Segmented wire stator ESL's

Everything else is two dixie cups and a string 😎

Greetings all from the Jazzman,

It still amazes me that an average Joe with practically no electronics experience can build a speaker at home that rivals the high end commercial offerings and bests most of them.  And building the actual ESL driver from scratch takes the cool factor to a whole new level.    

My latest speaker shown above would not have been possible without ESL gurus Bolserst and Golfnut sharing their extensive knowledge on the DIYAudio Forum.  Golfnut’s white paper [1] and segmented speaker provided the inspiration and Bolserst’s Segmented ESL Calculator and gift for explanation made it easy to derive the segmentation scheme and resistor values.  These guys are the best!  

I am most happy to share with you my DIY electrostatic loudspeaker projects.
Charlie Mimbs
Savannah, GA

[1] Wide-Range Electrostatic Loudspeaker with a Zero-Free Polar Response, D. R. White, JAES Volume 57 Issue 10 pp. 822-831, Oct. 2009

How do they work?

If your hair has ever stood on end while unloading a clothes dryer, then you've felt the same force that drives an electrostatic loudspeaker (ESL).

An ESL is a push/pull sonic motor consisting of a thin plastic diaphragm suspended between two conductive screens called stators.  
A separate power supply puts a DC biasing voltage on the driven diaphragm, and the output from an audio amplifier, routed thru a step-up transformer, puts the driving AC voltages on the stators.  And the ultra-light diaphragm responds with instant precision to reproduce the music with exquisite fidelity.

All of my ESL projects to date are bi-amplified hybrid designs using conventional woofers for bass with an active digital crossover upstream of the power amps.  

My first ESL’s used flat perf-metal stators, which gave great slam and imaging but beamed like crazy.  The next generation stators used segmented welding rod conductors on plastic light diffuser grids.  These had switch-selectable wide and narrow dispersion modes and a nice balanced sound but were visually just butt ugly.  Finally, my newest panels are more finely segmented for optimal dispersion, they are practically immune to arcing, and they have lovely oak lattice supported copper wire stators that both look and sound like fine musical instruments. 

Flat perf-metal panels are by far the easiest to build and they sound fantastic in the sweet spot.  They are also directional to the extreme, prone to arcing if the stator coatings aren’t applied to perfection, and they require a very stable amplifier to drive them, as their load is all capacitance.  

Welding rod panels can be segmented for wide dispersion and balanced response, and when segmented they are a much easier load (part resistive / part capacitive).  The downside is they are time consuming to build.  
Insulated wire stator ESL’s like the one shown on this page are my preferred choice.  They can be electrically segmented for wide dispersion, balanced response, and easier load.  And they are so well insulated that arcing is not a concern.  However, they are time-consuming to build; requiring a support lattice and a stout jig to stretch the wires.  

More to come on electrical segmentation but first; the basics of my newest speaker design:  

Dimensions:     66.5”H x 15”W x 19.75D
Bass cab:         4ft3, 9ft transmission line, V-section beam splitter    
Woofer:            10” Aurum Cantus AC-250MKII
Stators:             Segmented wire conductors on oak lattice, 10.5” x 46.5” active area 
Wires:               20 AWG solid copper, .010 XLPVC, 11/inch spacing, 43% open   
Diaphragm:       6-micron Mylar C with Licron Crystal ESD coating
Bias supply:      2.7kVDC from 115V/230V TFMR into a diode/capacitor cascade  
Transformers:   (2) 50VA 230V/2x6V toroidal, 76:1 ratio 
Crossover:        Behringer DCX2496, 48db/octave LR filter @ 228Hz 

Wavelengths shorter than a speaker’s radiating width tend to beam rather than spreading out.  And flat panel ESL’s are the worst for beaming because they require a large radiating area to offset their limited excursion and dipole rolloff.  The resulting “head-in-a-vise” sweet spot is great for solo listening at the focus but not so good for entertaining guests, where wider dispersion is needed.  

The typical way to widen an ESL’s dispersion is to curve its panel, and thereby it’s wave front.  Segmented ESL’s curve the wave front electrically, using discrete stator conductors driving discrete zones on the diaphragm.      

My segmented panel uses symmetrically arrayed, discrete vertical wire groups driven by stepped-frequency signals; acting as a line source projecting a cylindrical wave front.                  

Each stator has 90 insulated copper wire conductors arrayed in 15 groups of six wires.  These 15 wire groups are apportioned into eight discretely powered electrical sections; consisting of one wire group in the center of the stator comprising section one, and seven left/right paired wire groups arrayed symmetrically on either side comprising sections 2-8 (see Schematic).  

The section one wire groups connect directly to the amplifier interface and receive the full audio bandwidth above the crossover frequency.  The left/right paired wire groups in sections 2-8 are powered thru an RC transmission line that progressively steps down frequencies toward the panel edges.  The RC line consists of resistors inserted between the wire groups which couple with the wires capacitances to form a series of low pass filters.   

