Monster Massive

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
Patrick,

Why is it you want to build a horn with the HF dispersion limited to only about 15 degrees, then transitioning to omni ?

Art

Because a pile of sims that I did seem to indicate that "in real life" the directivity is broadened.

IE, if you have two horns, one with a big roundover, and one without, the horn with the roundover will have wider directivity. And this isn't just at low frequencies; it appears that it broadens it at high frequencies.

The easiest way to determine if this hypothesis is true would be to look at polar plots of LeCleach or spherical horns. In both those horn types, the first half of the horn is about 25 degrees wide.

Unfortunately, that data is hard to come by. John Sheerin published a single plot on his site. Besides that, a lot of the polar plots on LeCleach seem to be sims, not 'real' measurements.

So I'll find out the hard way, by building and measuring!

Plus, I've built nothing but waveguides for years, and it might be fun to try a horn for once.
 
Because a pile of sims that I did seem to indicate that "in real life" the directivity is broadened.

IE, if you have two horns, one with a big roundover, and one without, the horn with the roundover will have wider directivity. And this isn't just at low frequencies; it appears that it broadens it at high frequencies.

The easiest way to determine if this hypothesis is true would be to look at polar plots of LeCleach or spherical horns. In both those horn types, the first half of the horn is about 25 degrees wide.
In all the "real life" measurements I have seen, the HF follows the initial throat angle, unless the throat is pinched below a wavelength wide and then diffracted, generally avoided for Hi-Fi use.

A large roundover will reduce diffraction effects, but won't change the HF dispersion.
By increasing the mouth dimension, it may provide pattern control to a lower frequency.

Take a look at the E-JMLC-1000 ( the first one I happened to find), goes from a -6 beamwidth of around 20 degrees vertical at 10,000 Hz to 80 degrees at 2000 Hz. At 20 kHz, you tell me, 8 degrees?
Even horizontally, looks like about 30 degrees at 16 kHz, 40 at 10 kHz.

Have fun!
 

Attachments

  • Beamy HF.png
    Beamy HF.png
    419.3 KB · Views: 1,107
Last edited:
Initially I was going to 'approximate' the shape of a LeCleach curve, by combining a conical horn with a humongous roundover.

Cosmetically, it looked like crap. I don't have a table saw, and trying to do the angles for a conical horn with a bandsaw didn't work out so hot.

So I went back to the method that has always worked well for me:
I'm making a mold.

IMG_20131103_185405.JPG

I'm using seventeen 1" layers of foamular.
The height and width of the horn follow the golden ratio.


IMG_20131103_185301.JPG


monster-massive4.jpg


IMG_20131103_185405.JPG


IMG_20131103_185536.JPG


IMG_20131104_170824.jpg

If you look at the mold, it has stair steps like a flower pot in a Devo video.
I had to use a rasp, a file, and some sandpaper to make it match the curve we see in Hornresp.

IMG_20131104_170846.jpg

Here's a pic showing the size of the horn in comparison to my Vandersteen 2Cs.
I plan on orienting the horn vertically, similar to what JBL does with their Synthesis 1400 array.
 
I would be very tempted to approximate the LeCleach horn using conical sections. Maybe something more like this? Would this make it easier to build?
Much easier to build, but the HF dispersion is still going to be determined by the initial throat angles to around one wavelength of the throat width.

Each horn "break" also causes some diffraction, though sanding the transitions takes care of most of that problem, leaving you with a beamy horn :).
 
Unfortunately, that data is hard to come by. John Sheerin published a single plot on his site. Besides that, a lot of the polar plots on LeCleach seem to be sims, not 'real' measurements.

A search for user jzagaja and JMLC might show some good info... He's the one from the Polish made horns, among others they run a few JMLC profiles :).
Usually he provides some (real) measurements.
 
Much easier to build, but the HF dispersion is still going to be determined by the initial throat angles to around one wavelength of the throat width.

Each horn "break" also causes some diffraction, though sanding the transitions takes care of most of that problem, leaving you with a beamy horn :).

