What X-over for a hybrid ESL?

[markusA]

I find myself with a little extra time to consider the finer details of my current project, a hybrid ESL. (Check my blog for project updates)

A while back Calvin was nice enough to walk us through the development of a passive crossover for a ESL.
Today I find myself asking a pretty relevant question.
What kind of XO should I choose?
The cross will be somewhere in the 250Hz +/-50Hz area.

The options are active, passive or digital?
The easiest would without a doubt be to buy a commercial active x-over.
The most difficult to build is probably the passive one.

My question is which one is the better and why?

A good amp will cost money, passive could end up being the cheap solution.
Digital would offer a lot of fun parameters to tweak.
Maybe a passive XO with a digital EQ?
The floor is yours...

What to build and why?

 

[Chris Daly]

I find myself with a little extra time to consider the finer details of my current project, a hybrid ESL. (Check my blog for project updates)

A while back Calvin was nice enough to walk us through the development of a passive crossover for a ESL.
Today I find myself asking a pretty relevant question.
What kind of XO should I choose?
The cross will be somewhere in the 250Hz +/-50Hz area.

The options are active, passive or digital?
The easiest would without a doubt be to buy a commercial active x-over.
The most difficult to build is probably the passive one.

My question is which one is the better and why?

A good amp will cost money, passive could end up being the cheap solution.
Digital would offer a lot of fun parameters to tweak.
Maybe a passive XO with a digital EQ?
The floor is yours...

What to build and why?

The ESL presents many more design challenges than just choosing frequency or how that frequency is obtained. Here is a good start attached that shows a voltage multiplier coupling transformer loading choke, audio transformer and crossover components. The design was in production for over 25 years and is highly regarded as one of the worlds best speakers. It is known as the Quad ESL 57. it is very much a trusted design suitable for and designed around the ESL panels that Quad produced.

There are also many other excellent references on ESL DiY designs right here on the forum, but the ESL 57 is an excellent starting if not finishing point on the subject of ESL's

Opening up a Martin Logan Quest Z one will find enormous effort put into crossover and electronics. Martin Logan opt for the approach of the ESL panel reproducing upper frequencies, rather than full range. So their approach is different to Quad.

What this should say is that ESL's are a serious speaker, requiring serious design effort
with high voltage components, that is a symbiosis of many parts dedicated toward producing excellent sound.

Cheers / Chris

 

[markusA]

I'm not entirelly sure what to make of your post?
Are you an advocat of passive crossovers?

I'm very much aware of the fact that it'll be long and hard work to get a passive XO optimised.
I'm more concerned with which one is actually better and why?
There are some designers claiming that passive is better above 200Hz.
Some swear that active is the way to go?

 

[Calvin]

Hi,

I´d choose the Xover after having measured the drivers.
Basically I´d opt for as much bandwidth for the ESL part as sensible.
Sensible meaning here to HP-limit the lower bandwidth to at least 1/2 octave above the fundamental resonance of the panel, or a frequency point requiring no more than ~+6dB equalization for acoustic phase cancellation, whichever comes first.
The probabely biggest advantage of an hybrid against a FR is its largely increased dynamic range and low distortion figures. It results from the restriction to a frequency range where the diaphragm excursions become very small (<1mm) which is typically the range >150Hz.
Depending on segment sizes and mechanical diaphragm tension the fundamental resonance may be in the range rom 100Hz -200Hz. This suggests an X-over range from 150-300Hz, if the 6db-rule and the requirement for the dynamic range, i.e. panel size allow for such low Xover -frequency.
Since ESLs need some form of equalization, passive crossovers mean a loss of efficiency in every case, in extremes -as for example the old SequelII- up to 15dB. After my experience xovering actively is the only way to achieve top-notch results. Wether You xover with analog filters or digital filters doesn´t matter much sonically if OPamps are involved. Almost all commercial analog Xovers miss on the EQ-capabilities required for ESLs.
The digital filters offer alot more flexiblility, convenience, filter capablities and will probabely even cost less (see MiniDSP et al).
It may be possible to top the sonic results with elaborated discrete-circuits analog xovers, but as long as You haven´t found the final set of parameters I´d suggest You use a digital filter.

jauu
Calvin

 

[john65b]

I have a DCX2496 biamping a pair of ML CLS with a pair of Magnepan Tympanis. Currently crossed over at 250hz. I also want to make a passive to compare. But since I built the electronics for the CLS panels (bias circuit and stepup transformer), I have no idea what the impedance is of my ESLs are. I put a voltmeter on them and get just under 2 ohms, and the maggies are 8ohms. Would it be correct to design a passive xover around these two measured impedances?

Anyway, I will continue and watch thsi thread...

 

[Chris Daly]

I'm not entirelly sure what to make of your post?
Are you an advocat of passive crossovers?

I'm very much aware of the fact that it'll be long and hard work to get a passive XO optimised.
I'm more concerned with which one is actually better and why?
There are some designers claiming that passive is better above 200Hz.
Some swear that active is the way to go?

What I am referring to is that conventional active crossovers are designed for preamp type voltages - and they can do an excellent job when placed in that role, ESL's though are high voltage devices. The ESL 57 schematic shows the audio transformer stepping up the voltage to the passive crossover components. So here we have Quad purpose designing that there is some benefit in having the crossover components running at high voltage... You could look at the Wireless World articles from 1957, and I am confident in amongst the myriad of mathematics is an answer for why it is better to have the crossover of frequency done at HV

Your design possibly needs to consider this voltage step up and the exposure of components to that voltage. Very few active crossovers that I know of are designed for such voltages or could survive such voltages. Hence the wisdom of passive crossovers.

Relative to the ESL 57: The audio transformer is used to step the voltage up which comes from your amplifier. The transformer has a single primary and two secondary windings. The primary winding is what you connect your amplifier too. The secondary windings eventually connect to the bass and treble panels (after the signal goes through the crossover). This means that all the components in the crossover have to be rated at high voltages. The secondary windings are nominally 55:1 and 110:1 for the treble and bass panels respectively. This means that at the maximum input voltage of 36 volts you are getting about 4000 volts on the stators of the bass panels !! The transformer is totally encased in wax.

Cheers / Chris

 

[markusA]

I suppose that pretty much sums it up?
My initial thoughts of a digital XO is probably both the easiest as well as best way to go.
I guess there isn't much to do except start saving up for the DAC/DSP combo.
I predict a regular active crossover like the Behringer will be the safest bet to get the system up and running while I save up for the real deal. Lol
I got to get myself a good paying job, this is going to require a bigh chunk of $$$.

Like I stated in the beginning, work is gently moving along and I had a little time left over. I'm currently shopping around for a local powder coater for the panels.

 

[Calvin]

Hi,

maybe the ´wisdom´ in using the HV-side is simply marketing?
Up till today, the ill myth holds that a fullrange ESL is the superior ESL.
Now what should all the people think if they realized that their beloved Quad company didn´t built nor sold true fullrange ESLs but rather multi-way/multi-paneled ESL-speakers with secondary-side Xovers?
They could and do market such a speaker still though as xoverless speaker, thereby claiming to fulfill an ideal, that no other speaker could offer.
The same principle as with MLs XLS (which in contast relies on a primary side Xover), which isn´t a fullrange ESL like its predecessor the CLS.
If it weren´t for other than marketing reasons they surely had not opted for a clearly inferior ESL panel as bass part.
All issues and problems are solved if You just claim: "Its a Fullranger!"

jauu
Calvin

 

[Chris Daly]

Here is a reprint of the 1955 article by Peter Walker, showing development of what was to become the ESL57. Much wisdom on many fronts. the use of bass and treble panels and a crossover at HV from what I can ascertain, was to improve efficiency. As can be seen many design issues and criteria are proposed, including actual use, sound pressure levels, and PW answers each to propose the use of ESL panels as near perfect sound radiators.

Wide Range Electrostatic Loudspeakers P.J.Walker Wireless World May 1955 part 3

Also remember this was 1955 and production of the 57 led to arguably the best loudspeaker that was mass produced ever, with production numbers to over 25,000 pairs. 53,151 apparently the last one.

Cheers / Chris

 

[bolserst]

maybe the ´wisdom´ in using the HV-side is simply marketing?
Up till today, the ill myth holds that a fullrange ESL is the superior ESL.
Now what should all the people think if they realized that their beloved Quad company didn´t built nor sold true fullrange ESLs but rather multi-way/multi-paneled ESL-speakers with secondary-side Xovers?

In Quad's defense, I have never seen any documentation that would lead me to believe that they promoted their original ESL-57 as a crossover-less design. In fact, Walker and Baxandall provided a great deal of information on the ESL-57 and why they chose to use a 3 way design, how they implemented their crossover and how they dealt with the high-Q diaphragm resonances.

Now, I have not read any mention of why they chose to use a HV-side crossover. But, I would think the most likely reason was cost savings. If they had used primary side crossovers they would have needed 3 separate step-up transformers.

As to whether it was advised for QUAD to use ESL bass panels for their ESL-57 or cross to dynamic drivers, I could only say that having listened to some of the vintage woofers of the 50's I think they made the right choice.


Speaking of companies promoting full-range crossover-less ESLs....what about Acoustat?
The patent that their design was based on is in fact a method of combining output from HF & LF transformers thru a crossover.
How could they then promote this as a crossover-less design?

 

[bolserst]

I also want to make a passive to compare. But since I built the electronics for the CLS panels (bias circuit and stepup transformer), I have no idea what the impedance is of my ESLs are. I put a voltmeter on them and get just under 2 ohms, and the maggies are 8ohms. Would it be correct to design a passive xover around these two measured impedances?

The 2 ohm you measured is the DC resistance of the transformer primary windings and any resistance you might have in series with the primaries. This is not the AC impedance which would be needed to design passive crossovers. The impedance of an ESL varies wildly as does the phase. You would need to measure the impedance with something like the Dayton Audio WT3 tester. Many of the computer based audio measurement systems also have the capability of measuring impedance.

The maggies on the other hand will have an impedance that is pretty close to their DC resistance measurement. They do have a small amount of inductance the will cause an increase in impedance and rotation of phase as frequency increases.

Here is a posted example of what you might expect to see for the maggies.
The inductance is low enough, impedance doesn't start rising til >1kHz.
http://www.diyaudio.com/forums/plan...c-dc-resistance-vs-impedance.html#post2910039

 

[john65b]

The 2 ohm you measured is the DC resistance of the transformer primary windings and any resistance you might have in series with the primaries. This is not the AC impedance which would be needed to design passive crossovers.

Yes, this is exactly my issue. I know the actual impedance, if I were to measure as you indicate , would fluctuate wildly with frequency.

Now I think it was you who had recommended a 15ohm across the ESL Speaker binding posts (in parallel across the primary of the step-up tranny) and a 50uf cap in series with the speaker positive input would get me the 250hz cross...is this accurate? What ESL impedance was this based on?

My apologies if it was not you...

 

[CharlieM]

For those of us like myself who aren't skilled enough to design a passive crossover that works well with an ESL, and/or don't want the headache, an active crossover with adjustable gain provides easy, real-time tweaking capability to precisely balance the woofer and panel outputs, and I would say you get tighter/cleaner bass when you don't have a passive inductor between the amp and woofer. Active is the easy way to go. I've used an analog op-amp crossover and [now] a DSP crossover with my speakers and they both work great.

 

[bolserst]

Yes, this is exactly my issue. I know the actual impedance, if I were to measure as you indicate , would fluctuate wildly with frequency.

Now I think it was you who had recommended a 15ohm across the ESL Speaker binding posts (in parallel across the primary of the step-up tranny) and a 50uf cap in series with the speaker positive input would get me the 250hz cross...is this accurate? What ESL impedance was this based on?

