Announcement

Collapse

Midwest Audio Fest

It’s that time audio enthusiasts! Registration for the 2019 Speaker Design Competition is now open! Visit midwestaudiofest.com for details and to list your speaker project. We are excited to see all returning participants, and look forward to meeting some new designers this year, as well! Be sure your plans include a visit to the Parts Express Tent Sale for the lowest prices of the year, and the Audio Swap Meet where you can buy and trade with other audio fans. We hope to see you this summer! Vivian and Jill
See more
See less

More Ultimate Equalizer measurements - Polar plots

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

  • More Ultimate Equalizer measurements - Polar plots

    I made a set of 0-90 degree measurements for horizontal polar response this weekend. These have to be called preliminary, partly because I was looking more for the dipole response of the midrange and the tweeter response for smoothness in the polar response. My goal was to have the DXT appear to be something like the front half of a tweeter dipole.

    I also "cheated" a bit again. The impulse was windowed so that the lower limit was 500. This smoothed the response a bit from 500-1000, it's not that clean. There are some resonances in the midrange response from the open front that surprised me. The top panel of an H-frame dipole was left out in order to eliminate reflections to the mid and tweeter. It wasn't as good as I'd hoped, but with a top side to the H it's just as bad in other ways. Thick felt on it does improve things a bit. All this from just 1m, of course. I hope to make some changes to improve this. Looks like a damped cardioid bass section may be the better option, since driver alignment won't result in nearly as much diffraction/reflection to the midrange unit.

    The system is horizontally symmetric, so I used the 0-90 for both sides in the plots.

    I used HOLMImpulse to make all of the measurements. I must say that I find HOLM far easier and faster to work with than SoundEasy for multiple measurements. LAUD would be preferable, but I haven't found a way to measure the UE due to its latency. HOLM also has a 50 measurement stack that makes it easy to handle a large number of measurements, very helpful when doing polars. I exported those, edited out the first line in each, then imported them into SoundEasy for the polar plots.

    First up is 500-5000Hz:
    The transition to the DXT (1200Hz) is fairly smooth.
    This is the one that's not really so clean below 1K at the moment, the outer curves actually "wiggle" on all axes.

    Next is 5000-10000Hz:

    Next is 10K-20KHz:

    Finally here's the full 500-20KHz in one:

    Here are the horizontal (* not normalized) response curves, ignore everything below 500Hz:


    I don't yet know if I'll stay with this. The crossover is low, though as LR8 @1200, the DXT ought to handle it for my needs. Much higher than that and the midrange quickly starts to beam, centered around 2100Hz. You'll notice the response above the 1200Hz crossover point is very much like the native response of the DXT on a large baffle. I've sized the baffle and applied felt in a way that seems to have met my goal of a reasonably smooth off-axis to some degree. I'd hoped to cross at 2K or higher, but nothing could fix the 2K beaming.

    One last point. I used measurements with 1/6 octave smoothing for the input to the UE. I didn't want it to be flattening the on-axis response for every little peak/dip, since a fix on-axis may worsen the off-axis. Now that I see the results, I may change that. It looks like the DXT is fairly consistent to about 45 degrees with the felt treatment on the baffle, so the area above 4K could be further smoothed.

    dlr
    Last edited by dlr; 01-11-2011, 08:24 PM. Reason: Correction as noted to last graph comment with a *
    WinPCD - Windows .NET Passive Crossover Designer

    Dave's Speaker Pages

  • #2
    Re: More Ultimate Equalizer measurements - Polar plots

    I'd say that looks darn good Dave. It would be cool to see without smoothing applied, but my guess is it will still look very good.

    Thanks for sharing,

    Dan
    "guitar polygamy is a satisfying and socially acceptable alternative lifestyle."~Tony Woolley
    http://dtmblabber.blogspot.com/
    http://soundcloud.com/dantheman-10

    Comment


    • #3
      The Ultimate Equalizer was essential

      Originally posted by dantheman View Post
      I'd say that looks darn good Dave. It would be cool to see without smoothing applied, but my guess is it will still look very good.

      Thanks for sharing,

      Dan
      Thanks.

      It's not smoothing per se, but the short window has a smoothing effect at the low end. From 1K up there's very little smoothing effect from the window and primarily near 1K.

