Building the HVPS transformers (part 1)
High voltage transformer
The high voltage transformer is a 1:40 transformer converting 100Vp-p to 4000Vp-p, which is then rectified to give -2kV, +2kV, and +8kV (via a multiplier). Winding a transformer supporting voltages this high requires a special design — just overlapping turns as usual will place very large voltages across the thin enamel of the wire. I designed and 3D printed a special high-voltage bobbin, which has six sections separated by plastic fins.
The coil is wound in each of these sections in sequence, with each section having one sixth the voltage across it, and the fins separate each section from the last. When the wire needs to cross from one section to the next, a piece of kapton tape is placed over it to increase the insulation between the wire destined for the bottom of the section, and the wires at the top of that same section, as shown:
Two thousand turns are put on with a manual winding machine (333 or 334 per section):
Next, the wound bobbin is baked for half an hour at 80°C to drive off any moisture. It is dipped into Glyptal 1201 enamel, air dried and re-baked for an hour at 80°C, and this is repeated once more to give two coats of enamel. Note that both the temperature and the enamel solvent are harsh to the plastic bobbin, but it holds up well enough if handled gently while softened.
The primary is much simpler due to the low voltage. I simply made a bobbin from a few pieces of fishpaper (vulcanized fiber) and adhesive, then put 50 turns of magnet wire on:
In this pic you can also see the enamel coating that has been applied to the secondary.
This winding will eventually get dipped in Glyptal as well, but not until the power supply is tested: if the ratio needs adjustment, this winding is the place to do it.
This transformer will eventually be potted, but not until the power supply design is well tested — expect a post on that in a while.
The transformer was measured with an LCR meter:
- Primary unloaded: L=4.015mH Q=84.8 ESR=2.98 at 10kHz (DCR = 0.2)
- Secondary unloaded: L=9.736H Q=47.0 ESR=13.1k at 10kHz (DCR = 77.0)
- Lpri (sec shorted); 249µH; k = 0.968
A coupling constant of 0.968 isn't great, but this is somewhat intentional. It's a result of the two windings being on separate sides of the ferrite core rather than being wound concentrically. This structure allows easy, good isolation between the windings. A quick simulation confirms it will still produce the correct output. This will be tested on the prototype board soon.
A few measurements are made by applying waveforms to the primary using a function generator and measuring both sides. First, a ratio check:
With 19.77Vp-p on the primary, the secondary gives 845Vp-p. The computed ratio is 42.7. This is lower than the LCR-computed ratio (the square root of the ratio of the inductances) of 49.2 due to the effect of loading on the weak coupling.
To get a feel for the behavior over a range of frequencies, a slower square wave is applied:
This shows a ringing around 65 kHz. Because of the low number of cycles before ringdown, it's tricky to measure this resonance precisely using this method. To get a better measurement, the square wave frequency is increased to near the approximate resonant point. The sine wave generated by the resonant circuit superimposes on the square wave, and the frequency of the latter can then be adjusted to match the former:
This frequency measures as 59.7 kHz:
The filament transformer is a toroid, wound onto an off-the-shelf 1mH power inductor. It's wound twice: once with regular cheap wire for testing, then once the ratio is set, a second time with high voltage rated wire. This two-step process ensures good insulation, as the HV wire has silicone insulation that is easily damaged during handling. A layer of tape is wrapped around to hold the wire in place. No enamel dip is necessary.
Sorry no pics, I'm lazy.