Geert - Hybrid ESL

[GeSo]

Hello All I am new to this forum.
I ordered the books by Sangers and Wagner. I have been reading tons of sites while I wait for the books to arrive. I am very impressed with Geert Vij..... design and build.
I am interested in building a 5.1 surround system, 1 pair of hybrids for the front and for the center and surround just the panels if possible.

I understand why ESL's need to be away from the wall prferably 15" - 24". How are Magnepan and Final able to design wall mount panels? I would prefer that my surrounds are off the floor and preferably wall mounted if at all possible.

On one of the sites that I read (HTForum.nl) under d'underground? I believe there is a diagram with dimensions etc of Geert's build but I can no longer find it online. Does anyone know what the cut out is for the nail section and nail spacing?

Does the center channel have to be the same size as the front's to have the same frequency output, minus the cone driver of course.

I have also read that some builders uses perforated metal with different hole sizes to optimize full range performance. For example larger holes for the lower frequencies and smaller holes for the mid / high. With that said can that same therory be used with the wire stator? Using wider spacing for the lower frequency's and closer spacing for the mid / highs?

 

[Calvin]

Hi,

You know....sometimes a working link or a pic can be very helpful. Uncompleted names aren´t familiar to everyone.

The only idea of how to place a panel close to a wall that appears to me is to use reflectors on the backside of the panel to reflect the sound sideways. Without any waveguiding apparatu´ it´d simply sound terrible (to me the finals are very short of sounding terrible anyway.....but thats only personal taste)

I´m not into HT, but from what I know the sound of the centre channel should be the same as the main channel´s (though I would rather omit the centre channel at all and invest more into good main channel speakers). The problem with most panels now is, that they are intended to be used with the longest dimnsion in the vertical direction.
This will probabely disclude the use as centre channel. Alternatively the panel might be turned horizontally, but then it should be a large panel to accomodate for two or more listeners. I don´t know of any commercial product at the moment that features such a large panel.
Curving the panel might be a solution, but there is no high quality centre of this kind around.

I don´t know any good reason to use different hole sizes, especially not with larger holes for the bass and smaller ones for the midhighs.
The other way round would be correct efficiencywise! Smaller holes for the bass and larger holes for midhigh (but that would introduce other disadvantages). So the best solution is to use just one hole size with small holes. The same holds true for wire stators. Thinner wires with close spacing to each other for increased efficiency.


jauu
Calvin

 

[gvy]

Hi Geso

I think, that I am Geert Vij....
I have build "d'ondergrondse" , my home theatre including the ESL's and subs about 1,5 year ago.
The description of the building process of the complete theatre (in dutch) can be found here

http://www.htforum.nl/yabbse/index.php?topic=47163.0

page 15-20 handle about the esl's

or here for just the ESL's
http://www.zelfbouwaudio.nl/forum/viewtopic.php?t=3917

On the foto , you can see the system.
I do not use a center .
I do use 2 subs, and they are placed on the best acoustical location ( mid of the long walls, on the right and left of the listening place)

An externally hosted image should be here but it was not working when we last tested it.

greets
Geert

 

[GeSo]

I wasn't sure if I could post link and or pic since I am a new member. Figured a quick Google search would suffice.

The gents’ name that I am referring to is Geert Vijncke one of the links I have found of his build is http://www.zelfbouwaudio.nl/forum/viewtopic.php?t=3917

I have not listen to Final's but have and very much enjoy the sound of the Magnepan's.

So mounting them against the wall is not an option, understood.
After the front and rear channels are built I will determine if a center is needed. My next question is wire spacing; I believe Sanders recommends 10 wires per inch I have also read 15-20/inch. Does this all relate to the diameter of the wire? I have access to 24awg kynar wire. What would be the optimal spacing for such wire?

 

[GeSo]

Hello Geert and thanks for the reply,

Since I am new here my posts take a pit to show up on the site.

I am wondering if you have a CAD layout of your build? I am curious to know the HxWxD and spacing of the nail area. I have plans to build a SonoSub using a Shiva driver from www.DIYcable.com. I might as well use the same technique for the hybrid as well. I currently have a pair of Dayton Audio aluminum drivers from here.

http://www.partsexpress.com/pe/pshowdetl.cfm?Partnumber=295-335

Wondering if this would be comparable to the drivers you are using? Last question for now is how do you think your surrounds would sound if they were raised off the floor a bit without the help of the driver? I know they will not go into the low frequencies. I forgot where I read this but I thought there is a frequency cut off sent to the rear chanels? Thanks for now.

 

[Calvin]

Hi,

mounting the panel close to a wall without any reflecting/waveguiding device is no option.
You must assure that the ´reflected soundwaves´ are not passing the membrane. So a reflector with a triangular shape behind the panel might be the solution. (it won´t be as good as a freely positioned panel, but it should work sufficiently).

When designing a panel I start with thinking about a catalogue of goals.
a) what is the freq-range the panel should work over? -->Bandwidth of the panel should be ~1-2 octaves more.
b) what size is possible --> always use the largest dimensions for highest dynamics
c) what d/s is needed --> depending on the lower bandwidth limit and size of the panel´s segments --> use the smallest d/s possible
d) decide what stator material and concept You want to use
e) decide which type and dimension of materials you need
f) check about availability of materials and the possibilities of manufacture
g)....
h).... etc etc.

jauu
Calvin

 

[gvy]

Hi GeSo,

It is all in the dutch description :
Some dimensions:


panels outer dimensions : h = 1240mm x w = 250mm
membrane free moving dimensions : h = 1160mm x w = 170mm

wire incl isolation : diam = 1 mm (24AWG wirewrap)
there are 72 wires next each other
acoustical opening = 55% : ( 45%= wirewidth, 55%= opening)

wire-membrane distance = 1,5 mm ( used spacer material = PS plastic , polystyreen 2.5mm) 2.5mm(spacer) - 1mm(wirediam) = 1.5mm

Geert

 

[GeSo]

Full range panels

I have read the armature should not attempt to build a full range panel for their first project. After reading Wagner's book, reviewing The Audio Circuit site and going over Geert build I am confident that I can build a full range panel. The question is sound quality, electronics and efficiency. Mind you this is going to be for a home theater. The stator size I was thinking about would be 10” x 60” for the front channel and 10” x 30” for the rear, 12 wires per inch 30awg Kynar wire. I plan on using 3M tape for both the spacer and glue. For the diaphragm I was thinking about ordering DuPont Mylar. My goal is to try and cover as much of the audible range as possible, with the subwoofer handling the lower tones. I am clueless when it comes to the electronics. Which ratio would go best with my panels and who should I buy them from? I am also stuck on the coating as well.

FYI, the build process for the panels will be identical to the way Geert built his.

 

[Calvin]

Hi,

please do Yourself a favour and listen to the ol´chaps and build a hybrid first. Wagner´s FR and most of the projects from the audio circuit leave a lot to be desired, especially regarding dynamics and the demands of HT. The chances to build a good midhigh-panel with the first shot are low already. The panel will work of course -hey, nothing is easier to build than a ESL that puts out ´some sounds´.
It could be a revealing experience and it could even be that the panel outperforms everything you have listened to yet...but it won´t be a really good panel. The best panels of today are soo much superior to any dynamic driver that not just a few listeners doubt their ears on first contact!
It´d be rather the beginning of a long journey with increasing capabilities of the panels the more you´ve built.
Beginners are usually overwhelmed when they first listen to a ESL-panel even when it is a rather bad one. And usually Beginners make serious failores within the building process.
The most common failures are:
- trying to build a FR-panel
- low resonance Fs
- ignoring efficiency issues
- thinking of beeing able to do better by doing different (Why the *** did he do that? It looks soo weird, so I think I´ll better do it different)
The results are panels with horrible frequency response, abnormally low efficiency --> crappy dynamics...and low instability treshold.

Since You intend to use the panels for HT you need panels that can put out serious levels of SPL. This can only be achieved with small d/s and large membrane areas. For a Hybrid working from ~200Hz on upwards dimensions of 10"x50" and 1/25" (d/s) will give good results (my panels of this size reach a stunning 110dB@4m with app 50W of input power with exceptional low values of distortion).
But don´t expect to get anything useful below 200Hz! Add a proper dynamic bass here. A wire stator won´t reach the high levels of dynamics like a sheet metal stator, but with the recommended dimensions You could expect more than 100dB@4m without the need of excessive drive power.

