Hi @mabat
Kindest regards,
M
Could it be made square to prevent a 'pattern flip" or does the equal size of the horizontal and vertical dimensions cause a problem and, consequently a rectangle is preferred? I vaguely remember that Tom Danley had some problems with the early versions of his (square) Unities.
Kindest regards,
M
Pattern flip, to describe it in a simplified way, results from the 'too short' length of the vertical walls for the asymmetric waveguide. Solutions:
For a baffle-mouted device with a conventional baffle (i.e. no laeral cutout), the only option to avoud this is nest the wider horizontal pattern via a loading slot into the narrower vertical.
For a free-standing device, a non-beaming, ath-fied version of the Iwata-horn with a loading extension is the optimum solution in my opinion.
Making use of the parameter b in the R-OSSE formula. I tried to get there but seemingly could not direct ath to achieve the result that I want. That would include controlling the shape to look squareish from the front, and the termination not warping weirdly. And I never learned how to add an extension for loading. Maybe sometime.
For a baffle-mouted device with a conventional baffle (i.e. no laeral cutout), the only option to avoud this is nest the wider horizontal pattern via a loading slot into the narrower vertical.
For a free-standing device, a non-beaming, ath-fied version of the Iwata-horn with a loading extension is the optimum solution in my opinion.
Making use of the parameter b in the R-OSSE formula. I tried to get there but seemingly could not direct ath to achieve the result that I want. That would include controlling the shape to look squareish from the front, and the termination not warping weirdly. And I never learned how to add an extension for loading. Maybe sometime.
Dito, I don't seem to get notices anymore either.
It has been a while ago, I understand. Because of the outcome of your simulation that contradicted with what I heard, it motivated me to verify your model of my horn with real measurements.
I was not satisfied with optimization of the mid-high Le'Cleach horn and decided to do it again, this time I placed a table in the middle of the room, with the horn on top, to be able to use windowing analysis. The ceiling and floor were closest to the mic and speaker, so a window of 2ms start and 4ms end is used for the analysis. The mic is placed one meter from the throat, so 57cm from the mouth. I used some manual peqs to correct the curve (no REW EQ used). Pity that my horn is not hifi... still good to scare something 😉
...
As you can see, it is not so bad -in my opinion. Your prediction is pretty similar, except for the degree of "beaminess". Yes it is a bit beamy but not to the extent that you can say it is unusable. I choose for a more beamy horn, where others choose to use absorption at the first reflection points with wide dispersion horns. Also, I do not need "DI compensation" as such, just a few PEQ's here and there to flatten the curve.
You may view it as just another perspective of dealing with room acoustics and imaging -not better, not worse. It just fits me to take another path then the crowd and find out what it brings me. Up to now I am satisfied until I find something better -never stop learning, right? And I would like to remind you that many respectable horn builders also have taken this path, like Azure Horns or Bert Doppenberg, to name a few. And late Mr. Le'Cleach was an authority in the field himself. I do not claim it is the best design and that it cannot be improved, but it works for me and I share my experiences in the diy community. I used to have radial tractrix horns before, but they had their own problems, like pattern flip.
So... when you say it is not the driver, then what causes the anomaly in my measurements versus your model? Maybe your model of a Le'Cleach horn is less accurate, with all due respect?
Your SPL plot is 93 dB, mabat's is 40 dB (top right). The adjusted plot would reveal more than twice the beaminess. There is no anomaly.
Beyond that, what's the matter discussing this: the Le'Cleach is a very beamy horn and if you enjoy the presentation, that's good.
Are these measurements gated to exclude all reflections? If not, the 90° measurement is likely showing something closer to the power response which more-or-less agrees with Marcel's simulation (around -10dB from 1kHz to 10kHz).
That is not true either, he uses 5 degrees steps, I use 9 degree steps (90/10 steps). If you adjust for that, and counting dB's you see the difference.
The scale remains the scale, no matter the resolution.
What is this discussion for?
Doesn't look like it to me... Any chance you could show the impulse response along with the gate used?
If you didn't use a loopback or (separate) acoustic timing reference, REW may have locked on to a reflection rather than the direct sound. This tends to happen with high DI devices far off axis because reflections can be significantly louder than the direct sound.
I think you are right about the far off axis reflections. I found the measurements back and applied a window which results in the following picture. Note that I left out the 90 degrees and 80 degrees measurement, because indeed those contained rubbish. But up to 70 degrees it seems valid to me. Also, it is in 10 degree steps, not 9 degrees. What I try to find out is if mabats model above exaggerates the beaminess of my horn or not. It does not look very nice around 10kHz, but that is also more or less visible in the model, albeit at a lower frequency.
If you look at f.e. 20 degrees in the model (4 lines down) it shows way more attenuation in dB than below (the purple line). To be clear, if I make a mistake, I am man enough to admit. I am not very experienced with rew. That is why I post, to find out if mabat statement is true that it is an unusable horn when looking at his data (could be, yes). I am for now not willing to accept that statement and try to prove that in this case, his simulation is not correct.
I might be wrong, but I think that mabat's graph is in 10 degree steps too. How do you come to the conclusion of 5° steps?
In that case the fit is quite good. I made an overlay with matching scale.
The polar marked in red is 60° I suppose. It matches the 60° polar from your measurement (green).
In that case the fit is quite good. I made an overlay with matching scale.
The polar marked in red is 60° I suppose. It matches the 60° polar from your measurement (green).
Thank you for your effort! Because to the left of the graph in vertical text is stated 5 degrees normalized.
I see. Makes sense and inspecting the polar map more closely you seem right.
Weird, the ABEC simulation would be off by a factor of 2 with the angles. Something is not right.
EDIT: Your measurement was done pretty close, probably that is a factor and what is your reference axis?
Weird, the ABEC simulation would be off by a factor of 2 with the angles. Something is not right.
EDIT: Your measurement was done pretty close, probably that is a factor and what is your reference axis?
Last edited:
Exactly. I've seen people making this mistake a lot. If done properly, the measurements shall match the simulations quite well, that has been already verified so many times that it's hard to believe that this particular device should be an exception. The gated data are already much closer to the prediction.
Pretty close yes, maybe 50-75cm from the mouth as the measurements where done in the living room. The reference turning axis was the Z-axis in line with the front of the horn. Also the horn turns around the horn mouth vertical axis in the horizontal plane.
What about the 5 degrees normalisation in the graph? The gated data is not much closer to the prediction when you take that into account. Or do I need a new set of glasses..
The new graph looks good to me. I believe that the simulation output shows 10° steps, normalized to 5°. Using your 70° (cyan) measurement, there's about 18dB or so drop from 1kHz to 10kHz (following the general trend; i.e. ignoring the peak around 10kHz). The simulation shows about 17dB drop over the same frequency range in what should be the 70° curve. Add in the extra 1dB or so from the 5° normalization and it's a pretty close match.
Measuring close will change the results a bit, but I'm not certain what the effect would be in this case.
What is "unusable" is subjective.
Last edited:
Looking forward to rectangular waveguide for dfm2535.
There are some rectangular waveguides that were available few years ago (b52) and some that are available today (xt1018) but they all have directivity errors in their passband.
The steps are 5°, but for these small mic distances the actual mic position is quite dependent on the axis of rotation. My 20° can easily be Horneydude's 40°, as he obviously rotated the mic around the mouth. In the sims the origin is at the throat by default (which I believe was also the case here) with the mic farther away.
- I could measure a horn 10 cm from the mouth, rotated around the mouth and it would be virtually perfect from "0 to 90°". But that's not the goal, is it.
Last edited: