I guess I would look at it from the opposite point of view. The notion of taking a continuous audio spectrum and splitting it into sections, one section here and one section there, this is clearly an unnatural compromise forced on us by the inadequacy of drivers and the inability to do a very good job at radiating wavelengths of over a 1000 to 1 range from a single driver.
The art of speaker design, then become the practice of splitting the spectrum up and sending it to separate units that will recombine in space in the most seamless way possible. That being the case then I want to find the best compromise: splitting more ways makes the job of smooth recombination more difficult. At the same time you do get a few more degrees of freedom for controlling system frequency response with a 3-way, say, than a 2-way.
What is the optimum? Different designers may have different opinions but I find it easy to make a smooth and flat 2-way and can probably do a little better (with considerably more effort) with a 3-way. That is, the 3-way will have more uniform directivty or off-axis response. I can get slightly flatter response with 4 crossover corners rather than 2. Even with a 3 way, though, the job gets significantly more difficult since the lower corner of the midrange isn't independent of its upper corner. Things start to interact making crossover optimization difficult.
I certainly disagree with the notion that a 12 way system lets each instrument have its own driver. The instruments all cover a broad range of frequencies, so no matter how you slice it mutiple instruments must play over each driver.
If you are approaching this in an ideal way you must be giving each driver less than an octave of spectrum? Otherwise you must have multiple units covering overlapping frequency bands. I guarantee that approach will lead to a very messy 3 dimensional response. Can you show us polar curves or on and off axis frequency response?
David S.
The only 4-ways I've done were essentially 3-ways with a built in subwoofer. For power handling reasons the system had a 6 1/2 or 8" lower midrange. I recall that the subwoofer crossover was to be 100 Hz-ish and so a 3 way with 5" mid wouldn't have had the right output capability above the sub's range. Changing to 4-way made it work. (There was more than a little marketing input to this as well
)David S.
First of all your right in one point - the design causes more work than any other design.
second I´m sure that your experience in design of simple speakers might solve many issues -
but for sure there are more issues that you have not ever really touched - to go to the points that you stated / requested : the frequencyarea that a speaker is limited to operate is only 5 to 6 tones
and the crossovers used in the system are crossovers of forth order - a quality you probably might never see within a commercial system due to the fact that it is too expensive and its too much work to measure hundreds of capacitors and coils to find the correct fitting values - a task that no company would ever try to solve due to the fact that cost would explode..... and each speaker must be calibrated to a spcific given level of soundpressure emitted by the speaker to ensure the precise overall performance of all speakers in the system - and thats allready the next reason you´ll never find a company that would pick up such a task.
I´m sure you have viewed a lot of magazines that discuss the one or other issue of the task to design a speaker system and how to solve that one or other problem.
But allthough also a lot of issues are discussed without a given solution you might think of several problems to solve and at the same time
leave other problems beside due to the fact that a solution is related with to much efforts. Thats exactly the difference to this speakersystem.
This speakersystem was built under the focus to build only once a superior speakersystem and to carry out all tasks and to solve all issues and aim for only one goal : to get the best possible and technicaly availible speaker system
and to accept that that task forces the decesion that you must perform work and more work and spend hours for measuring and collecting parts that realy fit exactly with the needed values in the crossovers without
dealing with compromizes. And as explained: the measurements will be performed within the next 3 to 4 weeks - and the only reason that it needs that time is to get 3 points together:
first the transportation of the "fridges" ( my nickname to the speakersystem - bear in mind one maincase has weight of more than 90 Kg´s) to a studio in that a friend of mine usually is working as professional sound engineer,
second to get again that equipment boroughed from another friend, that has a job at Rhode&Schwarz as electronic engineer and that was allready used at that time when the speakers have been calibrated ( i guess you know that that is the highest quality of measurementequipment that can be used in such a task ) and third a free saturday or sunday that gives us (the friend of mine that is soundengineer and me ) the possibility to spend an entire day in the studio to use the echodead-room ....
but from the measurements performed during the process of speaker calibration i know one thing for sure: that graphs look nearly like drawn with a ruler....
the only reason that they must be repeated is,
that at the time the speakers have been calibrated I did not waist a single thought to publish a description of all the steps of the project...the first time thinking about documentation was two years ago after several friends asked me to perform a documentation.... and i just started that task one week ago and I´m just now searching for a bunch of other Photographs taken in those days when the soundsystem was built... and that was about some six or seven years ago...
Last edited:
Yikes!! 11 crossovers * 360 degrees = 3960 degrees of phase rotation. The group delay caused by phase rotation is generally inaudible with music and normal speakers but I'd be surprised if this one weren't audible. We're talking about the harmonics of an instrument arriving way before the fundamental.
