Yes, it is good approach to make things modular, and objectives you explain are fine, only knowing some particular context and exact ideal of sound might help refine further. My motivation here is expand perspective on things, not to take you or OP to any particular direction. I don't know any better.
Here is an interesting candidate for 6.5 inch midrange.
only US$380 , probably less than the ATC dome mid.
It's quite sensitive and ticks all the boxes except the looks. eek
see the data sheets
the raw data fr and imp can be downloaded
https://ptt.purifi-audio.com/shop/p...,4164,4165,4166,4167,4168,4169,4170,4171,7598
only US$380 , probably less than the ATC dome mid.
It's quite sensitive and ticks all the boxes except the looks. eek
see the data sheets
the raw data fr and imp can be downloaded
https://ptt.purifi-audio.com/shop/p...,4164,4165,4166,4167,4168,4169,4170,4171,7598
My impression is that on the 12cm tube the global dispersion goes from very wide below 1000 hertz to a constant 90 degrees above 1000 hertz very quickly.
Is this correct? If yes then Houston that is a real problem because of the sudden shift in dispersion even after normalised response will cause a huge shift in reflected sound in the pass band.
Is this correct? If yes then Houston that is a real problem because of the sudden shift in dispersion even after normalised response will cause a huge shift in reflected sound in the pass band.
Last edited:
Yes, you can read that form directivity index (DI) graph, which goes from ~0 to ~10db within the passband in the minimal baffle example. DI is the red line on center left image on the sixpack(s) I posted. Same information is basically visible on the normalized polar maps, and also readable from relationship of axial sound and power response, which together make the DI.
This is the reason why I wrote the two yield very different systems, the increase in DI is not at same bandwidth due to different size baffle. Coverage doesn't jump quickly but is smooth slope all the way which is not a problem itself, and which would happen with all systems that are not embedded into a wall at some frequency bandwidth and slope. Directivity on low bass is ~0 due to very long wavelength no matter the system, to ~10db+ for 20kHz due to very short wavelength, beaming of tweeter for any available tweeters (waveguides can widen response a bit on the top as well). For this reason the DI always increases somehow.
Sudden shifts in directivity happen with poorly implemented multiway systems where bigger transducer beams and crossover happens at frequency where the small one is still ~omni. When DI of these two are too different and it would make a jump in directivity through crossover.
Reasoning from these, if mid is on minimal baffle and DI goes from 0 to high within pass band, the tweeter should also have high DI at crossover = waveguide, and the woofer should also have low DI = narrow baffle as well. IF woofer was on big baffle, it could narrow and mid widen at crossover, which means DI won't match and jerk in directivity, same if there is dome tweeter without waveguide.
This is the reason why I wrote the two yield very different systems, the increase in DI is not at same bandwidth due to different size baffle. Coverage doesn't jump quickly but is smooth slope all the way which is not a problem itself, and which would happen with all systems that are not embedded into a wall at some frequency bandwidth and slope. Directivity on low bass is ~0 due to very long wavelength no matter the system, to ~10db+ for 20kHz due to very short wavelength, beaming of tweeter for any available tweeters (waveguides can widen response a bit on the top as well). For this reason the DI always increases somehow.
Sudden shifts in directivity happen with poorly implemented multiway systems where bigger transducer beams and crossover happens at frequency where the small one is still ~omni. When DI of these two are too different and it would make a jump in directivity through crossover.
Reasoning from these, if mid is on minimal baffle and DI goes from 0 to high within pass band, the tweeter should also have high DI at crossover = waveguide, and the woofer should also have low DI = narrow baffle as well. IF woofer was on big baffle, it could narrow and mid widen at crossover, which means DI won't match and jerk in directivity, same if there is dome tweeter without waveguide.
Last edited:
This kind of smoothly rising DI can integrate well with the right tweeter in the right baffle. As you say, VituixCad allows us to simulate the impact of baffle shape and driver spacing.
Here is an example of what can be done with VituixCad diffraction tool. I have been experimenting with sims of the Bliesma M74 driver with a small frame tweeter such as ScanSpeak D3004-6040, which has a flange diameter of 62 mm.
