Al Klappenberger

Just to verify the validly of this idea, I grafted the data generated by the 1-inch horn cut down to 2 inch eliptrac program run with the run of the 2-inch design data for the example. I did some grafting to put the segment height / width columns side-by-side. The results were almost too good to be true! This looks like a way to make an expandable horn. The "extender" could be made with the large end emulating a driver to just bolt right between the 2-inch horn and a 1 inch driver. There has GOT to be something wrong with this, but I don't see what it could be as yet! Al k. cut1vs2.txt

To compare the shortened 1 inch horn to the proper 2 Inch horn, I did the example again for a 2 Inch. It looks similar. There is a slight change in cutoff frequency though. Al K.

Guys, I did a search for the elliptic tractrix horn and found those. I never assumed that I had invented anything new! I have expanded the "eliptrac" design program so that it will design in a round area at the throat such that a horn designed for a 2 inch driver could be extended to use a 1-inch without disturbing the tractrix expansion. Just adding a conical extension of arbitrary length to a 2-inch horn to jury rig a 1 inch driver is just not the thing to do! The program asks for the diameter you want to be round and it finds the right place to put it. In the example I used before, it's a snitch over 3 Inches back from the mouth. The round area is in yellow. I also added the ability to specify the length of the round area at the throat directly. I'm not sure if that's good for anything but it seemed logical to do that too while I was at it. Al K.

Trey, Always test filters of any sort loaded with their designed impedance. I load the tweeter channel with 8 Ohms and the mid-range with 13 Ohms to simulate the K55V. The woofer channel gets loaded with 6 Ohms in series with 1 MHy. That's a slight fudge because the driver doesn't function like a series inductor to the transfer function. It just simulates the complex load the filter sees. The computer analysis does the same thing. Al K.

Chris, I forgot to mention the level thing. The -6 dB or so on the squawker is about right. The tweeter level is a matter of what tweeter you have. As it turns out, Bob's CT125 is just right for the wide-open tweeter channel of my Universal network requiring no external attenuator. I think his tweeter is less sensitive because of it's smaller magnet. That was necessary because a larger one would not have fit directly in place of the K77 under the Khorn top cover. That might also be why Klipsch made the K77m magnet square. It had to have more sensitivity to make up for the loss in the crumby AA tweeter filter. Al K.

Chris, I think your asking if the K77 needs help from the K55 above 6KHz. It's kind of a moot point considering it gets very little! The K55V poops out above 6 Khz, The Klipsch stock networks depend on that to form a crossover. The K55M doesn't go quite as high as the K55V. I think the K77 does fine by itself except above about 16 KHz or so where it poops out. I doubt that matters to most of us though. I can't even hear 15 Khz any more let alone anything higher! There is definitely a different sound when the mid-range high range is actually limited by a filter rather then allowing it to roll off naturally. That was driven home to me when I did a side-by-side comparison of a Forte I with the stock network A/B switched with one having my network upgrade. If that difference is "help" I suppose is a matter of opinion. I consider it driver interference from multiple sources since they are by no means in time alignment! Al K.

I am no horn expert. As I understand it, any abrupt change in the internal flare causes reflections within the horn, Even the square corners where the flat top joins the curved sides in the Edgar type Tractrix horns a discontinuity that will cause reflections. I know from my experience at microwave frequency that any transmission line with discontinuities cause energy to reflect back to the source. If the wavelength to the discontinuity is large enough the reflected wave changes the impedance seen by the source. You might want to look up a device called a "Smith chart". It will also set up standing waves inside the line, or inside the horn in this case. The hope here is that a smooth transition from round to elliptical will minimize the reflections. Al K.

It looks like that cutoff will make a horn that is just too small. Here's a run of a horn with a mouth of 4 X 3 Inches. It's 1.6 Inch long and has an Fc of under 1300 Hz! It's made of 14 segments if 1/8 inch thick stuff. Al K.

Dave, I think maybe 2400 Hz for the tweeter horn. I say that because the tweeter filter on the Universal network is 20 dB down at that frequency. 3700 Hz is 10 dB down. I think that would be fine too, maybe even better. I just sent you version 1.3a of the program. It makes AutoCad files and a text file of all the segments not just the 18 or less that show to keep from scrolling off the screen. Al k.

cfelliot, OOPS! I misunderstood yet again! The Tractrix expansion will force the mouth edges to be perpendicular to you. What I was saying was that the rate at which the ellipse at the mouth, as you look straight into the horn, turns into a round hole for the driver at the throat. I have it making the transition gradually, in a linear fashion, as you move down the horn from Mouth to throat. It could make that transition at any rate. I would have no idea what would sound best! Al K.

OH! I think that taper could be just about anything. I did linear because it was easy to do. Can you think of any reason to change the oval to round taper rate? Also, explain where the angle you mentioned would be. Al K.

cfelliot, I'm not sure I follow your question, but the material thickness determines the "step". I think that means the "step" must be linear. That is, number of steps = horn length / material thickness. In fact, I just modified the program so that you can input step thickness rather then how many steps if you want to. The last sample run shows a .585 in step. It would need to be made of .50 in think stuff to be practical. That would require 11.7 steps. You would need to do 12 and shave a bit off the throat end to get it right. The internal "loop" will overrun "round" just slightly. Is that what you mean? Al k.

