Announcement

Collapse
No announcement yet.

Xsim and Phase Questions

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

  • Geoff Millar
    started a topic Xsim and Phase Questions

    Xsim and Phase Questions

    Hello from Australia

    I use Xsim for mucking around with crossover designs, but am curious about 'phase' and how it's measured, particularly as it's so important in speaker/crossover design. What I've read suggests that you need to use measuring equipment and software to get accurate and useable data.

    Xsim has a check box in which to derive phase from driver frd and zma files when you "Tune" the drivers, but I'm not sure how accurate the phase figure is, given that information about 'acoustic centres' isn't entered.

    Is it possible to measure or assess phase without measuring equipment, to a standard from which you can design a reasonable crossover?

    I have built a speaker using Xsim to design the crossover (with much help from PETT members) but that was an adaptation of an existing commercial DIY design - presumably with correct phase already calculated in the original crossover..

    Your thoughts greatly appreciated!

    Thank you

    Geoff

  • 4thtry
    replied
    Originally posted by bwaslo View Post
    Be a little cautious with Xsim3D, as it has some bugs in it. It never got past the beta stage, too many complications about handling off-axis responses and baffle effects, and since kimmosoto had done a lot of that for his sim app, I just gave him Xsim code to do arbitrary circuit configurations. (I still prefer the ease of tossing together Xsim simulations, but then I'm more used to it, too).

    On the subject of 'derived' (or Hilbert) phase responses, I usually prefer to use measured phase as it isn't so subject to noise problems. Hilbert phase depends on the response of the magnitude response shape over ranges rather far from each frequency it calculates phase for, which too often falls into where driver response drops down into noise and starts to become unreliable. Measured phase in strong enough response areas isn't bothered by out of band weak responses. But when using tools like baffle simulators to modify response data files, Hilbert phase is about the only way to go, just make sure that the 'tails' you use to come up with 'clean' magnitude response curves in the out of band areas are reasonably accurate.
    Thanks, Bill. Going forward, I'll try to stick with the standard version of XSim. I like its speed and simplicity, allowing me to quickly load measurements and build crossovers without the need to extract minimum phase. Polars slow me down a little, but the modelling accuracy is always very good, even when I measure and model odd looking baffles with tapered roundovers (i.e., my SideTower's baffle mod). Thanks much for giving us such a great tool.

    Bill.

    Leave a comment:


  • bwaslo
    replied
    Be a little cautious with Xsim3D, as it has some bugs in it. It never got past the beta stage, too many complications about handling off-axis responses and baffle effects, and since kimmosoto had done a lot of that for his sim app, I just gave him Xsim code to do arbitrary circuit configurations. (I still prefer the ease of tossing together Xsim simulations, but then I'm more used to it, too).

    On the subject of 'derived' (or Hilbert) phase responses, I usually prefer to use measured phase as it isn't so subject to noise problems. Hilbert phase depends on the response of the magnitude response shape over ranges rather far from each frequency it calculates phase for, which too often falls into where driver response drops down into noise and starts to become unreliable. Measured phase in strong enough response areas isn't bothered by out of band weak responses. But when using tools like baffle simulators to modify response data files, Hilbert phase is about the only way to go, just make sure that the 'tails' you use to come up with 'clean' magnitude response curves in the out of band areas are reasonably accurate.

    Leave a comment:


  • 4thtry
    replied
    Originally posted by hoxuanduc View Post
    4thtry,

    Your tutorial is most excellent!!! Thank you!!!

    I've always done 2-way as I've been struggling with how to measure the woofers...

    Here is another tip... If you Xsim3d which supports active filter you can 'fake' a target curve by setting up idealized dummy drivers in parallel with your real drivers, That idealized driver has flat response @ 80db. Then you can apply the active filter (eg LR24) and compare the response of the filtered idealized driver with our filtered real driver.

    Thanks,

    Duc
    Thanks for the tip, hoxuanduc! I've been using the standard 2D version of XSim and was not aware of this feature. So, I loaded XSim 3D and set up two "fake" target curves using idealized dummy drivers. Once I had the two idealized dummy drivers and active filters set up, I could quickly switch target curves back & forth from high to low pass, change the frequency, change the filter type, etc., as I worked. This works quite well.

    Bill

    Leave a comment:


  • dcibel
    replied
    You can reconstruct the frequency response when a cap is used, basically apply an inverse transfer function of the series cap with your driver impedance. But as Jeff stated above, it's really not worth the effort, at the low level needed for a decent frequency response, a protection cap generally isn't needed, just be diligent at turning off your amp when you change cables and connections.

    I'll also add that Omnimic software provides some response files to reconstruct the correct response when the bass-removed or bass and mid removed sweeps are used.

    Last, when using a system with relatively accurate absolute SPL like Omnimic, the sensitivityat 2.83V can be easily determined when measuring at any SPL, all you need is a multimeter to measure the voltage of the measurement level (use 50Hz tone track).

