Announcement

Collapse
No announcement yet.

Offsets, Asymmetrical Slopes, and Mysticism - revisited *PICS*

Collapse
X
 
  • Filter
  • Time
  • Show
Clear All
new posts

  • #31
    Re: Practical factors considered

    Originally posted by Jeff B. View Post
    Actually you're both right. The session I sent Pete sometime back used 20mm, and my most recent ones used 25mm, which is what I told Deward. As I also told Deward, this is a difficult parameter to measure. I use a method of taking a summed measurement and adjusting the simulation till it matches. You can look for the reverse null too, which is sometimes easier to see. What you are actually deriving when you do this is the relative acoustic offset between the two drivers and it will vary depending on how you mount your drivers too. So which is right? Hmmmmm....Deward came up with 24mm doing it the reverse null method and my latest sims used 25mm. In the one I sent Pete my sim used 20mm. But the difference between 20 and 25mm translates into 10 degrees of phase shift at 2kHz, and this won't alter the summed response much at all, so it is difficult to know for sure, and probably not consequential in the long run. I think though the actual value is closer to the 25 though, but it could be 23........
    To be precise, the relative acoustic offset determined is for the driver models as created in the attempt to match the measured response. At best the models are approximations and the offset is appropriate only for those approximations.

    dlr
    WinPCD - Windows .NET Passive Crossover Designer

    Dave's Speaker Pages

    Comment


    • #32
      Re: Practical factors considered

      Originally posted by Jeff B. View Post
      You still need the right knee shape to sum correctly though. Jeff
      That seems to me the critical observation (and I'd include phase tracking within the summation region . . . otherwise you get the ripples clearly visible in jkim's summed responses). The difficulties that arise when using phase shift for offset correction show mostly as summation errors, made worse because when the splitting filter is "relaxed" (in slope) the critical summation region becomes broader and previously suppressed out-of-band artifacts again become issues. There is a "way out".

      The same phase shift can be introduced by making the crossover asymetric *at a higher order* . . . by "enhancing" it rather than "relaxing" it. And it turns out that when you do that not only does the summing region (however one defines that) become *smaller*, but it seems to be *easier* to preserve the desired (lower order) knee shape. The price, of course, is crossover complexity . . . typically an additional couple of components on one driver. OTOH since the original crossover slope is preserved (or "enhanced") it may not become necessary to add things like notch filters to deal with out-of-band artifacts. It's not as good a solution as the delay filters that one can easily implement at line level, but it accomplishes much of the same thing while remaining "passive".
      "It suggests that there is something that is happening in the real system that is not quite captured in the models."

      Comment


      • #33
        Re: Practical factors considered

        Some notes have been added to my initial post.

        -jAy

        Comment


        • #34
          Re: Practical factors considered

          Originally posted by jkim View Post
          3. In theory, you can also apply this technique to LR type crossovers of higher orders. For example, you can "relax" the rolloffs of Linkwitz-Riley 6th or 8th order filters to compensate a phase mismatch due to a driver offeset.
          I noted that one can acomplish an offset compensating phase shift by *adding* additional shift to one filter element . . . ie. going from a 4th/4th to a 4th/5th (4th+1) or some approximation, depending on how much phase shift is needed.

          You want to call that 6th-2/6th-1, apparently so you can keep using the term "relaxed" ? ? ?
          "It suggests that there is something that is happening in the real system that is not quite captured in the models."

          Comment


          • #35
            Re: Practical factors considered

            Originally posted by dlr View Post
            To be precise, the relative acoustic offset determined is for the driver models as created in the attempt to match the measured response. At best the models are approximations and the offset is appropriate only for those approximations.

            dlr
            Yes, that's pretty much the case. If someone was tasked with determining the true acoustic center of a driver it would be daunting, especially since it may not even be a fixed location. (Let's not go too far down that road). However, if the task is simply to determine the relative offset between two drivers in the crossover region, that's not quite as difficult. But, you are correct, it is still only an approximation at best. For modeling though, the approximation can be more than adequate for an LR4 crossover which is actually reasonbly immune from small offset mis-matches. If you are working with second order acoustic it's a lot more critical.