Sounds familiar; seems the working bandwidth of a driver is proportional to frequency. All the four way designs I've fooled with have essentially ended up with a sub for covering the excursion limited region, crossing off to a woofer above that to limit IMD and PMD.
Depends on the listener; haven't seen any double blind studies, though I did come across one single blind result where a third of listeners didn't discriminate, a third prefered warped phase (nonzero group delay), and a third preferred linear phase (zero group delay). Didn't save the link though.
Personally I scored 100% correct between linear and warped phase LR2, LR4, and LR6 within the limits of my ability to approximate double blind A/B testing last time I tried it; I always prefer linear phase.The crossover in this 12 way may not be implementing what Linkwitz refers to as the "correct" topology and, even if it is, the earlier filters in the chain won't have achieved a full 360 degree rotation before the next one comes in (neither will have the last one by 20kHz). I'd have to sim it, but my guess would be the total group delay winds up closer to LR8 in the "correct" topology.
the difference between guess and argument
The last quotes in the thread mostly cover the crossover - and reading them, it seems to me that a lot of the objections are resulting from the fact
that none of the people there and duscussing about the basics have ever really done experiments by themselves in practise - neither did they really see the tables with the values of the components in the related webpage - but they seem to claim beeing experts on the issue because they have read something about the issue.....
that is really fascinating - what is understood by a lot of people within the word "expertise".....
First point: If you have ever really calculated filters and then tried to build them, you should bear in mind what happened in reality:
you had some component values that where not realy within the row of values of the components and you bought the components in the store with the closest values ( but not realy the ones mathching exactly ) and therefor you have been forced by the availible components to carry out a compromize with very bad results depending to the size of not hitting the correct values calculated for the filter ( in the worst case 20% !)and therefor shifting the phases.As I explained in the web ( what you seemly not did read .... ) if you want the crossover to really operate exactly within the parameters that have been calculated by math you MUST pick up hundreds of coils and capacitors and measure them to find those with the EXACTLY MATCHING VALUES ( this is of course a lot of work and I really doubt, that any of the people discussing the issue ever carried out that task.... ) to get a precisely operating filter !
The next I doubt that any of these people has really set down for hours and done measurements with a dualchannel oszilloscope and examining filters by themself !
There is a fascinating effect within the filters that you might only find out, if you really examine what happens within a filter - instead of reading about it:
the phase of turn within the filter is dependent to the frequency.... did you ever really think about what that really means ?
In reality the frequencies the filter has been designed for to pass, pass the filter with nearly no phaseshift ! And thats the true fact ! The phaseshift occurs in fact to those frequencies that the filters was designed to cut off - i.e. the point where the filter is designed to operate as lowpass or as highpass !
Within the area of the frequencies where the filter is determined to operate and are determined to block the frequency !
This you only recognize by experimenting by yourself with the oszilloscope and realy understanding whats happening within the filters.
That phaseshifts that are viewed behind the filter happen within a crossover are 99% resulted by the lack of using not components that have been exactly calculated and therefor two filters siding each other cause overlaping of the frequncies due to tolerance of the components ( up to 20% ) and the resulting shift of the point where the filters have been calculated to start operation...... and that mess can bee avoided .... if you start using exactly measured and matching values .....
As explained within the webpages ( you remember that i mentioned that they should be really read previosly before issueing arguments instead of guesses and objections ? ) the spread of tolerance of the components even cause within series of good commercial speakersystems that much difference, that the bandwidth of quality reaches from unacceptable to brilliant and that that has to be compensated with other tricks like stuffing and several of the speakers even then can´t be rescued and the PCB of the crossovers needed to be changed to get acceptable results......
And finally: in German language there is a sentence used by technicians : "Wer misst , misst Mist." that is to be translated "who measures, measures often garbage"
The reason for that sentence is that measurements are very dependent to the circumstances and the point of measurement and that slight differences might lead up to large difference of resulted values of the measurement. This is a very interesting view to remind, when looking at measurements made within speakersystems. The measurements mostly made, have been made behind the filter and not within the filter....
whats the effect to the results ?
Well - the mainpart of the viewed disturbance don´t come from the filter itself but from the mix of the output of two filters and the fact that in the two things hit together: the one frequency that was calculated to pass the filter and the mix of disadvanteges caused by lapover of the two filters in an operating area ( where they have been designed to block the frequency ! and of course where the tolerances of the used components add together with disadvantages in missing the calculated values of the filter ) with phaseshifting - the result is audible - and tends from bad to worse....
and its the true result of not really paying attention to use exactly calculated values in a filter....