With such a tall and thin panel as You suggested phase cancellation must be corrected for below 800Hz(!). The level of equalization becomes so high below 200Hz, that this alone kills everything what is needed for good sound, dynamics and HT-usage. This is a problem every open baffle system faces, regardless of its dimensions. The only thing that ´saves the day´ is the huge high-Q bass resonance.
But -and it is a very big BUT- this is a one note bass featuring a high Q-factor, hence a soft sounding bass with low precision and a serious suckout in the upper bass region. Nothing one would accept with a dynamic bass, especially since this goes with thin and anaemic sounding vocals.
You would have to correct for the high Q and need considerable equing too. All of which kills output and dynamic headroom, especially when executed passively (it costed the old ML Sequel roughly 10dB of output and efficiency and this was crossed over at >250Hz!!). Staying above 200Hz eases the effort You have to put into the crossover too. A simple 2nd order Highpass (2 components) and a parallel Notch (needs 2 or 3 components) may be all You need. Without the possibility and knowledge(!) to measure the speaker´s behaviour it´ll be impossible to equalize a FR-panel.

A FR-ESL that could meet the dynamic requirements of HT needed to be much wider than Your panels...40" or more. Besides much reduced demands in building and greatly reduced size a hybrid-ESL would still reach better results with regards to sonics, dynamics and drive requirements.

jauu
Calvin

 

[GeSo]

Thank you Calvin,

I do not think I can build ESL’s better nor am I trying to reinvent the wheel here. My goal is to have panels that sound good and that are aesthetically pleasing. If I must go hybrid then I will. All four panels will be 10”x50”, what specs should I looks for with the driver? I noticed the NSS 0.3a uses a 7” woofer, as stated above I have a pair of Dayton 7” woofer. That can be found here http://www.partsexpress.com/pe/psho...tnumber=295-335 will this be ok? I imagine it should be fine since I am going to us a subwoofer as well.

Now that we agree that I am building a hybrid system can I have some advice on electronics? What size transformer bias supply, crossover etc would you suggest and from where? I am also having difficulty finding coating and diaphragm material in the states. Can anybody help with this?

 

[Calvin]

Hi,

nothing is a must! It´s just the knowledge and decades long experience that leads to the recommendations.
The great secrets of how to build a Hybrid that could fulfill highest demands with regard to dynamics and even sonics could be found in the design concept of the -unluckily now obsolete- ML Statement II.
To my opinion the most stringent and consistent design concept around. You may find points which leave something to be desired -even in this costly product and mainly in the execution of constructional details- but its the conceptual beauty, the basic ideas behind, the way how different design aspects neatly fit into a magnificent flawless whole that made this speaker truly outstanding.

So why not orientate at this masterpiece and take a look at the details?
- A large well executed panel with small d/s. This gives the efficiency and the dynamic headroom to handle extreme SPLs with lowest distortion above 250Hz. The panel exhibits a dipolar cylindrycal distribution character.
- A thin multi-driver bass tower featuring 8 rather small drivers in a dipolar casing. Ranging from app. 60Hz to 250Hz the drivers are not overly stressed (remember that a dipole bass needs more excursion for the same SPL as other casings!). Besides increasing dynamic range and reducing distortion by the use of multiple drivers this too creates a distribution character similar to the panel´s. At least in a frequency range of 1 to 2 octaves around the crossover frequency it shows the dipolar cylindrical shape eventually transitioning to a dipolar lobed character towards lower freqs.
- A large CB subwoofer. The dipole is limited with regards to bass extension especially in smaller rooms (though the simple theory and practise aren´t the same here). Still though a more ´classical´ woofer will be much more suitable for HT demands and will reach much greater dynamics and/or will cost much less effort.
You don´t necessarily need the subwoofer for music when You use eight 7"- or larger drivers with longthrow capabilities per side, but I´d recommend the use of a sub for HT. You may fudge a bit by using 4 or more drivers in CB or BR in a stacked configuration (tower) thereby omitting the need for a subwoofer alltogether. This would look like a smaller incarnation of the first ML Statement.
The transition from panel to bass may be slightly noticable though, whereas with the use of the dipole bass the transition will not be recognizable at all.
- A well executed crossover. The easist would be to use a digital crossover. There are some very priceworthy models on the market using IIR-filters. Some of them even fit in a slot of a partnering class-d amp (which is probabely the cheapest and one of the best ways to drive the panel apart from a beefy SE-Triode!). Additional EQs allow for some room correction. Configuration of the filters via ´filterblocks´ is quite easy and straightforward.
Costlier models featuring FIR-filtering and more filtering options (some of which are impossible with the analog and IIR filters) are superior sonically, but the latency time (especially when correction for group delay) can become critical with HT. In rooms with critical acoustics these filters are best.
Sonically the best (to my ears) is still an elaborated active analog crossover using simple discrete JFET-Buffers. But the implementation of additional filtering for equing and room correction quickly raises the complexity and part count. So this is just for rooms with good acoustic behaviour and a rather low filtering demand/complexity.