            Jeff
            Click here for Jeff Bagby's Loudspeaker Design Software

            Comment


            • #36
              Re: Offsets, Asymmetrical Slopes, and Mysticism - revisited *PICS*

              I've been following this thread for awhile now, unsure if I should speak up or not. However I have now just finished an after work screwdriver so I'm no longer fearful!

              I don't know how many of you are familiar with Vandersteen speakers, but he, Richard Vandersteen, touts his ability to provide time and phase coherence. I've had a pair of 2c's for 27 years and to my ear he's done a pretty good job.

              Anyway, the visible artifact is that he offsets his drivers linearly, apparently to align their acoustic centers. Check out this website:

              http://www.vandersteen.com/pages/Model5alit1.html

              I would assume the offset baffles he has must corrupt something and needs to be somehow accounted for somewhere.

              Does anybody think this is worth further discussion?
              By the way these are $17,000 a pair.

              JohnZ

              Comment


              • #37
                Re: Offsets, Asymmetrical Slopes, and Mysticism - revisited *PICS*

                Originally posted by johnzimmerman View Post
                I've been following this thread for awhile now, unsure if I should speak up or not. However I have now just finished an after work screwdriver so I'm no longer fearful!

                I don't know how many of you are familiar with Vandersteen speakers, but he, Richard Vandersteen, touts his ability to provide time and phase coherence. I've had a pair of 2c's for 27 years and to my ear he's done a pretty good job.

                Anyway, the visible artifact is that he offsets his drivers linearly, apparently to align their acoustic centers. Check out this website:

                http://www.vandersteen.com/pages/Model5alit1.html

                I would assume the offset baffles he has must corrupt something and needs to be somehow accounted for somewhere.

                Does anybody think this is worth further discussion?
                By the way these are $17,000 a pair.

                JohnZ

                I am familiar with Vandersteens. He uses the impluse response to align all of his drivers for precise time-alignment and then uses first order networks to maintain the phase alignment. With respect to phase and time alignment; he does a pretty good job. The downside is that his baffles, as you mentioned, contribute a lot of diffraction effects, and if you ever see measurements of his speakers they typically have frequency response plots that are far from flat. The one you mentioned above was reviewed not too long ago in Stereophile and the response looked terrible, with a deep on-axis null developing. In my opinion he sacrificed the more important thing - frequency response for the thing of much lesser importance - flat phase response.
                Click here for Jeff Bagby's Loudspeaker Design Software

                Comment


                • #38
                  Re: Offsets, Asymmetrical Slopes, and Mysticism - revisited *PICS*

                  Originally posted by johnzimmerman View Post
                  I would assume the offset baffles he has must corrupt something and needs to be somehow accounted for somewhere.JohnZ
                  Nope . . . that's how Thiel does it too. There's no other choice for first order crossovers, except for adding "active" delay. With stepped baffles one has to be careful to avoid diffraction . . . it looks like there's a lot of felt on that baffle to do that. Spica used a sloped baffle on the rather well regarded T-50 two-way.
                  "It suggests that there is something that is happening in the real system that is not quite captured in the models."

                  Comment


                  • #39
                    Re: Offsets, Asymmetrical Slopes, and Mysticism - revisited *PICS*

                    Originally posted by Deward Hastings View Post
                    Nope . . . that's how Thiel does it too. There's no other choice for first order crossovers, except for adding "active" delay. With stepped baffles one has to be careful to avoid diffraction . . . it looks like there's a lot of felt on that baffle to do that.
                    I agree, from what I see there ought to be enough felt to make diffraction a minor issue. Stepped baffles with felt can work well. If there's a null as Jeff mentioned, it's from poor crossover work/construction or the Stereophile measurements were not made on the same axis as the one for which the design was intended.

