I don't think Burst Decay measurements should be gated, at least not a short gate. That will introduce its own artifacts. Better would be to build a small 'anechoic chamber' around the mic and the horn with pillows, foam, fiberglass, etc.
Or just use Jean Michel's settings for the CSD. I tried them and they worked well for me except I preferred it with the contour box unchecked. Move the cursor (no marker) on the impulse to get it to start where you want. Once you get JM's settings working well, change them one at a time to see what happens. If you change the settings, you'll probably need to move the cursor to catch the rise.
Or just use Jean Michel's settings for the CSD. I tried them and they worked well for me except I preferred it with the contour box unchecked. Move the cursor (no marker) on the impulse to get it to start where you want. Once you get JM's settings working well, change them one at a time to see what happens. If you change the settings, you'll probably need to move the cursor to catch the rise.
Attachments
Hello,
I calculated wavelets for the impulse responses. These are 2 cycle transforms, to maximise temporal resolution.
511 + 802:
511 + 806:
811 + 802:
811 + 806:
You are right, 802 driver is producing more garbage than 806
Combination 811 + 802 does not look very good, what do you think?
- Elias
I calculated wavelets for the impulse responses. These are 2 cycle transforms, to maximise temporal resolution.
511 + 802:
An externally hosted image should be here but it was not working when we last tested it.
511 + 806:
An externally hosted image should be here but it was not working when we last tested it.
811 + 802:
An externally hosted image should be here but it was not working when we last tested it.
811 + 806:
An externally hosted image should be here but it was not working when we last tested it.
You are right, 802 driver is producing more garbage than 806
Combination 811 + 802 does not look very good, what do you think?- Elias
And here's the same as above but now normalised to 0dB at every frequency, to eliminate the differences in frequency responses and thus for better comparison.
511 + 802:
511 + 806:
811 + 802:
811 + 806:
Everyone can make their own conclusions But I think there is a heck lot of reflections/diffractions seen in these plots.
See how the room reflections carry all the horn artefacts in it (of course) and they get 'amplified' in perceptive sense! Maybe this is leading us somewhere to the origin of Horn Honk??
- Elias
511 + 802:
An externally hosted image should be here but it was not working when we last tested it.
511 + 806:
An externally hosted image should be here but it was not working when we last tested it.
811 + 802:
An externally hosted image should be here but it was not working when we last tested it.
811 + 806:
An externally hosted image should be here but it was not working when we last tested it.
Everyone can make their own conclusions
But I think there is a heck lot of reflections/diffractions seen in these plots.See how the room reflections carry all the horn artefacts in it (of course) and they get 'amplified' in perceptive sense! Maybe this is leading us somewhere to the origin of Horn Honk??
- Elias
Sure – but this was an easy example – but what about the measurement of the 60deg included conical without any treatment - I've posted *.pir file ?
Here I really would liked to have some order put into – no ?
Michael
Hello Michael,
Your Honker Horn looks interesting I have to try to analyse those impulse responses too.
What comes to ARTA, yes the time axis is in periods, and I believe the reason is, as stated also in somewhere in ARTA manual, that the intended purpose for such a time axis presentation is to make more 'perceptionally comparable' results for resonances. I.e. it's not suited for reflection/diffraction analysis as it is now.
Should be no problem for Mateljan to convert the time axis. If he calculates the wavelet transform 'conventionally' that is with mathematical equivalency of convolution, the time axis automatically yields linear time. From this it's easy to do the axis transform to periods. If, however, he does some 'tricks' in wavelet transform to achieve directly the period time axis, then going back to linear time may not be possible easily.
Also the normalisation to 0dB at every frequency is straight forward. At least in Octave it takes only one line of code
- Elias
Your Honker Horn looks interesting
I have to try to analyse those impulse responses too.What comes to ARTA, yes the time axis is in periods, and I believe the reason is, as stated also in somewhere in ARTA manual, that the intended purpose for such a time axis presentation is to make more 'perceptionally comparable' results for resonances. I.e. it's not suited for reflection/diffraction analysis as it is now.
Should be no problem for Mateljan to convert the time axis. If he calculates the wavelet transform 'conventionally' that is with mathematical equivalency of convolution, the time axis automatically yields linear time. From this it's easy to do the axis transform to periods. If, however, he does some 'tricks' in wavelet transform to achieve directly the period time axis, then going back to linear time may not be possible easily.
Also the normalisation to 0dB at every frequency is straight forward. At least in Octave it takes only one line of code
- Elias
Also the normalisation to 0dB at every frequency is straight forward. At least in Octave it takes only one line of code
Great work !
Last edited:
Your Honker Horn looks interesting
LOL
I'll probably will have to repeat measurement with a foam plug ?
Michael
Hello,
In CSD the 'ringing' continues as long as the last reflection dies out, because in CSD there is no temporal reolution finer than the FFT window width. What is seen in the CSD plot, it may not be a real resonance at all, but appears as such because of limited temporal resolution of the method.
I think CSD is a good method to mask the data from the observer Well, to be fair sometimes it can give some information, too.
- Elias
In CSD the 'ringing' continues as long as the last reflection dies out, because in CSD there is no temporal reolution finer than the FFT window width. What is seen in the CSD plot, it may not be a real resonance at all, but appears as such because of limited temporal resolution of the method.
I think CSD is a good method to mask the data from the observer
Well, to be fair sometimes it can give some information, too.- Elias
Jack Bouska noted that it's informative to plot on a linear frequency scale to see the throat-mouth reflections. I don't know if this shows that or not but there does seem to be somewhat even spacing going on starting with the strong one at 500. This is ungated (cursor set but not the marker.)
Attachments
Ok, yes – thanks for reminder – looks like artwork inspired by NY skyline that way - who would have thought that there's so much buried beauty in this honk thing – what you think ?
Michael
Last edited:
Thanks Elias for the really great plots! That shows a lot. Yes, the poor 802 with the Symbiotic fram looks bad in your plots, too. No matter how you measure or listen to it, it's bad. =(
In all the plots you did we can see a hot spot at about 5.5mS. I can easily see this in the impulse file. My guess is that it is a room reflection (side wall), tho it comes sooner than I would expect.
There is a smaller reflection at about 2mS on all my measurements too. It could be the mouth reflection, tho it does not really measure out for a round trip from mouth to driver and back. Close, but not exact.
So I will have to play with foam and other things as Michael suggests to see if those kinks in the impulse can be identified. Will be nice to know what they are.
************
So.... from those plots, do we see Honk? I can't identify it (yet). The driver makes more difference than the horn. Too bad I don't have my paper tractrix horns any more. They might tell another story.
Yes, the poor 802 with the Symbiotic fram looks bad in your plots, too. No matter how you measure or listen to it, it's bad. =(In all the plots you did we can see a hot spot at about 5.5mS. I can easily see this in the impulse file. My guess is that it is a room reflection (side wall), tho it comes sooner than I would expect.
There is a smaller reflection at about 2mS on all my measurements too. It could be the mouth reflection, tho it does not really measure out for a round trip from mouth to driver and back. Close, but not exact.
So I will have to play with foam and other things as Michael suggests to see if those kinks in the impulse can be identified. Will be nice to know what they are.
************
So.... from those plots, do we see Honk? I can't identify it (yet). The driver makes more difference than the horn. Too bad I don't have my paper tractrix horns any more. They might tell another story.
Here follows some more "beauty".
2 cycle wavelet responses normalised to 0dB at every freq.
Michael's 60deg horn:
120deg horn:
Well..
And here with longer time axis up to 10ms. See the room reflections hit after 4ms, carrying from every directions all that beautiful honk into your ears
60deg horn:
120deg horn:
One can also see from the room reflection pattern the difference in the directivity of the two horns.
- Elias
2 cycle wavelet responses normalised to 0dB at every freq.