Hey Art, here's an example of why someone might want to try a LeCleach profile for a Synergy horn:

cleach-11-8.jpg

This is the frequency response of my horn, with just two drivers. The midrange is a Dayton ND91. The tweeter is a Celestion CDX1-1425. The crossover consists of exactly one component: a 22ohm resistor.

Here's some things I see in the measurement:

1) The crossover from midrange to tweeter is so seamless it's ridiculous. I've been trying to achieve this kind of a transition for half a decade. The big 'Eureka' moment was when I realized that narrowing the wall angle 'fills in' the hole that had plagued all of my other Unity horn project. (My first Unity horn was 108 degrees by 72 degrees, back when I didn't grasp that Danley uses fifty degrees of coverage for a reason... it works better.)

2) Upper frequencies are clean and extended out past 20khz, much better than I could do with a DE250 or a D220Ti.

Maybe I'm getting carried away, but this is the first Synergy horn I've built where all the pieces just fell into place. Right now I think that all that's needed is another midrange to raise the output below 900hz, some frequency contouring on the tweeter, and I'm d-o-n-e.

Have been listening to this thing for a few hours now, in mono, duct tape and all. It sounds magnificent.
I posted a video to Youtube without any EQ on the horn; but I've been listening to it with EQ to flatten out the response.
 
In the last post I mentioned how a 'hole' in the midrange is a problem I've seen with a bunch of my Unity horns. Here's an example of what I mean. This is the frequency response of my first unity horn. One measurement is with the drivers in phase; the other is with the drivers out of phase. You'll notice a few things:

1) There's a 'depression' in the midrange
2) At the time I thought that the only way to 'fill in' the hole was to raise the level on the midranges, raise the level on the tweeter, or both. I think that Paul Spencer ran into similar issues with his Synergy horns; note that he used six midranges on one of his projects.
3) Adding midranges or tweeters works; just ask Tom Danley, some of his products use over a hundred drivers. This Unity horn was for my car, and I simply didn't have space for any more midranges. Four was all I could fit, and even that was an ugly mess.


Anyways, try narrowing the walls, or use a horn profile that's narrow at the throat. (tractrix, exponential, LeCleach, Spherical, etc.)


nov14-2.jpg


nov14-3.jpg


^^^ Again, these are seven year old measurements. This new project is proceeding a lot smoother than my old ones did.

newmid3.jpg

^^^ Another weird thing about that project was that I didn't understand that the length of the port adds a delay to the midranges. I think that's why both the in-phase and the out-of-phase response is so screwy. When the midrange and the tweeters were wired in phase, the ports were so long, it was adding about 180 degrees of rotation at the crossover. The net effect was a dip in the frequency response, even though the drivers were wired in phase. Reversing the phase filled in the hole, but added a peak lower in frequency.

Total no win situation!
 
Last edited:
Founder of XSA-Labs
Joined 2012
Paid Member
Patrick,
Very nice work with the LeCleach horn! The sound of the horn on the video is actually very nice - can you post some jazz with sax and bass to see how that sounds? You are the master of prototyping with whatever it takes to get it to fit. I love the tube duct and PVC chamber approach you did for your car - although the time delay was big. In my sims I see that the "dip" near 1Khz can be reduced by placing the mid driver closer to the mouth - about halfway down the length of the main expansion. Can you please provide detailed dimensions of your LeCleach horn? I would like to try it in foam core with smooth bends.
Regards,
Xrk971
 
Last edited:
Hey Art, here's an example of why someone might want to try a LeCleach profile for a Synergy horn:

1) The crossover from midrange to tweeter is so seamless it's ridiculous. I've been trying to achieve this kind of a transition for half a decade. The big 'Eureka' moment was when I realized that narrowing the wall angle 'fills in' the hole that had plagued all of my other Unity horn project.
The midrange "hole" that could not be fixed by reversing polarity was due to the phase difference between drivers at the acoustical crossover, not the horn wall angle.
It could have been fixed with delay, or proper passive crossover implementation (very difficult to do without looking at phase).