    Leave a comment:


  • Jeff B.
    replied
    Originally posted by dlr View Post
    As an aside, I never use a cap when measuring a tweeter, but I use LAUD (also developed by Waslo) that is calibrated and normalizes to 2.83 @1M, so lower power is adequate to get a good measurement of the nominal 2.83V response.

    That said, if a cap is used when measuring a tweeter and you are measuring with a two-channel system, you should connect the feedback channel after the cap. This will bypass the cap in the same way that it bypasses all upstream components (pre-amp, amp, cabling, etc.) and provide an accurate driver response. The only difference will be that at lower frequencies, in the rolloff area of the driver, the signal/noise ratio will be affected, but is unlikely to be a problem.

    dlr
    Very good point on the connection. And I don't use a cap on tweeters either.

    Leave a comment:


  • dlr
    replied
    As an aside, I never use a cap when measuring a tweeter, but I use LAUD (also developed by Waslo) that is calibrated and normalizes to 2.83 @1M, so lower power is adequate to get a good measurement of the nominal 2.83V response.

    That said, if a cap is used when measuring a tweeter and you are measuring with a two-channel system, you should connect the feedback channel after the cap. This will bypass the cap in the same way that it bypasses all upstream components (pre-amp, amp, cabling, etc.) and provide an accurate driver response. The only difference will be that at lower frequencies, in the rolloff area of the driver, the signal/noise ratio will be affected, but is unlikely to be a problem.

    dlr

    Leave a comment:


  • jhollander
    replied
    True, my off handed recollection was with a 125 uF cap, but even there the tail moves enough to change the phase.

    Leave a comment:


  • Jeff B.
    replied
    Originally posted by jhollander View Post
    Bill, the phase problem can be corrected by deriving minimum phase if you still want to use the cap.

    Oh just be careful not to mix minmum phase files with as measured phase files...
    That's not really correct. If the cap is causing the driver to roll-off at 18 dB/oct instead of 12 dB/oct, then the extracted phase will be incorrect too.

    Leave a comment:


  • jhollander
    replied
    Bill, the phase problem can be corrected by deriving minimum phase if you still want to use the cap.

    Oh just be careful not to mix minmum phase files with as measured phase files...
    Last edited by jhollander; 08-10-2019, 06:32 PM. Reason: Added the bit about not mixing min ph with as measured

    Leave a comment:


  • 4thtry
    replied
    Originally posted by Jeff B. View Post

    My suggestion is - Unless it is a true ribbon (with a transformer) don't measure with a cap on the tweeter.
    Thanks, John & Jeff, for your feedback & tips regarding my use of a protection cap in series with the tweeter. Your comments prompted me to set up OmniMic & test a tweeter with and without the 33uF cap to see just how much of a difference this really makes (see attached screen shots).

    Using ohm's law for the capacitance, the half power point for a 33uF cap, assuming 8 ohms, is 603Hz. But it reduces the response in the 1-2k area by roughly 2dB with a phase shift of about 20 degrees! I also tested 50uF (-3dB of 398Hz), 80uF (-3dB of 248Hz), and 125uF (-3dB of 159Hz). As you can see in the 2nd graph, I need to increase the cap to at least 125uF for it to have little to no effect on my measurements.

    Going forward, I think I need to modify my testing procedures to eliminate this cap completely, except, as Jeff noted, in the case of a transformer coupled ribbon.


    33uF VS No cap No Cap VS 33, 50, 80, or 125uF

    Leave a comment:


  • Jeff B.
    replied
    Originally posted by Murphy-Pie View Post
    Which means you can never get good phase data from Omnimic. If accurate phase is important to you then using something like ARTA in its semi-dual channel configuration or Holm Impulse and its "time lock" function are better ways to go. But most people find the phase data in Omnimic close enough for government work.
    Omnimic's phase is a form of minimum phase, but you still have to find the relative offset to design a crossover, unlike a system like Holm Impulse that can set a fixed time marker.

    Leave a comment:


  • Jeff B.
    replied
    Originally posted by jhollander View Post
    Just remember that the cap with the tweeter shifts the tweeter phase. The as measured phase will be shifted.
    My suggestion is - Unless it is a true ribbon (with a transformer) don't measure with a cap on the tweeter.

    Leave a comment:


  • jhollander
    replied
    Just remember that the cap with the tweeter shifts the tweeter phase. The as measured phase will be shifted.

    Leave a comment:


  • hoxuanduc
    replied
    4thtry,

    Your tutorial is most excellent!!! Thank you!!!

    I've always done 2-way as I've been struggling with how to measure the woofers...

    Here is another tip... If you Xsim3d which supports active filter you can 'fake' a target curve by setting up idealized dummy drivers in parallel with your real drivers, That idealized driver has flat response @ 80db. Then you can apply the active filter (eg LR24) and compare the response of the filtered idealized driver with our filtered real driver.

    Thanks,

    Duc

    Leave a comment:

Working...
X