Well I suppose the shallow vs. steep argument will just go on and on

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
I didn't know someone who may be more analytical in his designs would prefer more shallow slope filters. I seem to find more and more evidence to support my case as time goes on.

Well from the man himself - Zaph.
Zaph|Audio - ZD5 - Scan Speak 15W8530K00 and Vifa XT25

Part of the magic of this system is simply due to the fact that it's a solid LR2 design. Accurate 2nd order slopes are hard to pull off, and only very wide bandwidth drivers with smooth response and low distortion need apply. Make no mistake that shallow slope crossovers sound better than steep slope crossovers. But doing shallow slopes right amounts to much, much more than just throwing a cap and coil on the tweeter and woofer. My Waveguide TMM design was more of a hardware solution addressing some typical LR2 design issues. This Scan Speak / Vifa system is more of an electrical solution.

In the end, an LR2 design has 180 degrees of phase wrap through the crossover while an LR4 has 360 degrees. The lower phase wrap directly equates to an improved midrange coherency. Honestly, most drivers and system designs require LR4 or greater slopes just to work properly. But when everything comes together for a LR2 system, it's the sweet spot in speaker design.
 
Well I don't know about arguing about it but whatever works is fine with me. I'll never be able to low pass a metal cone bassmid with a single inductor only, and I surely won't put a 3rd order electrical on a nice paper or poly cone unit. I need steeper order HP on tweeter for phase alignment anyway so it's minimum 2nd order electrical, and I would hate to fry the delicate unit, so it stays 2nd order or higher.
 
We'd all like to use shallow filters, no doubt. Even none at all like your headphone driver which is just a glorified plastic tweeter. They sound good on detail.

But a loudspeaker needs big displacement in a room to do bass, so we end up multiway.

Problem is cone breakup, non-linearity caused by excessive excursion, and strain on tweeters, leading to distortion at certain frequencies. And some dispersion issues, I suppose. Not too hard with a good 5" bass and a 3.5kHz crossover. 12dB/octave LR2 will do. But don't forget the BBC did famously well with BW3 on the LS3/5A.

But these speakers don't go loud or bassy. Phase alignment is what we often want in speakers too. But phase linearity is not such a big deal. The main event is actually group delay, which worsens with higher orders and higher Q. Because in our Universe the ringing always follows the impulse. The arrow of time always points forwards.

That is what you hear. And the tradeoff is really that reducing distortion involves worsening group delay. Choose your poison! :D

Much incomprehensible stuff here for the interested student: Crossovers

Well, actually I did this at College. But I have forgotten most of it now. We now have speaker simulators.
 

Attachments

  • Klippel Metal Cone Breakup.JPG
    Klippel Metal Cone Breakup.JPG
    40.4 KB · Views: 1,188
  • Duelund a=4.JPG
    Duelund a=4.JPG
    60.7 KB · Views: 192
  • Duelund_3way_tg1.jpg
    Duelund_3way_tg1.jpg
    37.1 KB · Views: 1,492
  • S7 4th order LR4 Theoretical.PNG
    S7 4th order LR4 Theoretical.PNG
    20.8 KB · Views: 1,146
  • S7 Third Order BW3 Theoretical.PNG
    S7 Third Order BW3 Theoretical.PNG
    18.4 KB · Views: 1,797
  • S7 Second order LR2 Theoretical.PNG
    S7 Second order LR2 Theoretical.PNG
    17.9 KB · Views: 1,160
Last edited:
No-one uses images like I do. That is because I have a very visual imagination. Not mathematical, not wordy.

Here's something I am listening to right now, and that is a SEAS 19TAF/G 3/4" metal tweeter:

599660d1487267698-restoring-monitor-audio-r300-bookshelf-speakers-monitor-audio-r300-md-resurrected-cabinet-jpg


Well-behaved 8" paper bass on second order electrical has goodish natural 4th order LR4 rolloff, but is not without problems. Time-alignment is difficult. It breaks up around 3kHz, as you can see on the phase plot. IMO, this produces mushy and annoying 2nd harmonic distortion at 6kHz. So a notch filter (15R plus 0.68uF across 1mH) takes those out quite considerably. By 10db here. Which is a lot.

But you might as well use a steeper third or 4th order filter with the benefit of hindsight. As Harbeth do. LR2, BW3 and LR4 happen to be elegant solutions in filtering. As I see it, phase and group delay and dispersion issues are not distortion, they are more a transform or tonality with a slight ringing quality. Distortion or breakup however IS fatiguing. And you usually cure it with some steepness, and keeping drivers in their comfort zone.

So like Dave says, context is important. A 5" bass is an entirely different animal. The breakup is much higher, around 8kHz. So LR2 might sound OK with a well-behaved driver. And 5" polycones are the easiest example.
 

