throat mass and forms a low pass corner. To a degree, this volume also shifts the lowest resonant frequency downward slightly.
Also worth mentioning, at least with low frequency horns, my Mathcad program and Akabak both generally over state the Q’s one see’s in the computer models made with them and underestimate the extra acoustical losses one encounters with very small horns.
Occasionally, one will find a predicted hf feature, which is entirely absent in the measured result.
Someone found the patent application on line for the dts-20, a package shape Mark Seaton suggested and this shows the enclosure layout.
Here is what it does;
http://www.danleysoundlabs.com/pdf/DTS 20 Spec Sheet.PDF
Here is what is inside.
This cabinet has a low cutoff and as you can see, a very long path length.
http://www.diyaudio.com/forums/attachment.php?postid=1426385&stamp=1202828040
Best,
Tom
Hi Ian
I am not sure how you modeled the TH-115, but what I got was about 3dB down in sensitivity.
Also, as I said earlier, I am happy when a computer model for a horn is within a dB or two of what you measure when you build it.
In my case, we sell the actual thing and not the computer model so how it measures trumps the model (normally the model is a bit better than the real thing).
We measure at 10 meters most of the time because one would literally be in the mouth bubble at one meter with large woofers. This also compensates for the fact that the speakers aren’t buried in the ground firing up into half space.
We have a mic calibrator, Earthworks mics, use an HP3456a Voltmeter and send some stuff to an independent company for measurement also.
I have made a fair number of boxes, which didn’t work out for one reason or another so this appears to be “the way it is” so far as some computer ambiguity or at least indicates that something’s are still left out..
Your reference to excursion discrepancy is puzzling with David’s reply , yes the cones do move and the models I use tend to overstate the Q’s on its predictions like magnitude and excursion.
The best bet is to observe the motion while in operation.
Best,
Tom
Hi Patrick
I am not sure the concept of HOM’s applies when the horn dimensions are such a small fraction of the wavelength.
For instance, think about how big a wavelength is at 20KHz and then think about how large the horn that wavelength is bouncing around inside of is (many many wl’s).
Then think of a TH with an upper operational frequency of say 100Hz and how big that WL is compared to the inside of the TH, only in its length is does it even approach 1wl.
In other words, it is too small at low frequencies to support non-axial or not length wise modes.
Now, if you drive it higher up at say 2KHz or above, then sure.
To be sure there are several axial resonances, but that is how low frequency horns work and it is by manipulating the series that one can greatly reduce or eliminate the dip between the first and second peak. that “small bass horns” like the Lab sub exhibit when used alone.
With a bass horn it is very hard / impossible to make one that has a flat response curve to its low corner. They essentially always have a significant droop off and then at some point have a knee. With the Tapped horn, one can often make the system Flatter for a given size and cutoff. One drives the shape and proportions according to the predictions.
It has been my experience that bends are often necessary in horns.
A bend can be anywhere from completely harmless and only adds a tiny lump of extra mass at the bend, or can cause a large notch, or even produce scattering, HOM’s,.
Observationally I would say it depends how large the difference between the inner path and outer path length is compared to the wavelength your talking about.
A real example of bending, If you had a 10KHz source driving a 3 foot copper tube that has a bore of say .062”, you would find that you could literally wind the tubing around a coffee cup a number of times and not significantly effect the sound coming out the end.
Best,
Tom
The quarter wave resonance is what defines the low corner of the response for a TH.
Essentially, one finds that wherever the knee in the response is, is where the quarter wave resonance is.
As with a conventional horn, having a compliance volume at the throat combines with the throat mass and forms a low pass corner. To a degree, this volume also shifts the lowest resonant frequency downward slightly.
Also worth mentioning, at least with low frequency horns, my Mathcad program and Akabak both generally over state the Q’s one see’s in the computer models made with them and underestimate the extra acoustical losses one encounters with very small horns.
Occasionally, one will find a predicted hf feature, which is entirely absent in the measured result.
Someone found the patent application on line for the dts-20, a package shape Mark Seaton suggested and this shows the enclosure layout.