The smaller Daytons look well, especially for the ´cheaper solution´ using 4 drivers per side in CB- or BR-cabinets. For a dipole cabinet You´d need min. 6 preferably 8 of them and some equing (basslift etc. ). You could equalize down to ~35Hz without killing dynamics but no more. This can be low enough even for HT but this depends on Your demands. Building the large solution with a big subwoofer crossing over at ~60Hz, the dipole needs nearly no bass lifting which increases dynamic headroom remarkably. You may start with the 8 driver dipole and upgrade later with a dedicated sub. The concept is quite flexible here.

jauu
Calvin

 

[Too Tall]

Hi,

snip-

You would have to correct for the high Q and need considerable equing too. All of which kills output and dynamic headroom, especially when executed passively (it costed the old ML Sequel roughly 10dB of output and efficiency and this was crossed over at >250Hz!!).

Staying above 200Hz eases the effort You have to put into the crossover too. A simple 2nd order Highpass (2 components) and a parallel Notch (needs 2 or 3 components) may be all You need. Without the possibility and knowledge(!) to measure the speaker´s behaviour it´ll be impossible to equalize a FR-panel.

snip-


jauu
Calvin

If you go with a passive crossover you will need a series resonance filter in shunt.

Any passive filters are affected by driver impedance. If the driver impedance peak is within a couple of octaves of the crossover point and/or is very steep you must first flatten the driver impedance peak to have any chance that the filters will work at all.

Note that my experience is with an early version of the Innersound 15"x45" flat panel that is crossed over 4th order at 400Hz with a DIY active xover to a 10" Peerless CSC-X.

At the time since I had all the measurements I tried modeling a passive xover and nothing would work. I kept getting horrible peaks.

So I emailed Roger and he told me I had to flatten the driver impedance peak, which for that panel was a huge peak around 600Hz. After that everything worked like a charm.

 

[Calvin]

Hi,

you will need a series resonance filter in shunt.

that is exactly what I meant with ´parallel Notch´, a RCL series connected circuit connected in parallel to the Input. So we just differ in semantics. The juice is the same

A good point.....impedance equing makes filter design easier and more predictable and sometimes even necessary.
It depends though on the situation. There are cases where no imp-equing is needed at all, many cases where just a parallel resistance is sufficient and some cases where a more elaborated 3-component parallel-notch/series-resonance-filter-in-shunt is needed.

Since some form of amplitude equing is needed with ESLs and the impedance is varying vastly it is imho impossible to design a good passive filter without the use of measuring stuff. Added the efficiency problem of passive filtering, I´d say there´s no way around active filtering if You want really first clas results.....but then ...active is the superior way to go anyway, isn´t it?

jauu
Calvin

 

[MJ Dijkstra]

Hi Calvin,

Regarding active crossovers, are you familiar with the stuff Thel offers?
Seems interesting to me.

http://www.thel-audioworld.de/

 

[Too Tall]

Hi,


that is exactly what I meant with ´parallel Notch´, a RCL series connected circuit connected in parallel to the Input. So we just differ in semantics. The juice is the same

Well that is the thing.
You and I know what you are talking about, but someone new to this could become terribly confused.
We need to be careful with terms.

The filter needed to flatten impedance is a "Series Resonance filter".
This is a resistor, an inductor and capacitor wired in series. (The order does not matter.)

To place it properly to flatten an impedance peak in a driver it must be wired parallel to the driver (from positive to negative often called a "shunt"). It must also be right next to the driver so that the 2nd order passive filter (or whatever you are using) sees a corrected "Flat" impedance.

A good point.....impedance equing makes filter design easier and more predictable and sometimes even necessary.
It depends though on the situation.

Agreed

There are cases where no imp-equing is needed at all, many cases where just a parallel resistance is sufficient and some cases where a more elaborated 3-component parallel-notch/series-resonance-filter-in-shunt is needed.

I know some people use a single resistor either in series or in parallel with the driver to give the passive xover filters “something to chew on’. The trouble there is the resistor can get very hot.

