Announcement

Collapse
No announcement yet.

VituixCAD v2 released

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

  • #76
    Originally posted by kimmosto View Post
    Here we have that kind of software. Have you considered testing or discussion about VituixCAD, or are you just defending your measurement gear, software and methods?

    We could also say that it is possible to design very good speakers without measured phase or without off-axis measurements or maybe without measurements at all. It is just designers choice how much accurate simulation information he/she requires or would like for see to get better forecast in each different case, or how much rely on luck or excellent acoustic basic design or experience or cloning from some other designs - or ears. Corrupting measured phase by default is one step towards blindness and ignorance.
    We have a basic disagreement with your last comment the focal point. There is nothing in using direct measurements that cannot be done equally as well with generated minimum-phase responses in the design process. Both have their uses and it is not blindness nor ignorance. My main system incorporates the Ultimate Equalizer that results in a linear phase system. I would not call the required alteration of the individual driver phase responses to be corruption.

    I will exit this discussion on that note. I can see it will not be constructive.

    dlr
    Last edited by dlr; 11-04-2018, 03:53 PM.
    WinPCD - Windows .NET Passive Crossover Designer

    Dave's Speaker Pages

    Comment


    • #77
      Here is one example. 3-way with 12" PA coaxial + 12" woofer. All far field measurements of higher frequencies are processed with minimum phase calculation for the right side simulation. Then far field is merged to near field measurements of lower frequencies which are processed with diffraction simulation to convert quasi far field. Difference between acoustic centers of coax tweeter and coax cone is defined with excess group delay method. That is simple and fast if separate sum measurement is not available (or shape is too bad for three measurement method). Difference result was 138 mm which is entered as negative value to Z coordinate of coax cone and woofer. Z of coax tweeter is zero.

      As we can see, axial/listening window to 0-20 deg is equal. Both methods are valid for that part. Minimum phase method gives the same magnitude and phase responses as with actual measured phase so difference detection with excess group delay method was okay. But problems start to rise step by step to off-axis, and 40 deg is already ruined at higher XO freq. The reason is that with simple geometry simulation tweeter has rotated to the same distance from mic/listener than acoustic center of cone when off-axis angle is 90 degrees. This error will even increase to rear sector with closed boxes because sign of distance differences swap. This is much better controlled with actual/measured phase.

      Bad result with minimum phase method is fully expected. I suppose experienced designer is able to ignore power response and directivity index simulation with this kind of driver geometry to avoid wrong choices in crossover design. But it would be unfortunate (though not my bad) if someone willingly throw useful information away due to limited measurement gear or method.

      Comment


      • #78
        Kimmo, I've read your documentation on producing polar measurements with ARTA. It would be really nice if we could produce a video documenting that process, because it is quite complex. There are a lot of things you have to pay attention to but I'm getting the hang of it.

        Right now I am developing an Arduino based turntable which will hopefully interface with ARTA to automate measurements. Your work with vituixcad is the inspiration for that effort.

        Comment


        • #79
          I think the first step would be updating document for 'Convert IR to FR' tool instead of polar response export in ARTA. It won't help much because both are easy and fast in that job. VCAD is just more visual and scanning+checking of all measurements is possible with IR and FR windows.

          Motorized table is nice project. I'm still using very simple manual table. That gives enough time to move cables and damping material on the floor between measurements, but ergonomics is usually quite bad. Biggest challenge is vertical plane if that is needed and cabinet is not simple smallish box. Setting of measurement distance is small but very important part when measuring with actual phase.

          Data processing turns more complex if directivity information at low frequencies is bad in gated far field measurements. I have made double merging in my latest projects; simulated directivity at low freq is merged with directivity in far field measurements at about 150 Hz. That process is about three times more complex than simple merging of near field + baffle loss and far field with directivity in far field. Fortunately that is not always needed. I'm using this method to make simulations as truthful as possible, because these are not diy projects for me.

          Comment


          • #80
            Originally posted by dlr View Post
            I would not call the required alteration of the individual driver phase responses to be corruption.
            Short reply to correct possible misunderstanding. Simulation does not give accurate result if phase is altered from reality before responses are loaded into simulator. FIR eq cannot remove that difference. Minimum phase extraction is typically partly guessing and assuming if measured band is not very wide. Measured phase might also have also some errors, but VituixCAD could help with that. It is possible merge cropped and extended minimum phase low frequency part to high frequency part with measured phase.

            Comment


            • #81
              I was planning to stay out, but I'll jump back in briefly.

              First to address this:
              Simulation does not give accurate result if phase is altered from reality before responses are loaded into simulator.
              Flat out wrong if done correctly. You may believe this, but I have done that for more than 20 years as have countless others. I get exceedingly close measured summed response to modeled sum. I will not debate that with you. Done correctly it works. If it doesn't work for you, you are doing something wrong. All one need do is add excess-phase to a minimum-phase model until the phases overlap, but perfectly matching phase is not a requirement to obtain a nearly perfect match in summed response between model and measurement. The latter is what proved to me the validity of this. This is, of course, for typical drivers, not special cases. Those always require special consideration, but that is separate from the issue as I am discussing it.

              John K showed another way to model a driver to generate phase that perfectly matches the measured phase that is a generated minimum-phase response with the required excess-phase added to match the measured phase. He was assisting the developer of SoundEasy and had a lot of influence in it. SoundEasy does make it easy to either measure or import files to then manipulate as needed. It's an iterative process in which the user works with highpass/lowpass driver extension and generates phase until the result nearly perfectly overlays the measured phase. I had never needed to do this because I could perfectly match my model summed response to measured, but I did buy SE due to its ability to audition designs with the digital emulator. I have used his method as well at times, however. Both yield the same results compared to using measured phase directly, which I also do when not wanting to do extensive modeling. I later bought and still use the Ultimate Equalizer which manipulates the phase of each driver to achieve a measurable linear phase, and I do mean measurable. I can show you measured results from 0-90 degrees with on-axis that is linear phase. In fact, these were using my measurements from LAUD, not SE.

              You can then take the results from this scheme, remove excess-phase as necessary, such as mic to baffle front surface, and have minimum-phase files for use in off-axis simulations. Of course it's always possible to find edge cases where there are issues such as with a wave guide or unusual driver, but with typical drivers the off-axis results for the crossover influence itself can be very accurate. Diffraction is always the bigger issue off axis, that's a known situation that has been covered here for many years. We've also known that using measured results to account for diffraction is always the best option, that's not a surprise to any here. I prefer to have that in a separate piece of software because all-in-packages usually complicate things in one way or another, but to each his own.

              FIR eq cannot remove that difference.
              The FIR in the Ultimate Equalizer makes absolutely no use of measured phase whatsoever and is not doing EQ for that. You're making invalid assumptions rather than attempting to understand what I'm saying. The UE has built-in measurement capability, though I have not used it. I still measure with LAUD and post-process the files prior to import into the UE. If you do not have physically aligned drivers (mine are), you would need to add some excess-phase time-of-flight to adjust tor the offset(s). The UE has numerous options for crossovers, both linear- and minimum-phase as desired. But the files of models I created do not have the errors that you assume. If you use the UE alone, you can create an SPL model that matches the measured SPL. The UE does the rest, so the accuracy of the SPL model is key. And since low frequency by the DIY community almost always is limited, splicing is usually a requirement. Matching the modeled phase to measured phase is the best way to assure a close SPL-spliced model.

              dlr
              WinPCD - Windows .NET Passive Crossover Designer

              Dave's Speaker Pages

              Comment


              • #82
                Dave, it seems logical to believe a single measurement on axis would not predict the off axis response due to the driver's construction and position on a baffle. I don't believe there's any disagreement that minimum phase derivation matches as measured phase on- axis or whatever axis you choose to derive the offset. It is the off axis response that this program is simulating from the additional off axis measurements utilizing "as measured phase" with a t0 marker.

