Uploaded February 2026 | Updated September 2026, 2 weeks ago
Had these parts set aside and also had this SiC fet so it is what it is. Built some other versions of these using Labcoatz schmitt trigger oscillator approach but those were just tuning a shunt cap value. When tuning the resonant network consisting of the DC blocking + shunt + primary to ground + secondary loading.. It's different, but same same. Might rework this for higher voltage but would prefer a 1200V fet.
My RFC is probably something close to 30uH
1nF 3s3p shunt cap using 1nF MLCC's
Series resonant cap is some rf silicon 330pF I had with a 1nF on top of it
Secondary natural resonance was somewhere around 3mhz
(to retune this so I could push ~60VDC at ZVS I would try increasing the series capacitor by maybe a couple hundred pF, then reduce the shunt cap from 1nF to maybe ~700pF, and then deal with the body diode doing the work at the lower voltages)
These are annoying to tune and I'm pretty sure this one isn't tuned that well.
It's not hard to end up with ZVS and very low output loading in the network to
where you have good waveforms but the output power is trash.
The tuning obviously has to account for a changing plasma streamer being
the 'resistive' load. When using a 1.5A linear regulator right next to the driver
IC the decoupling is not an option as the high speed peak demands will
annihilate the rail. This would also inherently wreck the 5V rail and jerk the
oscillator around. I would put at least 50uF low ESR at the drive loop and
some more on everything else.
The SiC is less forgiving in some areas and more forgiving in others, but
overall allows for some decent out of tune operation without burning up
in the losses. While it didn't seem like I was on the edge of destruction it
was like a game of chasing higher output power relative to soft switching.
One or the other was easy, but not both at the same time etc.
Had these parts set aside and also had this SiC fet so it is what it is. Built some other versions of these using Labcoatz schmitt trigger oscillator approach but those were just tuning a shunt cap value. When tuning the resonant network consisting of the DC blocking + shunt + primary to ground + secondary loading.. It's different, but same same. Might rework this for higher voltage but would prefer a 1200V fet.
My RFC is probably something close to 30uH
1nF 3s3p shunt cap using 1nF MLCC's
Series resonant cap is some rf silicon 330pF I had with a 1nF on top of it
Secondary natural resonance was somewhere around 3mhz
(to retune this so I could push ~60VDC at ZVS I would try increasing the series capacitor by maybe a couple hundred pF, then reduce the shunt cap from 1nF to maybe ~700pF, and then deal with the body diode doing the work at the lower voltages)
These are annoying to tune and I'm pretty sure this one isn't tuned that well.
It's not hard to end up with ZVS and very low output loading in the network to
where you have good waveforms but the output power is trash.
The tuning obviously has to account for a changing plasma streamer being
the 'resistive' load. When using a 1.5A linear regulator right next to the driver
IC the decoupling is not an option as the high speed peak demands will
annihilate the rail. This would also inherently wreck the 5V rail and jerk the
oscillator around. I would put at least 50uF low ESR at the drive loop and
some more on everything else.
The SiC is less forgiving in some areas and more forgiving in others, but
overall allows for some decent out of tune operation without burning up
in the losses. While it didn't seem like I was on the edge of destruction it
was like a game of chasing higher output power relative to soft switching.
One or the other was easy, but not both at the same time etc.