Since some form of amplitude equing is needed with ESLs and the impedance is varying vastly it is imho impossible to design a good passive filter without the use of measuring stuff. Added the efficiency problem of passive filtering, I´d say there´s no way around active filtering if You want really first clas results.....but then ...active is the superior way to go anyway, isn´t it?

jauu
Calvin

For EQing in the drop from dipole cancellation the only practical way is to use something active so at that point you might as well do the xover active too. Having the xover point higher, say up around 1kHz with a panel 12” wide you avoid having to EQ and a passive xovers makes sense, though you may still need to flatten impedance.
I use Praxis to measure the impedance curve, though you can measure it with a Multi Meter one point at a time if you have a free afternoon to do it. A real pain, but it is free.
Cheers

 

[Calvin]

Hi,

the Thel filter belongs to the so called class of Universal Active Filters (UAF).
The UAF42 of BurrBrown is a fine example for this. The data sheet and application notes and the associated sim-program of the UAF42 can clear the mist
Standard 2nd-order filter applications like SallenKey etc. are typically low parts count circuits. Ususally just one OP, seldomly two OPs and a few caps and Rs are all what is needed. While the low parts count is fine regarding cost of the final product the simplicity doesn´t come for free. Sensitivity to component nonidealities, component tolerances, difficulty of tuning and just one output are typical limitations.
These limitations may be overcome with the use of multiple OP filter circuits. The so called State variable and BiQuad filters are examples for these more complex circuits. They belong to the class of UAFs.
A UAF is a multi-OP filter which provides for more than just one output simultaneously!
It consists of three OPs (in special cases just 2), but often a fourth spare OP is implemented to generate special filter functions or to be used as Buffer or Inverter. A UAF can simultaneosly provide second order Lowpass, Bandpass and Highpass responses. The 4th OP can be used to generate the Notch response. The UAF makes use of integrators and because of this it can be regarded as an analog computer implementation of a second order nonhomogenious differential equation. Now doesn´t that sound sweet?
Analog computing crossover is - as far as I remember- the term Thel used to use for marketing purposes. Sounds terribly sophisticated, but basically the Thel filter is a Quad-OP with some caps and Rs.
The UAF42 comes with tightly tolerated C´s which eases the design and building process, but a good Quad-OP with a couple of discrete tightly tolerated caps and Rs will do the job equally well at probabely lower cost.
Since the different responses are ´computed´ using the same circuitry, the responses track each other, hence the claim of superior, even perfect phase behaviour.
While the Thel imho belongs to the better sounding OP-filters it nevertheless underlies serious restrictions. It s just a 2nd order (12dB) filter building block. Other orders, especially odd order filter functions are not considered.
The tracking of the amplitude and phase responses is ideal in theory, but it is impossible to tune the filter to driver nonlinearities. Which means that the resultand acoustic response of a two-way crossover with drivers can be worse than the response of a ´classical´ crossover and drivers, where the branches of the crossover are tuned independant of each other. You need to use drivers that ´behave´ at minimum +-1 octave linear around the crossover-freq, or use active equing (eg. with biquad-eqs).
If You for example design a 2-way 17/25 speaker with a CO-freq of 2kHz you have to make sure that the bass behaves linearly up to min 4kHz and the tweeter should be ´perfect´ down to at max. 1kHz. These are very serious demands!
A multi-OP-multi-output filter that is imho superior to the UAF is the so called Hawksford filter ("A family of circuit topologies for the Linkwitz-Riley -LR4- crossover alignement. 82nd AES Convention 1987). One of the incarnations (page 34 of the AES paper) of this filter uses simple inverting integrators whose outputs are summed up and distracted from the Input. Besides allowing for basically every order of filter steepness -not just LR4- the highpass signal is passing just one OP! The integrators (just one OP one R and one cap for each!) can be tuned individually, so there´s no need for tightly tolerated parts. If the quality of the drivers allow for ´tracking´ filters I´d prefer the Hawksford topology.
In all other cases I´d opt for ´classical´ filter structures using simple JFET-buffers instead of the OPs.

For EQing in the drop from dipole cancellation the only practical way is to use something active so at that point you might as well do the xover active too.

I don´t agree in fullness.
It is possible and easy to design a passive filter with a Q >1, hence with a peaking response. The problem is only that besides the freq-range around the peak everything else is lowered in level. That´s a trick especially ESLs allow for, because of their ´special´ input impedance (via transformer-->inductance)
A 2nd order filter (12dB) leads to such a peaking response that it even must be dampened by a resistor!
The ML Prodigy for example used this trick. The panels resonance (+10dB@250Hz) is followed by the suckout around 400Hz and the typically slightly raised response around 1kHz till it eventually rolls off above 2kHz. A peaking 12dB-highpass ~@400Hz fills the suckout and linearizes the panel down to 250Hz (besides the reduced electrical stress on the panel this leads to very high efficiency). So here the panels resonance is not ´notched out´ but instead used for good. The sonic fingerprint of the resonance (looong decay) can clearly be heard, so most speakers are crossed over way above the resonance and deal with it by the use of a notch filter. Circuit complexity, tuning problens and efficiency matters imply the usage of active filtering as superior here.


