High Order XOs

[daftcombo]

Am I the only one around here to think that the future will be made of active speakers with at least 96 dB / octave steep slopes as digital crossovers?

 

[617]

Am I the only one around here to think that the future will be made of active speakers with at least 96 dB / octave steep slopes as digital crossovers?

96db /octave doesn't give you much wiggle room for pattern matching between drivers. 24 db /octave is pretty standard.

 

[thewas]

I like the subtractive filter approach.

Interestingly my active motion feedback Philips loudspeakers from 1979 also use subtractive crossover filters.

 

[thewas]

Isn't the big downside of low orders is that the vertical directivity is really ugly?

Yes, unless they are coaxial drivers.

 

[Plcamp]

Interestingly my active motion feedback Philips loudspeakers from 1979 also use subtractive crossover filters.

I remember seeing a gent who investigated/made a subtractive analogue opamp crossover...but failed at higher order filters because the tolerance requirements tighten dramatically as order increases, and he realized that only after trying it.

On another note...Per above, I have my eye on coax drivers now. I have great trouble distinguishing what drivers will and won’t work well in open baffle, but the 10” eminence coax with a well chosen concentric compression driver might just be perfect, given I only need it to respond 250 (or higher) hz up (with dual 15” woofers supporting bass.) That crossover system will be interesting to design.

 

[puppet]

That paper refers to the xo examples as a 2nd order butterworth which should be 3db down. The examples are, in fact, a 2nd order Linkwitz-Riley as they're 6db down.

 

[BenB]

Here's something to consider regarding minimum phase and linear phase filters:

When you apply a minimum phase filter to a minimum phase device, you achieve a minimum phase result. This means that if you hit your target magnitude slopes, you have also attained your target phase characteristic.

When you apply a linear phase filter to a minimum phase device, you get something that is neither linear phase, nor minimum phase. In order to achieve the perfectly idealized linear phase result would require total compensation for the transfer function of the device to which the filter is applied (in our case, speaker drivers). It likely isn't necessary to perfectly achieve the idealized linear phase result. Once something is attenuated 40 dB, it doesn't have a big impact on the sound. It would be interesting to see a study on the impact of this, particularly for technologies that seek to achieve a particular result, like Keele-Horbach filters.

 

[puppet]

This might be an interesting read as well:
Zero Phase In Studio Monitors | audioXpress

 

[Deleted member 65]

My Gauder-Akustik Arcona speakers has this spec:

ARCONA 100
· 3-way floorstanding loudspeaker
· down-firing bass vented, high-/mid closed acoustically favourable cabinet
· symmetrized crossover with Mundorf parts
· ultrahigh slope crossover of 50 db/octave

 

[sigbergaudio]

You don't have to do all kinds of fancy things to achieve a linear speaker. Here's an active speaker we're currently developing. IIR DSP using assymetric 1.order filters for all crossovers combined with competent drivers. Note that the response in the low end isn't accurate since this is nearfield in a non-anechoic room. At least to our ears this results in a more natural sound and better crossover blend than ultrahigh slopes.

 

[UliBru]

That paper refers to the xo examples as a 2nd order butterworth which should be 3db down. The examples are, in fact, a 2nd order Linkwitz-Riley as they're 6db down.

Sorry for the confusion. It's now 15 years ago when I have written the paper.

I have created the XOs with the help of Audition function Effects - Scientific Filter - Butterworth 2nd order (this is also noted in the paper). And I have missed to check the resulting "Butterworth" for the -3 dB XO point.
Obviously the Audition function gives a wrong result. But it's my point, I have trusted Audition too much.

So here is a correction with proper Butterworth XOs:
The frequency responses and the sum look like

and the step response of the summed XOs is

This means that also the correct Butterworth XO of 2nd order does not change the content of the paper and its intention.