                    Edit: After looking at the Vandersteen site and pictures, let me amend that a bit. They have a near-profile shot of the top section. The felt, to me, appears to be less than optimal thickness, so there may be more diffraction effects remaining that could be the case.

                    I still suspect that the main problem is one of axis and/or XO design. The measurement distance would need to be farther than the typical 50" or so used by Stereophile to get full integration of the drivers' outputs, especially if they chose the wrong axis to measure. The mid/tweeter Fc is 5KHz. With the separation they show, it's not a surprise to see a non-flat response at 50". A null, though, still seems to be a XO design issue as well.

                    dlr

                    WinPCD - Windows .NET Passive Crossover Designer

                    Dave's Speaker Pages

                    Comment


                    • #40
                      Re: Offsets, Asymmetrical Slopes, and Mysticism - revisited *PICS*

                      Originally posted by dlr View Post
                      I agree, from what I see there ought to be enough felt to make diffraction a minor issue. Stepped baffles with felt can work well. If there's a null as Jeff mentioned, it's from poor crossover work/construction or the Stereophile measurements were not made on the same axis as the one for which the design was intended.

                      Edit: After looking at the Vandersteen site and pictures, let me amend that a bit. They have a near-profile shot of the top section. The felt, to me, appears to be less than optimal thickness, so there may be more diffraction effects remaining that could be the case.

                      I still suspect that the main problem is one of axis and/or XO design. The measurement distance would need to be farther than the typical 50" or so used by Stereophile to get full integration of the drivers' outputs, especially if they chose the wrong axis to measure. The mid/tweeter Fc is 5KHz. With the separation they show, it's not a surprise to see a non-flat response at 50". A null, though, still seems to be a XO design issue as well.

                      dlr

                      The speaker I saw the review for was the Quatro Wood, which is just a wood-veneered version of the Vandersteen Quatro. Here is the graph I was referring to:




                      In the graph the measurements were taken at 50" like you said. The red line is on-axis with the woofer, the blue line, with the null, is on-axis with the tweeter. Vandersteen may have designed them to intergrate at a greater distance, but being on the tweeter axis it will still be hard to overcome a completely out-of-phase relationship by just moving out a few feet if you stay on the tweeter axis. It could be that Vandersteen doesn't intend on the listening axis being the tweeter axis, but based on its height, I'm pretty sure it's the right one. I believe the issue lies in the problem of dealing with the tweeter's natural roll-off and subsequent phase shift while using a first order electrical network. Whenever I model first order crossover with aligned drivers I get nulls at the wrong spots due to the tweeter roll-off.

                      Here's a link to all of the measurements for the speaker if you are interested:

                      Stereophile measurements of the Vandersteen Quatro

                      Jeff
                      Last edited by Jeff B.; 06-11-2008, 09:58 PM. Reason: Fixed a mispelling and added a thought.
                      Click here for Jeff Bagby's Loudspeaker Design Software

                      Comment


                      • #41
                        Re: Offsets, Asymmetrical Slopes, and Mysticism - revisited *PICS*

                        Jeff,

                        John Dunlavy had a pretty good handle on systems with first order XOs, though his stepped baffle wasn't anything like Vandersteens. Most of his speakers measured very well and the few that I heard (sc-VI and the Cantata) sounded very good.

                        Here are the measurements of the SC-VI done at 10 ft.

                        http://stereophile.com/floorloudspea...62/index9.html

                        I agree with you and don't think the vandersteens measurements would change much by extending the measurement distance out on the tweeter axis.

                        Regards,

                        Dennis

                        Comment


                        • #42
                          Re: Offsets, Asymmetrical Slopes, and Mysticism - revisited *PICS*

                          Originally posted by djarchow View Post
                          Jeff,

                          John Dunlavy had a pretty good handle on systems with first order XOs, though his stepped baffle wasn't anything like Vandersteens. Most of his speakers measured very well and the few that I heard (sc-VI and the Cantata) sounded very good.

                          Here are the measurements of the SC-VI done at 10 ft.