I found that a small trapezoid baffle worked well, a baffle just large enough for the two drivers and a 37 mm radius edge all around. The tweeter does not use a waveguide, but the baffle radius begins where the flange ends. This sim uses 2 x 10” woofers which would have a separate cabinet. Crossovers are 600 and 3k.
The DI is flat from 400 to 8k, +/- 1 dB. Of course this is a simulation using flat disk pistons, and real drivers will have a slightly different radiation pattern, but I have found this kind of preliminary simulation to be very helpful in the early stages of the design. It greatly reduces the risk of design flaw. These kinds of flaws, baffle shape and driver spacing, are not really fixable after the speaker is built. Generally, an entirely new cabinet has to be built.
I have not tried to simulate circular baffles for the midrange and tweeter (cylindrical mounting). My intuitive expectation is that it should be fairly similar to my trapezoid.
My main point here is that simulating in advance reduces our risk of unfixable design flaws. If we base our designs on intuition or “gut feel”, there is a higher risk of getting a poor result.
I found that a small trapezoid baffle worked well, a baffle just large enough for the two drivers and a 37 mm radius edge all around. The tweeter does not use a waveguide, but the baffle radius begins where the flange ends. This sim uses 2 x 10” woofers which would have a separate cabinet. Crossovers are 600 and 3k.
The DI is flat from 400 to 8k, +/- 1 dB. Of course this is a simulation using flat disk pistons, and real drivers will have a slightly different radiation pattern, but I have found this kind of preliminary simulation to be very helpful in the early stages of the design. It greatly reduces the risk of design flaw. These kinds of flaws, baffle shape and driver spacing, are not really fixable after the speaker is built. Generally, an entirely new cabinet has to be built.
I have not tried to simulate circular baffles for the midrange and tweeter (cylindrical mounting). My intuitive expectation is that it should be fairly similar to my trapezoid.
My main point here is that simulating in advance reduces our risk of unfixable design flaws. If we base our designs on intuition or “gut feel”, there is a higher risk of getting a poor result.
Yes, it is not hard to simulate a circular baffle shape. We can do one for the tweeter, and one for the mid, and combine the two in simulation.
What diameter, and what is the edge radius? A sharp edge we can model as a 1 mm radius. A gentle rounded edge can be modelled with as much radius as the tooling allows. For me that is 37 mm max.
@tmuikku has made the argument that a sharp edge radius is acceptable on the mid driver, as long as the circular baffle is no larger than the driver. For the Bliesma M74 that would be 121 mm diameter. For the SB Satori MD60N, that would be 130 mm diameter.
For the tweeter, I recommend as much edge radius as possible.
What diameter, and what is the edge radius? A sharp edge we can model as a 1 mm radius. A gentle rounded edge can be modelled with as much radius as the tooling allows. For me that is 37 mm max.
@tmuikku has made the argument that a sharp edge radius is acceptable on the mid driver, as long as the circular baffle is no larger than the driver. For the Bliesma M74 that would be 121 mm diameter. For the SB Satori MD60N, that would be 130 mm diameter.
For the tweeter, I recommend as much edge radius as possible.
Summary
The main premise for all this is that the off axis response should always track the on axis response. In terms of imaging and reflected and direct sound ratios nothing else matters.
Geddes has researched that the key area of interest is from 700-7000 hertz. The reasons are well beyond detailed discussions here.
In a two way system the idea is to match the woofer off axis response with the tweeter.
But in a three way with the mid starting at 400 hertz what really happens to the off axis response.
The main premise for all this is that the off axis response should always track the on axis response. In terms of imaging and reflected and direct sound ratios nothing else matters.
Geddes has researched that the key area of interest is from 700-7000 hertz. The reasons are well beyond detailed discussions here.
In a two way system the idea is to match the woofer off axis response with the tweeter.
But in a three way with the mid starting at 400 hertz what really happens to the off axis response.
For the 74mm dome, I assumed a flange size of 121 mm, and assumed the baffle is the same size circle as the flange, and the radius edge is a hard edge. Although the voice coil is 74mm, the radiating surface is closer to 80mm considering the Sd of 50 cm^2. I plotted a frequency range from 200 - 20k.