Attachments

  • s7 Notch filtered 6kHz 8 inch plus 1 inch.PNG
    s7 Notch filtered 6kHz 8 inch plus 1 inch.PNG
    17.7 KB · Views: 596
  • s7 Notch Filtered 6kHz Two way 8 plus 1.PNG
    s7 Notch Filtered 6kHz Two way 8 plus 1.PNG
    7.8 KB · Views: 1,234
  • s7 Electrical effect of 6kHz notch.PNG
    s7 Electrical effect of 6kHz notch.PNG
    14.5 KB · Views: 149
  • s7 Phase with 6kHz Notch.PNG
    s7 Phase with 6kHz Notch.PNG
    19.8 KB · Views: 129
To be fair there have been audio files posted where you can compare the difference between 2nd and 4th order filters.

These are files where an audio sample has been taken, digitally filtered and therefore split into two separate streams. One of the streams is a low passed version, the other a high passed version. The two streams are then recombined back into a single file.

Basically with the above the amplitude of the input and output remains exactly the same, only the phase shift, associated with the different filters, is left present.

Various filters orders were provided. I couldn't hear any difference between the different slopes. Nada, zilch, nothing. If there was any difference it was so minuscule as to be irrelevant.

This also ties in well with my own experience. Using suitable drivers I have directly compared 2nd order vs 4th order Linkwitz Riley xovers. This was done via a DSP where both sets of filters were stored in memory and could be swapped between at the touch of a button. Switching between the two took such little time that you couln't even hear an interruption in the music except for a small 'click' or 'pop'.

Could I hear any difference? Short answer yes. But only because the vertical off axis, from one design to the other, affected the power response ever so slightly and this resulted in a tiny difference in tonal balance. You really had to listen hard to be able to hear it and without the ability to quickly swap between one and another it would have been impossible.

My conclusion from all of this is basically along the lines of - use whatever filters are necessary to get the job done. Striving for certain filter orders, because they somehow contain audio magic, when any kind of compromise is going to introduced by lowering the filter order, is just silly.
 
Well, I suppose we'd like to know the exact drivers, but an interesting test by 5th element.

I built 3kHz all-pass op-amp filters years back. We could never hear the group delay overall.

642341d1509050153-suppose-shallow-vs-steep-argument-duelund_3way_tg1-jpg


The current big wheeze is time-alignment: 18W-8434G00

It seems to make sense to recess the top end to even out the group delay as much as is possible. And this must help phase, because as a lot of people know, ideal LR2 (acoustical 12dB/octave slopes) is a negative polarity solution, and LR4 (acoustical 24dB/octave slopes) a positive polarity one when time aligned.

Troels Gravesen said:
Second order, flat baffle with notch and positive polarity near LR2:
Setting up V1 crossover was easy and overall tonal balance was good, but the transition from upper mid to treble appeared rather fizzy and blurred. Something just wasn't right despite the ability to play all kinds of music without any distress to the ear.

Same but third order tweeter, which phase-aligns better with asymettric slopes:
The V2 crossover solved this problem and kind of made the sound well known from many other speakers I've made, but not up to the level of stepped baffle/LR2 speakers I've made ever since the launch of the first Jenzen speaker - or the ScanSpeak Illuminator with its all-pass filter to the tweeter. The quality of the W18 driver strongly suggested there were more goodies in store.

Same idea, but negative polarity on stepped baffle and second order tweeter:
V3 crossover: Making the stepped baffle and implementing a true LR2 filter made a world of a difference. Suddenly music started flowing and sense of depth and perspective improved vastly. I never heard the difference between a flat baffle/2nd-3rd order crossover and a stepped baffle/true LR2 filter so clearly. It's night and day. It also made me think I have to do the Ellam XT once more with an easy stepped baffle because people continue building the Ellam XT speaker rather than the much better Ellam FLEX with its more complex front panel layout. There are more to be gained from these Ellam XT. And I would love to do the 9800 tweeter again in such set-up... well, well, back to the Discovery here.

4th. order electrical:
V4 crossover (stepped baffle): This has the well known 4th order characteristics. Even power response and a balanced presentation with a noticeable presence character. Good for monitoring! I know some people like this a lot but to my ears it sounds kind of flat and without the sense of depth and 3-dimensionality of the LR2 filter. Matter of taste in the end.

It's all good fun, IMO. To me, the cleanest, loudest sound is always higher order filters. There is something spooky about a speaker with no colouration or distortion and narrow dispersion, IMO. And I have heard a few, even expensive Koss electrostatic headphones. They actually sound, er, spooky! I can't describe it. But I think it's the lack of room acoustic or spatial context. Sort of hollowness. It's slightly uncomfortable. :D
 
My conclusion from all of this is basically along the lines of - use whatever filters are necessary to get the job done. Striving for certain filter orders, because they somehow contain audio magic, when any kind of compromise is going to introduced by lowering the filter order, is just silly.

I completely agree. I use a steep filter on the woofer and a shallow one on the tweeter, but the bottom line is whatever works. This idea of using "name brand" filters is absurd - especially in the electronics, because the acoustics will dominate the end result and there nothing is simple.
 