As driven by the sectioned stators, the diaphragm radiates the highest frequencies from only a narrow vertical zone at its center, and the left/right paired zones on either side radiate progressively lower frequencies toward the edges.  In this way, the width of the radiating zones is always less than the radiated wavelengths; resulting in all frequencies spreading out rather than beaming.  

·       Ideal (practical minimum) diaphragm-to-stator gap (d/s) is .062” (+.015/-.000)  
·       Wire diameter (insulation included) should not exceed the d/s 
·       Gap between wires (insulation-to-insulation) should not exceed the d/s
·       Ideal (max output) open area is 42%
·       Span between diaphragm supports: 70-100 x d/s 
·       Max span between wire supports (wire gauge/inches): 22/2, 20/3, 18/4
·       More/narrower wire groups = wider/smoother dispersion but no advantage < 12mm
·       Bias voltage: 2.5-5kHz
·       Transformer power/ratio: 80-120VA/50-100:1

Once the panel design is set, the next step is building a jig to stretch the stator wires. 

Stretching the copper wires to plastic deformation renders them perfectly straight and re-aligns their metallic structure such that they then remain straight when relaxed.  Experiments showed that stretching to 1% elongation is sufficient, so my wire loops were stretched from 47.5” to their 48” final length.   

Stretching ninety 20 gauge wires at once requires a strong jig and about 4,500 pounds of force.  My jig is a ¾ MDF platform mounted in a stout frame cut from yellow pine 4x4’s.  The wire loops wrap over .063 diameter x .250 length steel pins inserted half depth into drilled 3/16” aluminum plates.  The pins are angled 4 degrees from vertical to hook the wires.  One pin plate is stationary and the opposite moveable pin plate bolts to a 3/16 steel sub plate that’s welded to two 3/4 x 12 all-thread jack rods.  Turning the coupling nuts on the jack rods pulls the movable pin plate to stretch the wires. 

Note:  For my stators' wire diameter, correct spacing required using 1mm pins.  However 1mm pins were not strong enough and bent over when stretching the wires so I had to increase the pin diameter to .0625".   The 11 TPI all-thread rod comb guides corrected the wire spacing during glue-up.  

Below: Stator wires on the stretching jig

The wires are supported by an interlocking oak lattice which is assembled and glued down over the wires, in the stretching jig.  Before stringing the wires, the jig platform was covered with wax paper to prevent gluing the wires to the jig.  And before gluing, most wire tension was relaxed to prevent preloading and warping the stator.  Yellow wood glue was used for the wood-to-wood bonds and E6000 glue for the wood-to-wire bonds.

The vertical lattice rails were laid down into the jig first, and then the interlocking horizontal slats were glued down one-at-a-time, over the wires.  During this process, lengths of 5/8-11 TPI all-thread rods were placed over the wires as comb-guides to maintain correct wire spacing. 

CAD drawings for the cabinet, stator lattice, and stretching jig are available upon request.  Just email

Below:  All thread rods hold wire spacing during glue up

Below:  Oak support lattice assembled over wires

Below:  Completed stator 

The spacers bond the diaphragm to the lattice rails and set the diaphragm-to-stator gap (d/s) at .063”.  The vertical lattice rails are flush with the wires and its spacers are (1) layer of .063 x .075 wide 3M double-sided urethane foam tape.  The horizontal end rails run under the wires and anchor their end loops with glue bonds, and its spacers consist of (1) .047 x .075 polycarbonate shim bonded onto the wires with E6000 glue, plus (1) layer of .015 x .075 wide 3M UHB double sided foam tape over the shim (0.062 total).  The double-sided foam tape spacers bond the diaphragm to the stator instantly with minimal fuss. 

The diaphragm is vertically sectioned into equal thirds for stability.  It’s made of 6-micron Mylar C tensioned to 1.25% elongation using a pneumatic bike-tube jig.  The jig is an MDF platform sized two inches longer and wider than the stator and 2 inches high with all edges rounded over to .50” radius, sanded smooth and dusted with baby powder to prevent snagging the delicate diaphragm film.  A 700mm x 35mm Schrader valve type bike tube is stretched around it’s perimeter.  

The Mylar film is wrapped over the jig and secured on the back side with double sided tape.  Inflating the bike tube with a hand pump tensions the diaphragm.  Tension is gaged by first marking reference points on the diaphragm exactly 12 inches apart using a fine tip felt pen.  As the tube is inflated, the target elongation is reached when the distance between the reference marks reaches 12 and 5/32 inches.   The stator is then pressed into place over the diaphragm to affect the bond.  
Next, the periphery edges of the diaphragm were masked off with painters tape and the Licron Crystal ESD coating was spray applied in one “just wet” coat and allowed to dry for eight hours before assembling the panels.  The coating dries to a pale blue-gray, almost clear coating about 2-microns thick.      

Below: Bonding stator to diaphragm on bike tube jig

Below:  Bonded diaphragm ready for conductive coating

The charge ring is ¼ inch wide copper foil tape applied to the periphery of the rear stator, centered on the foam tape spacers.  The wire lead from the DC biasing power supply is soldered to it and when the front and rear stators are mated together the charge ring contacts and conducts the biasing voltage onto the diaphragm. 

