I have been working on scripts to speed up development of midrange horns in AKABAK3. They are in an early stage of development but I have decided to share them as they might be useful to people. At the moment they can generate meshes for horns based on horizontal and vertical guiding curves. This means that you can create scripts to draw any rectangular horn. Everything is very crude atm!
https://github.com/kipman725/OCTAVE_HORNS
The reason for this is that I am trying to make a horn for the M200 compresion driver with the following specification:
90x60 dispersion (-6dB)
400Hz - 4kHz operating band
maximum 0.9x0.5m mouth size
'good' loading, I'm not quite sure what would be acceptable but it seems a normalised real part of throat impedance of 0.2 is posible at 300Hz
Attached is some data on various exponential flare sections joined to a 90x60 conic-conic horn simlar to the EV HP640:
I have been adding the feature to push back the start of the vertical conical flare into the exponential horizontal flare. This is similar the the Altec Mantaray horn in that it allows the horizontal and vertical dimension to both be maximized rather than the maximum horizontal constraining the vertical (see attachment). This will decrease the frequency at which the horn has vertical directivity control and thus provide a better directivity match to the midbass horn; however the sudden change in the area expansion rule (arround z=0.1m in the attachment) I think will cause ripple in the throat impedance. Looking at the attached quadratic throat waveguide document the solution to the change in vertical wall angle is to use a circular arc so I will try that.
Another interesting point of starting the vertical expansion early is that the area expansion is not exponential for as long as before due to the additional expansion added by the vertical flare. This could be compensated for in the horizontal plane but would push the angle matching horizontal side wall point further towards the mouth. It would also have the effect of reducing frequency at which the horizontal started to beam.
Another area I have been looking at is the addition of fins in the exponential section to reduce beaming. This site has some info: https://www.araihorn.com/p/horn-design.html
There is also this Klipsch patent on putting inclusions into the horn that could be interesting in terms of correcting the area expansions:
https://patents.google.com/patent/US7686129B2/en?q=klipsch&oq=klipsch&page=2
(the focus of the patent is on directivity control but how one would calculate the bumps on that basis I don't know).
Another interesting point of starting the vertical expansion early is that the area expansion is not exponential for as long as before due to the additional expansion added by the vertical flare. This could be compensated for in the horizontal plane but would push the angle matching horizontal side wall point further towards the mouth. It would also have the effect of reducing frequency at which the horizontal started to beam.
Another area I have been looking at is the addition of fins in the exponential section to reduce beaming. This site has some info: https://www.araihorn.com/p/horn-design.html
There is also this Klipsch patent on putting inclusions into the horn that could be interesting in terms of correcting the area expansions:
https://patents.google.com/patent/US7686129B2/en?q=klipsch&oq=klipsch&page=2
(the focus of the patent is on directivity control but how one would calculate the bumps on that basis I don't know).
Attachments
I got a pair of horns from user ZXPC on ebay
11" x 17" ABS 2" Bolt-On Long Throw Horn 90° x 40° For Many 2" Exit Driver
...which I extended.
(1) added a roundover at the mouth
(2) mounted into their own cabinets (guttet JBL "cabaret" boxes).
The M200 worked pretty well on that horn (SPL plot below is of this combo). The operating band I got was a little bit wider than what you want.
I didn't do polars / serious testing ...but it seems like this combo would do almost exactly what you are after, and is pretty cheap & easy.
Note - I sold my pair of M200 because I preferred the sound of a 6" on the same horn - but my experiments were far from exhaustive.
11" x 17" ABS 2" Bolt-On Long Throw Horn 90° x 40° For Many 2" Exit Driver
...which I extended.
(1) added a roundover at the mouth
(2) mounted into their own cabinets (guttet JBL "cabaret" boxes).
The M200 worked pretty well on that horn (SPL plot below is of this combo). The operating band I got was a little bit wider than what you want.
I didn't do polars / serious testing ...but it seems like this combo would do almost exactly what you are after, and is pretty cheap & easy.
