Hi
ESL63 transformers are not suitable - that is a different class of ESL again
Essentially there are three classes of ESL depending on their impedance.....
Single segment ESL (e.g. perforated plate)
These radiate as a rectangular source (typically) with poor polar response (narrow listening position)
ESL load is single pure capacitor
best transformer (for maximum bandwidth) has high capacitance and low leakage inductance
larger (50VA) toroids good - commercial ESL transformers good
Leakage inductance a few mH, capacitance several 100 pF - see the Amplimo transformers for example
LC Transmission line (ESL 63 family)
This is multi-segment ESL connected as LC transmission line. The LC transmission line uses annular segments radiating with LC delay sections to simulate a point source = spherical wave front and wide listening position - LC transmission lines have a resistive input impedance, hence
ESL load is pure resistor over audio band
Best transformer has low capacitance and high leakage inductance
(ESL63 transformers are about 1.5 H and 30 pF I think)
RC Transmission line (multi-segment PCB or wires, with resistors)
This uses the RC transmission line to simulate a line source = cylindrical wave front= broad listening position.
ESL load is equal mix of resistor and capacitor
The best transformers have middling leakage inductance and middling capacitance, below 200-300 mH and below 100 pF respectively.
15VA transformer has 16 mH and 100 pF (unloaded bandwidth 130 kHz), when stacked the inductances add in series and capacitance add in parallel (fall) mostly.
ESL63 transformers are not suitable - that is a different class of ESL again
Essentially there are three classes of ESL depending on their impedance.....
Single segment ESL (e.g. perforated plate)
These radiate as a rectangular source (typically) with poor polar response (narrow listening position)
ESL load is single pure capacitor
best transformer (for maximum bandwidth) has high capacitance and low leakage inductance
larger (50VA) toroids good - commercial ESL transformers good
Leakage inductance a few mH, capacitance several 100 pF - see the Amplimo transformers for example
LC Transmission line (ESL 63 family)
This is multi-segment ESL connected as LC transmission line. The LC transmission line uses annular segments radiating with LC delay sections to simulate a point source = spherical wave front and wide listening position - LC transmission lines have a resistive input impedance, hence
ESL load is pure resistor over audio band
Best transformer has low capacitance and high leakage inductance
(ESL63 transformers are about 1.5 H and 30 pF I think)
RC Transmission line (multi-segment PCB or wires, with resistors)
This uses the RC transmission line to simulate a line source = cylindrical wave front= broad listening position.
ESL load is equal mix of resistor and capacitor
The best transformers have middling leakage inductance and middling capacitance, below 200-300 mH and below 100 pF respectively.
15VA transformer has 16 mH and 100 pF (unloaded bandwidth 130 kHz), when stacked the inductances add in series and capacitance add in parallel (fall) mostly.
15 inch is pretty wide. its ok but stability of the mylar can be an issue. there is a reason why acoustat and quad used more smaller panels.. they used multiple to get the desired spl, and they used smaller ones because a resonance of 10 hz is not needed and not desired, and will result in mylar being stuck to one stator. 
if you want to reach 50 you could go sligthly bigger then a quad panel for instance.
if you want to reach 50 you could go sligthly bigger then a quad panel for instance.
what is the difference for the transformer if there is a delay line ???? quad uses resistors as well. the the bass panels for instance are segmented(it has a high roll off). adding a delay wont change the fact the trannies need to drive the whole surface area full range.. in the schematic there are resistors. on almost all rings and bass panels. as far as i know. when you listen it is quite obvious the center ring emits most of the high frequency from 12khz and up. especially when you know it should increase 6db oct (or was it 3?) so the high frequency are heavily damped by some sort of filter... if it was only a delay i should be able to measure it still.
well enlighten me please on how they are capable of driving 1:250 without problems, since i dont get it. the delay line itself and the wave front is not the reason so what is it ?.. no offence i am just curious!
you could still use audiostatic transformers if needed
Last edited:
Hi Wrinex
In a true LC transmission line, all of the segments are driven at full voltage , as you say. However, the phases are different. At low frequencies, all of the segments are driven in phase.
