No announcement yet.

Learning Horn Res

  • Filter
  • Time
  • Show
Clear All
new posts

  • Learning Horn Res

    Trying to move forward. I have been using WinISD since I gave upon my White and Bullock lookup tables from Speaker Builder year one. It has usually been pretty close for sealed and ported. As BP-alignments are a bit more complex, trying to master HornRes.

    So, rather than figure out why measured BP-4 of the RSS Dayton's did not match WinISD, thought I would crawl before running and just model a closed box. Loaded the same parameters ( well, they use different ones, but the same WT-II measurement session) into the two simulators.

    I get a match on impedance plot showing tuning frequency does match ( 52 Hz) , but HR is almost double the value. ( 25 Ohms, 42 Ohms)
    Shape of acoustic power is basically the same, HR giving me box reflections, but the center of the hump ( The Titanic 10 has a humped response) for WinISD is 110 Hz, and 70 Hz for HR.

    Now, tomorrow I am going modify the old box to the parameters I have loaded and measure reality, but I am looking for reasons, what I did wrong, what variables may be different etc. as the difference is huge. WinISD is pretty close to the published graphs. Have I screwed up the parameters in HR?

    Click image for larger version

Name:	Capture3.JPG
Views:	182
Size:	161.6 KB
ID:	1461306 Click image for larger version

Name:	Capture1.JPG
Views:	169
Size:	72.6 KB
ID:	1461307 Click image for larger version

Name:	Capture2.JPG
Views:	157
Size:	55.2 KB
ID:	1461308

  • #2
    I can't see the hornresp input box well enough to say much, but it looks like you modeled in full space. WinISD models in half space.


    • #3
      Yes I did, and that would explain the difference in dB, but why woudl that shift the "hump"?


      • #4
        Changed to 2 pi. Changed to 2W to match the 2.86V from HR, and the amplitude is then the same. Curious, @ 20 Hz, the model amplitude is the same. Peak is the same, but 40 Hz different. Subtle shape of the curves are different, left side of HR straighter, where right of Win straighter. So using different parameters. Differences are enough one might pick a different initial crossover point. Most curious. I will see how they compare with ported and then with BP as I slowly figure out where to set what in HR. One of those programs that assumes yo know how to use it so it makes sense only to the experienced where WinISD is idiot level.


        • #5
          Put it into Bagby. Same tuning frequency. Amplitude of impedance peak half way between HR and WinISD, but does not show the HF acoustic roll-off, but flat which I know is incorrect.
          Nothing like finding your stand by tools that one thought were all based on the same equations curtesy Mr. Theil and Mr. Small are not the same. One thing I know differs is the loss factors Small provided. That was one of the guess-lookups from the old table based Bullock method. Bagby does have entry for guesstimates of leakage and absorption loss. I remember another parameter we used that was a rough guess based on the size of the box. Going to have to go back to Smalls AES paper I guess.


          • #6
            I can't comment on the shift of the hump as I don't have your data, as I said before your screen shots are illegible. Where HF roll off is concerned software is only accurate within the range of pure pistonic function, so the only roll off it can show is that from the voice coil inductance. It can't consider any of the other factors, such as radiation by the dust cover or break up mode. Loss factors Ql, Qa and Qp can be considered by WinISD, although IME most users are unaware of that.


            • #7
              Did you click on my attachments using a real computer, or trying to see them on a phone? They are perfectly clear when selected on a computer screen.

              There was another loss factor that had a larger effect related to box size. We guessed small, medium, large to select which column of the W/B tables we used. It was 40 years ago, so I have to go find my notes. I may even still have my copy of Bullock on Boxes somewhere. I have the AES papers from Small somewhere too.

              Leaks can make a small difference, stuffing a bit more, more the Q which is small differences to the slope of the LF roll-off. But same data three tools, three different answers. 70 vs 110 Hz is not a small difference.

              I would just dismiss HR except I really want it's ability to plot the box and port resonances and have a strong suspicion it will give the clue why the RSS as-built is so far off simulated, though in the RSS project, the tuning frequencies were way different too, where all three tools show the same impedance peak frequency even if radically different values for the Titanic sealed case.

              OK, news is over. Second cup of coffee is ready so I'll move to the shop and put the driver back in a real box. Not raining, so I may even measure outside to get a cleaner acoustic response.


              • #8
                Reality Check

                HR Z-Max 40.8 Ohm @ 44.89 Hz It have no Ff or Qtc I could find.
                Peak of response, 66 Hz, 89.4 dB. 20 Hz, 73 dB, 200 Hz, 84.5 dB

                Z-Max 28.9 Ohms @ 45.4 Hz. Fc 47.55 Qtc .622
                Peak of response 110 Hz, 89 dB, 20 Hz 72 dB, 200 Hz 87.1 dB

                Z-Max 33.12 Ohms @ 45.4 Hz Fc 45.55, Qtc .580
                Peak response > 100 Hz ( shows flat line) 89 dB 20 Hz 73 dB

                Z-Max 33.9 Ohms @ Fs 44.6 Hz, Qtc .705

                See graph for actual acoustic response

                Who is correct? None. Who is closest? Well not any of them really. HR on the "hump" center, Bagby on Z max, None of them on HF slope at all believable, though I guess the assumed flat line from Bagby to be closest. So, they are clearly making different assumptions as the T/S parameters are the same.

