Hello everybody!
since John Watkinson's "Celtic Audio" website is down and the essay "Putting the Science Back into Loudspeakers" is no longer available online I decided to attach it here at diyaudio.com
Because I think it is worth reading and I noticed that it was only discussed in neodymium/magnetic threads so far. And the question of neodymium is quite marginal in that essay by the Author of "The Art of Sound Reproduction"
fortunately the essay, "a thought provoking article" in the words of Siegfried Linkwitz, is short and the file is not too big to be attached here
best,
graaf
since John Watkinson's "Celtic Audio" website is down and the essay "Putting the Science Back into Loudspeakers" is no longer available online I decided to attach it here at diyaudio.com
Because I think it is worth reading and I noticed that it was only discussed in neodymium/magnetic threads so far. And the question of neodymium is quite marginal in that essay by the Author of "The Art of Sound Reproduction"
fortunately the essay, "a thought provoking article" in the words of Siegfried Linkwitz, is short and the file is not too big to be attached here
best,
graaf
Thanks Graaf.
I think everyone should read this, It will provoke some interesting discussion and research.
Using variable bit rate encoding to evaluate speaker quality is brilliant... But only if people's ability or training to detect compression artifacts are equal. Still great for one person benchmarking their speakers.
I wonder how many of the highly rated headphones are popular only because they are good at masking compression artifacts.
Now to find some free software that can do that on-the-fly provided CD is actually "blameless" that is.
so rectangular reflex boxes = bad. Now how to define what a good box/baffle is.
I think everyone should read this, It will provoke some interesting discussion and research.
Using variable bit rate encoding to evaluate speaker quality is brilliant... But only if people's ability or training to detect compression artifacts are equal. Still great for one person benchmarking their speakers.
I wonder how many of the highly rated headphones are popular only because they are good at masking compression artifacts.
Now to find some free software that can do that on-the-fly provided CD is actually "blameless" that is.
so rectangular reflex boxes = bad. Now how to define what a good box/baffle is.
Celtic's Cabar is of course a speaker where all off John W's claims are taken into consideration. There are no sharp corners and it has a transient perfect crossover. The woofer "box" is of cylindrical shape and offers therefore more inherent stiffness than a conventional box.
Since their homepage is down for days already (anyone knows what happened ?) one has to look elsewhere to see the design. Fortunately it is on the front cover of one of John's books:
http://books.elsevier.com/bookscat/coverslarge/9780240515120.jpg
It shows the right third of the speaker. The small wideband metal-cone driver is the mid-high. AFAIK is there another one on the backside as well. Together with the proprieatary signal processing an exactly predetermined radiation pattern is achieved.
There are only smooth shapes involved as one can see.
The metallic-looking "ring" is in fact the grille for the woofer which is hidden from sight and which is using the part of the tube to the left of the grill. The center section of the tube carries the electronics and the left part is the mirror-image of the part shown on the picture.
One that I know of that would also fulfil many of John W's claims:
http://www.pupazzo.page.ms/
Since both the Cabar and the Pupazzo have limited SAF in my case - I use a conventional box that is still better regarding transient reproduction than > 99.9 % of the speakers out there. And yes my mid-tweeter has a rare-earth magnet as well !
Regards
Charles
Since their homepage is down for days already (anyone knows what happened ?) one has to look elsewhere to see the design. Fortunately it is on the front cover of one of John's books:
http://books.elsevier.com/bookscat/coverslarge/9780240515120.jpg
It shows the right third of the speaker. The small wideband metal-cone driver is the mid-high. AFAIK is there another one on the backside as well. Together with the proprieatary signal processing an exactly predetermined radiation pattern is achieved.
There are only smooth shapes involved as one can see.
The metallic-looking "ring" is in fact the grille for the woofer which is hidden from sight and which is using the part of the tube to the left of the grill. The center section of the tube carries the electronics and the left part is the mirror-image of the part shown on the picture.
One that I know of that would also fulfil many of John W's claims:
http://www.pupazzo.page.ms/
Since both the Cabar and the Pupazzo have limited SAF in my case - I use a conventional box that is still better regarding transient reproduction than > 99.9 % of the speakers out there. And yes my mid-tweeter has a rare-earth magnet as well !
Regards
Charles
I think you are actually doing the constructor wrong. His design is older than B&O's space station. And it is omnidirectional, a thing that the B&O isn't.
And it is also transient-perfect another thing that the B&O isn't (which is quite a shame taken the B&O's massive built-in digital signal processing into account).
Regards
Charles
Hmmm, I like the gist of what John Watkinson says in that article, but he seems to avoid details about the Heisenberg Inequality (re: time and frequency accuracy being mutually exclusive).
