Skip to content
Benjamin Broce edited this page Aug 6, 2022 · 5 revisions

Power

Heat Test @ 24V, 3" gap, 10 turn primary (modified rev 1 board)

Ran for 20min continuous arc (will be much less power dissipation with max 50% duty cycle music) with the high-voltage side (coils, arc gap) enclosed in 1/4" acrylic cylinder with lid, with 1/4" ventilation gaps around the top and bottom. For reference, the melting point of acrylic is 160 C (we were nowhere near this). Results:

  • Acrylic walls - ok
  • Acrylic top - very warm but ok
  • Heat sink arc electrode - very hot but ok (we don't think it affected wood underneath)
  • Flyback coil - very warm, primary coil very hot but no sign of wear
  • MOSFET heat sink (very large) - very hot, probably needs a fan
  • Power shunt resistor - hot but ok (not used in final design)
  • Power caps - hot after 20min (with upgrade to 8 caps)
  • Resonance caps - very hot
  • 12V regulator - hot enough to need heat sink
  • Screw terminals for primary and power - hot

Conclusion - Can be run for at least 20min continuous arc, or 30-40min continuous music, preferably with the following upgrades:

  • Use a fan or more ventilation to avoid uncomfortable heat on enclosure lid (for those who touch)
  • Potentially use a larger heat sink electrode? (currently using 3"x3"x1.5" sink with fins @ top)
  • Increase the wire thickness of the primary windings on the flyback coil (AWG 14-16 stranded max)
  • Add a fan to cool the MOSFET heat sinks and boards (caps)
  • Use at least 6 power caps in parallel evenly distributed with respect to the primary coil with large PCB traces
  • Get resonance caps with a lower series resistance (ESR) to maintain resonance specs while decreasing heat loss
  • Potentially add small heat sink to voltage regulator
  • Use higher rating (30A) screw terminals for primary coil and power

Current Test @ 12V, 1/2" gap, 9 turn primary (board rev 1)

  • 300mA off
  • 4.7A on
  • 85mV peak-to-peak audio input trigger

Transformer

22.5kHz square wave on primary test -> Unloaded (no arc) secondary voltage (small CRT flyback, 9 turn primary)

  • 100mV peak-to-peak on primary (square) -> 1.72V peak-to-peak on secondary (sinusoid)
  • 500mV peak-to-peak on primary (square) -> 8.65V peak-to-peak on secondary (sinusoid)

Say we have a roughly linear gain of 17.25x on output voltage at this frequency. This means that our 12V primary square wave put us at ~207V peak-to-peak unloaded on the secondary (no arc), and a 24V primary square wave would give us ~414V peak-to-peak unloaded on the secondary (no arc).

If the flyback follows the 1:1 (voltage step : turn ratio) properties of a traditional transformer, this would mean that our secondary has a turn ratio of ~0.05797. With our 9-turn primary, this would mean our secondary has ~155 turns.

Does a commercial flyback transformer exist with this ratio?

Audio

As our driver circuit outputs a square wave at the frequency of the input as it exceeds a fixed reference value (ideally 0V), normal audio files work as input, but are very noisy. We've found that the arc sounds best when the driver input is given a single sawtooth wave or midi "chiff" voice with no sustain (midi "square waves" are less reliable).

Resources

Tone Generator for Testing: https://www.szynalski.com/tone-generator/

Online MIDI synth: https://signal.vercel.app/

Experimental Audio -> Midi File Converter: https://github.com/spotify/basic-pitch

1-2 voice music files:

We need a simple way to split these into different audio channels (left/right) if there is more than one note at a time, so we can split up the different notes onto our two coils.

Clone this wiki locally