At high frequencies, all of the segments are driven at different phases (due to the L between each C), so some of the sound from one segment will be cancelled by the audio from another. At high frequencies there is a lot of cancelation going on.
Another way to look at it ...You might recall Walker's equation, which says that the on-axis SPL is proportional to stator current. If you drive a pure capacitor with ac, the current increases with frequency. - so you would expect the SPL to increase with frequency. The LC transmission lines fiddles with the phase of the stator current in the different sections in such a way that the vector (i.e. include phase effects) sum of all of the stator currents is independent of frequency => and therefore the SPL is independent of frequency.
The extra resistors do something extra - they slightly attenuate the signal towards the edges of the ESL so that it does not have big diffraction effects. Also, as you note, the bass sections are more like RC network too.
Re step-up ratios
Single segment ESLs are very sensitive to inductance, including amplifier output inductance. The amplifier inductance as 'seen' by the ESL is the amplifier OP inductance times the square of the step up ratio. So 10 uH OP inductance with a 1:100 transformer, looks like 100 mH in series with the ESL. With a (medium sized) 1 nF ESL, the cutoff frequency would be 16 kHz. If you were to increase the step up ratio to 1:200, the bandwidth would fall to 4 kHz - intolerable. By the time you include the transformer inductance etc with a big ESL, the maximum step up ratio is about 1:80 and you must avoid high inductance transformers. The problem is the high capacitance of the ESL.
The Quad ESL 63 has a pure resistor load impedance, which means that inductance is not so important. If the transformer and everything else is made with as small a capacitance as you can, the inductance can be very large, and the step up ratio can be large For example a 1:250 step up ratio on 10 uH OP inductance means the ESL would see 0.625 H in series - this only makes a small difference when the capacitance is very low and the total transformer inductance (two transformers) is 3H. Hence the ESL 63 favours transformers with low capacitance and high inductance.
As above, the RC transmission line ESL lies midway between these two cases, requires medium inductance, medium capacitance, tolerates medium step-up ratio (200 is high unless you know your amp has low OP inductance).
regards
Rod
In a true LC transmission line, all of the segments are driven at full voltage , as you say. However, the phases are different. At low frequencies, all of the segments are driven in phase.
At high frequencies, all of the segments are driven at different phases (due to the L between each C), so some of the sound from one segment will be cancelled by the audio from another. At high frequencies there is a lot of cancelation going on.
Another way to look at it ...You might recall Walker's equation, which says that the on-axis SPL is proportional to stator current. If you drive a pure capacitor with ac, the current increases with frequency. - so you would expect the SPL to increase with frequency. The LC transmission lines fiddles with the phase of the stator current in the different sections in such a way that the vector (i.e. include phase effects) sum of all of the stator currents is independent of frequency => and therefore the SPL is independent of frequency.
The extra resistors do something extra - they slightly attenuate the signal towards the edges of the ESL so that it does not have big diffraction effects. Also, as you note, the bass sections are more like RC network too.
Re step-up ratios
Single segment ESLs are very sensitive to inductance, including amplifier output inductance. The amplifier inductance as 'seen' by the ESL is the amplifier OP inductance times the square of the step up ratio. So 10 uH OP inductance with a 1:100 transformer, looks like 100 mH in series with the ESL. With a (medium sized) 1 nF ESL, the cutoff frequency would be 16 kHz. If you were to increase the step up ratio to 1:200, the bandwidth would fall to 4 kHz - intolerable. By the time you include the transformer inductance etc with a big ESL, the maximum step up ratio is about 1:80 and you must avoid high inductance transformers. The problem is the high capacitance of the ESL.
