Neat way to find delay, phase, and polarity at xover

[Viper_user]

Hi viper_user, I'm not really understanding what you've done.
What two divers, or sections, are those impulses from? What were the passband ranges they covered?

And what was the stimulus signal...looks like normal sine sweep?


The thing is, with linear view impulses, each single frequency gets one vote as to where peak impulse timing is. So HF/VHF dominates the timing of the linear view impulse peak, and that's not what we need to time lower frequency sections.
We need lower frequency sections impulses alone, representative of the passband they will be operating in....to see where they begin, not peak.

Ime, impulse responses that cover significantly more higher frequency bandwidth than the driver section will be used for, don't work well for time & phase alignments. Unless peering into the driver section's bandwidth-filtered impulse response that decently matches up with the intended passband.

The starts rising from zero, of those filtered impulses need to be time aligned.
Aligning peaks across driver sections causes misalignment........that's what i hoped to show with my prior wavelet post.



Sure, I can catch decent sine-swept impulses at distance in a room, with repeatable TOFs (time of flight, aka excess delay, etc) provided there is enough HF energy...like at least up to 1000Hz.
But for subs, or low mid drivers, repeatability suffers, more and more as passband frequency range decreases.

That's why I'm so thrilled with the single-cycle tone burst for time alignment across sections....its repeatability is rock solid.
I knew i could use the bursts repeatability for xover alignments, e.g. sub to low, with greater accuracy than either phase overlay or filtered-impulses. (aligning burst centers.)

But until recent experiments, i didn't realize how accurate and repeatable the bursts work for timing across driver sections. (aligning the bursts' starts).

Mark you are right this method need HF content. Its a HF burst from two drivers into one mic. But i think it also will set any sub bass waves at the right start?

 

[fluid]

Mark you are right this method need HF content. Its a HF burst from two drivers into one mic. But i think it also will set any sub bass waves at the right start?

It is easy to see the peak from HF energy, when you put a low pass filter on it the peak becomes lower and wider and it makes visual alignment difficult. Using a wavelet input or filtering both sides of the regular impulse to only show the crossover region makes it easier to "see" the alignment or lack of it.

 

[fluid]

Group delay is about delay differences between adjacent envelopes of frequencies. But the length of envelopes covering a given number of tones varies with frequency. I hope i state it clearly.

Mark's speakers have no real group delay apart from where the low frequency rolloff occurs. The group delay that is present in the recording is then presented through the speakers without additional group delay modifying it.

 

[mark100]

Mark you are right this method need HF content. Its a HF burst from two drivers into one mic. But i think it also will set any sub bass waves at the right start?

Like Fluid was saying, without HF content to anchor timing, it gets hard to nail down exactly where in time the impulse falls, as measurement repeatability blows down low, at any distance from the speaker.
And it's especially hard to nail down where the initial rise from zero begins.
Remember that impulse peaks are not what we want to align, so no, it's not like to give the right timings.

 

[JanRSmit]

Hi Jan,
I don't think the nature of music has much to do with how our reproduction system needs to work.
I think the reproduction system's task is to linearly reproduce whatever waveforms are fed into it.

Linear reproduction comes from a system such that a full range impulse, like a balloon pop, gun shot, lightning strike, etc, ....have every frequencies' wave start from zero time and amplitude, at the same time. So envelopes are a by product of that timing, not the timing determinant.

I think we aim for the same: the start of a sound (the "0" is physically the same distance away from the listener).
I am not sure if wavelets show the start of a sound. I thought it shows the group delay, which is not the start of a sound at frequency level. Need to look into that.