You have to take the varying impedances, impedance phase, acoustic phase, power response, frequency responses, physical offsets, acoustic offsets, cabinet alignment, baffle step correction and diffraction, listening axes, off-axis responses, position in the room, size of the room, distance to the listener, etc all into account to get a good summation between the drivers used in the cabinets and your ears at your position.

While that is said, there is nothing wrong with experimenting, learning, or simulating for a best fit of most of these criteria. There is always a set of compromises and goals for a speaker project....

-Keep the impedance-phase between the 40 degree regions if possible. More than 60 in the negative is very capacitive for amplifiers. Positive impedance-phase is inductive, and negative is capacitive. Blue-line lower right:



-Keep impedance above 6 ohms for an 8 ohm nominal design and above 3.3 ohms for a 4 ohm nominal design, unless for a **very minute** region of bandwidth. No lower than 2.5 ohms in a 4 ohm design ever, in a short area, IMO.

-Spacing drivers closer helps with off-axis response. If possible, try to xover the adjacent drivers below the CTC (center to center) wavelength equivalent frequency.



-Offset tweeters can help the diffraction ripple, and make the response less troublesome. Offset in mirror-imaged pairs.



-Acoustic physical offsets are where the sound is emitted from the drivers, and if the woofer and tweeter are mounted on a flat baffle, these are misaligned, and should be accounted for. Stepped/tilted baffles can alleviate this issue, but present more design tradeoffs. These are typical ROT figures for AC offsets, but measuring the drivers with waveform analysis can give you more accurate results. (Using these methods on woofers, I have fudged numbers down 5mm a few times to account for longer throw drivers and maybe a more forward acoustic origination point. YMMV)




-There are several types of xover optimization, and these are the most common;
phase-coherent
transient perfect
flat power response

*Phase coherent is the easiest to model in a simulation, and just overlapping the 'vertical section' (or phase-wrapping section) of the drivers' individual phase responses is the basic method, all other things mentioned above considered. See first simulation picture here for an example.

*Flat power response (PR) is not where the PR is a horizontal line, but a smooth curve at a downward tilt towards upper frequencies with no scoops, bumps, etc.

*I haven't designed a TP system, please ask Jeff B. or Curt C. to help there.

-As to the physical AC offset, the xover being asymmetrical in electrical order by one position allows a better phase coherence, as does using an LCR shunt on a midrange in a 3-way. Using a 12dB/2nd on a woofer and an 18dB/3rd on the tweeter for instance.

-For a cabinet response, there are several ways to do this.
*EBS (extended bass shelf) is for a vented extended alignment that has a troughed response and a low tuning. This is less plagued by room gain, and sounds a bit less boomy.

*Flat is flat to the knee of the rolloff, and then tapers to 12dB for sealed, or 24dB for vented.

*Some use a more gradual rolloff with a higher F3 (90-100), and a bit of a hilled-boost in the 120 Hz range to make them seem larger, but not tubby.

-FR taming-
Inductors (L) are logarithmic devices, and increase impedance above the freq of cutoff (Fc). Capacitors are antilogarithmic devices, and increase impedance below the Fc. Placing these components in series causes the typical 6dB first order filters of lowpass and highpass respectively, but when placed in a shunt, act just the opposite. This is why a cap in series and a coil following in shunt are both high-pass components that steepen the rolloff together. These components interactions are the reason why the drivers' respective ranges remain at their nominal impedance without resistances added into the xover mix. Resistors adjust damping and attenuation levels.

*In terms of frequency response, you have some abhorrations that need tamed to sound right. A peak in the woofer's response due to breakup, can ring incessantly like a bell, unless electrically damped. If this is not tamed enough, but still damped some, it can cause a dip in the tweeter's response at this same frequency. Best to tame to at least -25 dB below reference for best results.

*In-bandwidth problems typically use a 'parallel LCR notch' (placed in series), and the out of bandwidth uses a 'series LCR notch' (placed in parallel with the driver as a shunt). Sometimes impedance issues can be problematic with the shunt method, and a 'parallel LCR' can be used. R will vary with the amount of damping required.

*If you have a tweeter with either no ferro-fluid, or low damping in its chamber, it can sound very nasal, or cause a dip in the woofers' response if not tamed. More than likely, the shunt method is used here, but sometimes a simple parallel resistor can benefit enough to save in cost.

*A zobel on a tweeter can be used to tilt down it's rising response, as will an LR parallel filter. The opposite types will 'lift' a decreasing response.

*A small cap in parallel with the main lowpass coil will steepen the slope without more costly components. These are called 'tank' filters, as they tank the response due to the resonance between the L and C. Be advised to use a resistor in series with the cap when across the coil to avoid low impedance issues above the audible range. This can cause oscillations in amplifiers and premature failures.

*Sometimes just adding a 1-4 ohm resistor in a shunt with the previously placed coil or cap will be what is needed to make the knee of the rolloff orderly. This is a damped circuit, and is used in conjunction with the tank above in this following graphic for the woofer. Using a resistor in the shunt like this after the tank makes the series resistor from the tank less noteworthy:



How to tell if you have an electrical error or problem with your layout:

Just picture yourself as an 'inaudible high frequency', and run your circuit taking the path of least resistance. You can pass through caps, but you get resistance at resistors (or impeded flow), and rolled off or impeded at coils. Drivers are seen as relative frequency units. They can go through tweeters, but that is also resistive. If you cannot pass the entire circuit from positive to negative unimpeded, you are fine.

The same works for coils and seeing yourself as a DC signal, or very low frequency. You don't want to see yourself pass through nothing but coils from positive to negative. This will usually throw your amplifier into protect mode.




*Most use 'least resistance' (DCR) in the coils for the woofers' lowpass, but sometimes you can use a higher DCR coil to help tame a bump or wide-hill in its response, while keeping things less costly.

-There are 2 common types of xover network, and are as follows:

*Series crossover (SXO)*


*Parallel crossover (PXO)*


Both are feasible means to a common goal, but the SXO is harder to tweak. If you change one component in the woofer section, the tweeter response changes also. You can sometimes use fewer components in one form over the other, depending on the drivers used and their FR behavior.

On the subject of 3-ways, generally the ROT is for the xover spread to be 3 octaves; ie- 800 and (1600/3200) 6400 would be a 3-octave spread. In today's building simulations, you can get by with much lower, but lower than 2 octave spread is really just a filler-driver and a special application.

If you have trouble integrating a mid with a tweeter in a 3-way, you can flatten the mid's impedance with a zobel or tame its breakup with an LCR shunt. Even some mids require an Fs comp using an LCR to get them to integrate well. All of these could help integration in a 3-way.

SXO's more often than not 'short-around' the driver that is not involved in a range of frequencies, whereas, PXO's increase resistance where a driver is out of bandwidth.

--------------------------------------------
There are lots of things I could say in addition to this blog posting, but I have to draw the line here to keep it reasonable in length.

Hope this helps you...........