Note - I sold my pair of M200 because I preferred the sound of a 6" on the same horn - but my experiments were far from exhaustive.
Thats a good frequency response, its noticably extendend compared to the datasheet response with the SH494:
JBL also seem to think cone drivers are the way to go for this kind of application: https://jblpro.com/en/site_elements/tech-note-cone-midrange-compression-drivers-cmcd due mainly to increased excursion capability resulting in more output in the low end. There is always the BMS 4599ND though if the M200 isn't enough....
JBL also seem to think cone drivers are the way to go for this kind of application: https://jblpro.com/en/site_elements/tech-note-cone-midrange-compression-drivers-cmcd due mainly to increased excursion capability resulting in more output in the low end. There is always the BMS 4599ND though if the M200 isn't enough....
Nice, thanks for sharing! Not sure if @mabat can use any of your functions for rectangular horns in Ath, but it’s handy to have as a separate tool regardless.
I’ve not used Octave in some time, but MATLAB has a pretty tidy GUI builder. Graphical controls can be easily mapped to variables, should you get to the stage of wanting to add sliders and a ‘real time’ visual of the profile.
There are also a variety of FEA & BEM codes that can be run on a mesh within MATLAB/Octave without passing over to AKABAK. Perhaps it could even feed your curves into Kolbrek’s MMM Toolbox. That might allow for an easier iterative process via algorithmic parameter adjustment to meet your goals.
Code & scripting aside, and without wanting to throw too much of a curveball - I believe you’re keen on multiple entry horns? Community actually launched one of those using the M200 in 2019:
https://downloads.biamp.com/assets/...heets/biamp_data_sheet_community-lvh909as.pdf
It’s definitely ‘inspired’ by the layout of something else, but the coverage pattern of one variant meets your goal while not being too much larger than your maximum mouth dimensions.
If nothing else, the main flare profile might be interesting. They seem to be using a small chamber with slotted exits from the HF waveguide divisions to 'tap' the M200s into the main horn. It could work even if you have less driver density, or truncate the throat section length and mount the middle M200 of the three shown axially instead of a 90-degree entry:
There are a bunch more drawings in the associated patent US202/086563A, which is actually pretty cool in itself. They use thin slots that exit into small chambers filled with resistive open cell foam at the HF throat.
It seems to reduce the delayed midrange reflections from the 'small end' of the horn, and associated lobing error. The overall response of the MF section seems pretty good, although the plot provided is lacking scale numbering:
I’ve not used Octave in some time, but MATLAB has a pretty tidy GUI builder. Graphical controls can be easily mapped to variables, should you get to the stage of wanting to add sliders and a ‘real time’ visual of the profile.
There are also a variety of FEA & BEM codes that can be run on a mesh within MATLAB/Octave without passing over to AKABAK. Perhaps it could even feed your curves into Kolbrek’s MMM Toolbox. That might allow for an easier iterative process via algorithmic parameter adjustment to meet your goals.
Code & scripting aside, and without wanting to throw too much of a curveball - I believe you’re keen on multiple entry horns? Community actually launched one of those using the M200 in 2019:
https://downloads.biamp.com/assets/...heets/biamp_data_sheet_community-lvh909as.pdf
It’s definitely ‘inspired’ by the layout of something else, but the coverage pattern of one variant meets your goal while not being too much larger than your maximum mouth dimensions.
If nothing else, the main flare profile might be interesting. They seem to be using a small chamber with slotted exits from the HF waveguide divisions to 'tap' the M200s into the main horn. It could work even if you have less driver density, or truncate the throat section length and mount the middle M200 of the three shown axially instead of a 90-degree entry:
There are a bunch more drawings in the associated patent US202/086563A, which is actually pretty cool in itself. They use thin slots that exit into small chambers filled with resistive open cell foam at the HF throat.
