'' ideal step response '' vs crossover point

[BrokenEnglishGuy]

Hi.

There is some comparasion between the ideal step response vs 2-way and 3-way step response, sure the 3-way looks worse than the 2-way. If someone have the image please post it.

That being said, the step response in a 3-way system improve when the crossover point its lower?
for example 3khz and 450hz vs 3khz and 200hz, the 3khz/200hz will be better?
Thanks

 

[fpitas]

I'm not sure that has much to do with it. Generally, low order crossovers show a prettier step response. However, the higher orders like LR4 give a better polar pattern, and hence often sound better despite the odd step response. If you insist on having your cake and eating it too, use say LR8 and phase-flatten.

 

[NTK]

Here are the simulated frequency magnitude responses, phase responses, and step responses.
Full range - 2nd order roll off at 40 Hz and 24 kHz.

2 way - Same roll offs as full range, cross-over frequency at 3 kHz, LR2 and LR4 cross-overs

3 way - Same roll offs as full range, cross-over frequencies at 200/450 Hz and 3 kHz, LR2 and LR4 cross-overs

 

[ppataki]

Hi.

There is some comparasion between the ideal step response vs 2-way and 3-way step response, sure the 3-way looks worse than the 2-way. If someone have the image please post it.

That being said, the step response in a 3-way system improve when the crossover point its lower?
for example 3khz and 450hz vs 3khz and 200hz, the 3khz/200hz will be better?
Thanks

Here is the step response of the Nubert nuVero 140 in my room
(that is a 3.5-way ported speaker)

You can see that the drivers are separated in time (the peaks and dips within the first 1ms)

And the same using Dirac Live (I believe 2.x at that time, this was 3 years ago) - now the drivers are time aligned

 

[norman bates]

And you hear a difference ?

If not, try the sound of rain, keys gingling, clapping, a wood block, or a snare drum.

Without a live instrument to compare crossover slopes, you would hear a difference, but not necessarily what is the most like the original.

We hear freq response Variation and distortion more easily.

We would never tolerate a CD player or amplifier that cannot pass a square wave, yet we allow speakers that butcher it.

 

[LouB]

And you hear a difference ?

If not, try the sound of rain, keys gingling, clapping, a wood block, or a snare drum.

Without a live instrument to compare crossover slopes, you would hear a difference, but not necessarily what is the most like the original.

We hear freq response Variation and distortion more easily.

We would never tolerate a CD player or amplifier that cannot pass a square wave, yet we allow speakers that butcher it.

Interesting, I never thought of actually comparing a live instrument to a recorded sound. I'd like to try that with a drummer. I played for years standing a few feet from a drummer who hit hard. (and my right ear is worse off than my left for that reason) My home system seems to produce what I think is a real close reproduction of a snare drum. Now I'm real curious as to just how close it really is, but not curious enough to follow through with testing it out.

 

[Michel Payou]

Here are the simulated frequency magnitude responses, phase responses, and step responses.
Full range - 2nd order roll off at 40 Hz and 24 kHz.
View attachment 290652
2 way - Same roll offs as full range, cross-over frequency at 3 kHz, LR2 and LR4 cross-overs
View attachment 290653View attachment 290654
3 way - Same roll offs as full range, cross-over frequencies at 200/450 Hz and 3 kHz, LR2 and LR4 cross-overs
View attachment 290655View attachment 290656View attachment 290657View attachment 290658

What software did you use ? Very nice.
I'd love to try it myself.

 

[NTK]

What software did you use ? Very nice.
I'd love to try it myself.

I used Python for these simulations. Attached are the Jupyter notebooks.

 

[gnarly]

What software did you use ? Very nice.
I'd love to try it myself.

Another software method is to rePhase to generate each the bandpasses you want to sum together, then import them into REW. Use REW's trace arithmetic to sum them together.
Then step, impulse, spectograms, etc, are all available for a look.

 

[gnarly]

I think it's worth mentioning, if one is willing to use complementary linear phase crossovers via FIR,
step response for a crossover at any frequency and at any order....can look textbook ideal,
like a full-ranger or a speaker using only first order IIR crossovers.

And it gets even better.....doesn't matter how many crossovers are used in the speaker design.

 

[NTK]

Here are the impulse and step responses from a linear phase filter with the same frequency magnitude response those as in post #3. (Note: All of the configurations shown in post #3 and this post have the same frequency magnitude responses after summing, but phase responses are different.)

 

[gnarly]

Here are the impulse and step responses from a linear phase filter with the same frequency magnitude response those as in post #3. (Note: All of the configurations shown in post #3 and this post have the same frequency magnitude responses after summing, but phase responses are different.)

View attachment 340851

Hi NTK, Thx for showing that.

For real processing, I use IIR for a system/sub high-pass.
(and never see a need for a system low pass).
Ime, Lin phase xovers only work without pre-ring potential, when fully complementary. So system ends, without anything to sum with, need to be IIR.

Here is the step response using a 2nd order IIR 40Hz Butterworth (instead of lin phase)
And with the 450Hz & 3kHz 4th order xovers you showed in bottom of #3.

You can see step no longer dips below zero, prior to rise/peak.
This is a measurement of the three sections summed together electrically.....not a sim.

 

[Tim Link]

Hi NTK, Thx for showing that.

For real processing, I use IIR for a system/sub high-pass.
(and never see a need for a system low pass).
Ime, Lin phase xovers only work without pre-ring potential, when fully complementary. So system ends, without anything to sum with, need to be IIR.

Here is the step response using a 2nd order IIR 40Hz Butterworth (instead of lin phase)
And with the 450Hz & 3kHz 4th order xovers you showed in bottom of #3.

View attachment 340958

You can see step no longer dips below zero, prior to rise/peak.
This is a measurement of the three sections summed together electrically.....not a sim.

So if I'm understanding this right, the case where it dips below zero prior to peak rise is the case where there is linear time alignment of all frequencies - linear phase. The case where it doesn't dip before peak and then goes straight up is minimum phase, which means all lower frequencies come in progressively later. Is that correct? I always thought this asymmetrical kind of step response represented linear phase, and it was very confusing to me. The dip before the peak in the step response has nothing to do with pre-ringing, or ripple. It's just a consequence of time alignment of all frequencies. Bass are longer than treble, and if all frequencies are time aligned the edges of the longer frequencies will stick out in front and back when they're all time aligned. I think the confusion is that with the "ideal" check mark looking minimum phase result the waves are all starting their cycle at the same time, which means the lower frequencies finish much later. All this time I thought Dunlavy and others were shooting for linear phase with their first order networks, but they aren't. They're shooting for minimum phase. That is not perfectly accurate, but should be within our hearing tolerances to sound natural.