2K said:
Thanks for the info. I just checked out their measurements of some Thiel and Vandersteen models. Yes, though not surprisingly faithful, they indeed implemented time-coherence.
But as we know, to do so, they sacrificed some of important design factors. I still believe that a more important benefit of using lower-order xovers is a wider overlap between drivers and, as a result, more natural transition of the drivers' distortion profiles. For this purpose, I think a faithful LR2 implementatin provides a sufficiently wide overlap. And it gives more robust off-axis performance than the 1st order xover which is very sensitive in listening axis.
Another important consideration is driver choice. I've been looking at many 3rd-party measurements of loudspeaker drivers available to DIYers. I especially focused on their nonlinear distortions. My conclusion is that it is not impossible to implement 1st order rolloffs, but it seriously limits our driver choice. I can come up with some LR2 2-way and 3-way designs that would give excellent system harmonic distortions and also good performance in other important criteria. But I haven't found a single driver combination for 1st order xover that satisfies all of my standards.
They tested the new Thiel 3.7 in the German mag "Stereo" recently and the measurement data is more than just excellent !!!!
If you know how to "juggle" with transfer functions the easiest way to build a transient-perfect crossover is still by going active.
Another important thing is using drivers with wide overlap. In the aforementioned test of the Thiel they said that the midrange of the Thiel would theoretically go up to 20 kHz but is crossed-over at two and something kHz.
It is also possible to use drivers with not so much overlap but then you will have to do bigger efforts elsewhere.
Regards
Charls
If you know how to "juggle" with transfer functions the easiest way to build a transient-perfect crossover is still by going active.
Another important thing is using drivers with wide overlap. In the aforementioned test of the Thiel they said that the midrange of the Thiel would theoretically go up to 20 kHz but is crossed-over at two and something kHz.
It is also possible to use drivers with not so much overlap but then you will have to do bigger efforts elsewhere.
Regards
Charls
phase_accurate said:
I had the chance to do a short audition last fall. I thought they were quite good!
I agree that digital crossovers are the way to go for an active solution.
What impressed me was using aluminum cone drivers with shallow slopes and getting away with it. By giving structure to the cone they were able to move the cone breakup point high enough so as not to be an issue.
Jay_WJ said:
I call it transient OK
Are you talking about vertical or horizontal off axis? Power response on 1st order is about as good as you can get. Then comes 3rd then 2nd order.
No question about it. Driver selection is not easy but not impossible. You need a very robust tweeter. Seas in the lower price range Scanspeak in the upper. How about a 5" or 7" revelator for the mid? That would meet your non linear criteria. As for a woofer the peerles 850146 works. Mark K has measurements on this one. Regretably it is no longer a current model.
2K said:
Power response is an average concept. I'm talking about consistency across different axes. 1st order xover is very susceptible to phase tracking errors over a wide frequency range. Drivers' small beaming effect can easily affect the summation. Besides, I don't like an odd order xover's vertical lobing.
SS 5" or 7" revelators are not appropriate by my standard. They indeed have one of the most extended responses. Someone may cross them at 4 to 5 kHz with a tweeter. But due to their beaming, I don't want to do that. By my standard, no 1" dome tweeters can handle a 1st order slope at 3 kHz Fc. Perhaps, the SS D3806 1.5" dome? Maybe. But as you know, it has its own weakness.
Anyway, this TP design requires compromises as other designs do. But I'm reluctant to accept those particular compromises to achieve the single goal.
Here we can see the on- (red) and off- (green) axis responses of the new Thiel CS 3.7. Off-axis taken at 30 degrees.
The lower diagram shows the step-response.
The amplitude response is excellent and the step-response is wolds apart from what most other three-way speakers offer.
Regards
Charles
The lower diagram shows the step-response.
The amplitude response is excellent and the step-response is wolds apart from what most other three-way speakers offer.
Regards
Charles
Attachments
phase_accurate said:
Thanks for posting the good info. The performance seems outstanding, indeed, though I want to see some more off-axis graphs (both amplitude and step responses) with different angles.
Right, the use of a coaxial driver must be the only way to overcome the first order xover's weakness I talked about. About this Thiel design, I'm particularly interested in how low the midrange is crossed over with the woofer, and how good its low-end distortions are.
I am not sure about that but I have seen other measurements where the acoustical crossover was at 600 Hz. Which is a bit high for my taste BTW. They mention that these Thiels can take a lot of punishment and still play with authority. But they didn't publish distortion measurements.