A simple experiment that anyone can do with a music keyboard and/or software is to play two closely spaced sinusoidal tones, and vary the frequency of one of them to observe some effects based on one's own hearing.
What I've typically found is that a tremelo effect can be heard up to a difference of F1 - F2 = approximately 20Hz. Above that, the sound of 2 dissonant frequencies starts to dominate over the AM effect.
Contrary to what Watkinson appeared to be saying about the importance of reproducing transients with high accuracy, experimental results suggest that their importance starts to diminish substantially below around 25ms ( 1/ (20Hz times 2 peaks per cycle)). The thing is that 25ms is a huge amount of time compared to the <1ms accuracy that's often suggested for so called "phase accurate" crossovers.
A simple experiment that anyone can do with a music keyboard and/or software is to play two closely spaced sinusoidal tones, and vary the frequency of one of them to observe some effects based on one's own hearing.
What I've typically found is that a tremelo effect can be heard up to a difference of F1 - F2 = approximately 20Hz. Above that, the sound of 2 dissonant frequencies starts to dominate over the AM effect.
Contrary to what Watkinson appeared to be saying about the importance of reproducing transients with high accuracy, experimental results suggest that their importance starts to diminish substantially below around 25ms ( 1/ (20Hz times 2 peaks per cycle)). The thing is that 25ms is a huge amount of time compared to the <1ms accuracy that's often suggested for so called "phase accurate" crossovers.
Having read many articles written by John W I know for sure that he is aware of the time/frequency trade-off.
I don't understand what a phase-accurate crossover has to do with the beat between two frequencies.
But I am convinced that timing errors in the area of 1ms or even larger are too much if one takes into consideration how our hearing detects direction.
Regards
Charles
I don't understand what a phase-accurate crossover has to do with the beat between two frequencies.
But I am convinced that timing errors in the area of 1ms or even larger are too much if one takes into consideration how our hearing detects direction.
Regards
Charles
phase_accurate said:
Well if the line starts to get blurred below 25ms and instead of hearing one pulsating tone we start hearing 2 continuous tones, then why worry about delays that are 25 times smaller?
It seems a bit like insisting on a 2,000Hz refresh rate on a CRT monitor, because it might be possible to see a little bit of flicker at 85Hz. It doesn't make sense.
I understand that human ears can distinguish small phase differences at several kHz, but I'm not sure how that's meant to translate to timing differences, which may vary from a few degrees to several thousand degrees depending on frequency.
SL still has the paper (along with others) here:
http://www.linkwitzlab.com/links.htm
In particular, see this link:
http://www.aip.org/pt/nov99/locsound.html
http://www.linkwitzlab.com/links.htm
CeramicMan said:
In particular, see this link:
http://www.aip.org/pt/nov99/locsound.html
Lol, you know what I mean! 
There's no 'right' or 'wrong' regarding those mutually exclusive properties. Hearing two continuous tones could equally be described as an inability to resolve pulses above a certain repetition rate
Besides, I think we're talking about two different effects here. I agree that localization of sounds in the 500Hz~5kHz range demonstrates an ability to work out phase differences in the order of microseconds. It could be as simple as neurons firing from opposite sides of the head and combining near the middle, where slight lateral offsets enable a natural learning process where the brain literally maps the sounds and recognizes familiar patterns.
But I don't see how that's relevant to crossovers as John Watkinson suggested. By definition, low-pass filters reduce the timing accuracy at which transients can be determined. Lobing effects are well known and phase rotation merely changes the angles at which they occur - usually it's in the vertical axis anyway.
IMO the important point is that each ear gets equal treatment, so the loudspeakers have to have very similar performance on both sides.
There's no 'right' or 'wrong' regarding those mutually exclusive properties. Hearing two continuous tones could equally be described as an inability to resolve pulses above a certain repetition rate Besides, I think we're talking about two different effects here. I agree that localization of sounds in the 500Hz~5kHz range demonstrates an ability to work out phase differences in the order of microseconds. It could be as simple as neurons firing from opposite sides of the head and combining near the middle, where slight lateral offsets enable a natural learning process where the brain literally maps the sounds and recognizes familiar patterns.
But I don't see how that's relevant to crossovers as John Watkinson suggested. By definition, low-pass filters reduce the timing accuracy at which transients can be determined. Lobing effects are well known and phase rotation merely changes the angles at which they occur - usually it's in the vertical axis anyway.
IMO the important point is that each ear gets equal treatment, so the loudspeakers have to have very similar performance on both sides.
I have to basically agree that treating both ears (or the signals for both ears) the same way should theoretically be sufficient.
But time-smear (due to allpass crossovers) would change the perceived SIZE of a virtual sond-source. Apart from that: Why do we accept speakers whose output is a mere caricature of its input signal ?
Regards
Charles
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.