Sounds familiar; seems the working bandwidth of a driver is proportional to frequency. All the four way designs I've fooled with have essentially ended up with a sub for covering the excursion limited region, crossing off to a woofer above that to limit IMD and PMD.
Depends on the listener; haven't seen any double blind studies, though I did come across one single blind result where a third of listeners didn't discriminate, a third prefered warped phase (nonzero group delay), and a third preferred linear phase (zero group delay). Didn't save the link though.
The crossover in this 12 way may not be implementing what Linkwitz refers to as the "correct" topology and, even if it is, the earlier filters in the chain won't have achieved a full 360 degree rotation before the next one comes in (neither will have the last one by 20kHz). I'd have to sim it, but my guess would be the total group delay winds up closer to LR8 in the "correct" topology.
The last quotes in the thread mostly cover the crossover - and reading them, it seems to me that a lot of the objections are resulting from the fact
that none of the people there and duscussing about the basics have ever really done experiments by themselves in practise - neither did they really see the tables with the values of the components in the related webpage - but they seem to claim beeing experts on the issue because they have read something about the issue.....
that is really fascinating - what is understood by a lot of people within the word "expertise".....
First point: If you have ever really calculated filters and then tried to build them, you should bear in mind what happened in reality:
you had some component values that where not realy within the row of values of the components and you bought the components in the store with the closest values ( but not realy the ones mathching exactly ) and therefor you have been forced by the availible components to carry out a compromize with very bad results depending to the size of not hitting the correct values calculated for the filter ( in the worst case 20% !)and therefor shifting the phases.As I explained in the web ( what you seemly not did read .... ) if you want the crossover to really operate exactly within the parameters that have been calculated by math you MUST pick up hundreds of coils and capacitors and measure them to find those with the EXACTLY MATCHING VALUES ( this is of course a lot of work and I really doubt, that any of the people discussing the issue ever carried out that task.... ) to get a precisely operating filter !
The next I doubt that any of these people has really set down for hours and done measurements with a dualchannel oszilloscope and examining filters by themself !
There is a fascinating effect within the filters that you might only find out, if you really examine what happens within a filter - instead of reading about it:
the phase of turn within the filter is dependent to the frequency.... did you ever really think about what that really means ?
In reality the frequencies the filter has been designed for to pass, pass the filter with nearly no phaseshift ! And thats the true fact ! The phaseshift occurs in fact to those frequencies that the filters was designed to cut off - i.e. the point where the filter is designed to operate as lowpass or as highpass !
Within the area of the frequencies where the filter is determined to operate and are determined to block the frequency !
This you only recognize by experimenting by yourself with the oszilloscope and realy understanding whats happening within the filters.
That phaseshifts that are viewed behind the filter happen within a crossover are 99% resulted by the lack of using not components that have been exactly calculated and therefor two filters siding each other cause overlaping of the frequncies due to tolerance of the components ( up to 20% ) and the resulting shift of the point where the filters have been calculated to start operation...... and that mess can bee avoided .... if you start using exactly measured and matching values .....
As explained within the webpages ( you remember that i mentioned that they should be really read previosly before issueing arguments instead of guesses and objections ? ) the spread of tolerance of the components even cause within series of good commercial speakersystems that much difference, that the bandwidth of quality reaches from unacceptable to brilliant and that that has to be compensated with other tricks like stuffing and several of the speakers even then can´t be rescued and the PCB of the crossovers needed to be changed to get acceptable results......
And finally: in German language there is a sentence used by technicians : "Wer misst , misst Mist." that is to be translated "who measures, measures often garbage"
The reason for that sentence is that measurements are very dependent to the circumstances and the point of measurement and that slight differences might lead up to large difference of resulted values of the measurement. This is a very interesting view to remind, when looking at measurements made within speakersystems. The measurements mostly made, have been made behind the filter and not within the filter....
whats the effect to the results ?
Well - the mainpart of the viewed disturbance don´t come from the filter itself but from the mix of the output of two filters and the fact that in the two things hit together: the one frequency that was calculated to pass the filter and the mix of disadvanteges caused by lapover of the two filters in an operating area ( where they have been designed to block the frequency ! and of course where the tolerances of the used components add together with disadvantages in missing the calculated values of the filter ) with phaseshifting - the result is audible - and tends from bad to worse....
and its the true result of not really paying attention to use exactly calculated values in a filter....