The Quad ESL 63 has a pure resistor load impedance, which means that inductance is not so important. If the transformer and everything else is made with as small a capacitance as you can, the inductance can be very large, and the step up ratio can be large For example a 1:250 step up ratio on 10 uH OP inductance means the ESL would see 0.625 H in series - this only makes a small difference when the capacitance is very low and the total transformer inductance (two transformers) is 3H. Hence the ESL 63 favours transformers with low capacitance and high inductance.
As above, the RC transmission line ESL lies midway between these two cases, requires medium inductance, medium capacitance, tolerates medium step-up ratio (200 is high unless you know your amp has low OP inductance).
regards
Rod
After reading the above I'm thinking my tandem 50VA 230V/2x6V toroid setup isn't ideal for my segmented hybrid panels. And the smaller toroids I've seen advertised online also have the two 115V windings, which I gather we want to avoid due to capacitance. Any recommendations?
Regarding segmentation:
I'm about halfway through building a wire stretching jig for 90-wire panels. The jig will use pins rather than threaded rods. But I'm at an impasse insofar as I can't drill the plates for the pins until I decide on a segmentation scheme.
The panel has 2 vertical spacers dividing the diaphragm into 3 equal sections, each driven by 30 wires (90 wires total). I'm torn between these following schemes:
Config 1: 10-wire group + (16) 5-wire groups in (9) electrical segments
Config 2: (15) 6-wire groups in (8) electrical segments
I'm completely deaf above 10Khz so I'm not sure it makes any difference which scheme I choose. My hearing actually starts rolling off dropping off around 5Khz so I have to apply some boost to get the treble to my liking.
I figure that with my hearing, if one configuration gives brighter treble above 5Khz, that's the one I should go with. Any thoughts on which configuration I should choose?
Chaz
Hi Chaz
generally, the smaller the segments (the greater the number) the better, although there is little to be gained from making them much smaller than about 12 mm wide. There are several benefits...
1. Good smooth equalisation requires a minimum of about 10 segments. Play with the Excel simulator to see this.
2. Smaller segments improves the off-axis response a little bit. The segments must be less than 20 mm wide to avoid high-frequency zeros and phase reversals somewhere off axis, but there is little benefit to segments smaller than 12 mm wide.
3. More segments means more resistors, which means that the power is more widely dispersed, and the voltage drop across the resistors is less. For a good full-range ESL, say > 450 mm wide, you will need maybe 30-40 segments, so you can use resistors with quite modest ratings ~ 300 V 1W - good surface mount resistors make for a nice compact design. The power and voltage ratings are not a big deal unless you test them with pure sinewaves > 5 kHz.
regards
Rod
generally, the smaller the segments (the greater the number) the better, although there is little to be gained from making them much smaller than about 12 mm wide. There are several benefits...
1. Good smooth equalisation requires a minimum of about 10 segments. Play with the Excel simulator to see this.
2. Smaller segments improves the off-axis response a little bit. The segments must be less than 20 mm wide to avoid high-frequency zeros and phase reversals somewhere off axis, but there is little benefit to segments smaller than 12 mm wide.
3. More segments means more resistors, which means that the power is more widely dispersed, and the voltage drop across the resistors is less. For a good full-range ESL, say > 450 mm wide, you will need maybe 30-40 segments, so you can use resistors with quite modest ratings ~ 300 V 1W - good surface mount resistors make for a nice compact design. The power and voltage ratings are not a big deal unless you test them with pure sinewaves > 5 kHz.
regards
Rod
I would love to see pics of that rig. I've been thinking thru ideas myself for a setup that will allow me to make those stator panels quicker.
Hi Bengel,
I plan to build wooden lattice stators and assemble them azz backwards. That is; the supporting lattice pieces will be assembled over the wires, as opposed to stretching the wires over a pre-assembled lattice. The tensioned wires will lie on a flat jig surface covered with wax paper. The wooden wire supports will then glue down one piece at a time, onto the wires and spacers. This method should yield flat stators with glue lines perfectly flush to the wires. I tried this on a small scale already and it looks like it's gonna work.
