Is it OK for a three-way designed speaker to out put the same tone through two drivers?

[gnarly]

These things also have issues. You'll need just about perfect directivity matching to make this work. In a normal less steep slope, directivity will slowly be matched. With brick wall filters you can get a sudden step. Extra care has to be taken as well to create filters with little rining. This might be audible in some cases.

Good points to consider.

Ime, steep linear phase xovers have helped solve both those issues.


Since directivity matching is a function of the acoustical design....center to center spacing, cone and waveguide pattern vs frequency, etc,....
we know there is an optimum crossover frequency for directivity matching.

I've found the odds of being able to actually use that optimum frequency point go up! with steep xovers.
Because the critical frequency range of required well behavior is greatly reduced, and it's simply easier to get good behavior in a smaller range.
Shallow slope filters are simply more work to make behave over a wider critical region, and often are problematic further outside the the critical region.

I've found a little bit of forethought in driver selection that fits the acoustic design, makes steep filters that match directivity a snap.



Agreed on extra care to avoid ringing.
Care must be taken to achieve a fully acoustically complementary design, which eliminates ringing.
Which is easy enough to achieve on-axis, but not so easy off-axis due to lobing.
Where there is lobing, we know there is an accompanying loss of acoustic complementary.
So, there is the potential for ringing, with off-axis lobing; but this potential is only in the critical crossover region.

The key to solving this is again steep filters, I think.
Hey, just shrink the range of the critical crossover region, the range of the lobing region, the range of potential ringing, by going steep!!

 

[Kvalsvoll]

Steep slopes or brickwalls aren't at all pragmatic with passive crossovers or active analog crossovers, due to both their complexity to implement and the sonic degradation that occurs with their use. (excessive phase rotation is the sonic issue)

There is lot of misunderstanding about filters and slope and phase. It really does not matter if it is implemented in dsp or as a passive network, it is the chosen transfer function that determines the behavior. Very steep filters with excellent performance in the time domain are possible, and such filters have both measurable and indeed audible advantages compred to trad. But they are much easier to implement in dsp, and some implementations simply can not be done using passive only.

 

[gnarly]

There is lot of misunderstanding about filters and slope and phase. It really does not matter if it is implemented in dsp or as a passive network, it is the chosen transfer function that determines the behavior. Very steep filters with excellent performance in the time domain are possible, and such filters have both measurable and indeed audible advantages compred to trad. But they are much easier to implement in dsp, and some implementations simply can not be done using passive only.

I don't think we said anything different, did we?

My comment about passive and active analog not being pragmatic for steep filters, due to complexity and sonic degradation holds true, i believe.
And like you say, some implementations simply cannot be done using passive only.
I've never seen a very steep linear phase passive, or analog active, implementation. Have you seen any? (by very steep, let's say around 100dB/oct)

100% agree equivalent transfer functions are all that matter, and if equivalent no differences exist between types.
Other than passive's reactance (which isn't a good thing imo)

I've made grillions of electrical transfer function measurements, one filter topology against another using dsp. Particularly minimum phase vs linear phase vs maximum phase.
It's been interesting, that once a certain light bulb clicked, crossovers have totally become my easiest part of speaker design.

Sorry if swerving too deep... I just get enthused with topics like this, i guess..

 

[Kvalsvoll]

I've never seen a very steep linear phase passive, or analog active, implementation. Have you seen any? (by very steep, let's say around 100dB/oct)

I have used steep filters many years, dating from 1994 using simpler variants.

 

[gnarly]

I have used steep filters many years, dating from 1994 using simpler variants.

Hi, what do you mean by simpler variants.... lost to what you mean...?
And are you saying steep via passive or analog active?
If so, how steep, and were they linear phase? or at least with very low phase rotation?

Personally, i feel phase rotation sonically kills any attempts at very steep IIR implementations.
You?

If you have an example of a steep IIR filter you've used, would love to see it. Thx.

 

[Kvalsvoll]

Hi, what do you mean by simpler variants.... lost to what you mean...?
And are you saying steep via passive or analog active?
If so, how steep, and were they linear phase? or at least with very low phase rotation?

Personally, i feel phase rotation sonically kills any attempts at very steep IIR implementations.
You?

If you have an example of a steep IIR filter you've used, would love to see it. Thx.

I never completed any articles or papers on filter technology, there were some information in the presentation for some of the earlier speakers, but they are no longer available on the web site.

What those speakers show, is a flat frequency response, phase that may shift say 90 degrees, some quite a bit more, off-axis is improved, impulse (IR) looks like a FIR-filter - symmetrical in time, step response close to perfect, because all sound across the entire frequency range starts quite close in time.

It is no longer a IIR filter, it needs delay on the HF section to work, slopes can look more similar to digital FIR filters. For passive implementation it can work for speakers with some sort of horn on the HF section and direct radiator below, because then there will be a time delay on the HF. Since this delay is fixed once the horn is designed, there is no way to adjust this delay if it is too far-off to work properly. Fortunately, it does not need to be exactly flat phase at higher frequencies, so in many cases, it can work quite well.

