• WANTED: Happy members who like to discuss audio and other topics related to our interest. Desire to learn and share knowledge of science required as is 20 years of participation in forums (not all true). Come here to have fun, be ready to be teased and not take online life too seriously. We now measure and review equipment for free! Click here for details.

Differences between tweeter designs?

andreasmaaan

Major Contributor
Patreon Donor
Joined
Jun 19, 2018
Messages
1,017
Likes
437
#21
I think you’re making the right decision in not going with the ESL63s in this case.

Which coaxial drivers did you have in mind?

And which software have you been using for modelling the crossovers?
 
Joined
Jun 20, 2018
Messages
112
Likes
26
#22
I think you’re making the right decision in not going with the ESL63s in this case.

Which coaxial drivers did you have in mind?

And which software have you been using for modelling the crossovers?
No I don't want to use actual coaxial drivers. I'm aware of the errors of the concentric models for sale except for the way Geithain does it (but that probably still has some difraction problems? And at least great restrictions on the tweeter used. And I'm not going to attempt such a thing with my zero experience.)
I want to make a design which in practice behaves mostly like a dual concentric design by making the distance between the acoustic centers of the mid and tweeter as small as possible while still mounting the tweeter on a baffle above the mid, combined with a low crossover frequency this will give near dual concentric/coaxial radiation pattern in the vertical plain.

Here a picture of a Butterworth crossover radiation pattern:
n160fig1a.png


And here a picture of a Linkwitz-Riley crossover radiation pattern:
n160fig2.png


The text that comes with the pictures states that these are the radiation patterns for a 1700Hz crossover with the drivers 7inches apart (haven't calculated if this is correct but it seems about right).
Now of course in reality the situation is often even worse where a crossover between mid and treble is most often 3000 to 4000Hz giving more lobes / cancellation axises closer in angle to the on-axis.
And I belief most often the Linkwitz-Riley is chosen for the mid to bass crossover or bass to sub crossover and the Butterworth one for mid to treble crossover. (because all angles together/room reflections/reverb add up to flat for the Butterworth all pass and the Linkwitz-Riley doesn't add up to flat when you combine all angles but adds up with a dip of -3dB(I think?) except when the crossover is very low and the drivers close enough that the cancellation axis is more that 90 degrees from the on-axis.
And just look how finicky, even in the above in reality good case scenario, the Butterworth on-axis is, if you move just 5 degrees lower you already get an almost 2dB peak (and frequency ripple all around the crossover frequency, very audible as I tested with a simulation with headphones) and 5 degrees to high you get an even larger dip already.
The Linkwitz-Riley crossover has a more stable patch around the on-axis, you'd have to move 10 degrees or so up or down for an about 2db dip in the above example (and again, in many real world situations much less movement would be needed).
Now of course in a very live room with many diffuse reflections this evens out mostly. But I will be listening near-field in an almost completely dead room. And again testing this with my headphones and simulated off axis crossover sound I can tell this is all incredibly audible and totally destructive for soundquality / transpancy etc. This is a very big part of what makes speakers sound bad I can hear and recognize from the sound of it.

So my solution will be to use the Linkwitz-Riley crossover (perhaps 24dB/oct perhaps 36dB/oct) and place the mid and treble drivers so close together and put the crossover so low that the first cancellation axis up and down is more than 90 degrees from the on-axis so in effect the cancellation axis is never fully there anymore (virtually behind the speaker) and the on-axis response is within half a dB or so over a wide verticle angle.
The only downside to this is the low crossover point which needs to be at least below 2kHz (preferably 1.5kHz or so) depending on how close the mid and treble drivers are (10cm is about the max depending on crossover freq but even closer seems very doable).
I don't understand why most speaker builders don't already do this. That vertical phasing around the crossover is really really bad sounding and I now realize/ recognize it has been bugging me pretty much my whole life. Feel good about building a speaker that doesn't have this.