Of course, I shamelessly stole ideas from Bazukas, Bolserts and Golfnut for laying out the wires and pins. The pins will be set into 3/16 thick aluminum plates. A pair of 3/4 threaded rods with coupling nuts will serve as jackscrews to stretch the wires.
I hope I don't come to regret using wood rather than rectangular metal tubing for the jig frame but I'm trying to save some coin. Besides, I already had some hefty pieces of wood lying around; not to mention I almost had a stroke when I priced ($$$) steel tubing.
I will post some photos when it's done.
Jazz
Last edited:
I have zero experience with ESLs but if yal dont mind would like to add what ive experienced with a bunch of planer magnetic panel builds to perhaps help make decisions. Specifically with respect to the "curved" panels.
To avoid all the crossover issues etc I experamented for some time with curved. Built two types, one tru curve, the other multiple flat sections. The tru curve fist off is a very different animal largly as it does not really follow the typical very low mass possabilities of flat diaphragms. By this I mean that construction of the tru curved diaphragm required me to use very thick mylar. I think in the end I had to use 2 or 3 mill thick !!! thats typically about10 times as much mass as can easily be constructed with flat diaphragms. The reasion was I simply could not figure a way to mount thin films in a tru curve. With the tru curve arraingment I used the structural properties of much thicker film that heldits shape easily as you simply layed it carfully over the supporting frame work wich BTW was an exact copy of the Martin Logan design. Simple perf steel with foam strips. The foam strips acting as the support for the curved film. In such an arraingment the film becomes a riggid curved structure and the foam strips have some damping property terminating the film. It was very easy to construct once you had the suporting structure built. Now I dont know and maybe someone can educate us here but IMO this is all Martin Logan is doing. Its not really exotic and not elaborate tensioning of the diaphragm is going on at all. However the drawbacks were a heavyer diaphragm and the rigid film structures resonances were more obvious to my ears. Also very limited lower frequencys!! Need a huge panel to get to even 100hz.
Much better sound was the multi flat section unit using very thin film, as thin as u want. I used the same foam strips to divide up the vertical sections appox 3 inch wide over a curved perf steel / magnet structure. The hard part was tensioning the damn diaphragm over this. Fortunatly with very thin film the exact tension seemed unimportant so long as not too tight and I got a nice response down to 50 hz on a panel approx 3 ft tall and 2 feet wide. A tru full rage planer magnetic with at least a reasionable horiz directivity. It was my refrence but it was limited in SPL.
So the tru curve as I see it is limited in low freq response unless huge. The multi section is tricky to apply film. At least in my experience
To avoid all the crossover issues etc I experamented for some time with curved. Built two types, one tru curve, the other multiple flat sections. The tru curve fist off is a very different animal largly as it does not really follow the typical very low mass possabilities of flat diaphragms. By this I mean that construction of the tru curved diaphragm required me to use very thick mylar. I think in the end I had to use 2 or 3 mill thick !!! thats typically about10 times as much mass as can easily be constructed with flat diaphragms. The reasion was I simply could not figure a way to mount thin films in a tru curve. With the tru curve arraingment I used the structural properties of much thicker film that heldits shape easily as you simply layed it carfully over the supporting frame work wich BTW was an exact copy of the Martin Logan design. Simple perf steel with foam strips. The foam strips acting as the support for the curved film. In such an arraingment the film becomes a riggid curved structure and the foam strips have some damping property terminating the film. It was very easy to construct once you had the suporting structure built. Now I dont know and maybe someone can educate us here but IMO this is all Martin Logan is doing. Its not really exotic and not elaborate tensioning of the diaphragm is going on at all. However the drawbacks were a heavyer diaphragm and the rigid film structures resonances were more obvious to my ears. Also very limited lower frequencys!! Need a huge panel to get to even 100hz.