Below:  Rear stator with spacers & charge ring

Below:  Completed front & rear stators ready for assembly

The segmentation scheme and resistor values were derived using the Segmented ESLCalculator spreadsheet.  From the spreadsheet options I chose Symmetric Config 2, both stators segmented, in eight electrical sections.  For my panel the spreadsheet calculated 120kΩ for R, with R/9 series resistance on the section one wire groups, 0.75R on section two wire groups, and R on sections 3-8 wire groups. 

It’s common practice to move/reflect the section one resistances to the primary side of the transformer to protect against core saturation (see Schematic, damping resistor R1).  Reflecting section one's R/9 resistances across an ideal transformer would divide the sum by the turns ratio squared (4.6Ω).  However, placing this much resistance on the primary side of a real transformer with its winding resistance, leakage inductance and winding capacitance, would result in significant rolloff of high frequencies.

The transformer winding capacitance adds to the load capacitance and its leakage inductance combines with the load capacitance to generate an ultrasonic resonance peak in the frequency response and rapid rolloff above it.  Coincident with this response peak is an impedance minimum which can be a difficult load for the driving amplifier.  When series resistance is added on the primary side it dampens this resonance peak. Too much resistance on the primary side over-damps the resonance, rolling off the high frequencies.

The spreadsheet assumes an ideal transformer is used, so it doesn’t calculate the effect of resistance on the primary side of a real transformer. The general guidance is to omit the section one resistors, add a 1Ω series resistor on the primary side and give it a whirl. My panels sounded really good with this initial setup. 

From there the only tuning, if any, is adjusting the series resistance on the primary and/or the first two stator sections to dial in the treble response.  Less resistance increases treble and visa versa.  My old ears don’t hear the highs so well but I didn’t want less than 1Ω on the primary side and the section one resistors were already omitted, so I reduced the section two resistors from 0.75R (90kΩ) to 60kΩ to brighten up the treble, and that works for me.  

The schematic and parts list show the spreadsheet values except with section one resistors omitted and reflected as 1Ω on the primary.  I think this would be optimal for most listeners. 

All remaining resistors on the secondary side are 2W, 500V in series.  Wattage/voltage are highest across the first resistors and decrease down the line.  Multiple resistors are ganged to spread the load, as follows: 

Section 1:          none  (reflected as 1Ω on TFMR primary)
Section 2:          (3)  30kΩ
Sections 3, 4:    (3)  40kΩ
Sections 5-7:     (1)  100kΩ + (1) 20kΩ
Section 8:          (1)  120kΩ

Below:  RC network resistors



Parts list for two speakers:

Each stat panel has an interface to its amplifier; consisting of a high voltage DC power supply to bias the diaphragm and one or more step up transformers to convert the amplifier’s output into the higher voltage AC required to charge the stators.  

Each interface uses (2) 50VA 230V/2x6V toroidal transformers wired in tandem with the 6V windings in parallel as the primary and 230V windings in series as the secondary; giving a 76:1 winding ratio. The DC biasing supply uses a floating ground that’s center tapped between the transformers 230V windings.

The DC biasing supply is a simple half-wave rectifier and voltage multiplier outputting 2.7kV.  It’s powered by 115VAC mains current into a 115V/230V transformer and diode/capacitor ladder with a 20MΩ charging resistor at the output.  The charging resistor helps stabilize the charge on the diaphragm and limits the potential current that might otherwise sustain any arcing to the stators.     

Below: Amp/panel interface 

The woofer box is a single-fold, tapered transmission line stuffed with 0.5lbs/Ft3 of polyfil.  The line’s sectional area is 125% of the woofer’s cone area at the front, tapering to 100% of same at the terminus.  The cabinet is ¾” MDF sheathed in 5mm red oak plywood and the panel frame is solid red oak.  

To minimize the woofer box’s profile and footprint, its volume extends upward, behind the stat panel, and its frontal surfaces are angled to form a V-shaped “beam splitter” which deflects the panel’s rearward sound out the open sides of the speaker rather than back to the diaphragm.   

When designing the bass section I followed Roger Sanders’ lead and opted for a transmission line enclosure and a woofer with low moving mass, low QTS, and low inductance coupled with a very strong motor magnet.  The ideal matching woofer doesn’t exist of course but low inductance takes priority and the Aurum Cantus AC250 MKII I chose works pretty well.  

The speaker pair is vertically bi-amplified using a Behringer DCX2496 digital crossover feeding a pair of vintage Carver TFM-25, 225 watts/channel stereo amplifiers.

The general guidance is to set the crossover frequency is least two octaves above the diaphragm’s drum head resonance with a 24 db/octave filter slope or at least one octave above resonance with a 48db/octave filter slope.  My diaphragms resonate at about 90 Hz and the crossover is set at 228 Hz using the Behringer’s 48db/ocatve Linkwitz-Riley filter.  

Below: Beam splitter transmission line bass section


Below:  Bob Carver loves my new speakers!
               Cellphone Video from Carverfest 2016

Cellphone Video on Youtube (perf-metal panels)

Links to other pages: