I just finished my experiments with all-pass filtering using in ear monitors. This experience and my previous knowledge about filters and ringing are helping to grasp what @gedlee painstakingly told myself and maybe others about frequency response. Its one of those things that is very simple, so for no good reason it doesn't stick. It reminds me of my job at work, where things are pretty simple yet we complicate them for no good reason.
A smooth frequency response is the most important thing because whether an electro acoustical or mechanical acoustical filter, they operate generally the same. Smoothness might be described with Q once the relation to peak and low/high pass filters are seen. Relate the sensitivity response as another filter. Large and/or sharp changes in response create higher levels of group delay and ringing.
Source Signal, EQ, the Driver, Baffle, Waveguides, etc all apart of a system. Like the F of woofer changes once it joins an enclosure, a filter up stream, as in the response of the source becomes apart of the equalization down stream of the source, including mechanical acoustic impacts of driver, waveguide, baffle, etc. The resulting response is what matters, with a Premise of accurate reproduction of the source. The room is at the end of the chain, and I guess you could argue listening position and eventually your ears and the psychoanalysis, if wanting to go to the nth degree. Speaking nominally, the response had from the loudspeaker system plus room needs to be smooth and devoid of sharp changes.
Experimenting with the worst parts of my spectrum, I used an outside measurement, which are about 15ms at 167hz and about -10db from the fundamental level. I used an all pass centered on that frequency with a Q of 8.81 to achieve the same group delay, according to Vituixcad. I used a high pass to experiment with 167hz being -10db. I could hear the difference. The attack time wasn't what was making a difference that I could tell, it was the ringing. It was a similar event to reverb that had a fundamental resonance centered on the particular frequency. I am very interested to try this with the horn. With the high pass, it was just faintly discernable in vocals of J Cole, rapping. bypassed it the vocal as every so slightly dryer in the area of focus. Then the song has a transition to another part where there is a very punchy kick drum taking lead... this is more like an impulse response, and I can hear the Hf of kick drum ringing clearly how ever subtle the change is I can hear it distinctly.
If you look at the FR, it is not, smooth, nor a final voicing. I just set this up quickly to test headroom at that moment. A better voicing should create a better result and there's also the question of accuracy of my testing to what will happen with the loudspeaker. I have been doing critical listening for some time now and these changes very subtle. If I am going for utmost accuracy I still must consider addressing these issues.
The GD at the trouble area is 10ms in raw response, and with equalizer apo set to match that GD, everything is fine. The change is barely noticeable using tone burst of rectangular windowing, as in virtually non existent. In the measurement below is a minimum phase 48db/octave LF high pass at about 180hz. I think that is just more reason to think that I can do better with better voicing, but I won't be certain what needs to be done until I have the horn up and running again.
All and all a very informing experiment.
A smooth frequency response is the most important thing because whether an electro acoustical or mechanical acoustical filter, they operate generally the same. Smoothness might be described with Q once the relation to peak and low/high pass filters are seen. Relate the sensitivity response as another filter. Large and/or sharp changes in response create higher levels of group delay and ringing.
Source Signal, EQ, the Driver, Baffle, Waveguides, etc all apart of a system. Like the F of woofer changes once it joins an enclosure, a filter up stream, as in the response of the source becomes apart of the equalization down stream of the source, including mechanical acoustic impacts of driver, waveguide, baffle, etc. The resulting response is what matters, with a Premise of accurate reproduction of the source. The room is at the end of the chain, and I guess you could argue listening position and eventually your ears and the psychoanalysis, if wanting to go to the nth degree. Speaking nominally, the response had from the loudspeaker system plus room needs to be smooth and devoid of sharp changes.
Experimenting with the worst parts of my spectrum, I used an outside measurement, which are about 15ms at 167hz and about -10db from the fundamental level. I used an all pass centered on that frequency with a Q of 8.81 to achieve the same group delay, according to Vituixcad. I used a high pass to experiment with 167hz being -10db. I could hear the difference. The attack time wasn't what was making a difference that I could tell, it was the ringing. It was a similar event to reverb that had a fundamental resonance centered on the particular frequency. I am very interested to try this with the horn. With the high pass, it was just faintly discernable in vocals of J Cole, rapping. bypassed it the vocal as every so slightly dryer in the area of focus. Then the song has a transition to another part where there is a very punchy kick drum taking lead... this is more like an impulse response, and I can hear the Hf of kick drum ringing clearly how ever subtle the change is I can hear it distinctly.
If you look at the FR, it is not, smooth, nor a final voicing. I just set this up quickly to test headroom at that moment. A better voicing should create a better result and there's also the question of accuracy of my testing to what will happen with the loudspeaker. I have been doing critical listening for some time now and these changes very subtle. If I am going for utmost accuracy I still must consider addressing these issues.
The GD at the trouble area is 10ms in raw response, and with equalizer apo set to match that GD, everything is fine. The change is barely noticeable using tone burst of rectangular windowing, as in virtually non existent. In the measurement below is a minimum phase 48db/octave LF high pass at about 180hz. I think that is just more reason to think that I can do better with better voicing, but I won't be certain what needs to be done until I have the horn up and running again.
All and all a very informing experiment.
Last edited:
Tonight I did some more modelling. In horn resp I can ballpark the drivers characteristics but not the horn response. Horns/waveguides do not change the nature of excursion. It looks like large horns do help to curve excursion below cutoff more so than a pure waveguide but that is neither here nor there in the face of a high pass used for crossing above the F of a waveguide. Because of this, I should be able to use horn resp to ballpark excursion. I loaded up the raw response into VituixCad, and created a voicing. then I loaded the filters into Hornresp, in particular the ones near the roll off of the horn, that mattered most for this exercise. From there I could adjust voltage in vituixcad to achieve certain spl, and then try it out in hornresp to see where it might take me, excursion wise. 1db increments the graph, looks like the XO point is at about ~212hz. This isn't much different than what I've achieved with the rough voicing so far, which is probably about 215hz -6db.
112.5db
107db
Respectively this is what would be potentially asked of each horn for 115db and 110db peaks. From what I've seen 15db headroom is about standard over average and 95db average was the max safe average for something like an hour. This may explain why I have not had an issue with thd as the excursion is about 0.6mm during transients listening at 95db average, which is loud enough for me. As a system, things will keep transient excursion near half xmax and that is desirable.
112.5db
107db
Respectively this is what would be potentially asked of each horn for 115db and 110db peaks. From what I've seen 15db headroom is about standard over average and 95db average was the max safe average for something like an hour. This may explain why I have not had an issue with thd as the excursion is about 0.6mm during transients listening at 95db average, which is loud enough for me. As a system, things will keep transient excursion near half xmax and that is desirable.