 

[Digital_Thor]

I know this is an old thread. But I can't really get my head around KEF's idea of time smear. They seem to claim that 24dB LR filters are not good and that we should stick with lower orders and keep them fitting for the speaker drivers FR in a given construction - meaning that it is meaningless to talk about electric filter functions, if we do not include the rest of the components making the total/final loudspeaker - also keeping in mind that we should distinguish wisely between electrical and acoustical filters - or the combination of those.
https://assets.kef.com/pdf_doc/REF/REF-White-Paper-201219-LR.pdf
I can make out that they have to take into consideration that it's a passive construction - therefore saving on the components.
But how much is marketing? They usually seem to present their speakers as quite well-designed, which makes me wish to believe what they write.
It is just puzzling for me to understand the relationship between the "time smear" and the given filters, when I've heard several times, that 24dB LR is the filter that usually keeps the phase best controlled through the cross-over, while keeping good protection of the drivers.
Am I missing something here?

 

[BenB]

KEF is wrong to demonize high parts counts in passive crossovers. Considering each drive unit individually, they are approximately minimum phase devices, and the passive crossovers used to shape their power spectra are also minimum phase. By filtering a minimum phase device with a minimum phase filter, the result is also minimum phase. This means the response of each driver individually has an impulse response that is as sharp/compact as possible for the resulting (individual) frequency response.
The distortion associated with passive crossover components is very small in comparison to driver distortions, particularly if you stay away from electrolytic caps and iron core inductors.
As you have noted, power handling is a much more important concern than any audible effects of the "time smear" associated with higher order crossovers (assuming reasonable orders... probably less than 6th or so). For non-coaxials, the lower crossover frequency facilitated by a higher order filter has a longer wavelength, implying reduced lobing effects for the same amount of physical driver separation.
Although individual drivers shaped by passive crossovers exhibit minimum phase response, the combined output is not minimum phase, and higher order filters do result in more phase distortion or "time smear". However, the impact is completely inaudible above about 2 kHz or so for reasonable crossover orders. Although the time smear is most audible in the lower midrange and punchy bass (not rumbling bass) region of 100 Hz to 300 Hz, the impact is still very subtle. Best practice is probably to use nothing more than a 2nd order electrical crossover order in that region. For tweeters, 4th order electrical or even more is fine. Our ears simply aren't that sensitive to it.
Having said that, I do prefer to work with drivers that don't require much in the way of response shaping. I'd rather spend the money on a better driver than on a complex shaping network.

 

[Digital_Thor]

Thank you, Ben. Sometimes I re-read something, and still I can't fully make out the context or meaning of the message, which makes it great to come here and get some feed-back on other people's grabs on the same subject
And it is a good point that we should distinguish between a normal midrange/tweeter arrangement and a coax, since with the coax we no longer have the CTC in the equation.

I've read somewhere that they focused on getting the cross-over right, regarding the power response, since we could always correct the final response with EQ.
Which should be the reason why it is ok that the speaker is not perfectly linear, as long as it has a flat/smooth DI. And with the increasing number of DSP-pre-amps available on the marked today - it's easy to correct - if done with the right data though.
This leads me to a question. I run my system fully active with multi-channel DSP and a dedicated amp for each driver. Here it seems that if I want a smooth DI, I measure each driver unit alone, finding out where they have the response on all axis's, before adding filters and deciding on the slope/type. Because then I know that my final EQ of the combined drivers, will give me a linear problem to solve, which is the only problem a DSP can fix.
I do not believe that you can just put a microphone in the listening spot, and then let some algorithm smooth it all out and make the best compromise, if you don't know exactly what you have to work with, before you "introduce" the room. Because my understanding about higher order filters - above 24dB per octave - is, that you make too abrupt changes in the FR, making it more difficult to EQ the final result, since smoothness is the easiest problem to solve?

 

[Keith_W]

It is true that if you have two drivers with mismatched directivity, then using a high order slope will introduce an abrupt change in DI.

May I ask what type of DSP you are using? Linear phase FIR or minimum phase IIR? Because, if you are using lin phase you don't have to worry about time smear.

 

[Digital_Thor]

It is true that if you have two drivers with mismatched directivity, then using a high order slope will introduce an abrupt change in DI.

May I ask what type of DSP you are using? Linear phase FIR or minimum phase IIR? Because, if you are using lin phase you don't have to worry about time smear.

