@Teravolt: Have you finished your coil? An update would be nice.

I know this thread is a few years old, but the topic name does match my new achievement.
After getting rid of my flashover problems, I am now able to modulate the ramp to generate notes.

Here is a short preview at very low power (32 V at the bus).
The coil works with phase-shift modulation. A full power run is planned,
but at the moment the sunset is very late, and I do not want to disturb the neighborhood.

I do not use MIDI. The frequency of each note and the duration until the next note
is transmitted to the FPGA in the coil via an optical serial communication.

I love the idea to use the feedback comparator as self oscillation circuit. Nice idea!

The coil is up and running again. 

Some thoughts about the failure:

• I have readjusted the phase-lead (was off significantly) but do not know why. I own a Coilcraft SLOT7 kit and tested different values until I achieved the best result. Now the switching losses should be lowered.
• By raising the secondary coil and reducing the coupling, the primary current went up. My FPGA code is written as universal DRSSTC/QCW driver. One feature is a pulse skip mode. When the OCD trips, the current is limited by freewheeling cycles. During my experiments with lower coupling, this feature was active. Near the top of the ramp, the current hit the OCD threshold (160 A) and pulse skip mode was active for several milliseconds.
• The ramp was too long.
• No current waveform monitoring after changing the system.

The conclusion is: Ever change a running system, but double check everything.


To avoid flashovers I rotated the brass connection clamp. Now the screw head points downwards.


At the moment the secondary is only raised by 4mm (coupling: 0.375) instead of 10mm (coupling: 0.35).
With a 11 ms ramp from 20 to 85 % a current of 160 A is reached. The OCD trips sometimes.
No flashovers due to increased distance between topload and connection clamp.

I do not want to fry the IGBTs again, at least not today.
Tomorrow I will lower the secondary even more. The goal is to stay below 160 A with ramps up to 16 ms and of course no flashovers. Perhaps I have to isolate the connection clamp as suggested by davekni.


Some lucky shots:
Arc length is 130 - 140 cm (measured via pixel counting with 30 cm torroid as reference).
I hope to improve arc length by increasing the coupling back to the inital value.

Thanky you davekni for the thoughts and questions.

Yes, very impressive coil!  Congratulations on fixing your flashover issue too.  Thus you likely don't want to change anything.  Still, I have a couple thoughts and questions.  In general, I believe QCW coils work better with as high coupling as possible (without flashovers).  Thus I'm thinking about alternate solutions.

Thanks. My goal of the project was to implement a QCW without a buck-converter with my own FPGA based driver.
The size and electrical parameters are close to Loneoceans QCW 1.5 (<-- thank you). Mainly to have a safe starting point for my VHDL adventure.

Where are the flashovers hitting primary?  Is it always at the highest point where upper end of primary is bolted to lead wire?  That connection could be insulated, or even an entire horizontal plastic layer (big washer) could fit (and seal to) the primary coil form just above top of primary coil to block arcs.

The flashovers are mostly between the brass connection element (upper end of primary to lead wire) and
the secondary and somtimes directly to the top load. Insulation is a good idea. I will keep this on my to do list.

Do you get no top arc when a flashover occurs?  It is hard to imagine no arc forming for 75% of the ramp when it does form at lower Vbus.

I reviewed a video. The top arc is always forming, and my first thoughts were incorrect.

Would it work to test 15% to 60% at 333Vbus?  That should theoretically behave the same as 20% to 80% at 250Vbus.  If it doesn't, perhaps there's something about the high-frequency components of H-bridge output voltage initiating the flashover arc.

I have tested 20% to 50% at around 330Vbus prior opening the thread. The result was a flashover in 10-20% of the sparks.
With 10mm lifted secondary, 20% to 90% at even 360 V is possible.
Today I lowered the secondary 2 mm (lifted 8 mm from initial setup) and everything is still fine.

I have ordered acrylic disks with a height of 2 mm each. With the disks, I can adjust the height of the secondary easily.

Today I also tried 18 ms ramps with 10 Hz repetition rate. This was too much stress for the IGBTs.
All eight are shorted internally and disintegrated partially. Everthing else is fine (gate resistors, driver etc.).
Next monday the mouser delivery will arrive.

Hello,

a few years ago i build a QCW coil, but never finished the project completely.
This week I want to finish the project.

Short description of the coil:
* h-bride with 8 x FGH75T65SHD-F155
* V_bus: rectified 230V AC mains = 320-350 V DC
* Upper pole of the combination betteen primary and secondary: 406 kHz with a drop of about 30 kHz during spark growth
* Primary capacitor: 12.4 nF / 10 kV
* Peak primary current: 160 A
* Ramp form 20 % to 80 % of V_bus via phase-shift-modulation
* Ramp rise time 10 - 20 ms
* Ramp fall time 1 - 4 ms
* Ramp wick time 3 ms at 20 % (flat period prior rising ramp)
* Pointy Tungsten electrode as breakout point
* JavaTC file as attachment

The coil is running very well with nice straight sparks (approx. 160 cm), but only at bus voltages of around 250 V. If I crank up the voltage to the full 340 V, flashovers between primary and secondary coil are present (even with a ramp from 20 to 50% and 20 ms ramp rise time).

I measured the field with a nearby oscilloscope probe. The flashover occurs at around 3/4 of the ramp. It feels as if the spark at the top of the coil can not break out and therefore the voltage at the top of the coil rises to much.

Do you have any ideas? Should I reduce the coupling (calculated 0.391)?

Greetings
Gerd

Do you have a schematic of the controller and a manual?
Is the configuration of the CPLD available?

I have a short update.

A few days ago I received PCBs for a rogowski coil integrator.
Additionaly I wound a small coil onto the insulator PTFE of an RG178 coaxial cable.
The insulator has an diameter of 0.86mm and the used wire is 0.1mm. The windings are wound without space between the windings.
As a protective layer a heat shrink is on top of the wire. The total diameter of the coil is only 1,75 mm.

At the moment I only tuned the low frequency range (active integrator with opamp).
A test at 50 Hz and around 6 A peak shows the same respones as my professional coil.
Both coils are designed to have 5 mV/A output signal.

The next step is to tune the high frequency range (passive integrator).
For this task I designed a pulse generator (16 µs pulse with 150 A peak and flat top, rise and fall time around 800ns).
I found the calibration procedure/waveform in an official calibration document.

Thank you for your comments.
A connection to the internet is not planned. So very low risk because of the old XP.

Do you think 950 EUR is ok for a unit in very good condition (no cracks, good screen no scratches).

This thread is very inspiring. I like the concept to measure small voltages without common mode influence.

I have searched for commercial units and other DIY projects. During my search I found a similar project:
 https://hackaday.io/project/12231-fiber-optic-isolated-voltage-probe

The approach is similary, but they use other circuits to realise the transmitter (fet input stage) and a transimpedance amplifier as light to voltage conversion.
Layout and schematic is available.

Here is a picture of the windings of the Ultra Mini probe: Page 19: https://gmw.com/wp-content/uploads/2019/01/APEC-March-2015.pdf