Out of curiosity, is our vocal cord + diaphragm combo combined to be 1st, 2nd, 3rd, or 4th? I mean natural our voice has a high frequency rolled off so the freq. response would be like a 5.5 in. woofer unless you're a bird then it would be more like a tweeter. This would lead to another point I would like to make, but not now.
 
...

Various filters orders were provided. I couldn't hear any difference between the different slopes. Nada, zilch, nothing. If there was any difference it was so minuscule as to be irrelevant.

This also ties in well with my own experience. Using suitable drivers I have directly compared 2nd order vs 4th order Linkwitz Riley xovers. This was done via a DSP where both sets of filters were stored in memory and could be swapped between at the touch of a button. Switching between the two took such little time that you couln't even hear an interruption in the music except for a small 'click' or 'pop'.

Could I hear any difference? Short answer yes. But only because the vertical off axis...

My conclusion from all of this is basically along the lines of - use whatever filters are necessary to get the job done. Striving for certain filter orders, because they somehow contain audio magic, when any kind of compromise is going to introduced by lowering the filter order, is just silly.

I pretty much agree with this. In the low Khz range, phase just doesn't matter much, if at all. In the several hundred hz range, it matters a little. I did ABX testing with wav files with and without phase distortions like those from a crossover, and couldn't tell them apart if the crossover emulated was 2 Khz. At 200 Hz I could, but didn't really have a preference.

I know how to use DSP to fix phase distortions on axis (even did that as a senior design project) but I don't bother.

YMMV.
 
I was wondering if DSP is a valid way to test, since your signal is already has a lot of phase shift to begin with due to the ADC stage and DAC stage(if that is what you are using). Also, given most DSP do not have good ADC and DAC, the tests you're referring to is a bit of suspect. I used to have an old Kenwood receiver and I couldn't tell much from anything since the signal was so corrupted by the internal digital processing.
People who designed audio DAC had to deal with phase shift way back then. In the early 80's, DAC didn't have any fancy upsampling so they had to use very steep filter for the reconstruction stage after the DAC so early CD player sounded harsh because of the excessive phase shift especially in the higher frequency caused by steep slope filter. Now with better upsampling, they can use gentle slope on their filter which fixes a lot of the issues back then. Of course there are other factors in making good DAC but it was one of the main one. One of the reason behind SACD was to use very high sampling rate (MHz rate) so to shift out of band harmonic way out of the audible range which helps make the reconstruction filter much easier.
 
Actually, modern ADC and DACs have virtually no nonlinear phase shift in the audio band, as the filtering is mostly done with internal DSP processing. Easily shown with a square wave generator and an analog oscilloscope.

Audio band phase linearity is almost trivial to correct even with speakers (at least for one point in space) these days, using FIR equalization. But the importance of having linear phase appears to be mostly for the benefit of marketing magazines and speakers rather than higher-fi.
 
It has been argued all over the internet that phase differences are inaudible, even by experts.

My ears tell me a different story entirely. I first noticed a large improvement in sound quality back when I was installing car stereos in the 90s. Back then there were a few systems you could install which included a digital signal processor box which among other things could function as a digital crossover. Some of these as I recall had linear phase FIR filters.

The first of these systems I installed resulted in a night and day improvement in sound quality. I wanted one, but they were too expensive for my personal budget so I continued to tinker with inexpensive solutions.

I started building 2-way speakers as a hobby for myself, and began settling on designs which could get away with using the simplest crossovers that could still be properly called a crossover - the 1st order Butterworth, with just a cap on the tweeter and coil on the woofer. Driver selection is important because you're not getting as much attenuation as you would with higher order filters. Tweeters especially needed to be able to handle some energy at one or two octaves below the crossover point without being destroyed, and it also helped to have a low resonance frequency.

Many of these 2-ways had dome tweeters by Morel and Dynaudio, but my favorite was the Bohlender Graebener Neo 8, since it could handle a bit of lower frequency energy and a somewhat lower crossover point.

The results were so good that I could say they compared to some high end commercial speakers. The smoother phase across the audio band really did seem to have the effect of getting the speaker "out of the way" and let the music come out more true to form. These speakers are not capable of high power handling or volume compared to many other designs, but I never listened to music that loud anyway, and the smoothness and finesse of a properly implemented 1st-order system had me sold on the idea that phase response was important. Keep in mind my conclusions are after literally years and years of first installing car and then high-end home theater systems. I have a crate full of caps and coils from crossover experimentation all the way to 36 dB/oct and everything in-between.

The "holy grail" of crossovers to me is one that essentially "does no harm" to the signal, not just amplitude, but also phase. An "ideal" crossover, if you will, but that solution can only be done via technology.

Fast forward today and I am experimenting with a miniDSP 4x10, which has the ability to tailor the audio reaching your amps and speakers with surgical precision. Results have been fantastic. Never have I heard such realism from a pair of speakers except perhaps some of the exotic ultra-high end systems being demonstrated at boutique shops, which I could never afford anyway.
 
Last edited:
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.