I have been playing around with waveguides on my ribbons in an effort to use them in higher sensitivity and more controlled directivity systems with larger woofers.
I admit to being uneducated in this area and busy enough now that I cannot devote much time to study. So I decided to build my curiosity by just throwing a bunch of crude “wave guides” ( all straight sided “cone shaped” , together (hot melt and thin wood) and just play.
Yes I am quickly learning how critical small details can be here, however my first big impression is the roll off above about 10 kHz no matter what wave guide I use.
The wave guides can be tailored to get a response that is straight line linear down tilted slope from 1k out to 10k with 10k being -2db . From 10k to 20k the FR curve becomes slightly curved in shape and is down -5 db from 1k level.
This is an UN EQed response .
My question is this. I see this same effect (actually even more) on some other measurements of horn/CD driver arraignments that get some good press. So the question is, is this kind of response a real issue with systems using such controlled directivity??
I have not done listening tests yet and it will be a while before I do, so just curious if there is really a big issue here. In past designs using direct radiators with similar responses I always come away thinking that above 10k is not as big a deal as some are saying.
Anyway looking for wave guided wisdom……
I admit to being uneducated in this area and busy enough now that I cannot devote much time to study. So I decided to build my curiosity by just throwing a bunch of crude “wave guides” ( all straight sided “cone shaped” , together (hot melt and thin wood) and just play.
Yes I am quickly learning how critical small details can be here, however my first big impression is the roll off above about 10 kHz no matter what wave guide I use.
The wave guides can be tailored to get a response that is straight line linear down tilted slope from 1k out to 10k with 10k being -2db . From 10k to 20k the FR curve becomes slightly curved in shape and is down -5 db from 1k level.
This is an UN EQed response .
My question is this. I see this same effect (actually even more) on some other measurements of horn/CD driver arraignments that get some good press. So the question is, is this kind of response a real issue with systems using such controlled directivity??
I have not done listening tests yet and it will be a while before I do, so just curious if there is really a big issue here. In past designs using direct radiators with similar responses I always come away thinking that above 10k is not as big a deal as some are saying.
Anyway looking for wave guided wisdom……
Think of a horn as an expanding 1/4 wavelength tube. Anything that happens within that 1/4 wavelength will either form a good wavefront, or curtail/modify it quite strongly.
So if you calculate the 1/4 wavelength at 10 kHz, you get about 3/8". That's your "horn" at 10 kHz. Anything beyond that point in the horn controls the polars, but weakly.
So (and I would guess that you see this coming) what you have is a phase plug issue. You need to keep the edges of the diaphragm greater than that very small distance (3/8") from destructively combining at off-angles greater than just a few degrees.
I would recommend looking at the Beveridge loudspeaker designs and patent. The Beveridge loudspeakers themselves are way too expensive and esoteric to be viable (as well as not terribly reliable, as I'm told by a local owner), but the basic principle of the waveguide planar (monopole) design is still valid for your design problem. I'd be thinking of ways to implement the same type of waveguide vanes as you see below--because the patent ran out 25 years ago, and so infringement is not an issue...
Thanks Cask, I am not sure what you mean by following...
"You need to keep the edges of the diaphragm greater than that very small distance (3/8")"....
"You need to keep the edges of the diaphragm greater than that very small distance (3/8")"....
By the way, at 20 kHz, the 1/4 wavelength becomes 3/16ths inch--and I think you can see what's happening in that top octave.
"Waveguides" (i.e., horns) have the same issue. A lot of people believe that they have horn issues when they talk about issues about 6-7 kHz, but what they really have is compression driver phase plug issues. Now you know why.
Chris
"Waveguides" (i.e., horns) have the same issue. A lot of people believe that they have horn issues when they talk about issues about 6-7 kHz, but what they really have is compression driver phase plug issues. Now you know why.
Chris
Think of the width of the ribbon itself (if oriented vertically, as all ribbons are typically)--at the point where the width of the ribbon approaches 1/4 wavelength, you're going to get destructive interference at anything other than dead on-axis, normal to the ribbon itself.
