Audio transformer : digging into details

Audio input transformer is a step-up kind with ratios of 90 for treble and 270 for bass (quote from Sheldon Stokes says 96/270. diyAudio/bolserst says 87/290. diyAudio/Imazed says 96/289 … But this is based on voltage or turns measurement ? Probably first choice: voltage measurement.

You can read 55/110 here. This ratio is for windings themselves, and overall this makes 2*55 and (110+2*55+110): “110 / 330”. It may be related to turns ratio this time.

From diyAudio/bolserst

Maximum voltage at treble side should remain far below 3kV, which can be forced by an optional clipping board that clamps at 2200V. This 3kV corresponds to about 33V at input side.

Components used inside audio transformer do have to support up to 6kV. But this happens only with high power amplifiers such as Quad 303 (2x45W). With 5W, voltage inside audio transformer will stay below 100V. But who wants to use as little power as 5W with a low sensitivity speaker? Therefore, use high voltage components only! specifies 10kV for resistors and 4 to 6 kV for capacitors. Considering schematic (3 resistors  inserted in serie in the bass line), VR37 are probably good enough. VR68 will be overkill but are still good choice since same size as original components. Besides, 1W vs 1/2W.


An interesting discussion occured on diyAudio, here. See some extracts from contributor ‘bolserst‘ below:

For a complete understanding of the ESL-57 crossover as well as its acoustic design I would highly recommend reading the description by Peter Baxandall on pages 169 – 179 of “Loudspeaker and Headphone Handbook”, J. Borwick, Ed.

See attached ESL-57 schematic to which I have added the transformer leakage inductance and winding capacitance since they are directly involved in how the crossover works. The values were taken from the Baxandall description…actual production values may vary slightly.

Loading Choke
The purpose of the loading choke was to reduce the magnitude of the impedance peak that occurs at the low-to-mid frequencies of all transformer driven ESLs. It is my understanding that most tube amplifier output stages do not like to be driven at high levels without some amount of loading…internal tube arcing can occur. Without the loading choke, the impedance in the mid-bass region might have been high enough to cause problems with some tube amps of the day. The KLH Nine had a similar loading choke.

For the brief description below to make sense, you must understand that an ESL panel that is small relative to listening distance has a response that naturally rises 6dB/octave when driven from a voltage source; response is flat when driven from a current source.

Bass panel Crossover
Ignoring diaphragm resonance, the bass panels have a natural 6dB/oct rising response. The 6 x 180K resistors in series with the Bass Panel effectively provide current drive above about 300Hz and flatten the response of the bass panels. Looking at it another way, the RC filter puts a 6dB/oct LP filter on the voltage reaching the bass panels. Either way you look at it, the response remains flat until about 1kHz where there is a 2nd order LP network formed by the leakage inductance (46H) and the panel capacitance(400pF).

Below 300Hz, the response would still fall at 6dB/octave. The addition of baffling around the panels and small side wings keeps the response from falling quite this quickly. The diaphragm resonance with Q of 2 or 3 provides equalization for the bottom octave or so.

Tweeter panel Crossover
The sum of the winding capacitance and tweeter panel capacitance in combination with the leakage inductance(145mH) forms a 2nd order LP network just above the audio band at about 24Khz. The resonance is damped by the 150K resistor(junction 9 & 14). This is not immediately obvious, but at 24Khz, the 560pF caps and the mid panel capacitance are essentially short circuits provided the electrical path for the damping. Most other ESLs damp this HF resonance by including a small series resistance of 1-2 ohm in the primary circuit.

The two 560pF caps and 270K resistor were added for production models after serial number 16800. The purpose was to shelve down the drive voltage at lower frequencies by 6dB to avoid arcing the tweeter panel when used with amplifier power > 15W, which was the power level originally designed for.

Mid panel Crossover
The two 150K resistors provide current drive for the mid panel flattening the response as was described in the description of the bass panel crossover. The 560pF cap in parallel with one of the 150K resistors provides a bypass for this resistance so the right amount of damping is applied to the 24kHz tweeter panel crossover resonance. Without it, the response of the tweeter panel would peak up at the top end.