I think this was the post you were thinking of.
http://www.diyaudio.com/forums/plan...tin-logan-cls-power-supply-3.html#post2882393

The technique of putting a 10ohm - 15ohm resistor across the primary of the step-up transformer goes a long way to stabilizing the impedance magnitude and bringing the phase close to zero (ie resistive). Basically the low value resistance put in parallel with the primary swamps or overpowers the varying impedance, so it isn't terribly important what the verying impedance was. You are left with an impedance that is fairly constant at 10ohm - 15ohm and will work well as a load for a high-pass network made up of a single capacitor.

Lower values of resistance result in a more constant impedance. The trade off is that lower values of resistance will need to be able to dissipate more power than if you chose a higher value. In my experience, 10 - 15ohm is a good range to choose from. Much lower than 10 ohm and power dissipation becomes a problem. Much higher than 15ohm and the impedance isn't as constant and resistive as you would like for an easy passive crossover.

 

[AcoustatAnswerMan]

Speaking of companies promoting full-range crossover-less ESLs....what about Acoustat?
The patent that their design was based on is in fact a method of combining output from HF & LF transformers thru a crossover.
How could they then promote this as a crossover-less design?

I don't recall Acoustat ever waving the 'crossover-less' flag, or using that terminology in their marketing. The patented Magne-Kinetic design, of using a high and low frequency transformer with outputs combined into a full-range signal presented to the panels, is certainly NOT crossover-less in the strictest sense. But, those speakers did not use different drivers for different frequencies. That is, all panel area was identical, and played full range.

The later Spectra Series was a little different. Although the transformer implementation is completly different, the result is the same, with different sections of the transformers handling different frequencies. The big difference is that not all panel area produced full range. The panel was split up into vertical strips, with one strip playing full range, another area playing lows and mids only, and the final area playing lows only. This was done using all the same type of panel, regardless of operation. The frequency roll-offs are accomplished using simple resistors acting against the inherent capacitance of the panels. So it could be said that the Spectra Series actually had two crossover systems.

However, I believe this thread is getting off topic a bit, as I believe the original question related to how to implement a crossover between an ESL panel and a woofer. There is no correct answer to this question, as it all depends on the skill and inclinations of the builder. Certainly the cheapest way to go would be passive. Acoustat used passive crossovers for its hybrid designs, with a 6-dB/octave high pass for the ESL, and a 12-dB/octave low pass for the woofer. The down side to this approach is that it can take quite a bit of fiddling to get the values correct. You may also find, as I did when voicing the Spectra 11 and 1100, that the best results might be obtained by having the poles of the low-pass slightly mis-aligned. But this would depend on your design and the actual LF performance of your ESL.

Using a commercial or home-brew active crossover would offer the most flexibility for tweaking final values, but the down side is that it adds the cost of the crossover AND a second amplifier.

My suggestion would be to go with a passive design first, and see if you can get acceptable results.

 

[tyu]

I well be the frist to say passive crossover can eat power...an sound bad if all is not right but as for althings Active ..... i had the best crossover i could find for my Apogees in the 90s an it was a all KRell setup...an i could still hear the Active crossover..Just one more thing in front of the sound

Here you go
.My suggestion would be to go with a passive design first, and see if you can get acceptable results.

Vary good info...if you can get the sound you wont out of passive frist... good luck with all the other..

just my 5cent

 

[bolserst]

I don't recall Acoustat ever waving the 'crossover-less' flag, or using that terminology in their marketing. The patented Magne-Kinetic design, of using a high and low frequency transformer with outputs combined into a full-range signal presented to the panels, is certainly NOT crossover-less in the strictest sense. But, those speakers did not use different drivers for different frequencies. That is, all panel area was identical, and played full range.

Hello AcoustatAnswerMan,

My comment was meant as humor, not accusation. Sorry if it didn't read that way. I personally think that the MK transformer interface is an elegant solution for driving and providing EQ for full range ESLs. I particularly like the later Spectra series transformer arrangement where the series capacitors on the primary side can be dispensed with, and the high voltage AC across the mixer capacitors is half that of the pre-Spectra interfaces with the same step-up ratios. Also, as you mentioned, Acoustat was one of the first to implement resistor ladder networks to provide increasing driven panel area with decreasing frequency.