jauu
Calvin

 

[Too Tall]

Hi,

the Thel filter belongs to the so called class of Universal Active Filters (UAF).
The UAF42 of BurrBrown is a fine example for this. The data sheet and application notes and the associated sim-program of the UAF42 can clear the mist
Standard 2nd-order filter applications like SallenKey etc. are typically low parts count circuits. Ususally just one OP, seldomly two OPs and a few caps and Rs are all what is needed. While the low parts count is fine regarding cost of the final product the simplicity doesn´t come for free. Sensitivity to component nonidealities, component tolerances, difficulty of tuning and just one output are typical limitations.
These limitations may be overcome with the use of multiple OP filter circuits. The so called State variable and BiQuad filters are examples for these more complex circuits. They belong to the class of UAFs.
A UAF is a multi-OP filter which provides for more than just one output simultaneously!
It consists of three OPs (in special cases just 2), but often a fourth spare OP is implemented to generate special filter functions or to be used as Buffer or Inverter. A UAF can simultaneosly provide second order Lowpass, Bandpass and Highpass responses. The 4th OP can be used to generate the Notch response. The UAF makes use of integrators and because of this it can be regarded as an analog computer implementation of a second order nonhomogenious differential equation. Now doesn´t that sound sweet?
Analog computing crossover is - as far as I remember- the term Thel used to use for marketing purposes. Sounds terribly sophisticated, but basically the Thel filter is a Quad-OP with some caps and Rs.
The UAF42 comes with tightly tolerated C´s which eases the design and building process, but a good Quad-OP with a couple of discrete tightly tolerated caps and Rs will do the job equally well at probabely lower cost.
Since the different responses are ´computed´ using the same circuitry, the responses track each other, hence the claim of superior, even perfect phase behaviour.
While the Thel imho belongs to the better sounding OP-filters it nevertheless underlies serious restrictions. It s just a 2nd order (12dB) filter building block. Other orders, especially odd order filter functions are not considered.
The tracking of the amplitude and phase responses is ideal in theory, but it is impossible to tune the filter to driver nonlinearities. Which means that the resultand acoustic response of a two-way crossover with drivers can be worse than the response of a ´classical´ crossover and drivers, where the branches of the crossover are tuned independant of each other. You need to use drivers that ´behave´ at minimum +-1 octave linear around the crossover-freq, or use active equing (eg. with biquad-eqs).
If You for example design a 2-way 17/25 speaker with a CO-freq of 2kHz you have to make sure that the bass behaves linearly up to min 4kHz and the tweeter should be ´perfect´ down to at max. 1kHz. These are very serious demands!
A multi-OP-multi-output filter that is imho superior to the UAF is the so called Hawksford filter ("A family of circuit topologies for the Linkwitz-Riley -LR4- crossover alignement. 82nd AES Convention 1987). One of the incarnations (page 34 of the AES paper) of this filter uses simple inverting integrators whose outputs are summed up and distracted from the Input. Besides allowing for basically every order of filter steepness -not just LR4- the highpass signal is passing just one OP! The integrators (just one OP one R and one cap for each!) can be tuned individually, so there´s no need for tightly tolerated parts. If the quality of the drivers allow for ´tracking´ filters I´d prefer the Hawksford topology.
In all other cases I´d opt for ´classical´ filter structures using simple JFET-buffers instead of the OPs.



I don´t agree in fullness.
It is possible and easy to design a passive filter with a Q >1, hence with a peaking response. The problem is only that besides the freq-range around the peak everything else is lowered in level. That´s a trick especially ESLs allow for, because of their ´special´ input impedance (via transformer-->inductance)
A 2nd order filter (12dB) leads to such a peaking response that it even must be dampened by a resistor!
The ML Prodigy for example used this trick. The panels resonance (+10dB@250Hz) is followed by the suckout around 400Hz and the typically slightly raised response around 1kHz till it eventually rolls off above 2kHz. A peaking 12dB-highpass ~@400Hz fills the suckout and linearizes the panel down to 250Hz (besides the reduced electrical stress on the panel this leads to very high efficiency). So here the panels resonance is not ´notched out´ but instead used for good. The sonic fingerprint of the resonance (looong decay) can clearly be heard, so most speakers are crossed over way above the resonance and deal with it by the use of a notch filter. Circuit complexity, tuning problens and efficiency matters imply the usage of active filtering as superior here.


jauu
Calvin

I don't understand some of your post. The technical talk goes over my head at times so I can't comment on a lot of it.

