Announcement

Collapse
No announcement yet.

How to get phase data from a speaker without test equipment?

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

  • How to get phase data from a speaker without test equipment?

    Been searching for this information for a while & haven't quite got the answer for obtaining the phase data for these two drivers:
    SB Acoustics SB17MFC35-4 (6" poly woofer) and the Peerless XT25SC90-04 Tweeter.

    I assume you can get something off of the Response Modeler software, but is it accurate enough? I'm not familiar enough on how to do this, so if it is possible please share on how to do this. I'm just trying to get the two drivers in close phase. Should I just reverse the polarity of the tweeter and look for a smooth but pronounced dip at the crossover frequency to get a good ball park?

    Can the z-offset be measured physically or from drawings or do you really need to use testing equipment to get this? How accurate AND precise does the number need to be?

    Thanks in advance for your help?
    John

  • #2
    Yes a product like SoundEasy can do some fancy math, but it costs a LOT more than the test equipment,
    If you are going to do a decent job, get a DATS for the electrical measurements, the easy way, or use Limp and a little test gig you can make. Cheaper but works fine. So you get impedance and phase.

    Then get a real mic and IO box or a USB mic and use HOLM to get real response and phase data. If you measure both divers in the same setup with the same settings, then their relative SPL wil be accurate. I use HOLM. Import the data into Excel. Then pick a center band value and compute the offset ( HOLM references zero) with the calculated SPL from DATS for the woofer.( I have th older WT II actually) and export the file. Then I get a frd file with the 80 to 89 dB range PSD lite is expecting.

    The only true way I have measured offset accurately is with my old Old Colony Pulser and a scope. I wish companies would publish their baffle to AC as well as the back volume of the driver.

    Offset on that woofer will be about 2 inches. Roughly.

    Comment


    • #3
      I always have to think about the difference between precision and accuracy. Precision is basically how many places after the decimal (in meters) do you need to the nearest sub-mm, mm, sub-cm or cm. Accuracy is how close your numbers are to the actual real world drivers. This picture of this target, helps explain the difference.
      Attached Files

      Comment


      • a4eaudio
        a4eaudio commented
        Editing a comment
        Nice! I may have to save this image.

    • #4
      If you can measure frequency you have the phase, so I'm a bit confused by your comment about looking for a dip in the frequency response. You can use the Hibbert transform in Response Modeler or Response Blender to transform the frequency response to get the minimum phase of each driver. I prefer the Response Blender as you can set the tails and get what I consider a more accuracy derivation.

      I think what you are asking how to find the offset between the drivers. Again if you can measure the combined frequency response of the two drivers you can find the offset. If you can't measure then it's a bit of crap shoot. PCD will do the math if you put in x, y and guess at the z offsets. For XSim use the right triangle theorem and estimate the difference in the path length to the mic. The time delay between the two drivers is the offset in the minimum phase. PCD and XSim use change in distance to calculate the delay.
      John H

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

      Comment


      • #5
        Originally posted by jhollander View Post
        If you can measure frequency you have the phase, so I'm a bit confused by your comment about looking for a dip in the frequency response. You can use the Hibbert transform in Response Modeler or Response Blender to transform the frequency response to get the minimum phase of each driver. I prefer the Response Blender as you can set the tails and get what I consider a more accuracy derivation.
        What if I can't measure frequency because I don't have the equipment? And what if neither the frd or zma files have no phase data? Do you not need the phase from the zma to get an approximate of the right-hand side of the impedance graph showing the phase. Please educate me as I really don't know. Thanks!

        Comment


        • #6
          If you want to build speakers, you need tools and you need to learn how to use them. Otherwise, stick to kits or published designs.
          $70 for a USB mic. Mic stand, cables. Various small parts, like a big cap to protect a tweeter while testing, a dozen nickels, some amp you can use. Order a .1% 10 Ohm Resister for calibration. There is no excuse not to learn to do it correctly. I consider my Woofer Tester ( precursor to DATS) to be the best $100 I ever spent. I prefer my Beringer mic and Focusrite IO box as it gives me more flexibility in measuring at different levels, but the USB mics will do the job. Actually, Linkwitz shows you how to make a capsule mic that is actually more than good enough for just a few bucks. In reality, you only need it flat through the crossover range.

          Table saw, jig saw, router, dust collector, drills, truck to get lumber, etc. Now these get expensive. Kind of like a calculation a spouse did on the cost of deer meet. Came out over $2000 per pound. No one in their right mind ever said this was a cheap hobby. A big part of DIY is the definition of DIY. Like many here, I am also a woodworker, so those costs are amortized across many projects from picture frames to a couple complete kitchen and baths, a dozen pieces of furniture etc. Loke woodworking, there wil be tons of help, but you will find the quality of the help improves as your level of that big "Y" in diy increases.

          As J said, the Hilbert Transform can compute phase from the frd and zma files. Not very well and the OEM files will not be very accurate. I fiddled with SoundEasy and finally threw it all away. I measure.
          Of course, for all I know, my software is just doing it itself fooling me, but it works. Did you search the WEB and download the tools he told you about?

          Go READ the ARTA manual to understand measuring. Read the LIMP and HOLM manuals. I think you can even download the SoundEasy manual. It has tons of technical data in how it all works. Far better than wringing your hands or expecting some one else to do the work for you for free. If you do not have a background in electronics, spend a couple hours on an on-line free class in basic AC/DC electronics.

          Please, I am not picking on you, this posting goes for a lot of folks out there who think they can just buy some drivers and build a pair of $10,000 speakers for $100 with no effort of their own, or the many who believe they can build something smaller and louder than physics allows. " Design me a crossover, too lazy to assemble a kit, 4 inch subwoofer" The list goes on. It takes me a couple of months to build a speaker and I have been at it for 45 years.

          Or you can save all your time and money, go buy a pair of Vanderstein 2ce's and just listen to music. Not perfect speakers, but the quintessential "do no wrong" Cheaper in the long run, but not as much fun.

          Comment


          • kushtekaa
            kushtekaa commented
            Editing a comment
            Thanks for the reality check. Yes, been looking at software, mainly Limp by Arta and REW, plus all the options for mics, soundcards, ... most of yesterday. Sounds like for ease of use get the DATS package? $330 is a little hard to swallow as that will be more than my first pair of speakers. I've already got all the woodworking tools, so that's not a problem.

        • #7
          That PSD-Lite program can also remove excess phase from measurements or extract the phase data from just the amplitude measurement. That feature is somewhat hidden: it's in the "Select Drivers" form under "Tools" and "Response Editor".

          There are actually several methods for extracting phase. The Hilbert Transform is most commonly used, and the Cepstral Isolation is an alternative method that also requires conversions from the time to frequency domains. If you open up the "Processing Options" menu for the Response Editor, there are two links that describe the technical details of each of those methods. PSD-Lite only implements the Cepstral Isolation method--I never bothered to finish off the Hilbert Transform version. However, I have compared those two methods and they appear to give identical results, as they should.

          PSD-Lite also provides a different method for removing excess phase from measurement data. Minimum phase systems have minimum group delay. In PSD-Lite there is a tool that lets you adjust the signal delay and it calculates the group delay. The user can interactively enter the delay, or the code can automatically find the delay that results in minimum group delay. It is a pretty cool method that is very fast, but it requires high sample rates for high accuracy. PSD-Lite uses a fixed sample rate of 48KHz, so the accuracy is limited to delay steps of 1/2" or so.

          I'm not suggesting anyone use PSD-Lite--just pointing out that there are a lot of tools that will generate a phase file from amplitude data. The "accuracy" question is somewhat complicated. The Hilbert transform requires "reconstruction" of the frequency data outside the measurement window for high accuracy. This is done by extrapolating "tails" to the top and bottom of the frequency data. PSD-Lite lets you adjust the length of those tails and provides different smoothing and interpolation algorithms for the input and output data. But I've never done much experimenting with those features.
          Last edited by neildavis; 12-31-2020, 11:51 AM.
          Free Passive Speaker Designer Lite (PSD-Lite) -- http://www.audiodevelopers.com/Softw...Lite/setup.exe

          Comment


          • #8
            There are some articles on designing an x-o without measuring on another site. Some guys like to do this and then measure. The compromises you make are that the factory graphs are not that accurate as mentioned above, they are missing the baffle response and the phase and SPL match between drivers is unknown.

            The short steps would be to trace the frequency response and impedance of each driver to make FRD and ZMA files. Alternatively you can use a box modeling program to generate the impedance file of the woofer. Do the Hilbert transform to generate the minimum phase for the impedance/ ZMA files. Use the Response Blender to add the baffle response to the drivers FR, then generate the minimum phase for each driver. You can adjust the SPL in the Response Blender. In whatever crossover program you use, estimate the delay/ distance offset.

            For your particular project you can ask if anyone has the FRD and ZMA files and their baffle and box size to speed things along.
            John H

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

            Comment


            • #9
              I go through using manufacturer provided FRD and ZMA files or tracing graphs, processing with Response Modeler, then using WinPCD to model the results on my site here:
              http://woggmusic.com/advanced-speake...gn-simulation/

              To John's point above: this will be as accurate as the manufacturer's quality control and measurement accuracy. I've done 2 builds like that, and they were pretty dang close when tested afterward. But the best way to design is to use actual measurement data of your drivers in your design target, provided you get the measurement capability.
              Electronics engineer, woofer enthusiast, and musician.
              Wogg Music
              Published projects: PPA100 Bass Guitar Amp, ISO El-Cheapo Sub, Indy 8 2.1 powered sub, MicroSat, SuperNova Minimus

              Comment


              • kushtekaa
                kushtekaa commented
                Editing a comment
                Love your website! Excellent graphics, easy to read & understand. I'm learning a great deal from it. Thanks for sharing.

