The KEF uniQ units were a fairly constant directivity solution. The woofer cone acted as an effective waveguide at a beamwidth quite close to the woofer's beamwidth for likely crossover frequencies. Response smoothness in the top Octave due to surround reflections was the geater problem. There are other coaxial solutions that could be used if desired. Here is a clever JBL solution for a ceiling speaker. I see no evidence of the crossover point in its measurements.
http://www.jblpro.com/catalog/support/getfile.aspx?docid=1646&doctype=3
I wasn't offering the coaxial as the only answer to the problem, just as something available for those that are really worried about off axis response perfection in all directions.
Two things regarding power response: first if you accept quadrature summing you can have flat power response through the crossover region.
I fall more into the "in-phase" camp. While in-phase crossovers tends to have a power response hole, you would have to convince me that holes in the power response are problematic, when Lipshitz and Vanderkooy, along with Toole (at least his measurements) show that they are not.
Earl, I really sense that you are moulding some of your science beliefs around what best fits the marketing of your product.
Regards,
David S.
Most are, for sure. This one is a nice exception and works very well. But you don't see an 18" coaxial every day. It can be done, but usually isn't.
Attachments
I do its a major part of my designs, but the group delays from diffraction and HOM are muach great in time than anything from the crossover. That makes them more of a problem.
Some people claim that HOMs are not an issue, admittedly they are small, but I cannot find any other reason that a waveguide which minimizes them to thelargest extent possible, as well as all forms of diffraction, should sound so much more transparent (or whatever subjective adjective you want to use).
Dave
You can believe that if you want but the facts show that it precisely the other way around. I did my designs long before I ever sold them so my opinions on design and the science behind them were formed long before I was a manufacturer. OS waveguides go back more than 20 years now. You can go back as far as you want and you won't find that I changed my opinion to suite the designs. The designs fit the opinions that I have foremed from the science that I did or understood - virtually all of which precedes the marketing.
Last edited:
I am questioning that there are "no go" areas for crossover points, that coaxials can't be made to work, that crossover points can't be made invisible and that power response holes are disqualifying.
You should address those points.
David
This horn would be greatly improved by a lossy foam rectangle mounted in front of the baffle, right around the edge of the horn. That acts a lot like a roundover, but is lossy, so there's some LF loss (LF relative to the horn bandwidth) that is preserved with a roundover. But it's both effective and easy to do, I've done this with improperly terminated horns before and it works very nicely.
I clicked that link and darned near had to change my pants! If they're actually managing to get the cone profile of the mid to work the way it appears, that's a heck of a solution.
Hmm, I think you're actually arguing the same side of the argument as me there... I also said that group delay of the crossover filter itself (often obsessed over by some when choosing crossover slopes) is not a big issue compared to the potential for group delay elsewhere and further is unlikely, on its own, to come anywhere near the threshold of detection in any reasonable crossover design.
When I said to consider all sources of group delay I was thinking more along the lines of not ignoring the group delay Keyser pointed out that's due to the wave-guide having a much deeper acoustic centre than the woofer. What does your research on group delay have to say on that issue ?
I think you're probably correct there, and have answered your own question - its the time delayed reflections and diffraction artefacts that are damaging to the perceived quality, and the elimination of them that gives the improvement in transparency.
Some of that will be reduction in sharp ripples in the amplitude response, (I find a high frequency response that is nominally flat but has a lot of narrow band up and down ripples to sound somewhat harsh and ultimately fatiguing) but I don't think frequency response alone is the whole story.
There must be something about the time domain response when there are delayed reflections and refractions present that is perceived more negatively than the amplitude errors alone would suggest, and also why EQ can't fix it.
I've been doing my own experiments with tweeter diffraction reduction recently and its convinced me of just how important it is to eliminate as much as possible diffraction and reflections particularly from about 2Khz upwards where the ear becomes very acute.
Elimination of even small amounts of residual diffraction in this frequency range seems to improve the perceived smoothness and transparency out of proportion to the fairly small change in frequency response that goes along with it, and applying EQ to make a similar change affects the tonal balance but doesn't give the same improvement in smoothness.
This is one reason why I think a conventional dome tweeter with a flat face plate mounted on a flat baffle is a train wreck waiting to happen. No matter whether the tweeter is flush mounted or not, and where it is on the baffle, diffraction effects are going to be severe unless the cabinet edge has a rather large radius.
While large amounts of time delayed reflections and diffraction from HOM and other sources may affect the group delay at certain frequencies, I'm not sure that group delay is the best way to characterise the problem any more than looking at its effect in the frequency domain is.
For example a smooth overall shift in group delay from two time unaligned drivers is possible even when no reflection or diffraction is present, but does it sound as bad as an equal amount of group delay at certain frequencies from reflections ?
One of the great qualities of the Gedlee speakers is PRAT. Earl might shoot me for saying that, but it's true.
Drums are great, and I don't mean just dynamic, the Pace Rhythm and Timing are excellent. It brings the music alive, it swings. Is that due to a reduction in HOMs, good crossover work, the OS waveguide or some combo of all those things? I don't know. I just know that it works - you hear it right away.
Simon
Yes, I see we do agree on a lot of things.
As to the waveguide and woofer time alignment, this just seemed to take care of itself (In the Abbey and the Summa). The delay from the low pass filter exactl;y aligned the woofers response to the tweeters as the impulse response shows. I did not have to do anything in that regard. Now I don;t know that ALL two-way woofer/horn systems will act this way, but mine with third order LP filters do. The Second order ones do not quit make it.
