Photomultiplier Power Supply Transfomer
The transformer L1, L2, and L3 could be wound on an EPCOS ETD 29 core - the B66358-G-X187 (Farnell part number 1781876) is the cheaper N87 option. The centre-tapped primary - L1 and L2 - would be bifilar wound as a single layer of 10 turns of a twisted pair 0.85mm diameter enamelled copper wire - plus grade 2 enamel. The wire's maximum total overall diameter is 0.937mm, so 20 turns would fit as a single layer on the EPCOS flat-on-the-board B66359W1013T001 former.
The former's winding window is 19.4mm long, and 5mm high, and the bifilar primary, plus a single layer of 120 micron transformer tape, would use up 2mm of that, leaving 3mm for the secondary winding.
Even with a voltage doubler circuit on the output I need about 540 turns on the secondary to get the maximum output voltage I want.
The finest wire I could wind on myself is 0.1mm diameter copper wire, 0.129 mm OD with grade 2 enamel insulation. I could get 150 turns per layer on the former, which would mean a minimum of four layers, allowing me up to 0.75mm per layer. This would allow me to space each of the copper layers with four layers of 120 micron mylar transformer tape.
The area of each layer is it's width – 19.4 mm - by its length – 52.8mm (on average, from the former data sheet where it is listed as ln), or 0.001024 m^2.
The capacitance of a parallel plate capacitor C is the product of the permittivity of free space
multiplied by the dielectric constant of the material between the plates – here a mixture of mylar and an acrylic adhesive - which I'll guess to be three.
multiplied the area of the plates – here 0,001024m^2 - divided by the thickness of the material, here 0.48mm.
This comes out at 57pF per layer. Three such capacitors in series are equivalent to a single 19pF parallel capacitor.
If I went for 0.14mm copper wire, 0.176mm OD with grade 2 enamel insulation, I'd need five layers of winding. I could still put four layers of 120 micron mylar tape between each layer and still have room to put a single layer of 120 micron tape on top of the outer-most layer.
The interlayer capacitance is still 57pF but there are now four such capacitors in series, equivalent to a single 15pF capacitor.
It would be nice to “bank” the winding, splitting it into two segment, each 9mm long, halving the interlayer capacitance to about 30pF, and putting ten such layers in series, for an effective parallel capacitance of perhaps 4pF but this means a two-section former, which doesn't seem to commercially available for the ETD 29 core. Two section cores are commercially available for other core formats.
It is feasible to wind banked windings as a series of self-supporting cores, using wire which is supplied coated with a thin layer of thermosetting resin, which can be heated to the setting temperature after winding by a brief and carefully calculated burst of current to get the whole of each bank of the coil up to the setting temperature after it has been wound and before the supporting collar is removed.
It is a messy and labour-intensive procedure. You need a thin insulating washer (which ends up stuck to the bank) between the coil and the supporting collar to let you take out the supporting collar after it has done its job. This washer also insulates one bank from the next, and reduces the bank-to-bank capacitance.