                          http://stereophile.com/floorloudspea...62/index9.html

                          I agree with you and don't think the vandersteens measurements would change much by extending the measurement distance out on the tweeter axis.

                          Regards,

                          Dennis

                          Yes, I am familiar with Dunlavy's and they do, in fact, measure and sound very good. I also think what John Bau pulled off with the Spica TC-50 so long ago in phase alignment was an incredible accomplishment given the resources of the time. I probably should go into that one sometime. We've talked about this a lot in the past, so you know I am not opposed to creating these phase coherent systems, rather I have some affection toward them. I just haven't been impressed with the Vandersteens. I also find it interesting how incredibly flat the new Dynaudios measure in reviews, and they still claim to be using first order networks. I wonder if they are still using all-pass delays with them.
                          Jeff
                          Click here for Jeff Bagby's Loudspeaker Design Software

                          Comment


                          • #43
                            Re: Offsets, Asymmetrical Slopes, and Mysticism - revisited *PICS*

                            Originally posted by Jeff B. View Post
                            Yes, I am familiar with Dunlavy's and they do, in fact, measure and sound very good. I also think what John Bau pulled off with the Spica TC-50 so long ago in phase alignment was an incredible accomplishment given the resources of the time. I probably should go into that one sometime. We've talked about this a lot in the past, so you know I am not opposed to creating these phase coherent systems, rather I have some affection toward them. I just haven't been impressed with the Vandersteens. I also find it interesting how incredibly flat the new Dynaudios measure in reviews, and they still claim to be using first order networks. I wonder if they are still using all-pass delays with them.
                            Jeff
                            Jeff,

                            Yep, we have been in a number of these discussions over the years, especially back on MAD. I posted the reference to the Dunlavy speakers more for others that had not had exposure to a well done 1st order system as opposed to the Vandersteens.

                            In looking at some of the meaurements of the Dynaudios, they aren't time aligning the drivers and in the three ways, the mids are wired with reverse polarity from the woofers and tweeters. Since they aren't doing all the things they need to to produce a minimum phase speaker, the benefits of doing 1st order XOs aren't there either. They are doing a good job of getting flat frequency response, especially with 1st order slopes.

                            Regards,

                            Dennis

                            Comment


                            • #44
                              Re: Offsets, Asymmetrical Slopes, and Mysticism - revisited *PICS*

                              I had posted this brief discussion way back in the MAD hatday. You may find it of interest. Basically if toy use a flat baffle then you can never get a truely accurate transient unless there is electronic delay on the tweeter in addition to a transient correct crossover. If you align the ACs then it becomes a matter of using a transient accutate crossover. The game SPICA played was to move the tweeter behind the woofer to account for the delay of the woofer filter. This prought the upper range of the tweeter in time coherence with the midrange of the woofer but there are still problems in the crossover region with regsard to response flatness. Through optimization flat response and good transient behavior can be obtained this way, but only on axis.



                              http://www.geocities.com/kreskovs/John1.html
                              John k.... Music and Design NaO dsp Dipole Loudspeakers.

                              Comment


                              • #45
                                Originally posted by Jeff B. View Post
                                ........ For each crossover order, beginning with a first order network, there is 90 degrees of phase rotation per order. This is easily shown mathematically based on the poles of the complex response. This one factor is not coincidental, it is mathematical......Jeff B.
                                It is unfortunate that most of the pictures from this post are gone thanks for Photobucket. Such a valuable thread. Does an one have a copy of this Jeff's first post with the embedded pictures?

                                I also wanted to ask a question in regards to Jeff's statement above:

                                Does driver phase change/rotate based on the acoustic crossover order or the electrical crossover? For example, if there is 90 degrees phase rotation per order, would the the driver have 180 degrees phase rotation for a second order electrical, which lets assume resulted in effectively a 4th order acoustic crossover, or would it have 360 degrees of phase rotation (i.e., 90 degrees of each acoustic crossover order)?

                                Comment

                                Working...
                                X