Stepped baffle, but not time aligned..?WTP?

[eriksquires]

WTP = What's the point

I've seen Gravesen and Wilson both do something I thought was very weird. I understand the idea of staggering drivers so they have the same acoustical distance. That's fine. I also understand that crossovers must match in frequency and phase.

What I'm missing is this:

If you are going to go through the trouble of physically staggering drivers, why would you do so and end up reversing the midrange?

For instance, Vandersteen does NOT do this. Drivers are all in positive polarity, and produce a fairly close to ideal step response. On the other hand Gravesen and Wilson (please, no flames about who is better) make several stepped speakers but then end up with reversed polarity mid ranges.

In this case, what's the advantage to making a stepped baffle (or whatever you want to call the idea of having a different mounting plane) if you aren't going to get near-ideal step reponse? I can't believe it's just for looks. My ignorant self thinks "Well, if I am going to end up with a reversed mid-range, I might as well mount everything on the same baffle and save myself the trouble."

Any engineering insights that I'm missing would be greatly appreciated.

Thanks!


Erik

 

[CharlieLaub]

Here's a guess at one possibility: they used a second order crossover. In that case the phase of lowpass and highpass outputs are 180 degrees out of phase (and would create a null). The solution is typically to connect one of the drivers with the reverse polarity so that they are again in phase at the crossover point.

 

[eriksquires]

Here's a guess at one possibility: they used a second order crossover. In that case the phase of lowpass and highpass outputs are 180 degrees out of phase (and would create a null). The solution is typically to connect one of the drivers with the reverse polarity so that they are again in phase at the crossover point.

I double checked in XSim before posting this.

You are right, and I'm shocked, shocked I say! I thought based on the usual polar notation that one would lag by 180 and the other lead, therefore having exactly the same relative phase. I'm clearly wrong. I've been doing the late Dr. Leach wrong.

That is, one would go around the chart clock-wise, the other counter clock-wise and meet. I'm clearly not remembering something correctly.

Thanks!

Erik

 

[eriksquires]

I see, it seems in an ideal second order filter the phase only gone through about half of it's transition (90 degrees) at the meeting point, so they are in fact out of alignment where they meet. It's only long after this point that phase reaches it's maximum lag or lead, and compels the polarity inversion.

Well, I just re-learned something ! But I'm still not sure why I should go through the trouble.

 

[BYRTT]

Great you got that 180º 2nd order filter out of phase sorted out in XSim.

Here another fun study made in XSim based on build in textbook drivers that weird enough relate on being 180º out of phase : )
It's B&O filler driver principle where roll off slope created by deep null for LR 2nd order XO'ed woofer tweeter wired in same polarity create transient perfect system on axis when null is filled by a 1st order band pass mid also wired same polarity. All three drivers end always being 90º apart as with 2-way 1st order systems. In first plot graphs show the great data, then look second plot graphs when mid is pulled out of circuit. There is no system stop bands into this study which will make a difference on phase there, attach XSim dxo file if you want to play around pull mid in and out and study, also text in schematic window have guide how to setup filter slopes.

 

[TMM]

Thread about that (B&O filler) here:
http://www.diyaudio.com/forums/multi-way/88135-filler-driver-ala-b-o-27.html

I think the hardest part of implementing it are finding drivers which will hit the required slopes without bulk distortion, and preventing excessive peaks in the power response due to the drivers becoming in-phase with eachother instead of 90degrees as you move off axis.

 

[eriksquires]

Byrtt:

Very nice design!

Erik

 

[BYRTT]

TMM,
Think you right and also to keep slopes right ctc distance plus drivers relative acoustic center is not for fun, but in theory nice transient system performer from IRR filtering system.

 

[CharlieLaub]

Great you got that 180º 2nd order filter out of phase sorted out in XSim.

Here another fun study made in XSim based on build in textbook drivers that weird enough relate on being 180º out of phase : )
It's B&O filler driver principle where roll off slope created by deep null for LR 2nd order XO'ed woofer tweeter wired in same polarity create transient perfect system on axis when null is filled by a 1st order band pass mid also wired same polarity. All three drivers end always being 90º apart as with 2-way 1st order systems. In first plot graphs show the great data, then look second plot graphs when mid is pulled out of circuit. There is no system stop bands into this study which will make a difference on phase there, attach XSim dxo file if you want to play around pull mid in and out and study, also text in schematic window have guide how to setup filter slopes.

This kind of approach is not at all as practical as it would first appear. The first problem is that you would need drivers that have an extremely wide bandwidth because the slope of the filter (1st order) is so low. It looks like your modeling does not take into account the phase rotation of the driver itself - that's a giant oversimplification for a system that is trying to model phase/time effects. A driver is a minimum phase device. Around resonance and a few octaves on either side the phase is anything but flat, or zero. The same is true within a couple of octaves of the HF rolloff region - the phase starts to change rapidly. The phase rotation will change the phase of the output, which will influence the ability to stay time aligned (time delay and phase rotation are related). I am not sure if your sims include this, but the drivers themselves can have vastly different placement of the acoustic center behind the front edge of the frame. This is why some designers "tilt" the baffle in an effort to improve the (physical) alignment of the acoustic centers. So, while a purely electrical sim looks great, if you don't have those acoustic centers aligned then you won't get your nice time aligned waveform either. Related to this is the fact that as the listening axis moves away from the axis where you have physically aligned the drivers (e.g. on axis) the drivers are no longer aligned and get more and more un-aligned as you move farther away from that axis.

Because of all of this, going to great lengths to time align a system is really not worth it. From what I have read, research shows that the brain "hears" something like the magnitude of the Fourier Transform (the frequency domain representation) of the time domain signal arriving at the ears, and the brain also does a lot of interesting "lumping" of different "arrival times" together (e.g. the first 6 msec is all perceived as the same signal). Alltogether this seems to indicate to me that time alignment makes things look nice on a scope but has little to do with actual hearing of the sound from the loudspeaker, and listening tests with music signals (not clicks, etc) seem to bear this out.

The best of these "time aligned" systems can reproduce a square wave at SOME frequencies on the design axis, but not for other frequencies, or other axes, etc. I have seen designers going to great lengths, for example using stepped baffles, to try and achieve the holy grail of time alignment. Unfortunately they are more likely creating unwanted diffraction from the edge of each step, which will do more harm than the "good" of any time alignment effort. It might be better to look at the big picture and not get stuck over-optimizing this one aspect of loudspeaker design.

 

[eriksquires]

Charlie,

The question wasn't really if time alignment was worth it. Personally the more I read the more I want to go with a DSP crossover.

The question was more about, if I'm going to bother to use a stepped baffle, what's the point if I'm not going to end up with uniform driver polarity.

At the same time, I appreciate all the information you posted, thanks!

AFAIK, Thiel used to and Vandersteen still does attempt to worry about each of these issues with some amount of success. However to be quite honest, at a personal value thing, I have never been moved to buy either brand except out of reputation. Your point that the value of the final product are arguable is well taken.

Erik

(P.S. That's not a cut on modern Thiel, I just haven't followed them in decades)

 

[BYRTT]

CharlieLaub,
You right model is simply the perfect electric slopes and summing as in electric domain, mostly posted because sense eriksquires like playing with XSim and there was relations to what he posted about phase. Regarding phase i changed over time experimenting and now favor FIR corrected XO points so IRR phaseturn only visible at stop bands and the lower the IRR roll off reach is in LF section the better, admit it takes massive DSP gear steering and processing time but ended over time and many exercises to sound best i ever had. Thanks taking time share good points.

 

[Grasso789]

BYRRT, thou are right, computers can do a lot, maybe they should also do the English grammar for thee. Charlie, "we cannot hear impulses, we only hear frequencies" is old, worn, beaten and actually wrong, if thou think about it. Erik, baffle geometry is important, and a zero runtime difference between drivers on axis helps in particular when using crossovers of odd and in particular first order (because of right phase difference in case of all odd order crossovers, and because of lack of tweaking knobs in case of first order crossovers). However what those Danes you keep mentioning do, is more of a personal way thing, because the result is not better than most other stuff. It ain't HiFi anyway.

 

[TMM]

When you boil it all down, impulse response and frequency response (amplitude and phase) are the same thing - linear distortion - just viewed in the time and frequency domains respectively. To have a nice impulse response you need both flat amplitude and phase response, however research shows that flat amplitude response is far more important than flat phase response so a speaker with an ugly impulse response due to non-flat phase can sound better than one with a squarer impulse but not as flat amplitude response.

Impulse correction technique s will correct both phase and amplitude response, but beware that it only improves the sound on 1 axis

Sent from my Nexus 5 using Tapatalk

 

[wolf_teeth]

If it was stepped and LR4, all would be in phase.
Wolf