Dave
BTW, for decades the 4430's were my speaker of choice. but I came to dislike several features of them, most notably the "harshness" of the horn. And yet there were other things that they did extremely well.
All of my current designs derive directly from this speaker, but with improvments in the key areas that I deemed were a problem. The 4430's were a great speaker in their day. We have learned a lot since then (or some of us have anyways!).
BTW, for decades the 4430's were my speaker of choice. but I came to dislike several features of them, most notably the "harshness" of the horn. And yet there were other things that they did extremely well.
All of my current designs derive directly from this speaker, but with improvments in the key areas that I deemed were a problem. The 4430's were a great speaker in their day. We have learned a lot since then (or some of us have anyways!).
In my polar software where I find the polar modal expansions, I can show how it is possible to get precise angular data at virtually infinite angular resolution (up to about 10 kHz) from only about 12 angular measurements. They are not evenly spaced however.
But its easy to show that when one has enough data to get the model to converge, then adding more terms does not change the picture.
Very fine angular increments are unnecssary when the data is used properly.
This is an important statement. What do you consider 'typical' design and is it possible to quantify 'very reverberant' with something like reverberation time?
Trully pleased to hear that.
I do remember that there was a certain "texture" to the highs especially noticable on pink noise. I always assumed it was the regular periodic ripple in the bottom of the horns range. The short length of the exponential section and narrow diffraction slot area lead to fairly strong reflections that put bumps into the impedance curve. The passive network interacted with the impedance bumps and added a little more to the ripple. It wasn't real bad, maybe 2dB peak to peak. An active crossover would help or better yet (as I suggested on the Lansing Heritage site) a somewhat negative output impedance.
As you know the first generation JBL horns were totally optimized for polar curves and H and V beamwidth. Nothing surpases them in that regard (to this day) but they compromised frequency response somewhat to achieve it. Some horns were better than others and if you didn't go crazy about horizontal beamwidth to the highest frequency (very narrow diffraction slot) then you could have good response along with good polars.
HOMs? You would know better than me. In the end I thought their sound pretty well followed their frequency response.
David S.
With Markus's pink noise test I get the same thing no matter what: a large, vague center image when looking forward, eyes close; when turning my head I get more HF content; when leaning to one side, the image follows me(IOW goes toward the near speaker).
My speakers are 10' in front of me and 17 degrees from the center line.
Dan
Thanks Hum!
My speakers are 10' in front of me and 17 degrees from the center line.
Dan
Thanks Hum!
I did some tests on the pink noise samples Markus posted.
This time I used full range FE126En in a small 20*20cm closed box.
Generally I can say it is very hard for me to hear phantom images on pink noise.
My room is about 3.5 * 7 m and it is exactly symmetric shoebox. Bare reflective side walls. Heavy curtains at the front wall. Acoustic absorption at the back wall.
I tried several configurations:
The listening distances is varied from 1m to 5m and stereo base angle is varied from about 20 to 90 degrees to search for the best combination.
----
1) Normal stereo triangle, speakers aimed at the listener.
Unfiltered pink noise:
Low freqs are coming from somewhere in front but cannot say what is the exact location of the phantom. High freq hiss is coming from the exact locations of the two speakers. So the image is split in three, one for low freq and two for high freqs. Very bad.
Then I low pass 24dB/oct at 3kHz:
High freqs are not coming from the speakers anymore. Low freqs are coming from somewhere in front but hard to say from where.
Then low pass 24dB/oct at 6kHz and the high freqs are at the speakers again. So the limit is somewhere between 3 - 6 kHz to localise the speakers.
I even tried extremely close field 1 m stereo triangle with the same results, vague phantom at low freq and high freqs at the speakers.
----
2) Normal stereo triangle, speakers toed in up to 45 degrees.
Hard to say how this is better than the case 1) in terms of phantom image localisation, except high freq hiss is less due to off axis location. Still high freqs are in the speakers, low freq somewhere in front.
Low pass at 3kHz helps here too to remove the image from the speakers. The low freq phantom is not solid, but it is at least coming from the front
----
3) Stereolithic back-to-back
Unfiltered pink noise:
Low freq image is very good in the center. Quess what, the speaker is there too At Very high freqs the image is split in two located far behind the side walls in very wide angle, maybe 120++ degrees.
Then I low pass 24dB/oct at 6kHz:
Almost ok ! No side wall images anymore, and low freq image in the center. A good starting point !
----
4) Raised flooder at the ear height
Unfiltered pink noise:
The low freq image is a bit elevated, not much but just noticeable. At very high freqs image comes from the two side from the speaker locations.
I did not try the low pass here, but I'm sure it will remove the image from the sides as in other cases as well.
----
5) Beveridge placement at the ear height (this is equal to 90 degrees toe in and speakers at the side walls)
The contralateral reflection at very high freqs make the high freq image to appear almost on both side of me, about 160 degrees. Wierd.