Draw yourself a picture of the wavefronts combining from the sides of the ribbon, then destructive cancelling each other out at the 1/4 wavelength point. Also think about the shape and the slits of typical compression driver phase plugs. They are preventing destructive cancellations at higher frequencies by forcing the acoustic wavefronts into small pathways behind the plugs--to combine without destructive interference occurring.
Chris
Here's a visual on what is happening, if you consider the left/right edges of the ribbon now as the two acoustic sources in the center of the animation. The narrowing polars has little to do with the external boundary (horn wall) and a lot more to do with the size of the radiating diaphragm/ribbon itself.
A direct link to the animation since diyAudio is freezing it here: http://resource.isvr.soton.ac.uk/spcg/tutorial/tutorial/Tutorial_files/interfmonopcolcol.gif
The same thing happens with cone diaphragms--which is the reason why some woofers and midrange drivers use center "bullet" phase plugs to break up the interference if they are used at frequencies corresponding between 1/4-1/2 wavelengths to the cone diameter. This is usually called the "narrowing polar region" of cone-type drivers--as the acoustic wavelengths become about the 1/4 of the width of the diaphragm--up to half wavelength, at which point you get total cancellation of acoustic wavefronts.
For instance, that happens at about 800 Hz for a 15" woofer cone, but that distance becomes progressively shorter as the wavelength decreases (frequency increases) for higher frequency drivers, until you find yourself in the top-most octave...10-20 kHz...where the distances at which interference occurs is measured at less than an inch.
Chris
A direct link to the animation since diyAudio is freezing it here: http://resource.isvr.soton.ac.uk/spcg/tutorial/tutorial/Tutorial_files/interfmonopcolcol.gif
The same thing happens with cone diaphragms--which is the reason why some woofers and midrange drivers use center "bullet" phase plugs to break up the interference if they are used at frequencies corresponding between 1/4-1/2 wavelengths to the cone diameter. This is usually called the "narrowing polar region" of cone-type drivers--as the acoustic wavelengths become about the 1/4 of the width of the diaphragm--up to half wavelength, at which point you get total cancellation of acoustic wavefronts.
For instance, that happens at about 800 Hz for a 15" woofer cone, but that distance becomes progressively shorter as the wavelength decreases (frequency increases) for higher frequency drivers, until you find yourself in the top-most octave...10-20 kHz...where the distances at which interference occurs is measured at less than an inch.
Chris
Last edited:
Yeah I like that Beveridge solution. I wonder if Tom Danley had it in mind when he came up with his layered combiner. I wish I could think of a practical way to build such a thing.
Going back to the original post, that tilted down response slope you reported is typical of a constant directivity wavequide and is routinely EQed away. This isn't a problem for compression drivers because they are robust and high sensitivity to begin with; I don't know if it will be an issue for your ribbon or not,probably not for home use.
The wavequides you build without a phase plug,Beveridge style or otherwise, or perhaps a diffraction slot, won't improve the basic HF dispersion of the ribbon, which depends on the width and aspect ratio of the radiating surface, but might extend the low end and I could see it working to match the horizontal directivity of a large woofer mounted below the ribbon.
Going back to the original post, that tilted down response slope you reported is typical of a constant directivity wavequide and is routinely EQed away. This isn't a problem for compression drivers because they are robust and high sensitivity to begin with; I don't know if it will be an issue for your ribbon or not,probably not for home use.
The wavequides you build without a phase plug,Beveridge style or otherwise, or perhaps a diffraction slot, won't improve the basic HF dispersion of the ribbon, which depends on the width and aspect ratio of the radiating surface, but might extend the low end and I could see it working to match the horizontal directivity of a large woofer mounted below the ribbon.
Comparing responses of Fountek NeoCD3.5H to NeoCD3 with similar size ribbon makes these things clearer.
Fountek Ribbon Tweeters
The width of horn exit is important, because it determines the "highpass of loading" eg. enforcement of response. NeoCD3.H horn is only 60mm but horn is quite deep, so loading is efficient. Also the driver "cavity" is very narrow, about 15mm which makes it exceptionally smooth above 15kHz, for a waveguide/horn driver.
Fountek Ribbon Tweeters
An externally hosted image should be here but it was not working when we last tested it.