For my hybrid system the design goal was fairly simple.
My target was a 4th order LR acoustic response at 400Hz.
To get 4th order I ran two 2nd order Burr Brown op amps in series.
The circuits were all Salen Key based from John Pomann's DIY xover.

Speaking of losing people who have made a great contribution to the Audio DIY community John is an unsung hero.
When he finally received his Medical Doctorate DIY lost a great resource.

I used the emulator in LspCAD to massage the high pass filter to make a bump just large enough to fill the hole while taking phase into account.

i did make a very serious "rookie" mistake.
I used an "equal-component" design to so I only needed a few capacitor values.

As a result the 2nd filter in the high pass had a +9db gain. In addition I did not set up the gain structure to keep that filter from seeing as little gain as practical

I found out that opamp did not have enough headroom for peaks.
I reset the gain structure to help as much as possible, but I should really go back in and find a better answer even if I need to buy a lot more cap values.

With LspCAD and the response curves involved I do not see a need for separate EQ. Massaging the 4th order filter is enough.

 

[Calvin]

Hi,

since you quoted my text in completion, I don´t know which part You don´t understand and which point needs clearance.
It can be very helful to use a simulator proggy like Switcher Cad of TI or Circuit Maker. Especially when you need response curves that are a bit off of the usual textbook standards.
With ESLs you just need these ´off´ kind of filters quite often. Or you alternatively use the usual filter-blocks in cascades. But then the parts number count explodes and sonics go to hell.
In any case does the simulation help in understanding how a filter works and can be mighty tool when reducing parts number count.

Equal component KRC filters ease the design process and lower cost, but are not the way to go here. Unity gain SallenKey are more promising. The problem with these ´simple´ filter structures is that if you change one parameter -lets say crossover freq Fo- You change several paramaters at once -Q-factor and Gain eg.
For evaluation purposes it would often be useful if the filter´s parameters could be tuned independent of each other. This requires more complex filter structures. The State Variable structures allow for independent tuning of Fo and Q and offer all 4 possible responses, HP, LP, BP and Notch. The Biquad structures offer only 3 responses (LP, BP and Notch) but allow for independent tuning of F0, Q and Gain.

It is my experience that you can´t use a LR4 filter for ESLs (which is fixed to a certain response with a Q of 0.5). Since ESL response drops sharply below the resonance it is convinient to kill the resonance with a notch and filter the panel with a 2nd order highpass with Q>1 (response of the filter is with gain of value Q at the crossover freq. This way filling up the suckout of the driver). The combination of the Notch´s response and the Highpass adds to a steeper than 2nd order filter curve. The curvature can be nicely fine tuned.
If circumstances are lucky you might even use a single filter (asymmetric -notch) that implements both functions (Notch + HP).
Then You need only 2 Buffers of which only one is in the signal path (and even this one can be omitted under cirtain circumstances )

You have to keep in mind that differences in the acoustic filter function of Bass and ESL are a major reason of the typical Hybrid sound. Because of the natural steep falling response of the ESL below its Fs, the Bass typically needs a very steep electrical filter function to reach a similar acoustic response. This rules out analog computing filters like the Thel if the Bass and ESL are not equalized before!

If You are not familiar with filter structures and filter implementation I´d opt for using digital IIR-Filters which are simple to program. Especially when the implementation of analog OPamp based filters results in a large OP number count a IIR-Filter is sonically superior.


jauu
Calvin

 

[Too Tall]

Hi,

since you quoted my text in completion, I don´t know which part You don´t understand and which point needs clearance.

I was in a hurry when I read through your post. I should have just waited till I had enough time to read it properly. My apologies.
After going through it again I see where you are going.

It can be very helful to use a simulator proggy like Switcher Cad of TI or Circuit Maker. Especially when you need response curves that are a bit off of the usual textbook standards.
[/B]

I use LspCAD Pro-
www.ijdata.com
http://www.ijdata.com/LspCAD_XO.html

I many cases I use the “Target optimizer” to tweak a textbook electrical 4th order LR to create an “acoustic” 4th order LR response. (analog active or passive)

With ESLs you just need these ´off´ kind of filters quite often. Or you alternatively use the usual filter-blocks in cascades. But then the parts number count explodes and sonics go to hell.
In any case does the simulation help in understanding how a filter works and can be mighty tool when reducing parts number count.

