Help with bridged-t all-pass

Discussion in 'The Projects Forum' started by nlieb, May 23, 2013.

  1. nlieb

    Thread Starter New Member

    May 23, 2013
    6
    0
    I'm working on a rather complex passive crossover circuit for a pair of speakers. It has grown so much that, while I have equalized spl to within +-1/2 of a db, there are phase errors between simultaneously playing drivers up to 77 degrees according to my SPICE simulation. This would be extremely problematic: not only would destructive interference create dead zones for certain tones in certain parts of the room (**edit**I guess this would happen even if the drivers were perfectly in phase, since you'd have two waves emanating from different points in space, but I assume, although, I haven't done the math that there would be some phase difference at which the error would be minimum. That angle might not even be 0 if the drivers were displaced on the axis that the sound was radiating on (these speakers are highly directional). My instinct is that this phase angle would be low, however, assuming the wavelength to be short, and besides my drivers are not so displaced.**edit**), I'm told phase errors this large are audible too.

    I wanted to insert some sort of phase delay circuit, preferably passive, to prevent this from happening. I've been gravitating towards the bridged t delay equalizer since it appears to be most simple, but the circuit I designed with this topology both has an incorrect phase correction AND seems to have a resonance frequency close to the band I'm correcting to boot (although not at it, so I don't think I did something as stupid as inputting the correction frequency as the maximum usable frequency).

    Can anyone explain or tell where I can download a text explaining how to design one of these things?
     
    Last edited: May 23, 2013
  2. nlieb

    Thread Starter New Member

    May 23, 2013
    6
    0
    BTW, here is the circuit diagram. I apologize for how crowded this is, I didn't have the right value capacitors for a lot of these so I was taking note of what values I had decided to use. The speakers themselves are at L1, R3, R4, R8, and R10. The DC resistance on L1 is 3.5Ohms. Also note that all of the inductors except for L7 (which I was planning on using a toroidal ferrite core for) use non-magnetic cores and so have substantial resistances.
     
    Last edited: May 23, 2013
  3. Ron H

    AAC Fanatic!

    Apr 14, 2005
    7,050
    657
    I don't mean to be rude, but that is a horribly drawn schematic.
    Combine all series and parallel components of like type (e.g., series resistors, parallel capacitors). Straighten out unnecessary jogs in wires. Separate components so that text does not overlap.
    Then post it again. I don't have the patience to figure out what you have drawn.

    Attached are SOME of the things you could change to clarify the schematic.
     
    Last edited: May 23, 2013
  4. kubeek

    AAC Fanatic!

    Sep 20, 2005
    4,670
    804
    And also you should make some boxes around the complex parts, for example what part of the schematic are the speakers and what are the filter parts?
    Another thing is, do you have the correct models of the speakers, or are you guessing the values?
     
  5. nlieb

    Thread Starter New Member

    May 23, 2013
    6
    0
    Just realized it would probably be more useful to post my raw bridge-t than my rather messy notes on how to build the circuit it was going to go into.

    This is a SPICE variant, so AC 1 0 means AC voltage, 1 volt amplitude, 0 degrees phase shift.

    R1 is there as a load.

    I just realized it **does** work for the phase shift, I just reversed the polarity on the voltage by accident. The phase shift is correct as shown. I don't suppose there's any way of moving the resonance without changing the phase shift, is there?
     
  6. Ron H

    AAC Fanatic!

    Apr 14, 2005
    7,050
    657
    Thanks.
    When you post an LTspice schematic, it is usually a good idea to also attach the .asc file, so we can simulate the circuit without having to recreate the schematic. In this case, it's pretty simple.:)
     
  7. nlieb

    Thread Starter New Member

    May 23, 2013
    6
    0
    Yeah, I just copied my working model out of LTSPICE (I would use a big boy SPICE, but my Gentoo box isn't available to me right now and besides I never got around to fully setting it up).

    I suppose I should have just drawn one up by hand and scanned it: that's what I'll do next time.

    For two of the five drivers I have values that are frequency and angle specific; for the other three drivers I have virtually no information because they were made in the 1960's except work other people have done, from which I know their sensitivities to current relative to each other. I've done some calculations and it turns out even if I'm off by 3 decibels in my estimate for the sensitivity of the older drivers (which is a worst case scenario), it does not very much change the values of the components.

    This is surprising until you realize that there are five speakers and that at least two drivers will usually be covering any band of the frequency spectrum.
     
    Last edited: May 23, 2013
  8. Ron H

    AAC Fanatic!

    Apr 14, 2005
    7,050
    657
    I simulated your circuit. The peak-to-peak amplitude ripple is about 5dB! Will that work for you?
    Where did you get the design parameters?
     
  9. nlieb

    Thread Starter New Member

    May 23, 2013
    6
    0
    It doesn't. That's my problem. I got the design parameters from Network Theory: Analysis And Synthesis by Smarajit Ghosh, page 612. The section in question is available on the google books preview.

    I solved the four conditions he gives for the reactances on the network and the characteristic impedance, substituting in the desired phase shift and the angular frequency at which that phase shift takes place.

    Obviously I did something wrong.

    **Edit** My solution gives a characteristic impedance for the line of -1 ohm (which I assume means it has a purely reactive load). The load modifies the characteristic impedance, and thus the solution, thus the r-dependence of the phase shift. I then added on a one ohm resistor, which changed the solution. However, removing the resistor seems not to solve the problem - if you try it, you'll see it actually makes it worse. **Edit**
     
    Last edited: May 23, 2013
  10. nlieb

    Thread Starter New Member

    May 23, 2013
    6
    0
    I need to come back to this in the morning, I don't think I'm making very much sense any more.
     
Loading...