Mid panel & Tweeter Panel Acoustic Crossover
Both the mid panels and tweeter panel are rolled off on the top end (above about 2kHz) due to the curvature of the panels spreading out the radiated HF sound. So the LP for the mid panel is purely acoustic, not electrical. The tweeter panel response is flattened back out using the slightly under-damped resonance at 24kHz.
ESL 57 model (from diyAudio/bolserst)
Internal wiring, seen from left
Internal wiring, seen from right

As you can see on pictures above:

  • outer wires of the 2 small coils connect to outer terminals (1 and 5 for serie 2)
  • inner wires of the 2 small coils connect to inner terminals (2 and 4 for serie 2)
  • small coils are on the top side (near plots 1 .. 5) and are a priori identical

How many turns ? I got this from one source:

  • Small coils are expected 7850 turns gauge #39 (0.0889mm) !! not what I measured, see below !!
  • Audio inputs (B/R) make 74 turns, gauge #21 (0.7239mm)
  • Last, big coils, 2 times 3926 turns gauge #39, turns ratio 53

On my side, I measured for small coils:

  • for one transfo, 180 turns per layer (width = 1 inch, thus diameter<0.141mm), and computed 1715 Ω/km: gauge #37 (0.113mm)
  • for another transfo, I got this on just one layer (slightly less accurate): 34.7m for layer #4 and 64.7Ohm and 3.1g -> 1.86Ohm/m, 89g/km: definitely gauge #37
  • for one coil, 41 layers, first and last being partly filled, around 180 turns/layer, 180*40.5=7290 turns in total
  • for another coil, 40 layers only, around 182.5 turns/layer, 128 only for first and last, 7211 turns in total, 113g of wire
  • this makes turns ratio 97.5 .. 98.5, resistance close to 2.2 kΩ. Obviously only total number of turns was monitored

For big coils, and each instance of high voltage wiring:

  • 8 layers, first and last being half filled
  • 565 turns per complete layer: about 3955 turns in total (turns ratio 53)
  • target is 1.93/1.73 kΩ (note that inner and outer wiring have same # of turns but not same radius thus not the same resistance)

For big coils, low voltage wiring:

  • I confirm gauge 21, 74 turns, 0.58 Ohm, 13.25m, 1 layer only
Big coil, LV, 1 layer, 74 turns
113g of wire in a small coil

Additional notes:

  • 1st layer is half-filled, and starts towards left
  • seen from connection wires, small and big coils are NOT wired in the same sense: anti-clockwise for small ones
  • 3.5 turns of paper is used between layers for insulation of big coils, only 1.5 turns for small coils
  • big coils have additional external insulation layers (4m30!), less for small coils (1m)
  • not all coils measure the same !
Summary of measurements on one sample (ratio 53/98)

My summary:

  • apply turns ratio of 53/98 (this makes a voltage ratio of “106/302” before applying yield and loss, to obtain hopefully the famous “90/290” after)
  • small coils: layers of 180 .. 185 turns, gauge 37, 40 .. 41 layers, 7250 turns (first and last layers filled at 75%), set winding tool with a spacing for a gauge 2 steps below (25.4/180). 1.5 turn of insulation between each layer (1.5 inch), 5 turns for final insulation
  • big coils, HV: layers of 565 turns, gauge 39, 8 layers, 3950 turns (first and last half-filled)
  • big coils, LV: 1 layer of 74 turns gauge 21
  • once completed with external insulation, small coils should mesure less than 42x76mm (paper core is 21x56mm, height 36mm). One turn will vary between 19.3cm and 15.4cm (avg 17.35cm thus roughly 1265m in total and 2.2kOhm)
  • for big coils, height of paper core is 74mm (insulation can be 2×1.5 inches wide)

2 wires on the ‘ring’ (loading choke), that join in the middle (round the big coil on the right)
black: 2 grounds that go to the big coil, yellow/green: audio input
big coil on the right (serie 1): brown-black-black-blue (6-13-7) + audio input green/yellow
red (outer) connect to 1 on this side, 4 on the other (serie 1)

Pictures imported from other sites:

blue/brown 1/2, brown/blue 4/5 (serie 2)

 1,307 total views,  1 views today