Horn Honk $$ WANTED $$

Status
Not open for further replies.
Hello Elias,

There is surely one thing that can be improved it is the crossover. (This is not a Le Cléac'h crossover as we can see by the noticeable delay around 2500Hz on the CSD here attached).

Best ergards from Paris, France

Jean-Michel Le Cléac'h


Hello,

Here's ZilchLab's impulse response file converted to WAV. I assumed sampling freq of 48kHz.

- Elias
 

Attachments

  • zilchlab.gif
    zilchlab.gif
    22.1 KB · Views: 236
Hi,

I don't understand your description what you would like to do? Import what from where to where? 😀

- Elias

...importing ARTA wav file into an Octave variable.

Basically ARTA wav file is a plain text file - BUT -
There are headers in ARTA wav file that have to be removed
Those headers in ARTA wav file spoil the simple "load" command of Octave
I haven't found appropriate file handling / manipulation commands in Octave to manage those data (st)ripping from ARTA wav file
😉

Michael
 
Last edited:
?? WAV file is a binary file, not a text.

Maybe post one file and let's see 😀

- Elias

...importing ARTA wav file into an Octave variable.

Basically ARTA wav file is a plain text file - BUT -
There are headers in ARTA wav file that have to be removed
Those headers in ARTA wav file spoil the simple "load" command of Octave
I haven't found appropriate file handling / manipulation commands in Octave to manage those data (st)ripping from ARTA wav file
😉

Michael
 
Hello,

Are you going to tell us what is your device? 🙂

- Elias
These are wave guides that I've been trying to design. Basically exploring the possibility of optimizing directivity, loading, and stored energy. Most data have been posted in the Geddess thread because that's where I was first inspired to do so. It's probably been two years now since the first elliptical wave guide that was build. The wave guides are driven by a concaved dome driver. Incidentally, the curves are getting a bit closer to the OS wave guide curve, but still not quite the OS as Earl originally posted. The "back 1" is the previous version, and the "back 2" is the design before back one. The one you already did a "noodle" plot on is the latest one. All impulse data are from simulation using AxiDriver and VACs viewer for converting and view the results.
 
Last edited:
?? WAV file is a binary file, not a text.

Maybe post one file and let's see 😀

- Elias

Sure, sorry !
Octave "wavread" command already works fine for importing 16bit and 32bit *.wav files !

#########

Its the ATRA *.txt files I have troubles with...

Here is an example:

http://www.kinotechnik.edis.at/pages/diyaudio/GD_HighPass/ARTA_try.txt

Basically we only would have to remove the first 4 header lines - but I didn't find the right commands in Octave to do so (I think Matlab has more extended file manipulation skills).
Sure - I could do *externally* with any editor - but I'd like to integrate that format of ARTA in my IR file loading script - so I want to do it in Octave directly...

"load ("-ascii", (ir_file_name));" isn't the way to go
🙁

Michael
 
Ummmm,

- Am I being unfare to Zilch by posting these pics derived from the following parameters ?

- I imported the OLA WAVE file into ARTA ( that Elias had transposed ) / then went for a HiRez look-see and got the following .

- ( 91 ms Gate / Full Window / 256K FFT )

- Am I looking at a whole lot of diffraction effects here ?


attachment.php
attachment.php




- Zilches' Original FR Look :
attachment.php



<> EarlK
 

Attachments

  • OLA_XT1086_GD3.PNG
    OLA_XT1086_GD3.PNG
    33 KB · Views: 442
  • OLA_XT1086_FR.PNG
    OLA_XT1086_FR.PNG
    32.3 KB · Views: 438
Hello Panomaniac,

I'll do a parallel with horns. I used to try different cut-off frequency for the high pass crossover for many Le Cléac'h horns having different acoustical cut-off ferquency. (The summation LP + HP of the used crossover has low phase distortion).

What I noticed is that the sound was improved when the cut-off frequency of the HP crossover was such that the corresponding group delay of the horn was lesser than 100mm.

So most probably the phase distorsion responsible of those 65 mm delay at 2500Hz will not be much noticeable but reducing that phase distortion will improve the sound... (at least for trained ears...).

See a comparison of the quasiwavelet graphs of a le Cléac'h crossover (left) and of a classic Butterworth 3rd order (right) at:

http://www.diyaudio.com/forums/atta...audio-measurements-what-how-wavelet_phase.gif

Best regards from Paris

Jean-Michel Le Cléac'h

I sure see the delay shift. But how bad is it really? Looks like a shift of about 2.5" to me, or 64mm. Could be better, but not that bad. Is that the right distance?
 