Equal component KRC filters ease the design process and lower cost, but are not the way to go here. Unity gain SallenKey are more promising. The problem with these ´simple´ filter structures is that if you change one parameter -lets say crossover freq Fo- You change several paramaters at once -Q-factor and Gain eg.
For evaluation purposes it would often be useful if the filter´s parameters could be tuned independent of each other. This requires more complex filter structures. The State Variable structures allow for independent tuning of Fo and Q and offer all 4 possible responses, HP, LP, BP and Notch. The Biquad structures offer only 3 responses (LP, BP and Notch) but allow for independent tuning of F0, Q and Gain.
[/B]

All this would be done in the modeling program

It is my experience that you can´t use a LR4 filter for ESLs (which is fixed to a certain response with a Q of 0.5).

It just makes a convenient target. I do not lock the electrical filter into a Q of 0.5. The optimizer moves the electrical response till the sum of the drivers actual real world response and the filters electrical response equals an ACOUSTIC 4th order response of each driver.

Since ESL response drops sharply below the resonance it is convinient to kill the resonance with a notch and filter the panel with a 2nd order highpass with Q>1 (response of the filter is with gain of value Q at the crossover freq. This way filling up the suckout of the driver). The combination of the Notch´s response and the Highpass adds to a steeper than 2nd order filter curve. The curvature can be nicely fine tuned.
[/B]

Since I went active I just massaged the high pass on the ESL panel to fill in the hole. No notch or EQ needed.
The only time I have used changing Q from typical to fill a hole was with the tuning of a bass woofer.
I was given a box where the ported alignment caused a peak in the response. Since I was not allowed to change driver or rebuild the box volume and port I worked with the baffle step and let the woofer peak fill the hole. I got at least another 3dB out of the design and the system was flat down into the 50Hz range.

If circumstances are lucky you might even use a single filter (asymmetric -notch) that implements both functions (Notch + HP).
[/B]

That is my first approach and I use it any time I can. For a passive 4th order use only two caps and two coils and accomplish what would normally require one or more parametric EQ in addition.

Then You need only 2 Buffers of which only one is in the signal path (and even this one can be omitted under cirtain circumstances )

You have to keep in mind that differences in the acoustic filter function of Bass and ESL are a major reason of the typical Hybrid sound. Because of the natural steep falling response of the ESL below its Fs, the Bass typically needs a very steep electrical filter function to reach a similar acoustic response. This rules out analog computing filters like the Thel if the Bass and ESL are not equalized before!

If You are not familiar with filter structures and filter implementation I´d opt for using digital IIR-Filters which are simple to program. Especially when the implementation of analog OPamp based filters results in a large OP number count a IIR-Filter is sonically superior.


jauu
Calvin [/B]

 

[Calvin]

Hi,

I liked LSP-Cad up to the 5.xx build since it was very intuitive in handling and the results seemed reasonable. It didn´t use a Spice-based simulator though. So the structure of the possible filter circuitry was restricted to simple ones (oh, how did I miss the bridged T-filter ). I have no experience with the 6.xx build apart from a quick handling test where it failed. Looks a bit too complicated to me. And since I concentrate rather on active speakers and have to work with Spice-simulators anyway, a ´real´ simulator proggy like LT´s works just finer for me

Since I went active I just massaged the high pass on the ESL panel to fill in the hole. No notch or EQ needed.

You may have been lucky here, because normally this is not the path to Endor Typical ESL-resonances peak at 10-20dB above normal level. Even with a filter of fourth order the crossover-freq should be placed no less than 2 octaves off the resonance freq to keep artifacts within the freq-response small enough. Since it is most often desired to have a lower crossover-freq, the demand for a linearization of the peak occurs. Besides the capability to shape the combined filter response of a 2nd order HP + notch (symmetric) its often sufficient to use a asymmetric Notch only, thereby reducing component count, filter structure complexity, size and cost!

That is my first approach and I use it any time I can....

Since You talk about a fourth order passive filter here -which is a cacade of two standard second order filters-, I´m not sure if You got the point I wanted to make. The point is, to look at different than standard filter-blocks to find something that suits the desired correction- and filter response better, preferably even with lower parts count. In the special case of an ESL this means the need to correct for the Fs-peak (which isn´t adressed -but just attenuated- by a simple HP).

jauu
Calvin