Much better sound was the multi flat section unit using very thin film, as thin as u want. I used the same foam strips to divide up the vertical sections appox 3 inch wide over a curved perf steel / magnet structure. The hard part was tensioning the damn diaphragm over this. Fortunatly with very thin film the exact tension seemed unimportant so long as not too tight and I got a nice response down to 50 hz on a panel approx 3 ft tall and 2 feet wide. A tru full rage planer magnetic with at least a reasionable horiz directivity. It was my refrence but it was limited in SPL.
So the tru curve as I see it is limited in low freq response unless huge. The multi section is tricky to apply film. At least in my experience
Last edited:
While curved panels are touted by some, at least in the case of the ML speaker they used to give up a lot in terms of smooth and even frequency response. They were the poster children for DSP room correction though. Haven't listened in ages, this may have improved.
Having said this, a smooth, near ideal frequency response may not be what you value most. Extended listening is in order, especially covering a variety of music.
Best,
Erik
Haven't listened in ages, this may have improved. Having said this, a smooth, near ideal frequency response may not be what you value most. Extended listening is in order, especially covering a variety of music.
Best,
Erik
So been working thru my design a bit and want to get some thoughts on the segment size....
20mm (or about 3/4") segments seems a bit small to me (but what do I know?). What are the audible consequences of going bigger?
I am starting to circle down on a ESL panel size of about 46 inches X 14 inches.... (hybrid design). If I use 3/4" width sections, that gives me 19 sections..... which seems overkill.
Here is what I have from the spreadsheet thus far... 150hz LF is not really going to happen but... hoping to maybe crossover in 200's
1 ==> h: 46.00 in height
2 ==> w: 14.00 in width
3 ==> d: 0.0625 in gap
4 ==> r: 5.0000 m distance
A: 0.4155 m^2 area (panel)
h: 1.1684 m height
d: 0.0016 m gap
Vpol: 3,175 volts Vbias (optimal)
Vsig: 6,350 volts Vsignal (max)
5 ==> N: 19 # sections
6 ==> fL: 150.00 Hz LFbreak pt
fH: 216600.00 Hz HFbreak pt
R: 48.22 Kohms Feed Resistance
Ctot: 1158.12 pF Cpanel
C: 60.95 pF Csection
w(sec): 0.74 in width section
Thanks for any advice.
Hi bengel
Your sums look about right, except for
1. The formula in the paper applies to the asymmetric ESL. For the symmetric version with the transformer connected to the central segment, the intersegment resistance should be 4 X your value.
2. Then the value needs to be split in two - half on one stator, half on the other stator (both in series with the stator capacitance).
3. Because of all of the holes in the stator, the capacitance will not be quite as high as for an unperforated plate/PCB/whatever. Fringing effects will make up for most of the holes, but if you assume the capacitance is 90% of that for a solid stator, you wont go too far wrong. I've not done measurements with wire stators, so I'm not sure what a 1.6 mm spacing means when the wires are thick.
regards
Rod
Your sums look about right, except for
1. The formula in the paper applies to the asymmetric ESL. For the symmetric version with the transformer connected to the central segment, the intersegment resistance should be 4 X your value.
2. Then the value needs to be split in two - half on one stator, half on the other stator (both in series with the stator capacitance).
3. Because of all of the holes in the stator, the capacitance will not be quite as high as for an unperforated plate/PCB/whatever. Fringing effects will make up for most of the holes, but if you assume the capacitance is 90% of that for a solid stator, you wont go too far wrong. I've not done measurements with wire stators, so I'm not sure what a 1.6 mm spacing means when the wires are thick.
regards
Rod
If you can tell me what fL you are targeting and width of each of the groups of 30 wires, I can create the directivity sonograms like I did for your previous build and see if one has a particular advantage over the other. My guess is that Config 2 will have slightly better dispersion in the top octave if you are planning on using fL=200Hz as you did previously.