Regards
Charles
Looking at the XO, I wonder if the above Thiel is of first order. My 4th order acoustical LR XO is slightly simpler than that. I am not talking about the notches, just the main RLC network. Of course, my drivers are different. But I don't think the Thiel network would give anything close to 1st order.
I didn't think the xover tinitus posted was one in a Thiel model, but a design that used Thiel DRivers. Am I wrong?
And right, electric order doesn't necessarily match acoustic order. But I haven't seen any drivers that require using greater than first order electrical filter to achieve a first order slope. Considering a driver's natural rolloff, using greater than first order electrical MAY give a first order slope near the xover freq but it will give a very fast phase turn a little far from the xover point, which can completely ruin the intended summation. For this reason, if there's a FR rise, it's usually a better idea to use a notch filter rather than a higher order filter to implement low order acoustic rolloff.
And right, electric order doesn't necessarily match acoustic order. But I haven't seen any drivers that require using greater than first order electrical filter to achieve a first order slope. Considering a driver's natural rolloff, using greater than first order electrical MAY give a first order slope near the xover freq but it will give a very fast phase turn a little far from the xover point, which can completely ruin the intended summation. For this reason, if there's a FR rise, it's usually a better idea to use a notch filter rather than a higher order filter to implement low order acoustic rolloff.
A bit of compromise helps a lot. Thiel speakers long ago used first order through the crossover region, and then added a second pole about one octave away from the crossover frequency. This requires a driver that is flat for two octaves past the xo frequency, which implies a pretty expensive driver. Nonetheless, it is more realizable with real world drivers (a few) than a pure electrical first order.
We're using a 10" SS and 4H 52 06 13 Skaaning/AT(concave dust cap) and 2905/9000 SS Revelator. (I understand the AT mid has evolved slightly since we bought ours.) The woofer absolutely requires a notch at the high end breakup frequency; we feel that a hard breakup needs to be at least 40 dB down, as the power integral over the ring time makes the breakup more audible than 40 dB attenuation would suggest.
We began with a multislope design for the 150 (?) Hz bass/mid xo, but implementing that obviously required a midrange series capacitor, which, for anything we could afford, limited the mid quality. (Best was 10 X 10uF, 5X5 antiparallel to balance out colorations, NorthCreek Crescendos (?) (Harmonies?). Instead we pulled out the cap, and used the sealed cabinet response to provide the 90 Hz low freq crossover, 2nd order of course. The HF crossover point is about 3800 Hz, higher than we intended, but it works very well. The tweeter is non-stressed, and the mid does astonishingly well.
As a practical matter, the skill, patience and care used in implementing any reasonable design are apt to outweigh the theoretical design differences. We heard a midbass improvement when we went from the first multislope to second order; the mid seemed to respond well to the improved damping without the series cap, and the midbass tightened up a bit. That sort of result is not considered in the standard discussions. I'd suggest being prepared to experiment a little, to see what really suits the drivers best.
We're using a 10" SS and 4H 52 06 13 Skaaning/AT(concave dust cap) and 2905/9000 SS Revelator. (I understand the AT mid has evolved slightly since we bought ours.) The woofer absolutely requires a notch at the high end breakup frequency; we feel that a hard breakup needs to be at least 40 dB down, as the power integral over the ring time makes the breakup more audible than 40 dB attenuation would suggest.
We began with a multislope design for the 150 (?) Hz bass/mid xo, but implementing that obviously required a midrange series capacitor, which, for anything we could afford, limited the mid quality. (Best was 10 X 10uF, 5X5 antiparallel to balance out colorations, NorthCreek Crescendos (?) (Harmonies?). Instead we pulled out the cap, and used the sealed cabinet response to provide the 90 Hz low freq crossover, 2nd order of course. The HF crossover point is about 3800 Hz, higher than we intended, but it works very well. The tweeter is non-stressed, and the mid does astonishingly well.
As a practical matter, the skill, patience and care used in implementing any reasonable design are apt to outweigh the theoretical design differences. We heard a midbass improvement when we went from the first multislope to second order; the mid seemed to respond well to the improved damping without the series cap, and the midbass tightened up a bit. That sort of result is not considered in the standard discussions. I'd suggest being prepared to experiment a little, to see what really suits the drivers best.
Jay_WJ said:
Jay
On an infinite baffle I would agree. For real world baffles you need more response shaping than a single component will provide.
As I have stated before you are doing very well to mantain a 1st order response 1 1/2 to two octaves past Fc with real world drivers but from what I have observed it's enough to maintain a very low phase rotation overall which produces good step response.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.