            • #10
              Not so many years ago, Zaph's site was the go-to for many people evaluating drivers. Countless speakers were designed using his data and the appropriate software. He has measurements for the XT25SC90. He also measured the 8 ohm version of the SB poly 6.5" in his blog. That might be enough for you to extrapolate the response of the 4 ohm version. You could simply tweak the relative difference in efficiency in the driver setup section of most modeling applications. You'll have to make an educated guess at the driver offset, but a little searching and you'll probably get close enough to build a very nice sounding speaker.

              If you haven't already read Paul Carmody's write-up on this topic, have a look - https://sites.google.com/site/undefi...d-measurements

              * wogg - I hadn't seen this on your site. Very helpful, well written article!
              Co-conspirator in the development of the "CR Gnarly Fidelity Reduction Unit" - Registered Trademark, Patent Pending.

              Comment


              • #11
                Originally posted by kushtekaa View Post
                Thanks for the reality check. Yes, been looking at software, mainly Limp by Arta and REW, plus all the options for mics, soundcards, ... most of yesterday. Sounds like for ease of use get the DATS package? $330 is a little hard to swallow as that will be more than my first pair of speakers. I've already got all the woodworking tools, so that's not a problem.
                Unless you plan to make this a hobby and design more speakers, you do NOT need that package. Note, that it is DATS plus the OmniMic microphone plus the Omni Mic software.

                DATS - This is a plug and play solution to measuring Thiel-Small (T/S) parameters as well as impedance. T/S parameters are used for the box modeling and the measured parameters may differ significantly from the manufactured specs listed on Parts Express product pages. (1) You can do some Google searching and find out if particular drivers are known to be close to mfg specs or have a reputation for being pretty far off and (2) you may be able to find others who have used particular drivers and measured them which might be more accurate than mfg specs. (The reference to Zaph's cite is a good example but he stopped updating his site years ago.) There is a "sticky" on the main page of this forum on how to build your own "jig". I have heard that it really isn't that hard and is cheap, BUT I went with DATS because I could afford it, it is really simple, and I plan on building more speakers.

                OmniMic Software - I don't have this, buy my understanding is that it makes many of the steps in designing a speaker pretty simple. Many people choose to use a portfolio of free DIY programs to design their speakers. I use WinISD, PCD and WinPCD, Response Blender, FPGraphTracer, REW, etc.

                Microphone - You "can" get away with using the $20 imm-6 microphone. It is made for phones so you need to buy some adapter cables to get it to work with a laptop. I know Wogg has used this microphone and I think he still does. The UMM-6 is $70 (USB mic, you likely don't need an additional sound card as most laptops will be adequate). Depending on how much you want to invest, if this is the only speakers you design you could buy both DATS and the UMM-6, use them, and re-sell them on eBay or on this forum's classifieds and probably get half your investment back.

                The quality of T/S parameters listed by manufactures vary a lot so you may end up with a poorly chosen box size and tuning without actual measurements. Do some research on the specific drivers to see if this is likely the case.

                Manufacturers FRD files may be more consistent than T/S parameters, but these are NOT measured in YOUR baffle so also may not give good results. You can use FPGraphTracer and put them into some existing tools out there to simulate your actual cabinet and extract minimum phase, but using mfg frd files and simulation may or may not get you close to what actual measurements would tell you.

                Without actual measurements you would have to estimate acoustic offset of the drivers. You could do this with mfg spec sheet drawings if you are just determining the feasibility of the design, or actual physical measurements if you have the drivers. Simply ignoring acoustic offset in your design may lead to poor results, but putting forth some effort and getting "pretty close" then you are unlikely to hear a difference at your actual listening position.

                Comment


                • #12
                  DATS 3 $129 these days. Keeps going up. I told you you can use LIMP and a couple of precision resistors instead. Cheap.
                  I see they are packaging the Dayton mic with a $20 stand for an insane profit. Search the WEB as there are still out there for $70 or so. Search for "calibrated usb microphone."
                  COME ON , this is DIY.

                  Yes, PAY TO PLAY. Speaker design is an actual engineering discipline. Tools, skills, knowledge.

                  Comment


                  • #13
                    I never bother with phase on impedance data. I'm not sure if it really matters. Some people worry about phase angles / swing.... maybe it causes their amplifiers headaches.

                    In anycase - Jeff Bagbys Frequency Response Modeler / blender does the derivation of phase from an imported FRD (frequency response) measurement.

                    You really should apply your baffle step and diffraction response first to the raw manufacturer files. I think Bagby's tools above can do this too. I use an ancient Excel 2000 tool called Baffle Diffrraction Simulator and Frequency Response Combiner to generate target FRD files based on my intended baffle. the FRC tool allows you to subtract the source conditions (source enclosure / baffle) and add in your target (derived from any woofer modeling tool and baffle diffraction simulator). Then it can derive minimum phase on the target curve... and you are ready for crossover simulation.

                    In terms of Z (driver acoustic centre depth) offset... .you can get very close with physical measurement approximation.

                    Basically - if you are designing your typical box speaker (sealed, vented, PR) and you have a standard baffle (4 corners with chamfers, square or roundover edges) you can use the above tools and method to get close to actual measurements. It all comes down to knowing your source data and measurement conditions and not making errors along the way in deriving the target slope.

                    Comment

                    Working...
                    X