Jean Michel on LeCleac'h horns

[soongsc]

Originally posted by Lynn Olson
Soongsc, more data is needed. It would be nice if the pictures weren't so tiny (being able to read the frequency would be very useful), and there were 3 dB plotlines on the contours to get an idea of the actual magnitude of the falloff with angle. JMMLC horns are not constant-directivity, so the degree of falloff is useful to know.

Without magnitudes, frequencies, knowledge of the horn profile, or diameter of the throat, the pictures don't convey much information. If the simulated driver is a physically unobtainable compression driver with zero throat length, that alone throws off the simulation.

If you have access to the original data, I'd take advantage of the DIYAudio 1000 x 1000 pixel limit and post them one by one.

Hi Lynn,

I'm really trying to obtain as consistent roll-off at as much frequencies as possible. The driver and horn dimensions I cannot reveal at this time, but if members read what I have typed in this thread and in the Geddess thread, it should be possible to grasp the idea and start playing around.

Frequencies I generally use in these 6 pack pics are something like 1K, 2.5K, 4K, 10K, 17K, 22K. The exact frequency sort of changes depending on what the software decides to split the frequencies at.

Originally posted by Lynn Olson
Well, I abused the images in Photoshop and Genuine Fractals, and this was the best I could do. At 3X enlargement, the horn appears to have zero throat diameter - although that would certainly give nice HF dispersion (at 22 kHz!), it would also seem a bit unrealistic.

No apparent throat depth either - is the compression driver assumed to be a point source? That might be valid for an Ionovac or the Acapella ionic driver, but for physical compression drivers with physical diaphragms ...

Well, it's the display limit of the software. It's necessary to zoom in on the throat in the software so get a look at the throat, and even that way there are artifacts in the view that are really not presentable. I don't like compression drivers. I bet if I really conducted a design review of those, none would be designed according to what I consider good engineering sense.

The concept and type of driver I have used had been explained in the Geddess thread. Hopefully for those whom are interested will be willing to read through my previous postes there.

Polar plots and impulse response of LeCleac'h horns had inspired me to look into getting some response between LeCleac'h horns and OS wave guides. If one studies in general how the dispersion and wave guide shape I have changed so far should give those interested a feeling what's going on. I think Jean Michel did notice it when he posted a comparison between one configuration I had posted and a LeCleac'h expansion of T=2.7 Fc=800. That is what prompted me to look into LeCleac'h expansion for parts of the wave guide I last posted here.

Things seem to be looking good, I think I will test build one wave guide after a few changes and see what real data shows.

[tinitus]

Nice looking CD at 80EUR

[soongsc]

After numerous simulations, it seems the last post has the smoothness of dispersion. If I try to increase compression using a smaller thoat, a little more HF dispersion is accomplished, but at some frequency there will be a sudden unsmoothness in a very narrow frequency band. So it seems the right amount of compression for a particular horn guide if important. I'm not sure whether there is any mathematical way of estimating the right match without numerous simulations or not. The LeCleach expansion of T=1.9 seemed to provide a good control at the mid frequencies before limited by the mouth size.

Below is a dispersion contour. Some small gliches still show. The drop around 20K is really great because this is where the driver breakup is.

[sqlkev]

hi everyone,
I'm highly interested in making a LeCleac'h horns in my next system.
What pointers could you give to a new guy like me in modeling and acquiring the data of the compression drivers I need to use for modeling?

excellent thread!

[Lynn Olson]

Soongsc, I'd also have a close look at the power-factor curves as well. There seems to be a sharp tradeoff between smooth loading and constant-directivity - I'm hardly an expert, but from the data I've seen, you don't get both at once - you have to choose. Even the modest T = 1.34 ratio has a noticeable shift in power-factor, which has to become part of the overall decision where you're going to put the highpass crossover. The power-factor ripples that appear in the conical horn, for example, do show up in the impedance curves of the physical device.

The other thing that has a big effect is the throat depth of the compression driver - if you're using one, that is. If you're using a dome tweeter or an ionic driver, I have no idea how that would affect the wavefront going into the BEM simulation - it would be different than the wavefront coming from a phase plug. If you're using a half-inch or 3/4" dome tweeter, the 22 kHz BEMs would start to make sense.

Note to moderators: I've made a request that the JMMLC content of the "Geddes OS" thread be moved over here, where it will have more context and extend the present discussion.

[soongsc]

Hi Lynn, In one run I accidentaly made the diaphram too small, and the results were terrible. So I think smaller diaphrams really need everything to be adjusted accordingly. The sim power curve pretty much follows the SPL curve in simulation. I don't think the BEm sim is going to be accurate, but after I build one and compare it against the sim, we can see how the results compare. Constant directivity is not first priority, but to have some directivity control to allow as common directivity pattern over a frequency range is the current design objective. I do not know what other LeCleac'h horns look like at 20KHz, but what I've seen posted in the Geddes thread way back is not what I'm after either. The whole idea started to evolve after I tested a large elliptical wave guide in the Geddes thread. Some data had been posted there concerning actual measurements on that, and a smaller wave guide that I did. One thing that I find in the audio world is that most people are to open to new ideas. Most just take a position and don't budge. Only recently I've been able to find a few mom & pop shops that will work with me to explore in a way that satisfies my needs. The bad economy seemed to have helped in this respect.