                Whether there is value in the extra work is debatable, but the concept using the measured changing phase off axis seems valid.
                John H

                Synergy Horn, SLS-85, BMR-3L, Mini-TL, BR-2, Titan OB, B452, Udique, Vultus, Latus1, Seriatim, Aperivox,Pencil Tower

                Comment


                • #83
                  Originally posted by jhollander View Post
                  Dave, it seems logical to believe a single measurement on axis would not predict the off axis response due to the driver's construction and position on a baffle. I don't believe there's any disagreement that minimum phase derivation matches as measured phase on- axis or whatever axis you choose to derive the offset. It is the off axis response that this program is simulating from the additional off axis measurements utilizing "as measured phase" with a t0 marker.

                  Whether there is value in the extra work is debatable, but the concept using the measured changing phase off axis seems valid.
                  I understand that. If the argument is to be that off-axis measurements are optimal, I fully agree and said so earlier. But the comments go farther than that, or so I read them, which seems to be that generated minimum-phase response is faulty from the start, that the OminMic is not sufficient, etc. I categorically reject must of that. My experience over years confirms that as does that of others.

                  The off-axis will change phase due to both distance (excess-phase) and driver directionality. That's a given. Everyone knows that. But a program that is modeling off-axis based on the on-axis only, it matters not whether or not the phase is measured or generated, only that it's accurate, relative phase between drivers. The argument put forward is that generated phase is deficient altogether. At least that's how I read it when I read "Simulation does not give accurate result if phase is altered from reality before responses are loaded into simulator." This is not a valid working assumption if the user knows how to create an accurate, generated minimum-phase response. This applies to single, on-axis measurements and any off-axis measurements. It's more work, but a large number of DIYers likely aren't interested in the efforts to make an entire set of measurements at off angles. Programs such as mine, PCD, and numerous others are useful as indicators of issues, not as absolutes to off-axis response. They are tools and any tool has it's limitations, they just need to be understood. I've always pointed that out beginning with the first time I included off-axis response to WinPCD.

                  I also do not accept the statements related to position on the baffle. This relates not to driver per se, but to acoustic center and relative offsets. Issues with driver directionality and baffle diffraction are valid, but separate issues. Each can be broken down into their constituent problems. Unusual drivers can always be used to try to make a case, but that does not make the problem an issue for all cases. Quite the contrary, for typical drivers the influence due to driver position is rather easily handled. The off-axis directionality is always an issue, so knowledgeable designers take this into account. Hence why I agree that off-axis measurements are best to understand what's happening there. But if a program takes on-axis measurements and extrapolates to the off-axis, as does PCD and WinPCD among others, driver position (acoustic center, this is) is the smallest of issues and easily modeled. If the minimum-phase responses are done adequately. Diffraction changes will swamp this, but the change due to angle, position and crossover are valid. Experienced users recognize this and find ways to handle that. Just as CALSOD in the 90's as a DOS program had that capability, limited as it was.

                  dlr
                  WinPCD - Windows .NET Passive Crossover Designer

                  Dave's Speaker Pages

                  Comment


                  • #84
                    Originally posted by jhollander View Post
                    Whether there is value in the extra work is debatable, but the concept using the measured changing phase off axis seems valid.
                    Yes. It should be extremely easy to understand that single acoustic offset with Z coordinate cannot serve off-axis simulation in all cases. For example tweeter's acoustic center is 50 mm deeper than woofer's due to wave guide. Those values are detected with any valid method and entered to drivers' locations in the simulation. Loaded frequency responses are processed with minimum phase having excess GD=0ms ie acoustic center approximately at 0 mm. When speaker is rotated to 180 deg in the simulation, tweeter will be 50 mm closer than woofer if simulator plays with simple 3D coordinate system i.e. without calculation of path length from the throat and around front baffle etc. Radiator might also have some non-minimum phase features due to frequency dependent acoustic impedance. In reality, tweeter is still further than woofer if we look whole path of sound wave (instead path of X-rays).
                    So minimum phase extraction could be an option somewhere else. Not here while discussing off-axis because crossover simulation in VituixCAD plays with simple 3D coordinate system and all baffle effects with total path length to origin of each driver should be included in measurement data. Path length calculation exists in Diffraction tool only. That is one possible (but slow and stupid) workaround if user is adding excess delay of close to actual path length to minimum phase processed response to get better results with simple geometry simulation.
                    Once again Measured phase is more accurate, simpler and faster method than minimum phase + path length simulation around possibly complex radiator and baffle construction.
                    Last edited by kimmosto; 11-06-2018, 06:03 AM.