Last edited:
You really don't understand filter design at all! Inductors can be unwound to get an exact value. It's not rocket science! Slightly altering capacitor values to stay within preferred values only slightly alters the filter Xover point and does to both the high pass and the low pass. The higher order the x-o the harder it gets but it is still not rocket science.
Some of us have been doing this for years, a few as their livelihood! What you want us to believe is that you have applied your mathematical mind and whoa ....... you found the holy grail!
Good for you! Let's see some measurements to back up your so called science! Building without measurements ..... pah! I don't mind you deluding yourself! I don't care for your attitude!
Don't try and delude the noobs with your nonsense, it's not fair on them.
Your claims so far are nothing but unsubstantiated male bovine droppings.
Some of us have been doing this for years, a few as their livelihood! What you want us to believe is that you have applied your mathematical mind and whoa ....... you found the holy grail!
Good for you! Let's see some measurements to back up your so called science! Building without measurements ..... pah! I don't mind you deluding yourself! I don't care for your attitude!
Don't try and delude the noobs with your nonsense, it's not fair on them.
Your claims so far are nothing but unsubstantiated male bovine droppings.
pheonix358 said:
What`s up Terry, not amused anymore?
Apparently, as me, You`ve to upgrade Your "male bovine meter" with some kind of (a very) "soft-start"
In many years of more or less professional speaker design I have learned a thing or two. For example: Component values with five percents of tolerance are absolutely fine, this is mostly far better then speaker parameter tolerances.
And if i look at the... well, not very high value drivers used here, I doubt that it is possible to build two somewhat identical speakers at all.
And no, you can't compensate speaker parameter deviation with changing crossover components (at least in most of the cases).
And if i look at the... well, not very high value drivers used here, I doubt that it is possible to build two somewhat identical speakers at all.
And no, you can't compensate speaker parameter deviation with changing crossover components (at least in most of the cases).
This being a DIY site there are a number of accomplished speaker designers here, both amateur and professional.
I've visited the site a couple of time (per post #9) and the English links seem to be broken?
As Phoenix says, its easy to take a few turns off of a coil and I've done that many times. For mass production I've always been able to adjust the design to standard values without compromise, but perhaps my standards are rather low!
No, what does it mean? The filters are 3 terminal devices. We care about what happens at the output. In fact our primary interest is the final combination of filter and driver, and how one acoustical section mates to the next.
Phaseshifts are due to component tolerance? No, the phase shifts are inherent in the topologies and values used. "Errors" in those values can have proportionate effects on corner shape and frequency and lesser effect on phase rotation.
So if we try every Chevrolet off the line some will be Skodas and some will be Ferraris?
Stuffing is a good trick.
Yes, yes, I'll have to start measuring within my filters. Above and below them too.
True words. This is the cause for most of our scepticism, that you can configure a system that smoothly blends together 12 sections, especially with a seemingly random assemblage and placement of drivers. Crossover component tolerances aside, most of us would start with a theoretical modeling of such a complex system to see if a series of sub-octave sections can be effectively combined and what the phase consequences of narrow band 4th order (8th order?) bandpasses would be. (We are getting close to adding the outputs of multiple high Q resonators here, generally not the recipe for a transparent system.) If the system configuration doesn't work well in modeling, then a years worth of crossover component selection isn't likely to fix it.
We look forward to seeing the measurements of your system.
David S.
Last edited:
What`s up Terry, not amused anymore?
Apparently, as me, You`ve to upgrade Your "male bovine meter" with some kind of (a very) "soft-start"
If it's a joke it goes too far. I am thinking that he is for real.
No one can design properly without measurements, not and come up with a design that beats the entire world and not from a collection of miss- matched drivers that rightly belong in a speaker museum. My sense of humor is overwhelmed by the attitude within these posts. My meter is reading far to many fluids and little if any solid material.
Terry
Here are pix off the site. my concern with this speaker is all the electronics used to integrate all the drivers with eachother. In my mind, it seems there could be a loss of something in the music.....maybe not.
Attachments
"So if we try every Chevrolet off the line some will be Skodas and some will be Ferraris?"
I would rephrase that since Skodas, just like Audi, use VW technology but are better built than either.
BOT how do you even design a 12 way passive xover? I find even 3 or 4 ways fiendishly complex, one of the many reasons I use active ones. As far as I understand it one cannot just cascade a number of passive filters without them interacting on each other (could be wrong on this though since IME actives simply sound better). Btw I have built myself a 4way active system and it just so happened that each driver covers around 3 octaves. Works for me…
I would rephrase that since Skodas, just like Audi, use VW technology but are better built than either.