The width of horn exit is important, because it determines the "highpass of loading" eg. enforcement of response. NeoCD3.H horn is only 60mm but horn is quite deep, so loading is efficient. Also the driver "cavity" is very narrow, about 15mm which makes it exceptionally smooth above 15kHz, for a waveguide/horn driver.
Well...flat response beyond 10 kHz is another subject entirely. I'll take a stab at answering that question...
Having demastered a lot of music tracks thus far (about 15K of them), I can say that flat response is definitely a challenged concept (i.e., virtually all stereo recordings have been significantly EQed during mastering, and many people call that "good"). But, if you agree that what you're really trying to achieve is to match the frequency response of the monitors/room acoustics of the mixing and/or mastering studio--at least the on-axis response of those monitors. This is all we can do to "reproduce accurately".
It appears that wide polar coverage is the biggest factor--especially at the frequencies where the reflectivity of room furnishings is low-because reflections up high are so difficult to hear otherwise. That means that flatness of SPL anywhere within the approx. 90 horizontal coverage sector of the loudspeaker is also important (...vertical coverage is another subject...). Sean Olive's patented decision model on the subjective factors predicting loudspeaker preference puts smooth and consistent power response of a loudspeaker as the most important factors, followed closely by smoothness and flatness of on-axis SPL. Most studio monitors nowadays have wide power response--like dome tweeters and wide-polar compression drivers/horns. So the bar for reproducing high frequencies well is actually pretty high.
However, high frequency extension is I believe something that is heavily offset by presbyopia, which is a great factor for anyone over 50-60 yo and especially male. If you're young or female, HF extension would also be important, but probably not nearly as important as a lot of "hi-fi audiophiles" think it is. I know that in my case the type of music transients that are most important in this regard is high percussion transients--snares, percussion hits, ride and finger cymbals, etc.--followed closely by (surprisingly) vocal timbre, which is quite strongly present above 10 kHz in terms of vocal fricatives, etc. and is a big deal in producing a "live...in the room" effect, but typically, these highest frequencies become important as the SPL of the playback is rises, like at concert level acoustic instrumentation, but not amplified instruments. For horn-loaded loudspeakers, this isn't an issue. For direct radiators--well, it's a big problem.
I've found that bullet tweeters with their characteristic narrow polars are much less engaging and natural sounding than tweeters that have wide polars (horizontally, at least) like domes and wide polar/horn-loaded compression drivers. YMMV.
Bottom line: evenness of power response is the number one requirement (implying wide but controlled and smooth coverage), and evenness of SPL on-axis is number two. So, yes, response at high frequencies is important, but the wide consistent coverage is more important than on-axis SPL...which can usually be EQed flat again.
Chris
Having demastered a lot of music tracks thus far (about 15K of them), I can say that flat response is definitely a challenged concept (i.e., virtually all stereo recordings have been significantly EQed during mastering, and many people call that "good"). But, if you agree that what you're really trying to achieve is to match the frequency response of the monitors/room acoustics of the mixing and/or mastering studio--at least the on-axis response of those monitors. This is all we can do to "reproduce accurately".
It appears that wide polar coverage is the biggest factor--especially at the frequencies where the reflectivity of room furnishings is low-because reflections up high are so difficult to hear otherwise. That means that flatness of SPL anywhere within the approx. 90 horizontal coverage sector of the loudspeaker is also important (...vertical coverage is another subject...). Sean Olive's patented decision model on the subjective factors predicting loudspeaker preference puts smooth and consistent power response of a loudspeaker as the most important factors, followed closely by smoothness and flatness of on-axis SPL. Most studio monitors nowadays have wide power response--like dome tweeters and wide-polar compression drivers/horns. So the bar for reproducing high frequencies well is actually pretty high.
However, high frequency extension is I believe something that is heavily offset by presbyopia, which is a great factor for anyone over 50-60 yo and especially male. If you're young or female, HF extension would also be important, but probably not nearly as important as a lot of "hi-fi audiophiles" think it is. I know that in my case the type of music transients that are most important in this regard is high percussion transients--snares, percussion hits, ride and finger cymbals, etc.--followed closely by (surprisingly) vocal timbre, which is quite strongly present above 10 kHz in terms of vocal fricatives, etc. and is a big deal in producing a "live...in the room" effect, but typically, these highest frequencies become important as the SPL of the playback is rises, like at concert level acoustic instrumentation, but not amplified instruments. For horn-loaded loudspeakers, this isn't an issue. For direct radiators--well, it's a big problem.