It seems to reduce the delayed midrange reflections from the 'small end' of the horn, and associated lobing error. The overall response of the MF section seems pretty good, although the plot provided is lacking scale numbering:
Last edited:
I aquired more M200's so thought I should check the condition of my stock. This is the test setup with the mic 60cm from the horn mouth, SPL adjusted for 1m, results gated at 9.6mS to get rip of the worst room reflections. The horn is from a Community CSX70-S2 and is quite short, only realy working down to 700Hz but I don't have any larger 2" throat horns to hand. Driver "05969 (repair)" had broken lead in wires which I have fixed (hence the hot glue to seal the driver back up).
The most noticable differences is that 15795 and 15464 (the new drivers) display increased sensitivity 600-1kHz and a notch at 5kHz. They are however very well matched with each other and have far larger serial numbers than the drivers I have so are presumably newer.
looking at the datasheet frequency response of the RMG-200A which uses a single M200 a simple horn this 5kHz notch is also visible:
I next measured the impdances of the drivers in free air:
Not sure whats going on here but as a broad trend the newer drivers have a lower Fs and higher Q. Physicaly the newer drivers have a less deep 'rim' on the top plastic moulding so at least some in production changes:
The THD also displays different trends for the older and newer drivers.
05887:
05952:
05969:
15464:
15795:
so from this I would say that perhaps 15464 is working best (0.5-1% THD at 102-107B/1m!) and 15795 possibly has dried up ferrofluid.
05887:
05952:
05969:
15464:
15795:
so from this I would say that perhaps 15464 is working best (0.5-1% THD at 102-107B/1m!) and 15795 possibly has dried up ferrofluid.
For fun I also measured the 'melded' tweeter arrays that feature authentic Motorola CTS peizo tweeters:
Its all at a very early stage just this week I changed the internal horn data format twice. I will be updating with big improvements soon. But if anyone wants to use it the code its out there, I'm very keen on sharing.
this is a great idea I will look into it. The AKABAK approach can be a a bit fustrating due to the actual data you want been obscured by simulation artifacts due to its more universial applicability.
Great find! I have never seen compresion midrange drivers on a MEH. I do like MEH, using them as my home system but with this system I wanted to go all out with the 'big stack of horns' trying to stick with one horn for each frequency band but using modern DSP and considering directivity of the horns.
I noted that extra HF extension, and am unsure of the reason why - but maybe it is because it was a very "burned in" unit. The driver was a survivor from a rig that had experienced very heavy use.
I didn't go M200 --> cones because I needed thunderous output, I just found the cones extremely easy to implement and enjoy (I could run them without any crossover at all due to the smooth response and built-in bandpassing), plus I'm less worried about damaging them - they cost a lot less than a 2" compression driver diaphragm.
Why are you considering building a stack if you are used to MEH? Most people seem to prefer the point source thing and go from stacks to MEH. Migrating the other way seems like a bit of an anomaly 🙂
I've not yet built a full MEH but have tinkered with prototypes, and my not very advanced technique (make a simple midrange hole, then try a whole lot of random drivers on it) seems to work pretty well.
I'm not getting rid of the MEH they arestill my home system. These horns are to replace the 3 soundgear orbits we use per side on a horn loaded system (current state below):
I just want a big stack of horns because I can 🙂 they are less acoustically sensible than a large MEH but it can still sound great and there is the flexibility to optimise each horn for the frequency band its operating in. For the specific system it will sound much better with non arrayed tops and the horizontal dispersion will be well controlled from >400Hz with matching top horns.
Summary of new progress:
1) The horns are now generated with fine initial sampling along their length (E.G 1mm) with no attempt to place points intelligently at angle changes etc. This brute force approach is used because the previous approach of using points of interest gets too complicated when horns have independent horizontal and vertical profiles.
2) A points culling operating is performed on the points to reduce complexity for the .STL generation.
3) The start of the vertical flare can now be pushed back towards the throat in order to maximise the mouth size. This makes the area expansion less ideal but results in improved directivity.