BOT how do you even design a 12 way passive xover? I find even 3 or 4 ways fiendishly complex, one of the many reasons I use active ones. As far as I understand it one cannot just cascade a number of passive filters without them interacting on each other (could be wrong on this though since IME actives simply sound better). Btw I have built myself a 4way active system and it just so happened that each driver covers around 3 octaves. Works for me…
Nobody remembers this thread ?!?
http://www.diyaudio.com/forums/multi-way/169296-thinking-building-7-way-loudspeaker.html
It was the "hot" project for the summer !
Every season brings more...
http://www.diyaudio.com/forums/multi-way/169296-thinking-building-7-way-loudspeaker.html
It was the "hot" project for the summer !
Every season brings more...
Yes, of course provided this are about the RIGHT values to begin with (which means they have to approximate the acoustic filter responses together with those of the drivers) and not only some resistor & xover frequency values thrown in a textbook formula and "calculated".
Try and design a 15Hz xover and then tell me where the components are. Then do the same for 30, 60 and 120. Yea, that's going to work! I assume that's a pic of the "before I put it together" cause there is a lot of wiring missing> Those electro caps are going to work well. Just the thing for a world beater.
If it's a joke it goes too far. I am thinking that he is for real.
Terry
Well, I could say now: "I`ve told You so" .....but I don`t
This "design" and above all the attitude of it`s "creator" is the most unbelievable and hilarious thing I`ve read about in an audio internet forum EVER.
Ronin -- may I suggest that before you continue to berate people here for not having read all your documentation, you should first finish the translation? Most here don't speak German, and the links at Basisseite Final Solution 04 don't work!! The only thing that can be read is your description of yourself.
Or are only German readers allowed to comment on your speaker system?
Until that translation is available, perhaps you could describe how you handle the off-axis responses of the two tweeters down at the bottom of the baffle, down near the floor? You are aware, aren't you, the frequency response and level of drivers changes at different listening angles -- and dramatically so for horns like those?
Or are only German readers allowed to comment on your speaker system?
Until that translation is available, perhaps you could describe how you handle the off-axis responses of the two tweeters down at the bottom of the baffle, down near the floor? You are aware, aren't you, the frequency response and level of drivers changes at different listening angles -- and dramatically so for horns like those?
In fairness, most of the links do work, but you need to click on the numbers to the right rather than the text.
Frankly, I still regard this whole business as quackery. To each their own I suppose, but the waffle about drivers having very narrow optimal operating windows smacks rather too much of MIT wire's (give me strength) 'poles of articulation.' As far as I'm concerned, if a driver did indeed possess such a narrow useful BW, then it is clearly unfit for purpose and should be immediately replaced with a unit possessing acceptable performance.
Even if you did decide to buy into this theory, the attendent problems of multiple XOs, several of which are slap-bang in the middle of our critical hearing BW, with disparate types of driver seemingly scattered across the front baffle without regard for system polar response do tend to rather detract from the appeal. The arbitary statements presented as fact without a single measurement to support them, and the fact that this alleged speaker (since we haven't actually seen any evidence of it in operation, let alone of its reputed performance) has apparantly also been created without any such tests being made, lead me to regretfully conclude that this device, should it actually exist, is best assessed by my close friend attached.
Incidentally, are those woofers at the top old Sansui units? They look like they've been pulled from an SP3500 to me, or something akin. If they have, it's a crying shame. Those old speakers, despite their limitations, have quite a following.
Frankly, I still regard this whole business as quackery. To each their own I suppose, but the waffle about drivers having very narrow optimal operating windows smacks rather too much of MIT wire's (give me strength) 'poles of articulation.' As far as I'm concerned, if a driver did indeed possess such a narrow useful BW, then it is clearly unfit for purpose and should be immediately replaced with a unit possessing acceptable performance.
Even if you did decide to buy into this theory, the attendent problems of multiple XOs, several of which are slap-bang in the middle of our critical hearing BW, with disparate types of driver seemingly scattered across the front baffle without regard for system polar response do tend to rather detract from the appeal. The arbitary statements presented as fact without a single measurement to support them, and the fact that this alleged speaker (since we haven't actually seen any evidence of it in operation, let alone of its reputed performance) has apparantly also been created without any such tests being made, lead me to regretfully conclude that this device, should it actually exist, is best assessed by my close friend attached.
Incidentally, are those woofers at the top old Sansui units? They look like they've been pulled from an SP3500 to me, or something akin. If they have, it's a crying shame. Those old speakers, despite their limitations, have quite a following.
Attachments
Last edited:
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.