If I recall correctly, horns initially were used to help amplify sound. My interest in using a horn/guide is to provide the simplests wave front as possible. This all started when I was working on full range drivers, and realized the sound radiated from the diaphragm was really complicated. This resulted in measurements that had a good SPL, but much of the mids and highes just was not good enough when I listened to them. Then I had the opportunity to listen to some well known horn speakers and could hear the quality that I was missing even though they were not perfect. This made me realize I just had to look into this deeper. Up to now, the results seem rewarding.

Well' it's time to build. I'll post some measurements when available.

Jean-Michel seems to think the lip curl back is very important. I wonder whether anyone done testing showing the difference between integrating the lip into a baffle versus a fully curled back lip. I'd also like to hear about audible differences.

Lynn,
Do you have any data you can show that explains the shift in power factor that you mentioned?

[David McBean]

Originally posted by soongsc
The throat section compresses and expands, and consists of multiple types of curves, part of which Le Cléac'h horn type expansion might be used, it seems not possible to do in Hornresp.

Hi soongsc,

Sorry, I misunderstood your requirement :-). I had assumed that you simply wanted to match the throat tangent angle of a conventional axisymmetric Le Cléac'h horn to the exit angle of a given compression driver.

Kind regards,

David

[Lynn Olson]

Originally posted by soongsc

Lynn, do you have any data you can show that explains the shift in power factor that you mentioned?

All of the BEM's and Power Factor data is from Bjorn Kolbrek, and was used extensively in the development of the AH425. Here's an example that was made in May of 2008, showing 440 and 420 Hz horns with a T ratio of 0.707 and the Altec 288B compression driver. As you can see, they are similar, but not identical.

I should mention I was never trying to get maximum bandwidth from the horn system, but low IM distortion, maximum headroom, and freedom from internal diffraction, reflection and multipath. This points away from commercial PA or theater horns with nonuniformities (kinks) in the expansion, and made me wary of sharp boundaries at either the driver/throat interface or between the mouth and free air. These boundaries are troublesome enough in direct-radiators, and seem to be far worse in horns.

Since I was confident of the performance of the direct-radiator woofer (which is in piston band at 800 Hz) and the supertweeter, the region of greatest interest was between the LF limit of the driver and diaphragm breakup/phase plug cancellations at the top of the range. My own experience with direct-radiator tweeters and horn systems is that trying to "stretch" the LF bandwidth is a bad idea - distortion skyrockets, and listening fatigue goes way up. That's why the Ariel uses a Scan-Speak 9000 tweeter crossed over at an extremely conservative 3.8 kHz, for example. I just don't care for the sound when it is crossed over any lower. Same with horns. They sound good when used conservatively, not so good when trying to "stretch" bandwidth.

A flat power factor means the power is going where it belongs - out the horn. True, it says nothing about reflections, diffraction, and HOM's, but it does show if the diaphragm is seeing a resistive and "stiff" air-load. This keeps distortion down, one of my primary goals in the critical 800 Hz to 3 kHz region, where distortion reaches peak audibility.

It was JMMLC in private communication with Bjorn, Martin, and myself that mentioned that group-delay problems are indicated by deviations in the PF curves. This was news to me; it hadn't occurred to me that horns could have sharp GD deviations at the bottom of the passband. Since a complex GD deviation is nearly impossible to correct in a passive crossover, that had my full attention.

As a result, Bjorn, Martin, and I tried various iterations, balancing overall size, smoothness of PF and frequency response at a range of distances and emission angles, over a period of several months before settling on the AH425. The latest spreadsheet from JMMLC incorporating information about the internal construction of the compression driver also assisted in the modelling program. I thank Bjorn for providing detailed information on the internal throat of the 288 - that's not available from any Altec website.

Since it looks like one of your goals is reasonably even coverage over a very wide bandwidth, your horn must necessarily be different, and the driver must be smaller in order to attain good dispersion above 10 kHz.

[Lynn Olson]

Here's one of the reasons we didn't build a T = 1.34 version; this shows frequency response at emission angles from zero to 90 degrees off-axis. It looked fine in the BEMs, nearly indistinguishable from the T = 0.707 version. But the FR would be non-trivial to equalize.

The models presume theoretically perfect drivers; real drivers are worse, frequently much worse. Combine the two sets of responses, and you can start to see why 1/3 to 1/6 octave response smoothing is so common in the high-efficiency world.

[Lynn Olson]

I can't find the AH425 FR vs emission angle, but here's the response vs distance. Easier to equalize - a notch filter tuned to about 550~600 Hz, and possible shelf filtering, depending on listening angle. I plan to aim these at a point two to three feet in front of the listener, following classical BBC aiming criteria.

Anyway, I'd like to emphasize the BEMs don't tell the whole story. Horns are very complex devices, and alter the loading, emission, and time response of the driver in complex ways. Looking at the datasets, I can see why JMMLC has made the choices he has - large format mids, small format highs, and choosing the right T ratio for the application.