Exactly... and that also makes sense to me
It is an IIR DSP, since when I jumped the active adventure more than 10 years ago, FIR was not available in 8 channels and with a low noise analog section.
Now, I have still not fully been convinced that FIR can offer a true audible difference - a technical and measurable - but audible?
And which gear actually exist, where you can switch smoothly between IIR and FIR with the same setup, speakers, amps and room, where the absolute only difference is the phase-correction?
When using my system from both gaming, movies, tv, and youtube... then I would deny FIR any day, if there's just a hint of delay, so that lip-sync is a problem - I just can't live with the unnatural experience when there's a timing difference between sound and picture

 

[Keith_W]

Exactly... and that also makes sense to me
It is an IIR DSP, since when I jumped the active adventure more than 10 years ago, FIR was not available in 8 channels and with a low noise analog section.
Now, I have still not fully been convinced that FIR can offer a true audible difference - a technical and measurable - but audible?

Well, "is it audible" is kinda can of worms territory. @BenB has already stated that phase may be audible in some frequency bands. Toole and Linkwitz would disagree with him. They think that intrachannel phase distortion is inaudible. There is something called "Ohm's Acoustic Law" which says that you can't hear it. However, there is a discussion on ASR with JJ who unequivocally states that phase is audible, if the deviation is more than 15 deg per ERB and provided the ERB bands are next to each other. I made a post in that thread where I looked at several studies cited by Toole, and showed that they studies were done by adding phase distortion on minimum-phase speakers. In other words, time smear already existed in the loudspeaker - what if we time smear it even more. If you read Linkwitz's blog, that is exactly the same experiment that he did.

I use linear phase FIR's and I have conducted my own experiment comparing min phase vs. linear phase. I have gone so far as to straighten out the minimum-phase behaviour of my drivers and make them linear phase. In my opinion, it is easily audible. The difference is that the lin phase version has so much more clarity.

The problem with my opinion is that it is anecdotal, and there are no published studies that support it. I begged JJ to publish, but he said that he does private work now and no longer publishes. Regardless, he has several test tones in that thread that you can download and try. Decide for yourself if you can hear a difference.

BTW, there is an additional consideration - I also believe that placing the MLP within the critical distance is important if you want to benefit the most from lack of phase distortion. Again, this opinion is my own and it is anecdotal, although I came to this realisation by reading Griesinger's writings on proximity. He said that the sensation of "proximity" is created by loss of phase coherence due to contamination by reflections the further you travel from the sound source.

So right now my opinion is admittedly on shaky ground. I believe lin phase sounds superior, but without any publications and with people like Toole and Linkwitz saying the opposite, I realise that I have little persuasive power.

And which gear actually exist, where you can switch smoothly between IIR and FIR with the same setup, speakers, amps and room, where the absolute only difference is the phase-correction?

Any PC based convolver can easily do it. Create minimum phase and linear phase versions of the filter, then load it into a convolver. The best convolver for this kind of testing is Hang Loose Convolver since it is zero latency.

When using my system from both gaming, movies, tv, and youtube... then I would deny FIR any day, if there's just a hint of delay, so that lip-sync is a problem - I just can't live with the unnatural experience when there's a timing difference between sound and picture

Well, that is a problem. The solution would be to use FIR for music, and IIR for video. The issue is that switching from FIR to IIR is not a seamless experience for my wife. It is easy enough for me, but it isn't for her. In fact the whole PC based setup is so complex that she can't use it. That is a major downside. I have enough brains to make it work, but not enough brains to make it usable.

 

[BenB]

Well, "is it audible" is kinda can of worms territory. @BenB has already stated that phase may be audible in some frequency bands. Toole and Linkwitz would disagree with him.

I tried to dig up an email exchange I had with Siegfried Linkwitz circa 2004 (+/- a year or two) regarding the audibility of phase distortion, but everything prior to 2007 is no longer available. In our exchange, he was not staunchly averse to the idea of phase distortion being audible in certain conditions. I wrote SL to tell him I had passed ABX testing comparing an all-pass (phase distorted) wave file against the original using high quality headphones. If I recall correctly, I induced phase distortion associated with a theoretically perfect 4th order crossover at 200 Hz. SL was not entirely skeptical of this claim in our exchange. I have no way of knowing if he authentically believed me. I believe he challenged me to repeat the exercise on speakers (subject to room reflections). After much practice, I was able to pass ABX testing on speakers as well. The speakers I used were two-way MTM design, with no crossover close to the 200 Hz region. Our exchange ended with SL telling me he would incorporate this information into his knowledge base. Again, he may have doubted the results and decided to be polite and non-confrontational.

For context, I tried to pass ABX testing with a 4th order 2 kHz allpass, but failed repeatedly, despite my young ears ( I was in my 20s back then). When I passed the 200 Hz allpass using speakers I was able to discern a difference, but not a preference.

My senior design project in college was to utilize DSP to correct magnitude and phase distortions in my DIY speakers (the MTM's mentioned earlier). The results were certainly audible. However, those speakers had significant frequency deviations. I repeated the process years later on a subsequent project that had a flatter spectrum. I was able to switch correction on and off in realtime. It was immediately apparent that the difference was incredibly subtle and not-at-all impactful on the enjoyment of the listening experience. The corrected impulse response was pretty, and the uncorrected one not so pretty, but we don't listen with our eyes.

Anyone interested in attempting something like the experiment I wrote SL about, you can use the files and filters I posted in this thread:

Phase Distortion ABX testing

I am posting files to facilitate ABX testing for the types of phase distortions caused by crossovers. I have acquired samples that are royalty free, and that I find to be of good quality. I am not claiming that these are the ideal files to reveal differences, nor am I claiming that I have...

www.audiosciencereview.com

Admittedly, there was a units conversion error in the "group delay" plots I posted, which was quickly pointed out (I don't typically calculate group delay, and look at phase instead because that helps me see how things will sum). At any rate, the filters are correct (only the plots have an error).

 

[Keith_W]

Again, he may have doubted the results and decided to be polite and non-confrontational.

In my travels I came across an audio blog some time ago written by an Asian audiophile who visited Linkwitz in his home. He was chuffed to meet the great man himself. But being a traditional audiophile of the subjectivist persuasion, he tried to convince Linkwitz that he would get better sound if he used exotic capacitors in his XO's. He said that Linkwitz smiled at him and entertained his idea but said "maybe he was being polite". It was hilarious, I could almost imagine the look on Linkwitz's face. If I find the blog again i'll post it here.

I am not suggesting that he was privately thinking "what a nutjob" when he was speaking to you He definitively states in his blog that phase is not audible.

And BTW of course there is a chance that Linkwitz may be wrong. He was an engineer, not a psychoacoustician. I would be more interested to hear what Moore or Glasberg has to say about it. Matter of fact, I should go look it up.

 

[BenB]

In my travels I came across an audio blog some time ago written by an Asian audiophile who visited Linkwitz in his home. He was chuffed to meet the great man himself. But being a traditional audiophile of the subjectivist persuasion, he tried to convince Linkwitz that he would get better sound if he used exotic capacitors in his XO's. He said that Linkwitz smiled at him and entertained his idea but said "maybe he was being polite". It was hilarious, I could almost imagine the look on Linkwitz's face. If I find the blog again i'll post it here.

I am not suggesting that he was privately thinking "what a nutjob" when he was speaking to you He definitively states in his blog that phase is not audible.

And BTW of course there is a chance that Linkwitz may be wrong. He was an engineer, not a psychoacoustician. I would be more interested to hear what Moore or Glasberg has to say about it. Matter of fact, I should go look it up.

FWIW, I believe that blog post predates my discussion with SL. I was well aware of his position, but also his interest in the subject, which is why I reached out to him. I was surprised and honored that he replied (he didn't have to). Reading the blog now, it doesn't come off to me as though he feels the science is definitively settled. Honestly there's little difference between our positions: His being that it was inaudible, and mine being that it's extremely subtle and not worth worrying about. Both of us want (or wanted ) others to experience the difference for themselves, in order to obtain an appropriate perspective.