      My last test before going to bed showed promise. It's a bit unconventional, but may be the only solution other than going to a U-frame cardioid for it's flat front. The resonances were largely damped.

      The better overall solution will be going to a 4-way that does not need the dual 10" drivers. Then the longer dipole baffle with the lower Fc of an upper bass driver will make the problem moot. John's NOTE is a prime example. My next system will be 4-way, already have the drivers sitting on the shelf except a tweeter. A 3-way like the one I show here has more limitations and a somewhat different set of issues.

      To comment again here, I would likely not be doing this were it not for the UE. I worked on this throughout the weekend. The experimentation of the off-axis was key. With the UE, I simply set up the mic at 30 and later 45 degrees and tested literally dozens of crossovers at each angle to find optimal at each angle. It was easy to test a range of Fc lot to high, 2nd order up to 8th, changes for W/M and separately M/T, all knowing that the on-axis response would always be essentially unchanged.

      Eventually the results converged for the 30 and 45 degree testing and was the opposite of my expectation and initial targets. The polar plots then confirmed that the 30 and 45 degree results largely carried over to the rest of the horizontal polar coverage. I could never have done this without the UE without a huge effort in time, especially since even most testing with models is limited in absolute accuracy with regard to the diffraction and driver directionality influence. I was able to do all of this in a weekend.

      dlr
      WinPCD - Windows .NET Passive Crossover Designer

      Dave's Speaker Pages

      Comment


      • #4
        Re: More Ultimate Equalizer measurements - Polar plots

        If you want to push your xover a bit further down in frequency with an LR8 phase response, or if you want a lower order with the same xover frequency, why not try an NTM filter? It'll provide better power handling than the equivalent LR filter and the same phase response, so off axis and lobe position etc should be as good or better.

        http://www.aes.org/e-lib/browse.cfm?elib=9232
        http://sound.westhost.com/articles/ntm-xover.htm

        In UE, it should be simple.

        Dave

        Comment


        • #5
          Re: More Ultimate Equalizer measurements - Polar plots

          Originally posted by DDF View Post
          If you want to push your xover a bit further down in frequency with an LR8 phase response, or if you want a lower order with the same xover frequency, why not try an NTM filter? It'll provide better power handling than the equivalent LR filter and the same phase response, so off axis and lobe position etc should be as good or better.

          http://www.aes.org/e-lib/browse.cfm?elib=9232
          http://sound.westhost.com/articles/ntm-xover.htm

          In UE, it should be simple.

          Dave
          Hi Dave,

          That could be a useful idea for something other than the UE. The targets are limited and fixed, Bessel, Butterworth and L-R, but that's not a problem. The UE EQs the phase, so as long as an L-R is used, even with the highest (LR8), the summed phase can be linear or minimum-phase.

          That kind of target could be used in the full SoundEasy as you can enter your own targets. It's a lot like an elliptic crossover, though the way that they degenerate into Allpass is interesting.

          The two issues that may arise is the off-axis summation. The other is that at some point in the stopband, the output is higher than for the actual Butterworth or L-R of equivalent Fc, though that point may be far enough down to be moot.

          I remember using CALSOD to optimize very early on and only showing maybe 30db down on the vertical scale. Allowing CALSOD to optimize at times seemed to bring it in better than other tries. Later I found the it was converging to a trap that was essentially an elliptic response. At the time I didn't recognize it and considered it to be a bad result, so I expanded the vertical scale and/or set the limits to prevent it. :(

          dlr
          WinPCD - Windows .NET Passive Crossover Designer

          Dave's Speaker Pages

          Comment


          • #6
            Re: More Ultimate Equalizer measurements - Polar plots

            Hi Dave,
            Off axis should be very close to LR as phase tracks the LR and the magnitude is quite close to LR until ~ 10 dB down.

            I've never used the NTM but I plan to figure out a passive implementation and then give it a shot.

            Surprising that UE uses canned targets, that's no fun. I rarely use canned typical acoustic targets. If it uses LMS or RLS techniques, it shouldn't care what the target is. Lets hope its a future feature.

            Edit: I guess if UE supports bulk delay, then canned targets are less a limitation.