I’m not sure if you have read through the build thread for CharlieM’s first segmented ESL.
If you haven’t, it may answer many of your questions on segment size selection.
A few posts of particular interest:
Rule of Thumb for number of segments for a given fL:
Post #23
Polar Response Comparison for different number of segments and configurations:
Post #39
Spreadsheet Electrical Segments .vs. Physical Segments for Symmetric Configuration 1 & 2:
Post #79
Post #103
For wire stators, the flat plate spacing of 1.6mm correlates well with wires spaced such that the diaphragm is 1.6mm away from the outer diameter of the wires. If wires are coated with high dielectric constant insulation like PVC, the spacing is to the outer diameter of the insulation not the copper conductor. Your 90% capacitance of flat plate approximation applies to wire stators as well, if the wire to wire spacing is kept similar are smaller than the spacing between wires and diaphragm.
Some pics and discussion of wire stator spacing here:
http://www.diyaudio.com/forums/planars-exotics/281348-sanity-check-wire-stator.html#post4492349
Indeed I have read through CharlieM's and Ken Siebert's builds.... Thanks so much for the direct links. I am desperately trying to search and read thru other posts so I don't re-ask the same questions that have been answered already.
All this info is like trying to take a drink from a firehose
. BTW using that formula gives me about 10 segments. That are about 1.5 inches (38mm) each.
FYI, unless someone would suggest otherwise, I think I am down to building 45 inch X 14 inch panels crossing over around 250 hz. This would be a symmetrical segmented panel. I am using the "tig rod" with 13 TPI rod approach CharlieM/Ken used.
They will be in a 15W ft X 35L ft room with the seating position about 15 foot from the speakers.
In summary, I have a couple of outstanding concerns/questions:
1) If I've done the math right, the spacing between the wires exceeds the stator to diaphragm spacing by a little bit... (.07 space versus .06 d/S space). Is this a problem?
2) I don't understand the concept of the capacitance of the segments. Can this be measured? i.e. take my meter and test the segment? Doesn't the material composition of the wire affect this? Or is this capacitance something totally different (if this makes sense).
3) The with the number of segments at 10, the size is 1.5 inches wide. Too wide?
4) These ESL's will be used in a home theater room and as such I have a big subwoofer (that I built). Any danger of the SPL of that sticking the diaphragm to the stators? What happens when/if that happens (fires etc...)?
5) Need to figure out what step up transformers to use for this config.
Thanks for all the help....
Golfnut, could you elaborate a bit more on the transformer setup you have in your system? What transformers are you using and how are they connected and to what step up ratio ? How can you reach lowest frequency of 50 hz without compromising on low frequency bandwidth ? At what voltage can you drive the transformers at 50 hz before saturation ?
I am currently building a 55 wire esl for which I intend to use six 2x6/230 50va transformers , 3 pairs in series. But I am hesitant to use them if there's a better alternative.
I am currently building a 55 wire esl for which I intend to use six 2x6/230 50va transformers , 3 pairs in series. But I am hesitant to use them if there's a better alternative.
Description of transformer configuration was given in sub-paragraph "Mistake 3" of Post #1 here:
http://www.diyaudio.com/forums/planars-exotics/234975-another-segmented-esl.html#post3470845
Image of concept posted here:
http://www.diyaudio.com/forums/planars-exotics/234975-another-segmented-esl-2.html#post3472414
Basically you use the 230V windings of the toriods in series for secondary, but remove or ignore the low voltage winding.
The toroids are stacked and a new primary is wound through the entire stack.
http://www.diyaudio.com/forums/planars-exotics/234975-another-segmented-esl.html#post3470845
Image of concept posted here:
http://www.diyaudio.com/forums/planars-exotics/234975-another-segmented-esl-2.html#post3472414
Basically you use the 230V windings of the toriods in series for secondary, but remove or ignore the low voltage winding.
The toroids are stacked and a new primary is wound through the entire stack.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.