Low freq image somewhere in front, but not clear.
----
6) And then, the ultimate stereo test: Stereo triangle with plugged one ear with hearing protector plug !
Quess what: I can hear high freqs coming from the speakers even with only one ear 
----
At first I did try the high passed pink noise samples as well, but no need to explain what happens: The image remains in the speakers
From all of these 5 configurations, the stereolithic back-to-back provided best low freq image in this mono test. Seems to be that none of them virtued in high freq (> 6kHz) phantom image generation.
Should I feel delighted about the result? Of course it is no surprise to me.
What can I do? How to improve things? But wait, I think I know how to improve this! And it works !
- Elias
This time I used full range FE126En in a small 20*20cm closed box.
Generally I can say it is very hard for me to hear phantom images on pink noise.
My room is about 3.5 * 7 m and it is exactly symmetric shoebox. Bare reflective side walls. Heavy curtains at the front wall. Acoustic absorption at the back wall.
I tried several configurations:
The listening distances is varied from 1m to 5m and stereo base angle is varied from about 20 to 90 degrees to search for the best combination.
----
1) Normal stereo triangle, speakers aimed at the listener.
Unfiltered pink noise:
Low freqs are coming from somewhere in front but cannot say what is the exact location of the phantom. High freq hiss is coming from the exact locations of the two speakers. So the image is split in three, one for low freq and two for high freqs. Very bad.
Then I low pass 24dB/oct at 3kHz:
High freqs are not coming from the speakers anymore. Low freqs are coming from somewhere in front but hard to say from where.
Then low pass 24dB/oct at 6kHz and the high freqs are at the speakers again. So the limit is somewhere between 3 - 6 kHz to localise the speakers.
I even tried extremely close field 1 m stereo triangle with the same results, vague phantom at low freq and high freqs at the speakers.
----
2) Normal stereo triangle, speakers toed in up to 45 degrees.
Hard to say how this is better than the case 1) in terms of phantom image localisation, except high freq hiss is less due to off axis location. Still high freqs are in the speakers, low freq somewhere in front.
Low pass at 3kHz helps here too to remove the image from the speakers. The low freq phantom is not solid, but it is at least coming from the front
----
3) Stereolithic back-to-back
Unfiltered pink noise:
Low freq image is very good in the center. Quess what, the speaker is there too
At Very high freqs the image is split in two located far behind the side walls in very wide angle, maybe 120++ degrees. Then I low pass 24dB/oct at 6kHz:
Almost ok !
No side wall images anymore, and low freq image in the center. A good starting point ! ----
4) Raised flooder at the ear height
Unfiltered pink noise:
The low freq image is a bit elevated, not much but just noticeable. At very high freqs image comes from the two side from the speaker locations.
I did not try the low pass here, but I'm sure it will remove the image from the sides as in other cases as well.
----
5) Beveridge placement at the ear height (this is equal to 90 degrees toe in and speakers at the side walls)
The contralateral reflection at very high freqs make the high freq image to appear almost on both side of me, about 160 degrees. Wierd.
Low freq image somewhere in front, but not clear.----
6) And then, the ultimate stereo test: Stereo triangle with plugged one ear with hearing protector plug !
Quess what: I can hear high freqs coming from the speakers even with only one ear
----
At first I did try the high passed pink noise samples as well, but no need to explain what happens: The image remains in the speakers
From all of these 5 configurations, the stereolithic back-to-back provided best low freq image in this mono test. Seems to be that none of them virtued in high freq (> 6kHz) phantom image generation.
Should I feel delighted about the result? Of course it is no surprise to me.
What can I do? How to improve things? But wait, I think I know how to improve this! And it works !
- Elias
When looking people's responses the stereo leaves me cold. Dan, you allways like to point that your speakers are well performing constant directivity smooth type, I've seen your graphs. Is this the best they can do "a large, vague center image" ?
Where is the pin point phantom localisation ?
- Elias
I think Schupbach may have a small genious inside. Maybe he hears like us, who knows.
- Elias
Member
Joined 2009
With the pink noise test I get larger image for the first file and it gets smaller as more of the low content is filtered out. Other than that I the same result in all cases. There's a very strong center image, when I turn my head the image becomes smeared toward the direction. When I move to the side the entire phantom image moves in the direction of the closer speaker.
I'm sitting to form an equilateral triangle with the speakers with 2m sides. The speakers are the ones in my signature, pointed outward about 15 degrees (at 0degrees they would be parallel to the walls).
I'm sitting to form an equilateral triangle with the speakers with 2m sides. The speakers are the ones in my signature, pointed outward about 15 degrees (at 0degrees they would be parallel to the walls).
You can't have pin point with continuous pink noise. There are no sharp transients or even envelopes to provide new localisation cues every few mlliseconds.
Member
Joined 2009
Linkwitz provides a pink noise test file for evaluating one's stereo image.