            Dave

            Comment


            • #7
              Re: More Ultimate Equalizer measurements - Polar plots

              Originally posted by DDF View Post
              Surprising that UE uses canned targets, that's no fun. I rarely use canned typical acoustic targets. If it uses LMS or RLS techniques, it shouldn't care what the target is. Lets hope its a future feature.

              Edit: I guess if UE supports bulk delay, then canned targets are less a limitation.
              I'm sure that Bohdan could add more crossover types, but the usefulness of additional ones is a bit limited, especially since it is possible to simply go with a high slope and still get flat phase. That's limited to the L-R type crossover, of course. If you need to steer the lobe a bit, you can use the other crossover types, but to be honest, with a flat phase L-R, I can't see much reason to use anything else, steering notwithstanding, unless a slightly different power response is desired, since non-coincident drivers will mean that the power response will be different, despite the linear phase on-axis. But one then has to accept a minimum-phase response as a tradeoff.

              I also wonder just how much trouble any non-aligned drivers will create. I would think that an elliptic filter would be more sensitive to offset, especially as you move off-axis.

              Could you elaborate on what you mean by bulk delay?

              dlr
              WinPCD - Windows .NET Passive Crossover Designer

              Dave's Speaker Pages

              Comment


              • #8
                Re: More Ultimate Equalizer measurements - Polar plots

                Bulk delay would be delaying one driver a fixed amount across all frequency (eg the tweeter).

                It's all about the beam and off axis, on axis is easy. That's why I'd want more filter functions so that I can optimize off axis for my baffle layout.

                For example, I very often use offset BU3s and get superior results off axis to LR2 or LR 4. But, some of this (but not all) is because the drivers are not at equal group delays on axis; that's why bulk delay in UE would make these special types less useful, but certainly not below consideration.

                Comment


                • #9
                  More Polar plots - this time Normalized to on-axis

                  I've corrected a mistake in my original post. The horizontal response family is absolute, not normalized. These are all the same plots, only normalized here. None of the curves is smoothed, but as stated earlier, the window tends to smooth the very low end. Ignore all below 500Hz.

                  500-5000Hz:
                  Again, 500-1000 is not really this smooth.

                  5000-10000Hz:

                  10000-20000Hz:

                  The full 500-20000Hz:

                  Here are the horizontal normalized response curves, ignore everything below 500Hz:


                  Puts it in a slightly different light. There's something of a "broad" flare, but some of that is the measurement resolution at the low end.

                  The DXT rolloff is probably affected a bit by the felt on the baffle sides. Out to about 45 degrees, the tweeter rolloff is fairly shallow. It increases after that, probably due to the felt on the baffle. The direct line from mic to dome is totally blocked at about 70 degrees.

                  dlr
                  WinPCD - Windows .NET Passive Crossover Designer

                  Dave's Speaker Pages

                  Comment


                  • #10
                    Re: More Ultimate Equalizer measurements - Polar plots

                    Originally posted by DDF View Post
                    Bulk delay would be delaying one driver a fixed amount across all frequency (eg the tweeter).

                    It's all about the beam and off axis, on axis is easy. That's why I'd want more filter functions so that I can optimize off axis for my baffle layout.

                    For example, I very often use offset BU3s and get superior results off axis to LR2 or LR 4. But, some of this (but not all) is because the drivers are not at equal group delays on axis; that's why bulk delay in UE would make these special types less useful, but certainly not below consideration.
                    With the UE, you wouldn't need to do this. All you need do is use an L-R with phase EQ for linear phase. As John's pointed out, the off-axis for a linear phase system that degrades somewhat is better than the on-axis for a minimum-phase system.

                    I could post some measured responses for the off-axis that includes the phase, if you care. It only varies a small bit form linear, it's wavy, seems to be mostly a function of the deviation from flat FR only. There is no wrap of phase at all, at least with aligned drivers as I have.

                    dlr
                    Last edited by dlr; 01-11-2011, 11:27 PM.
                    WinPCD - Windows .NET Passive Crossover Designer

                    Dave's Speaker Pages

                    Comment


                    • #11
                      Re: More Ultimate Equalizer measurements - Polar plots

                      Here's an overlay of the on-axis response and the 45 degree off-axis. The off-axis phase only deviates slightly and appears to be a function of the FR non-linearity due to the rolloff of the DXT, though it "recovers" a bit above about 8K.



                      dlr
                      WinPCD - Windows .NET Passive Crossover Designer

                      Dave's Speaker Pages

                      Comment


                      • #12
                        Re: More Ultimate Equalizer measurements - Polar plots

                        Didn't john show that the waveform distortion was better at 45 deg with that one implementation of lineasr phase than the min phase versions on axis? I don't recall any claim about frequency response magnitude.