Here is what it does;
http://www.danleysoundlabs.com/pdf/DTS 20 Spec Sheet.PDF
Here is what is inside.
This cabinet has a low cutoff and as you can see, a very long path length.
http://www.diyaudio.com/forums/attachment.php?postid=1426385&stamp=1202828040
Best,
Tom
Hi Ian
I am not sure how you modeled the TH-115, but what I got was about 3dB down in sensitivity.
Also, as I said earlier, I am happy when a computer model for a horn is within a dB or two of what you measure when you build it.
In my case, we sell the actual thing and not the computer model so how it measures trumps the model (normally the model is a bit better than the real thing).
We measure at 10 meters most of the time because one would literally be in the mouth bubble at one meter with large woofers. This also compensates for the fact that the speakers aren’t buried in the ground firing up into half space.
We have a mic calibrator, Earthworks mics, use an HP3456a Voltmeter and send some stuff to an independent company for measurement also.
I have made a fair number of boxes, which didn’t work out for one reason or another so this appears to be “the way it is” so far as some computer ambiguity or at least indicates that something’s are still left out..
Your reference to excursion discrepancy is puzzling with David’s reply , yes the cones do move and the models I use tend to overstate the Q’s on its predictions like magnitude and excursion.
The best bet is to observe the motion while in operation.
Best,
Tom
Hi Patrick
I am not sure the concept of HOM’s applies when the horn dimensions are such a small fraction of the wavelength.
For instance, think about how big a wavelength is at 20KHz and then think about how large the horn that wavelength is bouncing around inside of is (many many wl’s).
Then think of a TH with an upper operational frequency of say 100Hz and how big that WL is compared to the inside of the TH, only in its length is does it even approach 1wl.
In other words, it is too small at low frequencies to support non-axial or not length wise modes.
Now, if you drive it higher up at say 2KHz or above, then sure.
To be sure there are several axial resonances, but that is how low frequency horns work and it is by manipulating the series that one can greatly reduce or eliminate the dip between the first and second peak. that “small bass horns” like the Lab sub exhibit when used alone.
With a bass horn it is very hard / impossible to make one that has a flat response curve to its low corner. They essentially always have a significant droop off and then at some point have a knee. With the Tapped horn, one can often make the system Flatter for a given size and cutoff. One drives the shape and proportions according to the predictions.
It has been my experience that bends are often necessary in horns.
A bend can be anywhere from completely harmless and only adds a tiny lump of extra mass at the bend, or can cause a large notch, or even produce scattering, HOM’s,.
Observationally I would say it depends how large the difference between the inner path and outer path length is compared to the wavelength your talking about.
A real example of bending, If you had a 10KHz source driving a 3 foot copper tube that has a bore of say .062”, you would find that you could literally wind the tubing around a coffee cup a number of times and not significantly effect the sound coming out the end.
Best,
Tom
Thanks for the drawing. I guess calling those passages "horns" is really stretching the issue in my mind. But I can see whare you might be getting some decent path lengths with all the folds.
The driver radiates front and back correct? And the front is delay when it reaches the back by the path length so periodically it will be in phase and out of phase.
I still have trouble seeing how such a smooth system results from what should be a chaotic mess of resonances and interferences. Perhaps, like room acoustics and Distributed Mode panels, you are relying on a large density of resonances to yield a smooth sum. But in both of my examples, there is a region where there aren't enough modes for good summation followed by a region of reasonably smooth response with a sort of random variation. I don't see that in any of your examples.
The driver radiates front and back correct? And the front is delay when it reaches the back by the path length so periodically it will be in phase and out of phase.
I still have trouble seeing how such a smooth system results from what should be a chaotic mess of resonances and interferences. Perhaps, like room acoustics and Distributed Mode panels, you are relying on a large density of resonances to yield a smooth sum. But in both of my examples, there is a region where there aren't enough modes for good summation followed by a region of reasonably smooth response with a sort of random variation. I don't see that in any of your examples.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.