30 years ago, this tech could have some value, but today, it is better and more cost efficient to implement the filter in DSP. So for a high-performance speaker of today, it is possible to implement very sharp cut-off filters, and have flat phase at the same time.

What is important here, is whether this effort has any audible advantages. An it has. What I found, was that the very sharp and large attenuation of out-of-passband radiation has a significant impact on perceived sound quality. I have not found any indication that flat phase is important or even audible, at somewhat higher frequencies, as long as time delay is kept within reasonable limits. Which leads to filter requirements having less strict limits for phase linearity, and more strict requirements for frequency response.

With several implementations of speakers using those steep filters, I see no correlation between sound quality and more accurate/linear phase. Other properties of the design is more important, such as driver selection, and for the filters it is the steep attenuation than makes a difference.

Using asymmetric slopes and delay on the HF, similar results with flat phase and GD can be achieved. If the slopes are say at least 4. order (acoustical), that is a decent attenuation of out-of-passband sound from each section. This is how a crossover between bass-system and main speakers always should be implemented, as this crossover is in a part of the frequency range where smaller phase errors cause audible timing problems. I have many examples of measurements from such systems, where GD can be totally flat through the crossover region.

 

[fluid]

I think you mean 6dB there.

6dB is the point for a Linkwitz Riley style filter, 3dB is for a Butterworth, so both can be right depending on the topology.

 

[abdo123]

6dB is the point for a Linkwitz Riley style filter, 3dB is for a Butterworth, so both can be right depending on the topology.

Yes but a butterworth crossover will cause a +3dB boost at the crossover frequency which contradicts the statement i was replying to.

 

[fluid]

Yes but a butterworth crossover will cause a +3dB boost at the crossover frequency which contradicts the statement i was replying to.

With one side phase inverted and no delay there will be a peak with Butterworth, if delay can be set the combination can be flat.

Linkwitz Riley will also need a delay to be flat but much less.

 

[LTig]

Not silly at all
In fact, more and more advanced active speaker designs are beginning to use crossovers that have very steep slopes, even to the point of using brickwalls.

Steep slopes or brickwalls aren't at all pragmatic with passive crossovers or active analog crossovers, due to both their complexity to implement and the sonic degradation that occurs with their use. (excessive phase rotation is the sonic issue)

However, DSP FIR filter crossovers, can easily implement steep/slopes, and have no sonic degradation when properly implemented. (FIR crossovers can be constructed without phase rotation. Doing that does create a fixed constant delay, just like increased listening distance from speaker, that may or may not be an issue)

AFAIK the Neumann KH80DSP sports a slope of 48 dB per octave.

 

[gnarly]

I never completed any articles or papers on filter technology, there were some information in the presentation for some of the earlier speakers, but they are no longer available on the web site.

What those speakers show, is a flat frequency response, phase that may shift say 90 degrees, some quite a bit more, off-axis is improved, impulse (IR) looks like a FIR-filter - symmetrical in time, step response close to perfect, because all sound across the entire frequency range starts quite close in time.

It is no longer a IIR filter, it needs delay on the HF section to work, slopes can look more similar to digital FIR filters. For passive implementation it can work for speakers with some sort of horn on the HF section and direct radiator below, because then there will be a time delay on the HF. Since this delay is fixed once the horn is designed, there is no way to adjust this delay if it is too far-off to work properly. Fortunately, it does not need to be exactly flat phase at higher frequencies, so in many cases, it can work quite well.

30 years ago, this tech could have some value, but today, it is better and more cost efficient to implement the filter in DSP. So for a high-performance speaker of today, it is possible to implement very sharp cut-off filters, and have flat phase at the same time.

What is important here, is whether this effort has any audible advantages. An it has. What I found, was that the very sharp and large attenuation of out-of-passband radiation has a significant impact on perceived sound quality. I have not found any indication that flat phase is important or even audible, at somewhat higher frequencies, as long as time delay is kept within reasonable limits. Which leads to filter requirements having less strict limits for phase linearity, and more strict requirements for frequency response.

With several implementations of speakers using those steep filters, I see no correlation between sound quality and more accurate/linear phase. Other properties of the design is more important, such as driver selection, and for the filters it is the steep attenuation than makes a difference.

Using asymmetric slopes and delay on the HF, similar results with flat phase and GD can be achieved. If the slopes are say at least 4. order (acoustical), that is a decent attenuation of out-of-passband sound from each section. This is how a crossover between bass-system and main speakers always should be implemented, as this crossover is in a part of the frequency range where smaller phase errors cause audible timing problems. I have many examples of measurements from such systems, where GD can be totally flat through the crossover region.

Thank you for a detailed and considerate reply!
My experiences largely mimic your own.