                        Why would the phase linearization result in better off axis magnitude response? Does it apply different group delay eq to each driver in achieving phase linearization?

                        Comment


                        • #13
                          Re: More Ultimate Equalizer measurements - Polar plots

                          Originally posted by DDF View Post
                          Didn't john show that the waveform distortion was better at 45 deg with that one implementation of lineasr phase than the min phase versions on axis? I don't recall any claim about frequency response magnitude.
                          You're right, he was showing the impulse response. But I would think that it should translate into better frequency domain response as well.


                          Why would the phase linearization result in better off axis magnitude response? Does it apply different group delay eq to each driver in achieving phase linearization?
                          I believe that John said that all EQ was applied individually to each driver so that each matched their target. Any changes added also are applied to each affected driver rather than to the summed response, such as if a BBC-type dip were added. This should improve the off-axis as well since the degradation at that point will be related only to driver directionality, diffraction and driver alignment. That's pretty much what you see in the measurements above.

                          dlr
                          WinPCD - Windows .NET Passive Crossover Designer

                          Dave's Speaker Pages

                          Comment


                          • #14
                            Re: More Ultimate Equalizer measurements - Polar plots

                            Originally posted by dlr View Post
                            You're right, he was showing the impulse response. But I would think that it should translate into better frequency domain response as well.



                            I believe that John said that all EQ was applied individually to each driver so that each matched their target. Any changes added also are applied to each affected driver rather than to the summed response, such as if a BBC-type dip were added. This should improve the off-axis as well since the degradation at that point will be related only to driver directionality, diffraction and driver alignment. That's pretty much what you see in the measurements above.

                            dlr
                            Equalization as well as phase linearization is applied separately for each band pass. If you don't apply phase linearization then you will achieve either a "perfect" LR type (even order) or a perfect "Butterworth" type (odd order) and any driver offset can be corrected with a delay. With phase linearization the driver offset is part of the phase linearization, provided the drivers are measured relative a single mic position. However, in either case, the lobing is the same, and tilting the main lobe can be accomplished by altering the delay. For example, for an LR4 crossover delaying the tweeter more than required will tilt the lobe up.

                            More filter function are not necessary. Only filters which sum flat in phase can be used for linear phase crossover. That means LR types. You could use Duelund type, but what would be the point? The Duelund claim is that all drivers are in phase at all frequencies but this comes at the price of a lower order bandpass for the midrange of a 3-way. And does it really matter if the phase of the woofer doesn't match the tweeter if the woofer response is down 30 or 40 dB? Plus, with linear phase all drivers are in phase at all frequencies anyway. Additionally, the Duelund claim only applies on the design axis so while an interesting concept, it really isn't anything out of the ordinary. Thiele notched crossovers are also in phase. But do you really want a notch followed by a peak in the acoustic response? Most design that use them only use them to get a steep initial slope and then compensate for the resulting incorrect phase by some other offsetting. With DSP yuo can have a high slow at will with an LR6 or LR8. People were using notched filter long before Thiel's. What Theile did was show that complimentraty pairs which summed in phase at the crossover could be developed.

                            FYI, I put together a spreadsheet for the Theile filters several years ago. It does 4th through 6th order and computes the passive components for the 4th order HP and LP. It can be downloaded here: http://musicanddesign.com/codes/4-6t...otched1.00.zip
                            John k.... Music and Design NaO dsp Dipole Loudspeakers.

                            Comment


                            • #15
                              Re: More Ultimate Equalizer measurements - Polar plots

                              Sorry to DLR for going a little off-topic.

                              John, isn't Duelunds claim of all drivers in phase also in error? When I look at the math for the Duelund filter it doesn't take the woofers low frequency- and tweeters high frequency roll-off into acount. Wouldn't that prevent the woofer and tweeter fom being in phase with the bandpass anyway?

                              Best regards,
                              Mogens

                              Comment

                              Working...
                              X