I've found that bullet tweeters with their characteristic narrow polars are much less engaging and natural sounding than tweeters that have wide polars (horizontally, at least) like domes and wide polar/horn-loaded compression drivers. YMMV.
Bottom line: evenness of power response is the number one requirement (implying wide but controlled and smooth coverage), and evenness of SPL on-axis is number two. So, yes, response at high frequencies is important, but the wide consistent coverage is more important than on-axis SPL...which can usually be EQed flat again.
Chris
Sorry, presbycusis--not presbyopia. Even though you might be told by your audiologist that you have excellent hearing (...for your age...), this doesn't mean that you're not suffering significant changes in HF hearing acuity as time passes--significant loss, especially if you're male. There are quite a few "audiophiles" in their 40s-60s in denial, I've found. This is a significant issue that divides the audio enthusiast population into two groups, and that's an issue when talking about the importance of the high frequency characteristics of loudspeakers--that's probably not going to be resolved.
Heres a quick pick I took running out door this morning. No smoothing, 1 meter , approx. 15 degrees off axis, with the larger "wave guide" 15 inches by 10 inches mouth about 5 inches deep, 90 degree wall to wall straight sided.ribbon is 3 in long by about 3/4 inch wide. Magnet depth is 1/4 inch deep to diaphragm surface so theres a 1/4 inch straight sided slot to wave guide wall thats at a 45 deg angle from there.
Whats interesting to me is this basic down tilted response is close to how I tailor the natural ribbon ( no wave guide) response in almost every single direct radiator system I have ever voiced this ribbon into. EQed flat always seems to sound unnaturally bright. This has proven true in 3 vary different rooms and on too many builds to remember.
Soooo, looking at the natural freq response with out wave guide its typical ribbon rising then falling response. It rises smoothly about 5 dB from 700 to about 10 k Hz, then smoothly back down 5 dB to 20 k. This upper octave response fall off from peak is a combination of a practical transformer and the mass of the ribbon. Lighter ribbons and or running direct (no transformer) can extend to 20k with less roll off, however this ribbon construction/ transformer easily gives +- 1.5db from 1k to past 20Khz in direct radiator use AND she is considerably more reliable than the lighter ribbons.
Anyway if you overlay the ribbons natural response with the wave guided response I see that the wave guide has raised up the low end to meet the level of the natural response at the 10 khz peak, BUT, after that the response follows the natural response ( roll off from 10k to 20k). I will try to get a pic of the overlaid response tomorrow.
So what Im thinking at the moment is that IF the ribbons natural response continued to rise or stay flat at least after the 10 khz peak, then I would see a flatter response out to 20k on the wave guided FR.
Wondering at the moment if the 3/4 inch wide ribbon and NO phase plug is less the issue and if its really just the FR of the ribbon driving the wave guide rsponse after about 10 k??
Can you show a close-up picture of the ribbon mounted in the horn at the throat? The reason for the request is to see the transition from ribbon surrounding structure to horn--if there are any gaps, etc.
I hear ya on this. What confuses me is two fold..
1- 30 years ago I started developing ribbons. I could hear out to 20k then, yet even then I would tailor FR to a down tilt to get what sounded right to me, AND even then I remember questioning the necessity of flat to 20k
2- today I have trouble hearing beyond 13k, However I am still tailoring to the same down tilt. If I am loosing high freq hearing why am I still tailoring to the same a down tilt ( approx -2db at 10k ?)
Initial testing was crude and quickly revealed the importance of this transition. From there I made other wave guides that were a closer match to the ribbon slot and then filled in the small deviations with clay or tape to smooth the transition.
In the end I did get smoother response up high with the matched transition BUT the basic overall FR levels were close to same.
perhaps the section of the "horn" I have now that would be responsible for above 10k is just not the right shape?
As mentioned its a parallel walled "slot" of the magnets thats 1/4 " deep to diaphragm, then transitions to a 45 degree straight wall guide. The transition is now smooth BUT it is an "edge" so to speak rather than a rounded transition.