However, in essentially every pre-Spectra brochure or manual I have ever seen(included Stricklands white paper) the term crossoverless appeared...see below. I am sure what was meant was that no acoustic crossover was used between different panels or panel areas. This opinion is supported by the fact that the term crossoverless was never mentioned in the later Spectra series brochures or manuals. But I'm not sure the average consumer would have picked up on the distinction.

From the Acoustat White Paper:
“All Acoustat speakers are, and have always been, crossoverless full-range-element design”

From an Acoustat Brochure:
“The Acoustat full-range-element electrostatic loudspeaker systems represent the culmination of over 20 years of design and development - beginning with the bold assumption that it was possible to build an electrostatic speaker system of no compromise, one that would be of full range design, capable of high sound pressure levels, have no crossovers, and be more reliable than any speaker of any type.”

From an Acoustat Manual:
“The Acoustat pure electrostatic speakers(1+1, 2+2, etc) utilize full-range element electrostatic panels to reproduce all frequencies in the audible range without splitting the frequency sprectrum with separate drivers or crossovers”

However, I believe this thread is getting off topic a bit.

Agreed.

 

[DjSinae]

I've used an analog op-amp crossover and [now] a DSP crossover with my speakers and they both work great.

Hi charlie, I was wondering if you think something like this would be nice
CO3 Three Band Phase-Linear Crossover Filter Kit (Bass, Mid-range, and Treble)_Filter_Accessories Kit_Analog Metric - DIY Audio Kit

I will use minidsp at first but i would really like to have linear phase and also lower latency than DSP...

 

[CharlieM]

Hi charlie, I was wondering if you think something like this would be nice
CO3 Three Band Phase-Linear Crossover Filter Kit (Bass, Mid-range, and Treble)_Filter_Accessories Kit_Analog Metric - DIY Audio Kit

I will use minidsp at first but i would really like to have linear phase and also lower latency than DSP...

I'm thinking something like that would probably work fine, once you've settled on the crossover frequencies.

About a year ago I decided to take the plunge and go digital throughout my entire signal path, with a single D/A conversion coming out of my new DSP crossover. I know this is blasphemy to analog enthusiasts who still spin LP's and shun digital, but I like the sound just fine and I've gotten so spoiled by the convenience being able to play my whole library via a digital music server without ever getting off the couch that I haven't even played a CD since. And while I liked the sound of my old DBX analog crossover, I also like the sound of my 24 bit/96khz DSP crossover and it gives me a whole lot of flexibility for tweaking and experimenting with different setups. Analog or digital, I think you will be happy either way.

Good luck with your project!

 

[Calvin]

Hi,

of course may something like this "CO3 Three Band Phase-Linear Crossover Filter Kit (Bass, Mid-range, and Treble)_Filter_Accessories Kit_Analog Metric - DIY Audio Kit" work fine....
....
....
....
but
....
....
....
and this is a major BUT
....
....
....

only if the circumstances are all right.

This means that such a filter will give good results only if the speaker behaves close to a theoretically ideal bandpass, which no speaker does.
In other words: a filter that dosen´t take the speakers own behaviour into account will not give good results.
In the end its the acoustical output that counts not the electrical.
The main difference between most passive filtering and (analog) active filters is the lack of equalizing functions with the latter.
Especially ESL always require some means of equalization.
So a filter without equalizers won´t lead to good output.
The great advantage of digital filters like the minidsp is that they allow for filtering and equing at the same.
As such they fulfill the requirements of the speaker.
Personally, I never came across a OPamp graveyard Xover that did music any good and I know lots of opportunities to spend $89 for better.
One may just spoil a thought or two about the signal path in a analog crossover vs a digital crossover.
Is there any reason to believe that a analog OPamped crossover featuring thousands of active parts will do any better than a AD-DA conversion and some calculus in between?
If the analog crossover were carefully build with discrete parts for one specific speaker (as such much more costly), the outcome might be more pleasing than with a digital filter.
In any other case I´d say forget analog filters that offer just filtering wo. equalizers.

jauu
Calvin