A 13 sided polygon is a good approximation of a circle. Actually anything with more than 8 sides is a good approximation, but 13 sides looks more like a circle.
For the 6" cone driver, I used the physical dimensions of the Satori MW16TX driver. The Sd is 119 cm^2 and the frame size is 165mm. I assumed a circular baffle of the same size and a hard radius edge. I plotted a frequency range of 100 - 10k.
In both cases, there is an on-axis hump in the response, but this is not manifested as a hump in DI... the DI smoothly increases over the range. I applied a gentle EQ to both cases to flatten them out a bit. This helps visualize how the driver might be used in a system.
A 13 sided polygon is a good approximation of a circle. Actually anything with more than 8 sides is a good approximation, but 13 sides looks more like a circle.
For the 6" cone driver, I used the physical dimensions of the Satori MW16TX driver. The Sd is 119 cm^2 and the frame size is 165mm. I assumed a circular baffle of the same size and a hard radius edge. I plotted a frequency range of 100 - 10k.
In both cases, there is an on-axis hump in the response, but this is not manifested as a hump in DI... the DI smoothly increases over the range. I applied a gentle EQ to both cases to flatten them out a bit. This helps visualize how the driver might be used in a system.
That's great. Cool.
The rounded edge certainly makes a different
I am wondering now how the overall curves will look if a low pass shelf filter is applied to correct (normalises the baffle step?
In practise I would expect the baffle area of the woofers to provide some loading below 1000 hertz.
The rounded edge certainly makes a different
I am wondering now how the overall curves will look if a low pass shelf filter is applied to correct (normalises the baffle step?
In practise I would expect the baffle area of the woofers to provide some loading below 1000 hertz.
Can you include an approximation of the baffle for the cylinders above and below with the software? The diffraction off the edges of the neighbouring cylinders is unlikely to be as benign but will it be a major or minor issue?
It would be nice if VituixCad had that capability, but it does not. The diffraction modelling ends after the sound wave diffracts 90 degrees at the baffle edge. There is no additional diffraction modelling along the sides or back edge of a cabinet. I think this means the cabinet is assumed to extend backwards for an infinite depth... ?
Certainly I have measured differences between shallow depth cabinets and deep cabinets when both have the same baffle layout. It is usually in the 200 - 500 Hz range. The diffraction off of the back edge is real, but small, and there is so much room effect at those lower frequencies that I don't think that accuracy in that range is important.
Again, that is hard to model, and beyond the capability of this tool. If this were my project, I would build a prototype using less expensive drivers, and make a full set of horizontal and vertical polar FR scans. Then I would compare the actual results to the VituixCad predicted results to understand how/where the simulation is limited.
For example, this simulation seems to indicate that a hard edge on the midrange cylinder is benign, and it may very well be so for the midrange... but perhaps a rounded edge on the midrange cylinder is beneficial for the tweeter located just a few inches above it? The best way to find out is to build a prototype and test.
Last edited:
You can't make the baffle shape a figure-of-8 to approximate the front of a neighbouring cylinder? It would seem an odd restriction if not.
You wouldn't want to use 3D sound simulation software like akabak to tackle the full problem in a cheaper, quicker and more flexible manner? (Subject of course to having the software and knowing how to use it.) I am not familiair with vituixcad but the diffraction is almost certainly a fairly approximate one based on diffraction from edges.
In the case of stacking cylinders we already know from the measurements of the speakers with the best performance (Revel Salon 2, Vivid Giya) that smoothly filling in the edges between the figure-of-8 is the way to optimise off-axis performance (which might not be a concern for the OP). The assumptions in the vituixcad diffraction tool are unlikely to be able to simulate this adequately in the way it cannot simulate waveguides. Of course if there is no filling in just sharp or rounded edged cylinders then it is likely to be closer.