Last edited:
I agree with you. Most of the commercially available recordings would sound artificial if the speakers are flat to 20K. Old people like us simply can't hear that, maybe.
One question I have here is, many of us can clearly hear the deference in the steep digital filter around 22K. Why we can hear it?
I think that the answer to that mystery has been identified: aliasing is very audible, in fact it's pretty horrible. So it's not a 44.1 kHz sampling problem so much as a 22 kHz anti-aliasing filter used during the recording process. It's less clear what is occurring during down-sampling from 48 or 96 kHz sampling rates, however, if any differences actually exist. The main issue (aliasing) varies widely between recordings, etc., and seems to have decreased in severity as time has passed since the 1983 introduction of 44.1 kHz/16 bit CDs.
If you hear issues today, it's probably due to phase scrambling of down-mix stereo tracks due to mastering processes, which are EXTENSIVE and very damaging to preserving the realism of the music itself. (Ask anyone that's heard the original multitrack mixes or the original downmixes before they go to mastering--assuming stem mixing is largely not present during the mixing process. They all report amazing fidelity in the downmixes and a huge loss of fidelity after mastering. Go figure. I'd like to have the original downmixes. I've found that I can do my own mastering much better--on average.)
I've also noticed that the popular music labels are maybe not as careful about paying attention to those issues than classical music record companies to maintain high quality recording and signal processing equipment throughout...even though claims to the contrary are often made.
I have consistently heard the difference between 96/24 sampled recordings and 44.1/16, and I would identify that it's in the decays. The human hearing system appears to be exceptionally sensitive to changes in decay rates/smoothness (much more sensitive than people seem to collectively believe). When you go to 24 bits of bit depth, you're actually hearing smoother decays.
YMMV.
Chris
If you hear issues today, it's probably due to phase scrambling of down-mix stereo tracks due to mastering processes, which are EXTENSIVE and very damaging to preserving the realism of the music itself. (Ask anyone that's heard the original multitrack mixes or the original downmixes before they go to mastering--assuming stem mixing is largely not present during the mixing process. They all report amazing fidelity in the downmixes and a huge loss of fidelity after mastering. Go figure. I'd like to have the original downmixes. I've found that I can do my own mastering much better--on average.)
I've also noticed that the popular music labels are maybe not as careful about paying attention to those issues than classical music record companies to maintain high quality recording and signal processing equipment throughout...even though claims to the contrary are often made.
I have consistently heard the difference between 96/24 sampled recordings and 44.1/16, and I would identify that it's in the decays. The human hearing system appears to be exceptionally sensitive to changes in decay rates/smoothness (much more sensitive than people seem to collectively believe). When you go to 24 bits of bit depth, you're actually hearing smoother decays.
YMMV.
Chris
Last edited:
Hmm interesting. Not sure its a good comparison BUT it wasn't until the ribbon was developed to point where I was seeing smoother decays ( not just reduced resonance levels) that the sound became noticeably better.
..if I'm following you correctly:
This is all about dispersion of the driver, and how it is "squashed" relative to the resulting gain or loss at a particular angle.
Above 10 kHz you have greater pressure loss off-axis due to the source's (ribbon's) dimensions. (..and particularly the vertical contribution (to both vertical and horizontal response) is depressed - providing much less gain over-all.)
As a result you have less pressure to work with and must squash the directivity of the horn further IF you want to get more pressure near the 0 degree axis, which will result in further pressure loss outside of a certain axial window centered on 0 degree. (..in this sort of example it's often like +/- 10 or 15 degrees horizontally, and far worse/narrower vertically.)
Perceptually however, we tend to associate higher spl with more directive pressure.
Additionally, as you move further away from the source at higher freq.s a line-source behavior (within the vertical "window") develops which objectively results in greater pressure relative to a point-source character at lower freq.s..
..here is a modeler:
ABEC3
Generating diffraction increases ripple in the summed-output (direct from ribbon interacting with diffraction). Less ripple = smoother decays.
The deeper the waveguide, the greater the diffraction from the waveguide. A deeper waveguide also tends provide a greater reaction with that diffraction.
Suggestion: try modeling with less deep waveguides (say, starting at half an inch in depth), and then scaling-"up" with increasing depth to note differences.
- Status
- Not open for further replies.