Testing a stereo system for accuracy
It's different than Markus' "Full spectrum stereo". With Linkwitz's recording the image is more diffuse or blurred. With his "fuzzy Stereo" I get a full room image with no defined center and the speakers are easy to locate. In his Mono recording the image is between the speaker but not as sharp as with Markus'.
Testing a stereo system for accuracy
It's different than Markus' "Full spectrum stereo". With Linkwitz's recording the image is more diffuse or blurred. With his "fuzzy Stereo" I get a full room image with no defined center and the speakers are easy to locate. In his Mono recording the image is between the speaker but not as sharp as with Markus'.
I also tried to modulate the noise with various envelopes, but it certainly did not improve the phantom localisation but worsened the images in the speakers.
- Elias
You'll need to read my posts again about my speakers. They are a broad pattern.
0-90 in 11.25 degree steps
Dan
Member
Joined 2003
4M equilateral triangle in a symmetrical room. Speakers flush mounted in an infinite baffle wall, increasing directivity with increasing frequency (not CD).
Pink_Noise_Full_Spectrum:
Extremely strong center image, regardless of head rotation.
Pink_Noise_HP_24dB_1500Hz:
Still a very strong center image, but somewhat "diffuse" sounding. Both characteristics are independent of head rotation.
Pink_Noise_HP_24dB_3000Hz:
Well defined center image, but more diffuse. Could easily hear each tweeter with ear turned toward tweeter, but not facing straight ahead.
Pink_Noise_HP_24dB_6000Hz:
Center image but very wide, diffuse, covering the distance between the speakers. Could easily hear each tweeter with ear turned toward tweeter, but not facing straight ahead.
Overall, always a center image but wider and more diffuse with increasing frequency. Never heard independent tweeters when facing straight ahead.
Pink_Noise_Full_Spectrum:
Extremely strong center image, regardless of head rotation.
Pink_Noise_HP_24dB_1500Hz:
Still a very strong center image, but somewhat "diffuse" sounding. Both characteristics are independent of head rotation.
Pink_Noise_HP_24dB_3000Hz:
Well defined center image, but more diffuse. Could easily hear each tweeter with ear turned toward tweeter, but not facing straight ahead.
Pink_Noise_HP_24dB_6000Hz:
Center image but very wide, diffuse, covering the distance between the speakers. Could easily hear each tweeter with ear turned toward tweeter, but not facing straight ahead.
Overall, always a center image but wider and more diffuse with increasing frequency. Never heard independent tweeters when facing straight ahead.
The mono parts in Linkwitz's file should sound exactly the same. The problem might be that he switches to different states without any gap.
In a small room you have to go to great lengths to make them not highly absorptive, especially at high frequencies. T60? That measure is totally invalid for a small room and its not unique anyways, it varies with frequency.
My room has a concrete floor and two of four walls are concrete, one is covered in stone. The concrete floor is covered in Oak more for appearance, but it floats a little for LF absorption. The other walls are rigid and reflective at HFs but float and are highly damped at LFs. There is nothing like typical absorption found in most rooms anywhere (typical in that it is very absorptive at HF but not at LFs - the exact opposite of what you want), except the back wall behind the speakers and a patch of futon on the floor to absorb the floor bounce. The ceiling is also solid oak, but there is a difusor right at the first reflection. The furniture is solid oak frame with leather apolstery to minimize absorption. These are extremes when compared to a "typical room". The reverberation is quite apparent upon entering the room. When music is played the reverb tail is quite evident and when one eyes are closed the feeling is of a room of a much larger size than it actually is.
Then show how
Markus - I have discussed this before and the explaination is rather lengthy. I'll write it up someday and post it on my Forum. It just gets tiring giving a lengthy explaination and then having it relegated to obscurity only to get asked again some weeks or months later.
The software on my web site does this, but I have not made it public. I may or I may not, I haven't decided. It's a very powerful technique, because in addition to very high polar resolution you get - for free - a reconstruction of the source that created that polar pattern. For example, I can reconstruct the waves reflecting off of the waveguide mouth at the frequency where there is an axial hole.
I hope to improve the resolution because if I can get the resolution high enough then I can find the HOMs. In my waveguides they are pretty hard to detect, so it takes a very high resolution measurement to find them. I'm not there yet. I may make an HOM generator - i.e. a diffraction horn - to use as a test since the HOMs should not be hard to detect at all.
Its really very interesting, but as with so much else, takes more time than I have.
These "fixes" are not global, they will only correct a single point that you choose to "flatten" the curve at. The negative impedance is a better idea, but the peaks are not exactly one to one with peaks in the sound radiation. The best idea is just to not have any impedance fluctuations. Control the directivity with the device and not with diffraction and then there are no extraneous peaks.
My waveguides have no extraneous peaks. The impedance curves are identical on a plane wave tube or the waveguide. The drivers resonance peaks are always there and they can be a pain, but they are truely lumped parameter and their "fixes" are global.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.