Yeah votuixCAD has no 3D modeling in, but one could do such with BEM or with prototypes. You can try estimate magnitude of effects in head as well 
Simplest is to consider wavelength, if it's much longer than dimensions of adjacent structure, the effect is minor. For example small mid-woofer tube is small for bass woofers sound, but the bass woofer tube would be relatively big for mid bandwidth wavelength. Then you could estimate how much then: take distance from center of mid cone to edge of the woofer box, which is rough approximation of delay the reflection from woofer box have before arriving to listening spot, which would make some amplitude interference pattern (combfilter). Lets estimate amplitude and bandwidth of combfilter then: If delay was say 20cm from mid cone center to woofer tube/box surface near the mid, so combfilter would start at frequency whose wavelength 20cm x 2, so from about 950Hz and up until the mid beams, for 3" dome perhaps up to 3" wavelength so any interference would be roughly between 900- 4kHz. Now imagine magnitude, sound reflects specularly like light, so if there was light bulb on the mid cone and the woofer box was mirror, would you see the light from the mirror? likely not ubless the woofer structure protrudes closer to listener than the mid. So, if there is interference it would be some secondary effect, like sound diffracting from backsides of the structure back toward listening window. These make perhaps db or few undulations, mainly somewhere else than on axis because it's already secondary or further effects with greatly diminished amplitude.
Simulation or real measurement would shownit more accurately though, but I'd say it's quite minor, less in magnitude than baffle edge secondary sound source makes.
Simplest is to consider wavelength, if it's much longer than dimensions of adjacent structure, the effect is minor. For example small mid-woofer tube is small for bass woofers sound, but the bass woofer tube would be relatively big for mid bandwidth wavelength. Then you could estimate how much then: take distance from center of mid cone to edge of the woofer box, which is rough approximation of delay the reflection from woofer box have before arriving to listening spot, which would make some amplitude interference pattern (combfilter). Lets estimate amplitude and bandwidth of combfilter then: If delay was say 20cm from mid cone center to woofer tube/box surface near the mid, so combfilter would start at frequency whose wavelength 20cm x 2, so from about 950Hz and up until the mid beams, for 3" dome perhaps up to 3" wavelength so any interference would be roughly between 900- 4kHz. Now imagine magnitude, sound reflects specularly like light, so if there was light bulb on the mid cone and the woofer box was mirror, would you see the light from the mirror? likely not ubless the woofer structure protrudes closer to listener than the mid. So, if there is interference it would be some secondary effect, like sound diffracting from backsides of the structure back toward listening window. These make perhaps db or few undulations, mainly somewhere else than on axis because it's already secondary or further effects with greatly diminished amplitude.
Simulation or real measurement would shownit more accurately though, but I'd say it's quite minor, less in magnitude than baffle edge secondary sound source makes.
This you could do yeah. In reality you could space them out a bit, not touching each other which likely reduces the interference because if the first edge diffracts, there is bit less amplitude left to diffract again, also the furthest edge (bottom of woofer, thinking sound that starts from mid), would be quite far and be quite short compared to "sound bubble" size so minor effect compared to the first edge diffraction effects.
Last edited:
Thank you for your advice. Since I was planning to cross midhigh and tweeter at 4kHz, and now am thinking of 4.5~5kHz, I guess I can equate the voice source to be just midhigh+midbass and the tweeter as a harmonics generator....
My FIR crossover slope would be 100+dB/oct, so hopefully I can avoid or mitigate some problems with this. It's even easier at such high frequency, consuming less computing resource.
Not at all. I am getting a lot of information and learning from you leading the discussion with other experts.
I am ditching 5" midbass due to its foam surround. My hand skill sucks and I don't want to to do the reconing of these expensive drivers with my own hands. ;=). I guess that makes the problem back to worse for vertically long voice source, but no efficiency/max SPL problem.
Maybe unrelated, but did you or someone propose to tilt the drivers' axes so the lines meet somewhere close in front of the driver's array? Something like what WAMM or other adjustable modular monitors do. Would it help addressing the large voice source problem?
Using DSP crossover, time alignment of the driver is nothing to worry about. If DSP time alignment can do the same as the tilting adjustment, I'd go with the DSP way.
I was wondering about it when I started this thread, but then now I got contaminated by the intrigue in 4-way. I have an ATC-based 3-way already, so I guess one of many points of this project is to do something different from the past....
I keep seeing the mentions of under hung motors. Could you explain a bit about it?
Thank you very much for another set of suggestions. I will try to digest them sooner or later.
Yes, except the middle one looking tilted. Do you suggest making it kind of diamond-y shape?