Tricks for best crossover phase behavior

[jcandy]

So far, using my chosen design software, I am finding it fairly straightforward to

(1) Build a sensible box, say in an MT configuration (considering baffle diffraction, bracing, all that stuff) whilst consuming tasty beverage;

(2) Measure the individual in-box driver responses (SPL, phase, Z);

(3) Design a crossover, using this data and CAD software, which gives a flat SPL curve;

(4) Measure the SPL of the final design to verify the flat SPL.

However, what I am not sure about is the importance of the phase response and what designers generally consider the best "tricks" to optimize it. For example, presumably a phase that is continuous but changes rapidly is somehow worse than one with nearly constant phase

[Loren42]

Originally Posted by jcandy

So far, using my chosen design software, I am finding it fairly straightforward to

(1) Build a sensible box, say in an MT configuration (considering baffle diffraction, bracing, all that stuff) whilst consuming tasty beverage;

(2) Measure the individual in-box driver responses (SPL, phase, Z);

(3) Design a crossover, using this data and CAD software, which gives a flat SPL curve;

(4) Measure the SPL of the final design to verify the flat SPL.

However, what I am not sure about is the importance of the phase response and what designers generally consider the best "tricks" to optimize it. For example, presumably a phase that is continuous but changes rapidly is somehow worse than one with nearly constant phase

Answering you phase question, I think that the importance is that the phase response is not necessarily flat as much as the rate of change is not excessive or sudden, particularly at the crossover frequency.

Now, a phase response that is 180° out from the driver that it crosses with at the crossover frequency is not bad. However, a 90° phase jump is going to be an issue.

As far as your sequence of steps go, I might embellish a little.

2: The SPL response should not only include on-axis responses, but off-axis responses. Transitions between drivers at the crossover frequency should ideally be the same both on and off axis. The composite response curve should be a smooth transition in the on and off axis response.

3: Flat is nice, but more important is smooth transitions. Variations that are sudden are much worse than variations that gently go up or down and in some cases may even be desireable.

4: Yes, but do not forget off-axis responses.

Lastly, getting accurate SPL response curves is a real challenge. You should always be alert for potential problems with the test procedure and always question the validity of data even when it shows what you think you want to see.

Problems with testing are usually things that you may not consider to be problems. It is important to keep an open mind and think outside the box.

[twest820]

For analog implemenations look to Steen Duelund's papers and John Kreskovsky's transient perfect crossover articles. A linear phase crossover is considerably easier with digital, though; the most recent discussion in this forum is in cuibono's Violet DSP thread.