This is precisely what our resarch showed. That at higher SPLs the group delay effects become "unmasked" temporally (the exact opposite of frequency domain effects, which is quite unintuitive). Thus as a "bad" (high GD problems) system gets turned up it starts to sound bad. Poeple invariably call this "nonlinear distortion", which it is, but they attribute it to the wrong root cause. Then, in the end, as they continue to lower the "THD" it just doesn't sound any better. "Oh well, ship it anyways."
When Lidia and I published our work no one paid any attention. Then when we published our work on group delay no one even read the paper. Today THD is still measured and talked about as if nothing has changed.
Today people like Floyd Toole, Sean Olive, Alex Voishvillo, Lauri Fincham and a few others all agree on nonlinear distortion. But around here, its still "THD" - business as usual.
Yes, I see we do agree on a lot of things.
As to the waveguide and woofer time alignment, this just seemed to take care of itself (In the Abbey and the Summa). The delay from the low pass filter exactl;y aligned the woofers response to the tweeters as the impulse response shows. I did not have to do anything in that regard. Now I don;t know that ALL two-way woofer/horn systems will act this way, but mine with third order LP filters do. The Second order ones do not quit make it.
This is precisely what our resarch showed. That at higher SPLs the group delay effects become "unmasked" temporally (the exact opposite of frequency domain effects, which is quite unintuitive). Thus as a "bad" (high GD problems) system gets turned up it starts to sound bad. Poeple invariably call this "nonlinear distortion", which it is, but they attribute it to the wrong root cause. Then, in the end, as they continue to lower the "THD" it just doesn't sound any better. "Oh well, ship it anyways."
When Lidia and I published our work no one paid any attention. Then when we published our work on group delay no one even read the paper. Today THD is still measured and talked about as if nothing has changed.
Today people like Floyd Toole, Sean Olive, Alex Voishvillo, Lauri Fincham and a few others all agree on nonlinear distortion. But around here, its still "THD" - business as usual.
1) there are better and poorer places for crossovers
2) I haven't measured any "good" coaxials and I never trust published data
3) I don't know what "invisible" means, I can measure them and "see" those results
4) Sean Olive showed that power response holes do make an audible difference in his testing of room EQ systems. The eystems that "fixed" this hole were deemed to "sound better".
You don't read enough audio reviews! Pace Rhythm and Timing. Something a lot of people deny even exists with playback. Once you hear a system that has it, you understand.
Yes, most "hi fi" loudspeakers are dreadful in that regard (..and of course most have 1watt 1 meter ratings of less than 90 db, so it's not a big surprise).
Hi Earl,
The issue here is that you’ve made a blanket statement that I felt was unworthy of a person of your intelligence and knowledge, that there were frequency ranges where crossover points shouldn’t be placed.
2) If you don’t trust any manufacturer’s data then there is nothing I can do to persuade you. Still, both the Genelec and the JBL examples seem to be credible and support the point that coaxs can be free from obvious crossover effects.
3) I gave my definition of invisible: that response errors due to crossover where at a lower level than the remaining response errors of even a good transducer or cabinet mounting. In my experience that is not that hard to achieve. If you want to look at it another way, you actually have more degrees of freedom for response control in the crossover region. Arguably, you can have flatter response through the crossover region than away from the crossover region. (I'm not saying everyone achieves that. Just that it is possible.)
4) I think you are oversimplifying anything that Sean said. Clearly with an electronic EQ system (unlike at the crossover design stage) your only chance to “fix” a power response hole is by putting a complimentary response peak in the direct sound. I guarantee that Sean would not support that as good practice. You may note at his website where describes the test that you refer to, that he and I had some correspondence and discussed the pitfalls of steady state room correction.
Sean has done some excellent fundamental research showing that axial response flatness and axial response smoothness (plus LF extension) are the big factors in listener ranking. Both he and Floyd look at power response as a means of revealing resonances but acknowledge that no particular power response curve shape is indicated.
Regards,
David
He does that a LOT (..or rather absolute, or seemingly absolute, statements).
Just in the past couple of days or so there were several statements I'd object to..
..but after a while you just let "them go" (or most of them), and hope that no one treats them as "gospel".
absolutes
like this....
kinda oxymoronic one would think... given all the data published about the Abbeys and Summas... especially wrt quality of data often "taken to task" from other posters
like this....
kinda oxymoronic one would think... given all the data published about the Abbeys and Summas... especially wrt quality of data often "taken to task" from other posters
Last edited:
One of the great qualities of the Gedlee speakers is PRAT. Earl might shoot me for saying that, but it's true.
I posted the impulse response from his old speaker company(don't know if he's still doing that) and they were more compact than any monitors I've measured(some of which he said were good). Don't know how much that has to do with it. I know he's modded the crossovers since then though. I bet it's not much different on that respect.
I'd think the reduction of early reflection likely helps that too though.
Dan
Hi,
This might already be in the “same old, same old” discussion but Toole has made some statements regarding what would be a good speaker:
http://www.harmanaudio.com/all_about_audio/acoustical_design.pdf
To me it is pretty clear a direction for a design with regards to the power response shape.
To simplify further, please pardon me if it is over simplified, the DI, taken at the “designed axis” not what is usually referenced as the “on axis”, of the resulting speaker should be flattish with the transition, in the general case, from quasi omni to CD (step) more or less high in frequency and the CD plateau more or less high depending on the use/taste.
Dr Geddes might like to transition at around 500Hz to a high plateau while B&O prefers a low plateau (the “designed axis” notion becomes moot is this case) and Toole somewhere in the middle depending on the intending use/market segment target.
Toole clearly separates three independent dimensions: use, taste, intended result by the designer.
Cheers
M
Last edited: