Musical Class-E SSTC [Help Tuning]

[TiagoBS]

Hey guys, I'm starting this new project and would like some help tuning and understanding scope waves for CW and music operation.

I'm following this guide made by Zach from LabCoatz.
https://www.instructables.com/Making-the-Ultimate-Class-E-Musical-Tesla-Coil-Sin/

Currently my setup looks like this:
-

-


I am using a 20A power supply with voltage adjustment between 17-31VDC.
I also wound this transformer on a old flyback ferrite core, didn't calculate anything just wound the primary and secondary and adjusted the number of turns until I reached 61vdc output after rectification.



Here's the first problem, the ferrite core is getting quite hot during operation. The transformer primary and secondary don't seem to get hot, just the core.
Any idea why?

Coil operation at 61vdc

Here are the coil specs:

-

-


JavaTC recommends approximately 17nF for Resonant Tank Cap.
However, with this value it seems that the operation is not close to ZVS. Using 3.3nF plus transistor capacitance between DS on the other hand seems to leave the coil running close to ZVS but JavaTC calculates the primary frequency almost double the secondary.
What's wrong here?
In all tests I used 3 turns on the primary.


These are the readings I took while running.
Here using 17vdc
-


And here using 61vdc
-


I did the tests using three different transistors.
For all of them I had the best result using a 3.3nF capacitor between Drain and Source.

-IRF640N
-FGA60N65SMD
-K40T1202

When the coil is working with the interrupter at low frequency the transistors do not heat up, however when I go at higher frequencies or duty cycle I start to have heating problems.
Even so, it is possible to keep the coil running in CW for a few seconds.

-What I would like to understand is the behavior of the gate wave, why does it have all this ripple and why does it get worse with increasing voltage?
Another point is the audio quality, but I believe it is first necessary to solve the problem of correct functioning in ZVS.

Edit: I'm a data analyst with no electronics background, so I lack some of the basics, so I might end up asking the same question in different ways or I might struggle if the explanation is too technical. However I am very curious, persistent and I like to solve problems!
Thanks to everyone who can help.

[klugesmith]

>> Here's the first problem, the ferrite core is getting quite hot during operation. The transformer primary and secondary don't seem to get hot, just the core.  Any idea why?

We don't know the operating frequency or flux density (Bmax) of your un-engineered transformer on flyback core.
Core loss would be proportional to frequency and sort of proportional to square of Bmax.   
You could reduce Bmax by increasing primary and secondary turn counts in same proportion.
That core might be from a design intended to run at 16 kHz, as in CRT displays. 

If you have a fan to blow air on heatsinks for semiconductors, it could blow air on that transformer.

Maybe hot core is OK.   Some ferrite core materials for SMPS are designed for best magnetic performance at 100 °C.  There are applications where if a core does not get that hot in normal operation, the design is too big and heavy and expensive.  Industrial electric motors got progressively more compact in 20th century, partly from development of wire insulation that could endure higher temperatures.

[davekni]

Currently my setup looks like this:

Nice work tightly pairing your primary coil lead wires.  This is important for high-frequency coils.

I am using a 20A power supply with voltage adjustment between 17-31VDC.
I also wound this transformer on a old flyback ferrite core, didn't calculate anything just wound the primary and secondary and adjusted the number of turns until I reached 61vdc output after rectification.

Another option is to use 20-31Vdc directly.  Requires a higher-current (and lower voltage) FET and fewer primary turns (one or perhaps two).

Here's the first problem, the ferrite core is getting quite hot during operation. The transformer primary and secondary don't seem to get hot, just the core.
Any idea why?

Kludesmith did a great job of explaining this.

JavaTC recommends approximately 17nF for Resonant Tank Cap.
However, with this value it seems that the operation is not close to ZVS. Using 3.3nF plus transistor capacitance between DS on the other hand seems to leave the coil running close to ZVS but JavaTC calculates the primary frequency almost double the secondary.
What's wrong here?

Analog simulation is great for learning how such things work.  Easy to try values and probe voltages and currents without frying anything.  My favorite one is LTSpice.  There are other free ones, both PC programs and on-line.  No need for paid programs.
The simple way to think about your specific question is this:  For much of each cycle the FET is shorting the 3.3nF capacitor.  A short circuit is infinite capacitance.  So average capacitance is an average of 3.3nF and infinity.  (Average 1/capacitance to avoid getting infinite result.)
Another way to think about this:  Resonant time is only while FET is off.  L/C resonance makes a little over a half-cycle while FET is off.  (Half-cycle above DC supply voltage and a bit more before and after.)

-What I would like to understand is the behavior of the gate wave, why does it have all this ripple and why does it get worse with increasing voltage?

That is likely resonance of FET lead inductance with FET capacitance.  (3.3nF cap lead inductance contributes too.)  Best to have FET body as close to ECB connection as possible, and very-short cap leads.  Of course, that precludes screw terminals, so makes changing FETs a bit harder.  ECB layout also matters, especially FET source connection.  Return for driver chip should be separate from power current path, joining as close to FET source lead as possible.  There are some 4-lead FETs (especially in SiC) where that source connection is made inside package, which is even better.
For now, you could keep screw terminals and move FET as close as possible and shorten cap leads.  That would make some improvement.

BTW, 74HC14 schmitt-trigger oscillator frequency is sensitive to temperature and power supply voltage.  That is a drawback of this design.

Good luck with your project!  Looks like nice success already.

[YSPACE Labs]

I don’t think the drain-source cap should have a value that resonates with the primary inductance at the secondary resonant frequency. Class E is weird, and the secondary probably somewhat acts like the output network in a class E PA, and it takes energy from that drain-source cap through the primary. It’s definitely not a simple resonant circuit because the optimal ZVS class E waveform has two slopes - one as the inductor charges up the capacitor, and another steeper downward slope as a resonant circuit discharges the capacitor. In my experience (from my Class E coil), the drain-source cap just has to be tuned by looking at the waveforms and seeing if you need less or more capacitance. I’m not 100% sure about the theory, so correct me if I’m wrong.

[YSPACE Labs]

Quick correction maybe: Actually, according to Steve Ward (on his class E schematic), the primary inductance and DS cap do form a tuned circuit.