Although I've never been able to use fixed delay to mitigate phase rotation with any degree of success. But that may be due to the fact i haven't tried real hard to do such, given today's DSP's, which like you allude can easily implement flat phase directly.
Anyway, my takeaway from my rather futile efforts, is that I think the idea that fixed delay can be used to offset group delay, simply doesn't work well.

I very much agree with the audible advantages of attenuation of out-of-band radiation. In fact, i take the effort to smooth out-of-band minimum-phase response prior to applying very steep flat phase filters. (Well i should say, that's the the two stage manual process. It's easy to do it all in one stage with target matching FIR filter generator software.)
This leaves me with an acoustic response, that looks very close to the electrical crossover response...and gosh does it make tying driver sections together easy.

With regards to audibility of phase itself...(which is really the only thing that matters in all this effort, huh?), i think the audible benefits of flat phase get greater the lower in frequency the flat phase extends. One of the big reasons I enjoy outdoor listening so much, is to be able to hear this more clearly without a room's influence. That said, it's audible for me indoors too, but on a lesser percentage of recordings.

Thanks again

 

[gnarly]

AFAIK the Neumann KH80DSP sports a slope of 48 dB per octave.

Yes. Good example of the growing trend towards steeper linear phase crossovers.
According to the manual, it uses a 48dB/oct linear phase xover at 1.8kHz

 

[Penelinfi]

The crossover of that speaker is supposedly 500Hz and 2kHz. Given that it's a "low cost" offering, likely they use shallow slopes on the bass and mid crossover. You may even hear output from either driver quite a long way either side of those frequencies, at reduced volume the further you move from the crossover frequency.

 

[Plcamp]

If the two speaker’s centres are less than 1/4 wavelength separated, this shouldn’t matter…as might be the case with two OB woofers, one in a u-frame and one not.

I am thinking of trying that, with uframe crossed out at 150-200 hz and non uframe crossed to high end at 600 hz (both are 15” woofers. Using dsp to control, you can phase align them with allpass filters.

 

[NewbieAudiophileExpert]

The crossover of that speaker is supposedly 500Hz and 2kHz. Given that it's a "low cost" offering, likely they use shallow slopes on the bass and mid crossover. You may even hear output from either driver quite a long way either side of those frequencies, at reduced volume the further you move from the crossover frequency.

Do you mean for my wharfedales? sorry but there's so many people in this discussion that I don't want to assume.

EDIT: I see that you are indeed referring to my speakers.

Indeed, I do hear both drivers outputting even higher or lower frequencies, but the really low bass frequencies are pretty much reserved for the 6.5 inch midbass driver, which is really good.

Though I'd bet that just the tweeter and 5 inch mid/woofer would sound great on their own..

Believe it or not, but this revision of the Crystal towers is a step up from the previous generation, which was a two-way system with a passive radiator - this is a true 3 way design, with a sealed tweeter/mid enclosure, and a ported, 21-litre bass driver enclosure..

 

[gnarly]

The crossover of that speaker is supposedly 500Hz and 2kHz. Given that it's a "low cost" offering, likely they use shallow slopes on the bass and mid crossover. You may even hear output from either driver quite a long way either side of those frequencies, at reduced volume the further you move from the crossover frequency.

If we are still talking about the KH80DSP, unless i'm misreading the specs and FAQs for it, it's a two-way which can conceivably be used alone down to 57Hz, but is made to go with the KH750DSP companion sub.
Per the literature, The KH80 itself has a single linear phase xover 48 dB/oct at1.8kHz.
The sub's default xover is 24dB/oct at 80Hz, which is a regular type xover (not lin phase)
FAQ's say linear phase from 170Hz up, using 2ms latency to achieve that. That's a good achievement for that low a latency ime/imo.

I guess you were talking about a different speaker?????

 

[NewbieAudiophileExpert]

If we are still talking about the KH80DSP, unless i'm misreading the specs and FAQs for it, it's a two-way which can conceivably be used alone down to 57Hz, but is made to go with the KH750DSP companion sub.
Per the literature, The KH80 itself has a single xover 48 dB/oct at1.8kHz.
The sub's default xover is 24dB/oct at 80Hz.
FAQ's say linear phase from 170Hz up, using 2ms latency to achieve that. That's a good achievement for that low a latency ime/imo.

I guess you were talking about a different speaker?????

He's talking about this speaker - (https://www.wharfedale.co.uk/crystal4-3/).

 

[gnarly]

He's talking about this speaker - (https://www.wharfedale.co.uk/crystal4-3/).

Thank you...it does get confusing

 

[Penelinfi]

Thank you...it does get confusing

I was replying to the OP about the speakers he bought, I too thought I was in a different thread when I saw all this talk of KH and linear phase 48dB slopes above my post

 

[NewbieAudiophileExpert]

I was replying to the OP about the speakers he bought, I too thought I was in a different thread when I saw all this talk of KH and linear phase 48dB slopes above my post

I'm glad that I started a thread topic that others would enjoy discussing - i know how much i love talking about the speakers that i bought (ad nauseum) so no hate.