An example of the new horn design with 200Hz initial exponential flare is shown:
Next is to smooth the first vertical flare angle change (as this is 3D printed) and to try a throat vane to reduce the polar narrowing of the horizontal. I also have two more M200 on the way of more recent production.
1) The horns are now generated with fine initial sampling along their length (E.G 1mm) with no attempt to place points intelligently at angle changes etc. This brute force approach is used because the previous approach of using points of interest gets too complicated when horns have independent horizontal and vertical profiles.
2) A points culling operating is performed on the points to reduce complexity for the .STL generation.
3) The start of the vertical flare can now be pushed back towards the throat in order to maximise the mouth size. This makes the area expansion less ideal but results in improved directivity.
An example of the new horn design with 200Hz initial exponential flare is shown:
Next is to smooth the first vertical flare angle change (as this is 3D printed) and to try a throat vane to reduce the polar narrowing of the horizontal. I also have two more M200 on the way of more recent production.
You have imo a wrong sfa expansion in your math. The impedance is disappointing for such a long horn. What is your scale for the area? It looks like log10? What expansion coefficient has been used to calculate the surface area (sfa)? I suggest that you reed the Keele papers how long you have to make the exponential section. Fo Szr such flat front horns the math is more than simple:
Sz = S0 * exp^(m*z); Sz = x*y
This is easy to solve for x with constant y. The disadvantage of your flat front construction wave front is that it is not isophase.
Sz = S0 * exp^(m*z); Sz = x*y
This is easy to solve for x with constant y. The disadvantage of your flat front construction wave front is that it is not isophase.
Yes I'm deliberately comprising the loading by not following the Keele optimal exponential to conic criteria and then further compromising it by starting the vertical expansion early*. This is because I am trying to optimise directivity and not minimise the reflections at the junction point, I also didn't want any negative expansion in x or y. It may be better to follow Keele and match the horizontal wall angle by having an initial vertical expansion, but my experiance is that this will restrict the pattern by 4kHz. I don't think the loading is too bad if you compare the frequency repsonse with a constant power source to the EV HP9040, taking into account the HP9040 is longer as it has a narrower vertical pattern:
My horn (green is 3m):
ASFIK the only disadvantage to seperate vertical and horizontal apparent origins are arraying issues? (I only intend this horn to be use singly)
*Its worth noting that peaks >1 in relative throat impedance have a lesser relative effect than dips <1 on efficiency
My horn (green is 3m):
ASFIK the only disadvantage to seperate vertical and horizontal apparent origins are arraying issues? (I only intend this horn to be use singly)
*Its worth noting that peaks >1 in relative throat impedance have a lesser relative effect than dips <1 on efficiency
I have two more M200 (58309 and 28385) which are of more recent manufacture matching 15795 and 15464 (the drivers also with the 4-5kHz notch).
THD looks good:
58309:
28385:
Having 4 drivers does also raise the posibility of using two on each horn....
THD looks good:
58309:
28385:
Having 4 drivers does also raise the posibility of using two on each horn....
After a pretty busy festival season its time to get back on it! I decided one M200 per 90x60 horn is not enough and to design a dual driver throat adaptor. Existing dual driver throat adaptors tend to suck resulting in less than the expected 6dB output gain (assuming drivers in parallel) from fully coherant summation. This is an example of a 'good' adaptor with 2dB output gain while you would expect 6dB for a perfect adaptor. This adaptor is actualy one of the better examples as it expands two 1.3" throat drivers to a combined 2" throat, some adaptors have negative area expansion as they do things like put two 1" throat drivers onto a 1" throat!
So I have designed a dual driver throat adaptor with the following features:
Half of it is show below:
full thing:
optimised path:
M200s represented by green lines.
At the moment I just have point cloud which is not ideal for CAM and simulation, mabat posted some path extrusion scripts for fusion so I will try fiddling with that next.
So I have designed a dual driver throat adaptor with the following features:
- 2" driver throat (round)
- 4" exit (square)
- exponential area expansion (this example has Fc=200Hz)
- length determined by area expansion and throat/exit size
- minimum bend for driver clearance
Half of it is show below:
full thing:
optimised path:
M200s represented by green lines.
At the moment I just have point cloud which is not ideal for CAM and simulation, mabat posted some path extrusion scripts for fusion so I will try fiddling with that next.
Finaly managed to sim half an adaptor in AKABAK. To do this I basicly made a fusion360 model and meshed it in gmesh, the details are however quite complex*. Simulated with an absorbing boundry at the mouth. Results are a bit different to what I would expect as its meant to be an exponential area expansion with fc=200Hz, so I would expect a sharp cut off.
*1) octave generate cross sections and planes to draw said cross sections on
2) export to text file of commands to run in fusion360
3) python script to process commands (thanks mabat)
Had to do a lot more work to make the adpator compatible with the rest of the horn as it had to go to fully rectangular at the end which was causing lofting faliures. To fix this I ended up making the cross sections of the adaptor for lofting out of 4 spline curves joined at the mid corner position, giving the loft command something to attach the corner to.
This is the full internal horn shape, a simple conic-conic horn has been added onto the exponential throat adaptors:
and in AKABAK an infinite baffle simulation is performed:
~0.6 @ 400hz
This looks pretty much as I would expect apart from the horizontal polars which are broader in the 1k-3k range and then suddenly collapse. I suspect something to do with the area where the two throat adaptors meet (the new common square throat) so will have a look at the fields there. The overall perfomance looks usable 400-3k so I think I'm getting close to a final design.
This is the full internal horn shape, a simple conic-conic horn has been added onto the exponential throat adaptors:
and in AKABAK an infinite baffle simulation is performed:
~0.6 @ 400hz
This looks pretty much as I would expect apart from the horizontal polars which are broader in the 1k-3k range and then suddenly collapse. I suspect something to do with the area where the two throat adaptors meet (the new common square throat) so will have a look at the fields there. The overall perfomance looks usable 400-3k so I think I'm getting close to a final design.
So looking into the horizontal pattern issues.
379Hz the wavefront in the throat looks planar, waves in horn look as expected for conical horn:
1600Hz somewhat odd, perhaps some cancelation in the initial conical flare and then weird interaction with the flare
4k the source size is causing beaming
so I would guess something to do with the throat having two acoustic centers that are offset? might work better if I rotate the throat adaptors 90 degree so the offset centers are driving the 60 degree horn.
379Hz the wavefront in the throat looks planar, waves in horn look as expected for conical horn:
1600Hz somewhat odd, perhaps some cancelation in the initial conical flare and then weird interaction with the flare
4k the source size is causing beaming
so I would guess something to do with the throat having two acoustic centers that are offset? might work better if I rotate the throat adaptors 90 degree so the offset centers are driving the 60 degree horn.
I rotated the adaptor 90 degrees and observe very little change in the horizontal directivity:
I think perhaps I'm seeing an issue with the flare (bellow 3.5kHz after that the narrowing is the throat). Fortunatly my freinds inform me making a curved flare is not very difficult so I will write a script to generate a 'modified tractix' flare as per the K402 horn. I think this is relativly straightforward just itterating the curve to match tangents at the join point to the conical section and then itterating overal horn length down until the flare and the horn fit within the dimensional constraints.
*there will also be always some degree of 'issues' with this horn as the area expansion isn't constant from the adaptor to the flare causing some reflections.
I think perhaps I'm seeing an issue with the flare (bellow 3.5kHz after that the narrowing is the throat). Fortunatly my freinds inform me making a curved flare is not very difficult so I will write a script to generate a 'modified tractix' flare as per the K402 horn. I think this is relativly straightforward just itterating the curve to match tangents at the join point to the conical section and then itterating overal horn length down until the flare and the horn fit within the dimensional constraints.
*there will also be always some degree of 'issues' with this horn as the area expansion isn't constant from the adaptor to the flare causing some reflections.