The detail one can hear from from very narrow HF dispersion horns can be very engaging on axis, things in the mix normally lost in room reverberation can be heard clearly.
Your new horn appears to have a smooth on axis response.
It would be interesting to see it equalized flat on axis (no smoothing), then see the curves at 5 degree increments out to 40 degrees off axis.

P.S. -Just listened to the video, the HF response just dissapears when you can't see the throat.
That's one beamy horn!

Art
 
Last edited:
On post 649 of 'Geddes on Waveguides', I posted a method to use hemp or jute instead of carbon fiber or fiberglass. Here's the post:

Since I accidentally turned this into the 'drugs and waveguides' thread, I'll try to get things back on topic.

While working in the garage today, I figured out a way to improve the throat of a waveguide, by using hemp. (See what I did there?)

Here's some background on the problem:

SP_000106_GP.jpg


At the throat of a waveguide, we want a really smooth transition from compression driver to waveguide. This is easy to do with a commercial waveguide, but when you build your own, it is very tricky.

Back when I first started building waveguides in fiberglass, I would make the throat the same size as the compression driver.

This doesn't work so great, because you always have to file down the throat, to make sure it's a perfect match. So when you file down a 1" throat, you end up with a throat that's a little too big.

The solution that I've used for the past couple years is to simply make the throat a little too small, and then I run a drill bit right down the middle. This gets you a perfectly symmetrical 1" opening.

Unfortunately, this solution isn't perfect, because you need a material for the horn that is both strong and substantial. Wood would work great, but wood isn't ideal for the complex shape of a waveguide, particularly at the throat.

Fiberglass is so so. My main gripe with fiberglass is that it takes a decent amount to build up enough thickness. Fiberglass mat works, but it isn't as strong as fabric.

Carbon fiber works great, but carbon fiber is expensive.

Ideally, I would want something like carbon fiber, but cheaper. Easier to work with would be a nice bonus.

In order to come up with a good solution, it's worthwhile to consider why carbon fiber works so well. In my opinion, carbon fiber works well because the strands are much much smaller than glass. Given a fixed amount of fiberglass or carbon fiber, the carbon fiber is going to have more strands, because the strands are smaller in diameter. ( 5–10 μm versus 9-13 μm)

Bottom line:

A throat made of carbon fiber produces a smoother finish than a throat made of fiberglass.

Here's an alternative to CF, that seems to work much better than I'd dreamed.

First, you go to a craft store and buy a ball of hemp yarn. The reason that we want hemp is because hemp is strong, and it has a lot of surface area, due to the rough microscopic nature of the strands.

Second, get yourself some gorilla glue.

At this point, we have about $12 invested. To do the same with epoxy and carbon fiber would require a trip to TAP Plastics, and about $100. (Admittedly, that'll buy you a lot of epoxy.)

Now take the hemp and run it under the faucet. We're getting it wet because moisture makes gorilla glue cure faster.

Now the messy part. Put some gloves on, put the hemp yarn into your hand, and pour some gorilla glue onto the yarn. Now really squeeze it! We want the glue to be thoroughly saturated into the yarn.

20120916184139_50750.jpg


If you've made it this far, you've now created the hemp equivalent of prepreg carbon fiber. (The way that most carbon fiber is used these days is with the epoxy impregnated into the carbon fiber, to maximize strength by minimizing the amount of epoxy.)

bamboo3.jpg


I got the idea from Calfee bicycles. This is how Calfee bonds bamboo to bamboo. (I believe they use epoxy, not polyurethane glue.)