Ummmm,

- Am I being unfare to Zilch by posting these pics derived from the following parameters ?

- I imported the OLA WAVE file into ARTA ( that Elias had transposed ) / then went for a HiRez look-see and got the following .

- ( 91 ms Gate / Full Window / 256K FFT )

- Am I looking at a whole lot of diffraction effects here ?

...
<> EarlK
Room reflections?
 
Hello Earl,

here is attached what I consider a better group delay curve for Zilchlab IR. (I think you didn't use any marker before the first main reflection arrival.) But your's and mine lead to the same intrepretation.

We can see that above 3 kHz the group delay curve is very flat. (so it is "less" flat below 3kHz)

Best regards from Paris

Jean-Michel Le Cléac'h


 

Attachments

  • GD_Zilch.gif
    GD_Zilch.gif
    13.3 KB · Views: 106
Last edited:
Wouldn't it be much easier to remove header manually?

Yes indeed - but I am on packing a modular code, that is usable for anybody that wants to get into it.

One of the modules is the file import module - up to now it support imüport of wav file (16 bit and 32 bit) and plain text, ether one column or two column formatted.

So adding CLIO export file and ARATA txt export would be "nice to have"...

Normalization feature is ready and skeleton noodles and wavelet analysis are on the stove.
🙂

Michael
 
Hi,

How to transfer impulse response file from ARTA to OCTAVE



This Octave script read ARTA .txt file, skips the header, and reads the data into Octave matrix.

One plot is generated for your convenience.


### Read ARTA txt file into Octave matrix

filename = 'D:/temppi/ARTA_try.txt'

fid = fopen (filename);
fgetl (fid)
fgetl (fid)
fgetl (fid)
fgetl (fid)
data = zeros (2^18, 2);
i = 1;
while (! feof (fid) )
data (i++, 🙂 = str2num (fgetl (fid));
endwhile
fclose (fid);

data = data(1 : i - 1, 🙂;

figure (1)
plot (data🙂, 1), data🙂, 2));


- Elias
 
Last edited:
Hello,

You'll find here attached quasiwavelets graph (amplitude then hybrid amplitude x sign of phase) I obtained using my Matlab routine on Zilchlab IR.

(eventually compare with the quasiwavelets obtained for 2 crossovers I previously gave)

Best regards from Paris, France

Jean-Michel Le Cléac'h
 

Attachments

  • zilch_a.gif
    zilch_a.gif
    41.3 KB · Views: 195
  • zilch_b.gif
    zilch_b.gif
    34.1 KB · Views: 129
Last edited:
Here's ZilchLab's impulse response file converted to WAV. I assumed sampling freq of 48kHz.
Yes, correct.

I sure see the delay shift. But how bad is it really? Looks like a shift of about 2.5" to me, or 64mm. Could be better, but not that bad. Is that the right distance?
Yes, the crossover occurs acoustically at ~1.425 kHz. The HF is connected inverse polarity which, in combination with phase delay in the filters, compensates for the offset of the acoustic centers, the compression driver being approximately that distance behind the woofer. I believe this measurement to have been taken on the forward axis as located vertically at the maximal-depth reverse-polarity notch.

The CLIO AES white paper illustrates a similar situation; the delay is required to bring the woofer and waveguide into phase alignment in the crossover region. Alternatively, the drivers could be run in the same polarity with the woofer delayed to better result, but as you suggest, this design is "not that bad." It was developed using Jeff Bagby's PCD (Passive Crossover Designer) beginning here:

Flex Your PCD Mettle: - Page 10 - Techtalk at Parts-Express.com

This iteration is detailed at Post #199; note the phase interference between 2.0 kHz and 3 kHz, and also the presence of a notch filter just above there, to flatten the response. The schematic is not yet posted, but will be soon.

SO, do we place 18-Sound XT1086 into the "Honker" cell here based upon these analyses? 😀
 
Last edited:
Hello Earl,

here is attached what I consider a better group delay curve for Zilchlab IR. (I think you didn't use any marker before the first main reflection arrival.) But your's and mine lead to the same intrepretation.

We can see that above 3 kHz the group delay curve is very flat. (so it is "less" flat below 3kHz)

Best regards from Paris

Jean-Michel Le Cléac'h
Here is the same thing plotted at 1/24 octave smoothing rather than 1/3 octave. The group delay curve is quite noisy but plotting it as excess group delay shows what is going on better IMO, removing group delay caused strictly by the frequency response.

attachment.php
 

Attachments

  • 1086-groupdelay.png
    1086-groupdelay.png
    8.5 KB · Views: 619
Status
Not open for further replies.