                    Comment


                    • #85
                      DLR and Kimmosto

                      I think that this argument probably has a lot of educational value but I am having trouble really understanding what each of you is saying. Is there perhaps a graphical way that you could illustrate your points to make it more clear?

                      Comment


                      • #86
                        Originally posted by skatz View Post
                        DLR and Kimmosto

                        I think that this argument probably has a lot of educational value but I am having trouble really understanding what each of you is saying. Is there perhaps a graphical way that you could illustrate your points to make it more clear?
                        Great idea, and even more useful to noobs (like me). At a minimum we need to understand the practical implications, what we need to do about it to compensate or achieve (if anything) as part of the design and measurement flow.

                        I'm continuing to solely use VituixCAD (with updates, which are often at present). I want a nice accurate package I'm comfortable with that has as many usable features as possible. The modular nature of this software is akin to how I prefer writing software. There's zero reason to use 15 programs if 1 program provides every feature while ensuring each feature unto itself is written properly, integrates perfectly, doesn't interfere with any other module accuracy. That's part of comprehensive modular software design.

                        Since VCAD provides for it I'm creating library blocks of my actual drivers but further, I'll be using DATS v2 to measure each R/L/C in my existing XO. I'll then create LP/BP/HP tunable blocks. Ideally the end result will be a true representation of reality. Ofc if it doesn't and I've done it correctly something is wonky but not there yet.


                        kimmosto Based on current usage I've got a small batch of recommendations to consider:
                        • Make it possible to actually rename the C / L / R instead of locking it down and only providing auto increment for them.
                          • Example: C1, C2, L1 - L6 etc. When creating an XO using a reference design, the reference (let's say from Xover Pro 3) has all components listed already from C1 through C6, L1 through L4 etc. Being able to rename how these show up inside VCAD diagram would really help cut confusion. As-is, if an exact sequence of creation inside VCAD isn't followed, the names of the components don't (and won't ever) match those in the reference. This makes it tedious in medium+ sized XO layouts to confirm properties, know the right one is in the right location etc.
                          • This is further complicated when components are deleted and recreated since their names auto increment.
                        • Add a rotatable Text / Label component. This way things like Band pass (or anything else) sections can be called out. This would also allow comments to be added directly to the schematic.
                        • Remove the minimum form size parameters you've placed on the UI. As-is, having vertical docked programs, such as VCAD and PDF are tedious to reference because VCAD takes majority of screen space at all times.
                          • It's too bad this wasn't written using WPF. Everything would have been Vector based so scaling wouldn't be an issue at all, among so many other incredible benefits you'd get from Xaml, especially component data binding where changing a single value automatically reflects in all other data context consumers (and calculations) of it.
                          • That's one of the biggest, most dramatic improvements in WPF vs legacy win forms.
                          • If this ever became or becomes open source community project, and I'm still alive, I'd probably personally tackle converting the entire UI to WPF. I'd then figure out how to plug up your existing calculation modules into data sources, and then data bind them to UI elements. The results would be absolutely incredible from a UX perspective.
                          • I'll digress on this though.
                        • If they'd be used / useful internally, provide entire input set of common properties.
                          • Example: Allow entering or loading all the T/S parameters for a driver, especially for lib driver block. Right now I figure it's not usable (or you would have done this) so I'm probably going to just manually create entries for them inside the block as new rows to have as references.
                        Will provide more recommendations as I use the software. The only goal is to make it a better UX and workflow. Loving what you've done already, obviously.
                        Feel free to rip my assumptions apart when wrong, or fix if close.

                        Passive Radiators:
                        All PR(s) Vd must at-least double all woofer(s) Vd. Calc = Sd x Xmax to get Vd for all PR(s) and all woofer(s). If all PR(s) Vd at-least double all woofer(s) Vd they'll work.
                        For woofer(s) with large Xmax vs Sd, all PR(s) with Xmax at-least double all woofer(s) Xmax will work.
                        A PR max weight is said to be its Mms x3

                        PR Systems - tight focus with key parameters.
                        PR Speaker Design - thorough coverage.

                        Comment


                        • #87
                          ^Please do not make long feature request lists. This project already takes too much time and I have daily work and need some time for life out of programmer's box too. Platform and environment changes are out of question no matter possible advantages in UI or UX.

                          Comment


                          • #88
                            ^ That's a shame, but also your choice.
                            Feel free to rip my assumptions apart when wrong, or fix if close.

                            Passive Radiators:
                            All PR(s) Vd must at-least double all woofer(s) Vd. Calc = Sd x Xmax to get Vd for all PR(s) and all woofer(s). If all PR(s) Vd at-least double all woofer(s) Vd they'll work.
                            For woofer(s) with large Xmax vs Sd, all PR(s) with Xmax at-least double all woofer(s) Xmax will work.
                            A PR max weight is said to be its Mms x3

                            PR Systems - tight focus with key parameters.
                            PR Speaker Design - thorough coverage.

                            Comment


                            • #89
                              Originally posted by skatz View Post
                              Is there perhaps a graphical way that you could illustrate your points to make it more clear?
                              This pdf is probably as bad and difficult as my previous explanations, but please do not mind too much.
                              https://kimmosaunisto.net/Misc/MinPhaseVsMeasPhase.pdf
                              As mentioned already dozen times, this story is valid if geometry simulation in crossover simulator is simple: drivers are rotated around common origin, and path of sound wave from driver's origin or acoustic center to virtual mic/ear is straight line (possibly through the cabinet) - not from throat around box edges to back etc. That kind of geometry simulation is the most common.

                              Comment


                              • #90
                                I would like to give some short instructions for users who are "doomed" to use single channel gear and minimum phase extraction. Situation doesn't have to be as difficult or even impossible that it looks according my earlier messages and previously linked pdf. Actions with minimum phase method depend on speaker construction and driver types because single universal method does not exist.

                                First we need to measure difference in acoustic centers. That is best to do off-axis to get better correlation to whole sector 0-180 deg. Next graph shows difference of acoustic centers of tweeter and woofer in 12" coaxial driver (BMS 12C262) by excess group delay measurement with Clio 12.

                                60 degrees off-axis seem to be good direction to detect difference in acoustic centers. Tweeter is 149 mm further than woofer.

                                I have already shown that difference in acoustic centers cannot be entered to location Z mm coordinate of driver instance if enclosure type is closed and both drivers have common non-stepped baffle - such as coaxial driver has. That would lead to crap result for sure due to error in path length calculation with simple geometry simulation. The one and only valid place for acoustic center difference is Delay [us] correction parameter in Drivers tab. That maintains mechanical distance from driver's origin to mic constant while speaker is rotated in the simulation, but adds delay due to difference in acoustic centers to frequency response measurements with minimum phase.

                                New comparison of simulations. Measured phase + zero offset method on the left, and minimum phase + acoustic center difference in Delay offset of driver on the right.


                                Now simulation looks good and truthful also with minimum phase extraction.

                                Next question is what to do if drivers are located physically at different distance with stepped baffle or comparable mechanism, and enclosure type is closed. Then we need to enter distance difference of baffle levels to Z mm parameter of driver instance to get proper geometry simulation. In addition, we possibly need to add some Delay [us] offset for measurements of one driver if acoustic centers of drivers are not at the same depth from baffle surface. Effect of different surface level should be eliminated or subtracted from the result.

                                Next questions are what to do with dipoles, and especially if some drivers in construction are unidirectional and the rest are dipoles. How on earth we should manage polarity inversion to rear sector because minimum phase extraction removes that, and how to delegate difference in acoustic centers for mechanical Z [mm] and Delay [us] offset parameters of drivers? At this point of message I'm willing to select the easiest path and recommend buying measurement gear valid for dual channel measurements and forgetting minimum phase extraction

                                Hopefully this is helpful also for dlr in case WinPCD needs some new parameters and features or instructions for users to get more valid results to off-axis.

                                Comment

                                Working...
                                X