There's some neat advantages to gorilla glue though:

1) The huge one is getting the perfect mix every time. In my fiberglass projects I've found that the strength of fiberglass varies a lot on whether the epoxy is mixed perfectly, and the perfect mix depends on what temperature it is outside! (ugh.)
2) I was astonished at how quickly the gorilla glue cures when you add water to the hemp. What would normally take 24 hours took about one.
3) When the water/hemp/gorilla glue cures, it's hard like concrete I simply couldn't believe how strong it was. It's insane. As demonstrated by Calfee, it can take a serious beating.
4) The volume of the hemp is a lot greater than carbon fiber or glass, which is handy when you're trying to lay up something quickly. (This could also be a downside of course, if you needed a very thin layer this is not the best method.)
5) Probably one of the neatest things that you can do with this trick is to leverage loops and knots and braids. Try doing THAT with carbon fiber! This is an important advantage, as companies like Ferrari and Lamborghini have demonstrated that the orientation of carbon fiber plays a large role in it's strength. With hemp you can take this to another level, by literally tying two pieces together. (For example, if you are bonding a mounting plate to a horn, you can drill holes through the mounting plate and tie the two pieces together, prior to adding the glue.)

Basically the ideal material for the horn throat is strong, easily sanded, and produces a smooth finish. Carbon fiber fits the bill. Fiberglass less so. Bond is fairly terrible, because bondo has virtually no strength. I can't count the number of waveguides where the bondo simply delaminated from the waveguide.

Hemp works pretty darn well, and is an interesting and inexpensive alternative to carbon fiber.
 
In the previous measurements I only had one midrange on the horn; now I have four.

monster-massive6.jpg


Now I have a ridiculous problem - my midrange array is more efficient than my compression driver!

What a strange problem to have. I may have to 'cut' the mids to bring their efficiency down to the tweeters level.

Some observations about the measurement:

1) The big peak at the cursor is the dipole peak. I'm not an expert on dipoles, but as I understand it, that peak will get lower in frequency as I increase the baffle size. So once the baffle is done, and the speaker is a full 40cm wide, that peak should be lower in amplitude and lower in level. (Right now the horn has no baffle at all.)

2) Even with the dipole peak reduced via a filter or via a larger baffle, the mids are still at least 6dB more sensitive than the tweeter. This is kinda amusing, because the efficiency of four of them on a baffle is something like 90dB. My tweeter has an efficiency of 107dB iirc. So we're getting a ton of gain from the horn.

The crossover is currently three components. A 1.5uf cap on the tweeter, and an L-Pad that's a 10ohm resistor in parallel with the compression driver and a series 2ohm resistor.

Although I don't need to attenuate the compression driver, I'm using an L-Pad mostly to normalize the impedance curve. (The 10ohm resistor offsets the rise in impedance near resonance.)

Without the L-Pad I found I was getting a peak at 2500hz as the impedance of the compression driver rose.
 
monster-massive5.jpg


Here's a measurement showing the combined response, and the midranges alone.
In this measurement, I wrapped the midranges with some wool to make it cardioid-ish
That seems to have reduced the dipole peak and also made the in-room sound more natural.

In the final speaker I'll probably use fiberglass batting to accomplish the same thing.

Also, this measurement illustrates the peak at 2500hz which I mentioned in my last post. I was able to eliminate that peak by replacing the series resistor with an L-Pad. Doing that 'normalizes' the impedance curve.
 
The midrange "hole" that could not be fixed by reversing polarity was due to the phase difference between drivers at the acoustical crossover, not the horn wall angle.
It could have been fixed with delay, or proper passive crossover implementation (very difficult to do without looking at phase).

The detail one can hear from from very narrow HF dispersion horns can be very engaging on axis, things in the mix normally lost in room reverberation can be heard clearly.
Your new horn appears to have a smooth on axis response.
It would be interesting to see it equalized flat on axis (no smoothing), then see the curves at 5 degree increments out to 40 degrees off axis.

P.S. -Just listened to the video, the HF response just dissapears when you can't see the throat.
That's one beamy horn!

Art

This isn't equalized, but here's what it looks like at zero, 22.5 and 45 degrees off axis:

monster-massive7.jpg


Basically we see that the beamwidth is narrowing at high frequency, just as expected.
Note that the vertical range in this measurement is compressed to 40dB to make the variations clearer. (The other measurements used a 50dB window.)

I'm kinda stoked that the dipole radiation doesn't seem to be creating and weird lobes off axis, with the possible exception of a dip at 1khz.
Obviously the response will change quite a bit once I add a baffle, and I plan to make it cardioid-ish also.
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.