Why 2nd Order Is best or not

[DBMandrake]

Originally Posted by john k...

In fact, there are many ways to achieve crossover that has flat phase and amplitude and are thus TP.

What's another example of a crossover using only an analog/causal high pass and low pass section that sums to a transient/phase perfect response ?

Or are you referring to a 3rd driver filler approach or DSP approaches using non causal filters ?

That is not correct. The correction filter will show preringing but the net response will not, at lease not on the design axis.

Interesting, I didn't realise that. I don't know enough about that type of filter to disagree with you so I'll take your word on it.

I would have agreed with you several years ago but these days with the ability to switch between TP and non-TP crossovers while maintaining the exact same frequency and polar response tells me that the phase nonlinearity of a typical crossover of at least 4th order and below is not generally audible except under very specific conditions.

The audibility of a lot of things such as transient perfect square wave reproduction seems to be taken as an article of faith by many but when you do careful testing (or read the research) it's sometimes quite shocking at how "deaf" our ears are to certain classes of defect...

As I was saying in another thread recently, above a certain frequency we have no ability to detect phase at all, (somewhere above about 1-1.5Khz) except by any amplitude summing changes it introduces or perhaps arrival time differences on impulsive sounds if the group delay is really bad.

So if your crossover is well above this frequency and the amplitude response is flat, you're not going to hear the phase shift of the crossover per se, all you might hear is the group delay if it's really bad, or perhaps the filter ringing if it's really bad. A well designed 4th order or lower filter shouldn't have problems with either of these...

Accurate square wave reproduction is one of those things I wouldn't even bother trying to measure or achieve, not when so many sacrifices must be made to come even close to achieving it. Ruining a lot of very audible parameters to optimise one virtually inaudible one is never a good idea IMHO.

[DBMandrake]

Originally Posted by Pano

Electrical, acoustical, or does it matter?

Only acoustic matters

[markus76]

Originally Posted by OnAudio

In nature the most occuring systems are vented enclosures and open baffles.

Human beings and most animals could be modeled as vented enclosures with an adjustable tuned port. Guitars and many stringed instruments also fall into this category of vented enclosures.

A harp among others could be modelled as open baffle.

Thus a sytem that is focused on proper mimicry will employ a variety of technics.

Modeling speakers after instruments is a very, very limiting approach in sound reproduction. What we want to reproduce is a huge variety of aural spaces from bird chirps in a snow covered field to water drops in a huge cave. You have to think "sound field" not "instrument". The latter is only relevant on the production side, not on the consumer side.

[john k...]

Originally Posted by DBMandrake

What's another example of a crossover using only an analog/causal high pass and low pass section that sums to a transient/phase perfect response ?

There is a hole class of symmetric, 2-way, 2nd order TP crossovers that consists of overlapping HP and LP filters and applying EQ. The 2nd order filter characteristics can be used acoustic targets and the eq is best applied using analog active eq. I wrote an Audio-Xpress article on this back in 2002.

Transient Perfect Designer Two Way

Then there are all the subtractive crossovers that are TP which are defined by HP= 1 - LP. These are generally asymmetric at for any order LP filter the HP section ultimately has a 1st order roll off. The inverse, LP = 1- HP is similar but the LP is ultimately 1st order.

These are all casual and can be implemented using analog or digital approaches though some may not lend themselves to passive realization.

[mondogenerator]

isnt pre-ringing due to the type of feedback used in the filter cct? Effectively speeding up the rise time of the attack envelope, u essentially induce resonance? Ive studied this before...to be honest it put me off the whole topology, doesnt it erode the advantage of such filters somewhat?

[OnAudio]

Originally Posted by markus76

Modeling speakers after instruments is a very, very limiting approach in sound reproduction. What we want to reproduce is a huge variety of aural spaces from bird chirps in a snow covered field to water drops in a huge cave. You have to think "sound field" not "instrument". The latter is only relevant on the production side, not on the consumer side.

I propose that the microphone captures incomplete information. If the waveform captured is reproduced, we are not guaranteed that it will have the same propagation pattern or behaviour. That is why we need to give it gentle cues about its true form.

The bird in a snow covered field can also be modeled .

[markus76]

Originally Posted by OnAudio

I propose that the microphone captures incomplete information. If the waveform captured is reproduced, we are not guaranteed that it will have the same propagation pattern or behaviour. That is why we need to give it gentle cues about its true form.

The bird in a snow covered field can also be modeled .

Try binaural recordings with mics inside your ear. The information a microphone can capture IS complete. The only thing inclomplete is the wave field created by the current reproduction chain.

[OnAudio]

Originally Posted by markus76

The information a microphone can capture IS complete. The only thing inclomplete is the wave field created by the current reproduction chain.

So where does it disappear to ?

[markus76]

Originally Posted by OnAudio

So where does it disappear to ?

There are a couple of "black holes" where relevant information gets lost and/or distorted. So current channel-based production techniques don't try to recreate a wave field like binaural recordings do because two-speaker stereo is inherently inadequate for such an approach. Even 5.1 or 7.1 multichannel is inadequate.

[OnAudio]

I am proposing that with proper speaker design, the lost information can be 'recovered'