For amp I think I'll indeed go with the Hypex FusionAmp, probably the FA123. I finished reading the manual and filter designer doc. It's great what it can do! Only no fir filters (yet, seems like this may come later with an update there's already a slot for it in the filter designer) but I wasn't planning on using those anyhow.
Still looking at drivers and several other things, will probably be a few weeks before I actually begin building it. Will post here once I've built them :) (so far only built a sperical speaker with a full range driver, which was a success though, learned a few things about woodworking and damping box resonances with leadbitumen, bracing and stuffing etc. So this is a step up but not that much different if I keep it sealed, port may be different story)

Edit: And a big thanks to amirm for saving this post! :D (had accidentally deleted it when trying to delete a duplicate attachment)
 
Last edited:
Joined
Jun 20, 2018
Messages
112
Likes
26
#23
This is something I am really interested in.
I built during last 30 years some loudspeakers for myself and friends, usually bookshelves, two ways, simple and cheap.
I learnt how much a "bad" tweeter can damage a good project: I hate resonances in high frequencies close to audio range.
I remember 80's Focal units, reverse dome, titanium, expensive, fast, extended range, but for me absolutely not sustainable, high listening fatigue.
Then a famous Peerless (the moustache tweeter, model KO-10DT), cheap, not extended range, but acceptable to my ears.
I found a good compromise with Dynaudio units (D28, D260 esotec, T-330 esotar): nearly perfect but deadly expensive.
I used a lot of Dynaudio tweeters, never found a Scanspeak that was sounding something better.

But the real jump would have been an AMT: I remember the original first device used in ESS Transar systems.
Wonderful: very fast, no distortion, deeply sweet, no resonance.
So good that it was nearly impossible to find a good matching mid-woofer.
Dipole emission, dipersion, speed, everything was so wildly good that a noraml piston cone sounded Neanderthal.

Now all these Mundorf AMTs: are they really good?
Maybe expensive, I never dared to try two units for a nerw project.
I never had a chance to listen to a DYI project using them in a good way: do you think it is worth a test?
Which one do you suggest? The AMT29, which is marked as the best?

Thanks for any suggestion...
Thanks for sharing your experience with tweeters.
I'm still looking at tweeters but have been looking at vertical behavior as well and the AMT's have the most narrow vertical response I've ever seen. As in rediculously narrow, for instance moving just a a few centimeter or so vertically at 1meter distance seems to already change the frequency response significantly.
This may well be their biggest difference in sound compared to dome tweeters?
This isn't shown in their hifi catalog but in their pro catalog you can see the vertical responses: http://www.mundorf.com/PDF/Mundorf_proAMT_Catalog.pdf
For the smaller tweeters with horns this effect is much less so, but for their bigger ones (like the AMT29 surely) without horn the vertical response is like a laser beam.
I'm guessing the special sound from these tweeters is due to their extreme narrow vertical dispersion which virtually eliminates any floor and ceiling reflections. Can understand the problems in matching them with a mid driver in this regard as well, would perhaps be better matched without mid driver but directly with an 8" woofer which is somewhat narrow in dispersion around the 1-2kHz crossover as well (though the same in both vertical and horizontal directions).
Having owned speakers with AMT drivers I can't say I noticed this back then, but I was young and unexperienced (more than 15 years ago).

edit: but the smaller ones look nice! Like the AMT 21CM2.1-C. Though not suitable for a <2kHz low crossover point it seems at first glance.,
 
Last edited:

andreasmaaan

Major Contributor
Patreon Donor
Joined
Jun 19, 2018
Messages
1,017
Likes
437
#24
So my solution will be to use the Linkwitz-Riley crossover (perhaps 24dB/oct perhaps 36dB/oct) and place the mid and treble drivers so close together and put the crossover so low that the first cancellation axis up and down is more than 90 degrees from the on-axis so in effect the cancellation axis is never fully there anymore (virtually behind the speaker) and the on-axis response is within half a dB or so over a wide verticle angle.
The only downside to this is the low crossover point which needs to be at least below 2kHz (preferably 1.5kHz or so) depending on how close the mid and treble drivers are (10cm is about the max depending on crossover freq but even closer seems very doable).
I don't understand why most speaker builders don't already do this. That vertical phasing around the crossover is really really bad sounding and I now realize/ recognize it has been bugging me pretty much my whole life. Feel good about building a speaker that doesn't have this.
I agree with you on this point in general. Vertical lobing is something that many commercial speakers pay inadequate attention to minimising IMHO. However, very few 1" domes play well at moderate to loud levels below 2KHz, and those that do tend to be expensive. One thing that tends to sound worse than vertical lobing is a small direct-radiating tweeter distorting because it's crossed too low. Moreover, performance sacrifices must be made to squeeze a dome onto a smaller faceplate, and these may themselves result in poorer LF performance that may outweigh the benefit of a closer C2C spacing.

A couple of suggestions for 1" domes that might suit your needs would be the Scanspeak 602010 (which has relatively good performance below 2KHz and a tiny 619mm faceplate, so it can therefore be placed very close to the midrange) and the Scanspeak D2608/913000 (which has a standard 104mm faceplate, but which can be crossed significantly lower, depending on max SPL requirements ofc).
 
Joined
Jun 20, 2018
Messages
112
Likes
26
#25
I agree with you on this point in general. Vertical lobing is something that many commercial speakers pay inadequate attention to minimising IMHO. However, very few 1" domes play well at moderate to loud levels below 2KHz, and those that do tend to be expensive. One thing that tends to sound worse than vertical lobing is a small direct-radiating tweeter distorting because it's crossed too low. Moreover, performance sacrifices must be made to squeeze a dome onto a smaller faceplate, and these may themselves result in poorer LF performance that may outweigh the benefit of a closer C2C spacing.

A couple of suggestions for 1" domes that might suit your needs would be the Scanspeak 602010 (which has relatively good performance below 2KHz and a tiny 619mm faceplate, so it can therefore be placed very close to the midrange) and the Scanspeak D2608/913000 (which has a standard 104mm faceplate, but which can be crossed significantly lower, depending on max SPL requirements ofc).
Thank you for the tweeter suggestions and insights!

Low cost drivers are not really the main issue anymore. I'm going to borrow some money for this project, I want to do it right if I'm putting in this amount of effort and I'll be working on/listening to them a long time :)

After my headphone tests with simulated vertical lobing I did a quick test with a JBL LSR305p mkII. Studio not nearly finished yet but did a quick setup where I have pretty much all direct reflections absorbed in the mids and highs (including desk). They have about 13cm spacing between woofer and tweeter and a crossover at 2kHz, so not as good as I plan to build but not miles off either. The tweeter is recessed a bit in a waveguide but I found it is not recessed enough for the acoustic centers of the woofer and tweeter to line up vertically. I found that at 1m the drivers are roughly in phase at the vertical level of the midpoint of the woofer. Now this is not a big difference in listening height it's also not anywhere near the cancellation axis yet if I listen at tweeter height. But I find it's enough out of phase allready to make exactly the type of difference that matters to me. Listening at the correct axis there is a coherence and depth and relaxed sound around the crossover that is messed up already when listening to the slightly out of phase tweeter axis. Our/my ears are apperently incredibly sensitive to this kind of phase distortion (it can't be the amplitude reponse as a little dip by EQ around the same frequency doesn't matter. It's the specific ripple effect this phasing gives which apparently even when small in amplitude is just sooo audible (though it is masked a fair amount on the LSR305 by other distortions, the ported bass and the not so stellar resolution of the drivers, I can tell it will be more critical with better speakers).
I do hear the increased distortion with these cheap drivers at the lower treble. I'll surely avoid it best as I can with my design, but I can tell that I would even prefer increased distortion in the lower treble over a slightly out of phase crossover. They are 2 different types of "distortion" and one just messes up the music for me for nearfield listening without direct reflections (for mid or far field listening with reflections / room reverb etc it would surely be the other way around where any lower treble distortion is more important as you write). But I can now also tell for sure that when I had the 3-way Klein+Hummel O300 speakers in my previous semi-anechoic room listening to them nearfield I very rarely managed to listen to them at the exact in phase vertical response (I wasn't paying enough attention to it then) and they sounded horribly unmusical / forward / harsh to me around the mid-treble crossover as a result (edit: un-transparent is the right word. It prevented looking "into" the music / depth, which is a thing I find perhaps the most important.) Btw, their crossover was at 3300Hz and I see that in their new version the O310 they lowered the crossover to 2000Hz probably because of this.

As for the rest of my design, I'm pretty sure it will be a 3-way (for better bass probably 8" closed and crossed over fairly low at 200Hz or so, and not least because of the smaller dimension of a midrange driver and closer placement to the tweeter because of this). And the amps and DSP crossover will be the Hypex FA253.
I'll also make the enclosure very rounded to minimise baffle edge diffraction. If correct this will help to make it suitable for a nearfield listening distance of 1 to 1.5 meter. (and I can set the delays for the drivers in DSP so they are all in phase at my listening postition / distance)
The limiting factor for SPL will likely be the bass through the closed 8" woofer. I hope to get it flat to 30-40 Hz (with DSP EQ of course). My very short listening distance will help a lot in still getting good SPL, but can also make a DSP preset where the deep bass is a bit less flat when I want to play really loud and simply switch to it when I feel like it with the Hypex remote.
So this is what my ideal system looks like now :) Total material cost around 2 to 2.5 thousand Euro though.. (not counting room treatment but I'm good at doing this inexpensively)

edit: oh and sorry for the long posts haha. I'm researching this all days long now and get kinda excited about it apparently ;)
 
Last edited:

andreasmaaan

Major Contributor
Patreon Donor
Joined
Jun 19, 2018
Messages
1,017
Likes
437
#26
Great that you’re excited :)

How are you measuring the JBLs and other speakers? I’d be wary of drawing the kind of conclusions you are without fairly rigorous measurements personally.
 
Joined
Jun 20, 2018
Messages
112
Likes
26
#27
Great that you’re excited :)

How are you measuring the JBLs and other speakers? I’d be wary of drawing the kind of conclusions you are without fairly rigorous measurements personally.
No measurements at the moment. Will get a measurement mic again later this month.
The conclusion I'm drawing regarding the JBLs phase is due to fairly easy spotting of the phase cancellation axis above the speaker and harder spotting of the cancellation axis below the speaker because it is at a further angle and the tweeter and woofer have dropped off too far in response by themselves already. Rest is listening and looking at the rough angle of the treble and woofer in the verticle plane.
And the logic afterwards is that it is indeed unlikely that JBL put an allpass filter in the crossover to align the phase precisely for the treble at this price point.. (it's already special they recessed the treble for the waveguide, most speakers this cheap will be flat baffle and no phase correcting allpass in the crossover it seems to me)
All in all pretty confident I'm correct about this. Also because my listening experience with the JBL phase corresponds to my headphone simulations (only a bit less clear).
 
Joined
Jun 20, 2018
Messages
112
Likes
26
#28
Btw, perhaps Amir will find it interesting to measure the phase correct angle of the drivers of the LSR305P mkII? :)
If I'm correct he has them aswell? I won't be able to do it as mine are going back, sending them friday and won't have a measurement mic till end of the month at the earliest.
Would be interested to see how this is measured as well, don't know how to do this currently (will this show up clearly in Arta with a single sweep?)
 

andreasmaaan

Major Contributor
Patreon Donor
Joined
Jun 19, 2018
Messages
1,017
Likes
437
#29
Ok sure. Actually, I find it harder to believe JBL has not properly aligned the drivers on the LSRs, and all the DIY measurements of the Mk I that I’ve seen show a correct summation of the outputs on the tweeter axis. But as I mentioned in another thread just recently, far more expensive speakers I’ve measured have been guilty of this sin.

Also generally I’m sceptical of claims like this based on anything other than measurements. I’m also sceptical, based on research into group delay audibility vs research into linear distortion audibility, of the idea that a phase error of the magnitude and frequency you’re talking about could be audible, let alone more audible than the resulting FR anomalies it would create.

But no need to get bogged down in this. This thread is about testing ideas for your own design, which is sounding like it will be very interesting at this point :)
 
Joined
Jun 20, 2018
Messages
112
Likes
26
#30
Ok sure. Actually, I find it harder to believe JBL has not properly aligned the drivers on the LSRs, and all the DIY measurements of the Mk I that I’ve seen show a correct summation of the outputs on the tweeter axis. But as I mentioned in another thread just recently, far more expensive speakers I’ve measured have been guilty of this sin.

Also generally I’m sceptical of claims like this based on anything other than measurements. I’m also sceptical, based on research into group delay audibility vs research into linear distortion audibility, of the idea that a phase error of the magnitude and frequency you’re talking about could be audible, let alone more audible than the resulting FR anomalies it would create.

But no need to get bogged down in this. This thread is about testing ideas for your own design, which is sounding like it will be very interesting at this point :)
I think merely looking at frequency response on axis doesn't show driver phase alignment with enough detail. a 1 or 2 dB dip and even less dB ripple around the crossover is easily missed because of other bumps and dips in frequency response (especially if its also averaged to some degree as these measurments often are). But the effect of driver phase misalignment is definately audible even with only a 1dB ripple. Much more audible than a 1dB dip by EQ which isn't really audible when narrow enough to me.
But I understand your scepticism :) But please test this yourself, if you don't have a reflection free place to set up speakers try it with a simulation with headphones which shows the effect clearly. You can do this by setting up in a DAW and split the audio to two stereo channels where each channel gets its own EQ with minimal phase/anlogue behavior, one channel for instance 12dB lowpass and the other 12dB highpass both set to the same frequency for instance 2000Hz (with normal EQ for instance Fabfilter pro-Q2 this will automatically give a flat frequency response when summed) and then sum the channels again. You can then delay one of the channels to introduce the phase shift and when it's shifted by 1000/(A*2) milliseconds (where A is the crossover frequency) you'll get the full phase cancelation axis sound of that crossover.
I can really recomend this test, it's very revealing. If you do it at different frequencies and with different slopes you'll hear a lot of bad speakers come by just by recognition of their bad crossover sound! And it'll also show just how audible and degrading of depth / transparancy even small phasing is.
I think our ear is perhaps especially sensitive to this type of phasing as it is related to the phychoacoustics of locating sound source? (edit: oh and you have to set the delay properly when using non multiples of 24dB. For instance 12dB is ofcourse exactly out of phase without any delay and in phase with 1000/(A*2) millisecond delay on one channel, as are 36dB slopes. For 18dB slopes it is 1000/(A*4) delay to get in phase, etc. edit2: what I said here is not correct, one has to flip one channels polarity for it to be correct for 12dB and 36dB. I don't even know what to do for 18dB for it to be correct on-axis.)

Also good to read up more on the Seas-DXT tweeter! Thanks again.
I found this page which seems to suggest that a waveguided tweeter has less tendency for baffle edge diffraction. If so this is something very much worth taking into consideration! https://heissmann-acoustics.de/en/kantendiffraktion-sekundaerschallquellen-treiberanordnun/ it could indeed change the balance of things in favor of such a tweeter.
 
Last edited:

andreasmaaan

Major Contributor
Patreon Donor
Joined
Jun 19, 2018
Messages
1,017
Likes
437
#31
Also good to read up more on the Seas-DXT tweeter! Thanks again.
I found this page which seems to suggest that a waveguided tweeter has less tendency for baffle edge diffraction. If so this is something very much worth taking into consideration! https://heissmann-acoustics.de/en/kantendiffraktion-sekundaerschallquellen-treiberanordnun/ it could indeed change the balance of things in favor of such a tweeter.
No problem :) With a dome tweeter in a waveguide, you will more or less eliminate baffle edge diffraction if the diameter of the waveguide is greater than the longest wavelength to pass through it, but you will also create diffraction at the mouth and (more critically) the throat of the waveguide. I think in most cases this is a worthwhile trade-off, especially considering the reduction in nonlinear distortion that a waveguide can bring, but it's something else to be aware of.

But please test this yourself, if you don't have a reflection free place to set up speakers try it with a simulation with headphones which shows the effect clearly. You can do this by setting up in a DAW and split the audio to two stereo channels where each channel gets its own EQ with minimal phase/anlogue behavior, one channel for instance 12dB lowpass and the other 12dB highpass both set to the same frequency for instance 2000Hz (with normal EQ for instance Fabfilter pro-Q2 this will automatically give a flat frequency response when summed) and then sum the channels again. You can then delay one of the channels to introduce the phase shift and when it's shifted by 1000/A*2 milliseconds (where A is the crossover frequency) you'll get the full phase cancelation axis sound of that crossover.
I can really recomend this test, it's very revealing. If you do it at different frequencies and with different slopes you'll hear a lot of bad speakers come by just by recognition of their bad crossover sound! And it'll also show just how audible and degrading of depth / transparancy even small phasing is.
I think our ear is perhaps especially sensitive to this type of phasing as it is related to the phychoacoustics of locating sound source?
I'm not sure I follow. By doing this, you're distorting both the phase response and the frequency response. To isolate the effects of the phase distortion from those of frequency distortion, you'd need to create the same effect using an allpass filter. You'd also need to test yourself blind to eliminate biases. Just my 2c.

EDIT: another way to test this (that might be easier with your current setup) would be to create a filter that has the same frequency response as the filter that results from your simulated crossover with delay, but done using minimum phase EQ. You could then use a blind ABX comparator (e.g. in Foobar) to compare this to the filter you've already created (which is ofc not minimum phase) to see if you can reliably discern a difference between the two.
 
Last edited:
Joined
Jun 20, 2018
Messages
112
Likes
26
#32
No problem :) With a dome tweeter in a waveguide, you will more or less eliminate baffle edge diffraction if the diameter of the waveguide is greater than the longest wavelength to pass through it, but you will also create diffraction at the mouth and (more critically) the throat of the waveguide. I think in most cases this is a worthwhile trade-off, especially considering the reduction in nonlinear distortion that a waveguide can bring, but it's something else to be aware of.



I'm not sure I follow. By doing this, you're distorting both the phase response and the frequency response. To isolate the audible effects of the phase distortion from those of frequency distortion, you'd need to create the same effect using an allpass filter. You'd also need to test yourself blind to eliminate biases. Just my 2c.
Thanks for the extra info regarding waveguide/DXT tweeters. I'll read up on this. I can make a baffle which has minimal to no diffraction for the tweeter in most directions but by mounting it close to the mid driver it'll surely do something there, if a DXT tweeter helps more than it hurts that'd be good.

As for the crossover simulation.
Yes, you're distorting both the phase and the frequency. Just like a real crossover does with two drivers. (except we're not modelling the off axis dropoff of the mid driver around the crossover frequency)
I personally don't care if the audiblity is from frequency distortion or from phase disortion, they're both there with real drivers off-axis and there's no getting around it (other than dual/concentric coaxial or single driver, or minimise it the way I plan on doing it with low crossover and drivers close together). I can tell that the overall smooth phase shift caused by an allpass (in this case the crossover simulation on axis response) is not audible to me for crossover frequencies 200Hz and up (haven't tested lower) even with 96dB slopes.
What I think is good about this simulation is that it shows just how audible the crossover is when listened to at various off-axis degrees. Perhaps if we model the exact frequency ripple with a linear phase FIR it will sound the same so perhaps it's the ripple that counts and somehow makes it so much more audible and nasty sounding than a single EQ dip wide or narrow. But this test should show clearly that it is indeed very audible and negative for audio transparancy. (btw I can suggest for instance hot mastered recent hit songs for this test as they seem to fill out the frequency spectrum to the max while still sounding musical and with a sense of depth and balance. To me out of a quick selection from my test list these made the effects most audible)
 

andreasmaaan

Major Contributor
Patreon Donor
Joined
Jun 19, 2018
Messages
1,017
Likes
437
#33
(btw I can suggest for instance hot mastered recent hit songs for this test as they seem to fill out the frequency spectrum to the max while still sounding musical and with a sense of depth and balance. To me out of a quick selection from my test list these made the effects most audible)
This is an interesting comment. I've read a few studies testing various kinds of distortion audibility in which it seems to be the case that - perhaps surprisingly - more compressed, artificial and/or repetitive music tends to be more revealing of distortion than high-quality recordings of acoustic instruments/events. I speculate that the repetitiveness may allow listeners to more easily lock onto a specific sound and more easily detect a change in its presentation.
 
Joined
Jun 20, 2018
Messages
112
Likes
26
#34
Btw, it just ocurred to me that perhaps even time / CSD has to be taken into account?
As we know a steep filter/EQ slope gives ringing. Everything steeped than a 6dB slope rings to some degree which is visible in time as well.
For the crossover if we use for instance 96dB slopes for the lowpass and highpass, there's no ringing at all on axis as the two cancel eachother out. But if we listen to it off axis the ringing becomes (very) audible.
Is it perhaps the case that the very minimal frequency ripple throughout the frequency spectrum (at any crossover slope) is so audible because the many dips in the ripple (because of minor to full cancellations depending on frequency and phase relation) have infact a very high slope? And therefore are mini ringing things rippling all over the crossover region? Thinking about it this seems likely to me, and the likely reason why they are so audible vs the low audiblity / very different sound of a single low-Q EQ dip.
 

andreasmaaan

Major Contributor
Patreon Donor
Joined
Jun 19, 2018
Messages
1,017
Likes
437
#35
I follow your logic, but am not so sure about your assumption/conclusion that the ringing is very audible. I'm not sure if you saw the edit I added to post #32, but this might be one way to investigate this further.

In general, studies of which I'm aware have demonstrated quite a high tolerance for off-axis ringing resulting from steep filter slopes before audibility thresholds are surpassed. I'm talking here about FIR filters (which exhibit not only post- but also pre-ringing) with hundreds of taps, for example.

But if you have the time and inclination it would be interesting to do further experiments on yourself and maybe also share some of your files. Your approach to learning about these things for yourself is clever and inventive.

In any case, you can't go wrong by ensuring that your drivers are properly time-aligned and your filter slopes within experimentally derived group delay audibility thresholds (whether obtained by you personally, or from the literature), so our discussion here is largely academic :)
 
Joined
Jun 20, 2018
Messages
112
Likes
26
#36
Oh I think we may have a few miscommunications here.

Off axis ringing of the crossover filter slope is a seperate thing. With some quick testing with headphones I did not have a preference for either linear phase FIR crossovers or minimal phase/analogue response (which is the same thing) filters.
I also found I prefer the sound of around 36dB filter slopes at the cancellation axis of the crossover as it struck me as a good balance between less bandwidth for the phasing around the crossover and the strength of the ringing. (would need to test it more to be sure I prefer it over 24dB or 48dB)

I didn't see your edit before but reading it I'm not fully understanding it.
I already did all my testing with minimal phase filter slopes (which is the same as the natural phase response of analogue crossovers). With Fabfilter Pro-Q2 I have the choice between zero-latency phase response (which is not natural or minimal phase), minimal phase response (which I used for testing) and linear phase response (FIR) which I also used for testing.
The difference between the minimal phase response and linear phase response is mostly in the the overal allpass phase response of the two filters combined and the moment in time of the ringing (not the length of the ringing). But for off axis-response both the linear phase response and the minimal phase response high and low pass behave the same regarding the (partial) cancellation ripple (and its associated ringing?) around the crossover and the amount of ringing of the low and high pass which occur at their cutoff point. So the difference is only time. Both linear phase and minimal phase crossover behave just as bad in any other aspect. But I'm probably telling you things you already know so I must misunderstand you, or you've misunderstood what I meant before :)

Yes discussion is indeed largely academic to find the cause of audibility :) But I can appreciate this too, and it wouldn't be the first time I learned something unexpectedly with practical results / change of mind / focus following such a discussion.
But indeed good idea to share files! I'll do this when I do more testing as to which crossover slope I personally prefer (I have to as I'll soon be selling my headphones too to fund my speaker build and this is possibly an easier way to A/B the results than with the speaker in place and switching between crossover settings.. though perhaps that wouldn't be too hard either with DSP presets and the remote.. hmm hope I don't get lazy)
 
Joined
Jun 20, 2018
Messages
112
Likes
26
#37
A quick correction.. I actually did my testing wrong for the 12 and 36dB crossovers.
What I did was lazy and listened to the cancellation axis by simply setting up the 12dB and 36dB crossovers and then listening without a delay thinking this is the cancellation axis. It is not. The cancellation axis for 12dB and 36dB is simulated correctly by switching the polarity of one channel and then delaying it according to 1000/(A*2) milliseconds. This will give a different result.
I'll do this correctly with my next test.
 

andreasmaaan

Major Contributor
Patreon Donor
Joined
Jun 19, 2018
Messages
1,017
Likes
437
#38
Yeh that's correct re: needing to invert the polarity for 2nd and 6th order filters (not 4th or 8th).

Reading post #36, I see what you're doing I think.

All of it makes sense, except that I still think you need to isolate group delay / ringing from any changes to the frequency response - and test yourself blind - to really draw any conclusions about (a) what effect(s) you're hearing and (b) what you prefer.

It's an ingenious way of "listening" to how a crossover behaves though. Would be keen for you to send some of your files when you get a chance :)
 

LarsS

Senior Member
Patreon Donor
Joined
Mar 23, 2016
Messages
498
Likes
163
Location
Sweden, Stockholm
#39
Really interesting thread, thanks for all the indepth discussions & inspiration.

Being a fan of AMT tweeters I've built two pairs of bookshelf speakers previously for my kids and couldn't be happier with them. I've got two stereo setups at home both with AMT tweeters (Gauder Akustik Arcona 100 & Evolution Acoustics MicroOne).

Thinking of starting another build of speakers. The idea being DSP & active speakers. Making the build easier with fewer pitfalls I'm thinking of basing it on this Troels kit excluding the passive crossover, http://www.troelsgravesen.dk/AT-SW.htm

1537432505379.png
 

BE718

Major Contributor
forum experimenter
Patreon Donor
Joined
Mar 1, 2016
Messages
2,107
Likes
857
Location
Perth Western Australia
#40
Thanks for sharing your experience with tweeters.
I'm still looking at tweeters but have been looking at vertical behavior as well and the AMT's have the most narrow vertical response I've ever seen. As in rediculously narrow, for instance moving just a a few centimeter or so vertically at 1meter distance seems to already change the frequency response significantly.
This may well be their biggest difference in sound compared to dome tweeters?
This isn't shown in their hifi catalog but in their pro catalog you can see the vertical responses: http://www.mundorf.com/PDF/Mundorf_proAMT_Catalog.pdf
For the smaller tweeters with horns this effect is much less so, but for their bigger ones (like the AMT29 surely) without horn the vertical response is like a laser beam.
I'm guessing the special sound from these tweeters is due to their extreme narrow vertical dispersion which virtually eliminates any floor and ceiling reflections. Can understand the problems in matching them with a mid driver in this regard as well, would perhaps be better matched without mid driver but directly with an 8" woofer which is somewhat narrow in dispersion around the 1-2kHz crossover as well (though the same in both vertical and horizontal directions).
Having owned speakers with AMT drivers I can't say I noticed this back then, but I was young and unexperienced (more than 15 years ago).

edit: but the smaller ones look nice! Like the AMT 21CM2.1-C. Though not suitable for a <2kHz low crossover point it seems at first glance.,
One of the reasons I chose the AMT 23 was due it having better vertical dispersion than many of the other models.
 
Top Bottom