                Fl and Fa, 10, 100. box tested with no leaks.

                If anything is to be learned, it is why one must prototype and test. I see so many threads about first time builders with beautiful woodwork and no data.
                I used this sub in my office for several years using just the plate amp crossover and a HP cap on the amp to the mains. It worked pretty well. Better than the simulations really.
                Actually, driver in-box is far better than the Dayton data sheet would suggest as it shows the steeper LP response. ( OBE driver but datasheet still posted is you search)

                Click image for larger version

Name:	Capture6.PNG
Views:	110
Size:	13.3 KB
ID:	1461372


                • #9
                  I entered your HornResp data and got a curve almost identical to your WinISD result, with the peak at 110Hz.


                  • #10
                    Was suggested to try UniBox. Results about the same as WinISD for ported and sealed. But going to BP, there is a HUGE difference if you include port resonance. So Unibox and HR seem to be much closer. Lesson is WinISD is not reliable for BP design. The difference is much like what I measured vs simulated for the two RSS drivers.


                    • #11
                      Originally posted by billfitzmaurice View Post
                      I entered your HornResp data and got a curve almost identical to your WinISD result, with the peak at 110Hz.
                      That is most odd then. Why would I get a different result? I would assume I am setting something else incorrectly. Loss on both now set to 30 & 100 respectively for comparison to be the same.

                      HR version 51.2, WinISD .50a7, Unibox 408 , Bagby 5.0


                      • #12
                        Major advance. Selecting "lossy inductor" is what caused the graph peak to shift. So lesson is HR is better suited for drivers with massive motors. Had not bothered me as I had not played with one.


                        • #13
                          Found an Excel spreadsheet for calculating Semi-Inductance over on DIY. It would seem it would resolve the difference in the LP slope where where the simulation shows about a 6 dB roll-off but measured is about 3. I was aware of a difference, but not to this degree. I had incorrectly assumed the effect was at higher frequencies. It is one of the few new topics since I first learned about designing speakers. Circa 2010 in AES, so not covered in the old Anthology or even by D'Apolitto. For sure not back in the mid 70's when I started.

                          Now most of the mysteries have been solved. One has to contemplate that there is a difference in the models but what is the significance?

                          So, we model for purchase decision. Close enough for our use? Even the OEM specs are good enough for that giving them some optimistic leeway.
                          We then measure the real drivers and model to design a prototype. The closer the better, but still, we prototype to see how reality fits.
                          Then we tune the prototype, reverify, and commit to good construction.

                          Considering the 10 inch Titanic driver, using OEM data, one can see it may be suitable for a small sealed box, but not for critical Q. It reaches IB before. ( maybe a candidate for an H-frame di-pole woofer?) I can attest that that is correct as I was able to get .65 Qtc in a stuffed 30L sealed box. One can model for a ported box and see it may be better suited for use with a PR. Looking at the simulation, one might pause if it was too massive and could reach the crossover point or if it would sum well. But by just understanding it is a big motor with high inductance, one could rightly assume it is better than the graph, so is suitable for a 100Hz crossover. So, even OEM specs let us decide: Small sealed box, Qtc in the .65 to .75 range a good fit.

                          By really understanding the physics and the tool's limitations, one can still make good decisions. I feel comfortable still recommending WinISD, UniBox, or Bagby as a basic design tool for sealed and ported designs. WinISD for beginner's as it is so simple and obvious to use. Then if you want accuracy, put the design into HornRes. It shows you the results more accurately. I do not have a conclusion yet for BP-4 or BP-6 alignments other than WinISD was way off.

                          I saw a comment on using S1 and S2 for the baffle size in HR when modeling a direct radiator. I have not looked to see if it has any meaningful effect. HR also lets you pick radiation angle so for SUB simulations you can pick 1 pi, instead of the assumed 2 pi the "simple" tools default to. Again, a warning for those who seem to think they want to build a system flat to 20 in the simulation!

                          HR lets you put in values for lining or stuffing, but where can one figure out the values? I do not know. What is the acoustic resistance of my default PE poly stuffing at what density per pound?

                          Vituix box model also can use semi-inductance. I have not had the time to play with it. Possibly great potential.

                          Does your brain hurt reading all this. Mine does from figuring it out. I have to think about the old song from The King and I, where he contemplated, "It is a puzzlement" Were things I knew were true, actually true?