Dave, Here's the first practical draft of the program. Check your email. I'll look into the HPGL interface next. This might take a bit longer! Al K.

Dave, Here's guts of the program. I just need to check it over and add the variables. Right now it is fixed at 15 steps and tapers from elliptical to round in 15 linear steps. AL K.

Dave, I'm off a running. I got the program doing a Tractrix area taper down to the throat. The next thing will be to taper the ellipse of the mouth down to a round driver. I have been doing programing so I haven't read the latest comments yet. Dennis, I could easily add a conical taper section at the throat. Can you point me to any guidelines for how long to make it and to what diameter. Is it worth doing? Al K.

Dave, I think a few sample files would get me started. Maybe an example of whatever data is required to do one segment or even just a single elliptical cut. Maybe you could send me one by email. AutoCadd script files is not a problem either. Trachorn.exe makes one for the curved side. There's no hurry though. I need to get my head into gear to do the Tractrix design part first. Another thought is the conical segment at the throat that djk suggested in another thread to improve the HF dispersion. I should be able to work that in too pretty easily. Al k.

Tiger.., The program I am thinking about would bypass the AutoCadd stage entirely. My initial thinking is the program would ask for mouth width and height plus the round throat diameter, just as the Trachorn.exe program I wrote does, but would also ask for how may segments. It would then output the data on how to cut each segment directly to the CNC computer. I don't know how Dave's machine handles the data though. A simple 3 inch computer disk to transfer a file might be the way to go, at least initially. Al k.

Dave, I think I can write the design program very quickly. Doing the interface will be a bit more trouble considering your CNC machine and I are 2000 miles apart! For starters, just to verify the validity of the elliptical Tractrix, the data could be input to the machine manually for a while. Al K.

Dave, I have had some experience with interfacing engineering programs to external instruments like the CNC machine computer. I have done several using HPIB (IEEE 488) buss. I have that sort of setup doing computer testing of the networks I make, I envision a horn design program that will dump the cutting data directly into the CNC. I also think that the next iteration might be a Tractrix with an elliptical mouth that will fit in the Fastrac K Khorn frame for a mid-range. I know I can write the program to design the horn. How about that for a fun project? Al K.

crd, "you certainly are an overly sensitive person aren't you." About certain issues. I certainly am! Al k.

Dave, VERY INTERESTING! That horn is a 2-inch but it's square mounting flange suggests a possible version that would fit into the Khorn frame along with the Fastrack K or K2 but would be a 1-inch and could be used with Selenium or B&C tweeter drivers as an option to the K77 or other popular tweeters. AL K.

Earlier djk said this about the high dispersion: "I think the only reason you have any off-axis HF is that it radiates only from the small inner stub inside the driver, and is not seeing the side walls of the horn for a ways. " I believe now that he's right. I set the other driver I have on the bench with no horn and the mike suspended above it on axis, 30 and 45 deg off axis. The dispersion is still there! It's the driver itself, NOT the horn!

Dkalsi, The time to get the network done will be pretty quick once I get the parts from Solen. Unfortunately, the first set is already spoken for though. It's up in the air as to how much of an improvement it will be over a separate tweeter with ES networks. How something sounds is so subjective that only time will tell. Even the unexpected high dispersion thing is unresolved! I can only hope that the Linkwitz-Riley network concept will solve whatever problems exists with coaxial drivers like this. I know so little about horns that I can't make any rationalizations based on theory about it. I can't see that type of network being useful for anything else but sub-woofers considering how much effect just a slight difference in propagation time mess up the summation! Al K.

Dennis, I looked at that pdf. It's displays are a format I have never educated myself about. It's looks like it might be worth looking into. That Emilar horn does look interesting. It looks like a T35 sitting in it! With all the different types of horns that are made you would think there would be one that would stand out as overall best to everybody. The fact that so may exist says that there is just no "right" way to do it! Al K.

Dennis, I would like to see the data, but I'll bet it will be over my head. I really don't know much about horns. It's al_k@alkeng.com When I get this network done and the other one built up, I plan to do some listening tests this time, which I usually don't do. I don't trust my ears, but if there a glaring difference between this and what I am used to, I should hear it. There is yet another issue that bothers me. I wrote a little DOS program that takes the computed amplitude and phase data out of each of the channels of a crossover and sums them up. I have the option to invert one driver or add in a delay to one or the other. I found that the outputs add up flat as a pancake if the delay a zero, but add just 1/4 inch different propagation delay to on channel and it results in a 2 dB dip. More delay screws things up even more. This is why I have never attempted a Linkwitz-Riley network before. I don't know just how closely aligned the acoustic sources are in the DCX50. All I'm sure of that it's a lot closer than a separate tweeter and mid-range! Al k.