Michael's 60deg horn:
An externally hosted image should be here but it was not working when we last tested it.
120deg horn:
An externally hosted image should be here but it was not working when we last tested it.
Well..
And here with longer time axis up to 10ms. See the room reflections hit after 4ms, carrying from every directions all that beautiful honk into your ears
60deg horn:
An externally hosted image should be here but it was not working when we last tested it.
120deg horn:
An externally hosted image should be here but it was not working when we last tested it.
One can also see from the room reflection pattern the difference in the directivity of the two horns.
- Elias
Hello,
Without seeing the measurement environment, I would say everything before 3ms comes from the horn+cabinet and after 3ms is room reflections convoluted with the horn+cabinet response.
See here: At 3ms there is a presumable a reflection from a small object because there is no low frequencies present (low frequencies bypass the object by 'bending' over it). Maybe a microphone stand?
By the way, there seems to be some pre-ringing artefacts around 3kHz in both of 802 measurements. What that could be?
- Elias
Without seeing the measurement environment, I would say everything before 3ms comes from the horn+cabinet and after 3ms is room reflections convoluted with the horn+cabinet response.
See here: At 3ms there is a presumable a reflection from a small object because there is no low frequencies present (low frequencies bypass the object by 'bending' over it). Maybe a microphone stand?
An externally hosted image should be here but it was not working when we last tested it.
By the way, there seems to be some pre-ringing artefacts around 3kHz in both of 802 measurements. What that could be?
- Elias
Elias said:
- Yes, the "usefullness" of CSD was hotly challenged here about a year ago by a subsequently banned member ( I think it occured within Lynn O(s) tome of a thread ) .
- The arguements against believing CSDs compromised resolution ( leading to no misleading conclusions / excepting pretty pictures ) were compelling enough .
- Perhaps we should stick to using GD / Impedance / Burst Decay Plots & of course Octave to show the way ?
<> cheers
Last edited:
I confirm, but if you listen to gated sinusoidal it rings there when the signal stops.
Historically, I believe much more is made of damping the horn itself than apppears to be warranted in its measured performance, though it's clearly "excitable" at high SPL....
Last edited:
Panomaniac said:
- Thirty years plus of SR work has left me with an indelible impression of what constitutes "Horn Honk" . For me, it's a frequency dependant, dynamic disconnection of signal, that no longer faithfully tracks the input signal. Like a solar flare, after it reaches critical resonance it just floats out there on its own , in all its' splendor . It generally takes some real input level to really get it going .
- So far all of the plots I've seen here show various forms of coloration / but no full bore honk . I suspect these tests are much too polite in level ( though I too don't want to crank out 110-120 db levels in my residence ) .
<< EarlK
ps : Hi Zilch !
Hello,
Do you have the impulse response of that data?
I'm almost most certain that everytime time passes a room reflection, the CSD drops a step
- Elias
Do you have the impulse response of that data?
I'm almost most certain that everytime time passes a room reflection, the CSD drops a step
- Elias
Hello Earl,
If horn honk is level dependent, it cannot be seen in the impulse response. IR only describes a linear system i.e. no level dependence.
Maybe it's distortion that you heard, and relate it to the honk?
- Elias
If horn honk is level dependent, it cannot be seen in the impulse response. IR only describes a linear system i.e. no level dependence.
Maybe it's distortion that you heard, and relate it to the honk?
- Elias
- Thirty years plus of SR work has left me with an indelible impression of what constitutes "Horn Honk" . For me, it's a frequency dependant, dynamic disconnection of signal, that no longer faithfully tracks the input signal. Like a solar flare, after it reaches critical resonance it just floats out there on its own , in all its' splendor . It generally takes some real input level to really get it going .
- So far all of the plots I've seen here show various forms of coloration / but no full bore honk . I suspect these tests are much too polite in level ( though I too don't want to crank out 110-120 db levels in my residence ) .
<< EarlK
ps : Hi Zilch !
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.