My first DRSSTC on bricks

[Laci]

I started the building of my first DRSSTC using two used SKM400GB126DH2 modules.Had some problems because not getting what I ordered,but eventually I got generously what I ordered.

The busbar is made out of two 135x22x2mm aluminium rails,with the large DC bus capacitor in the middle for good current sharing.Two 1uF 1250V snubber capacitors are coming,an aluminium heatsink soon and the gate drive components on perfboard/PCB left to do.I'm also trying to modify an ATX PSU for higher voltage and current for proper gate drive supply.

The MMC is rated 0.094uF,4000VDC using 2 in series,4 in parallel WIMA FKP-1 capacitors mounted on a painted piece of tough cardboard(hopefully good enough insulation).Soldering to aluminium was impossible,so soldered the capacitors together using a large solder joint,with a copper wire in the middle.This is probably the weakspot,we'll see how it turns out to be while operating.

Some pictures about the inverter and MMC:




I have a question regarding the inverter and another about the UD3. I see everybody hooking up the CTs between the inverter and the MMC.My MMC fits perfectly on the inverter and that would be satisfying and also really low inductance.Is it possible to connect the CTs to the other side,between the inverter and the primary?
About the UD3:does it uses internal comparators in the Cypress chip instead of external ones,like the TL3116?In this case the expensive comparators are omitted,but is the chip fast enough?(in my case about 150kHz)

[Hydron]

CT between inverter and primary is fine - the place to avoid is between primary and MMC due to high voltages present there.

As for the MMC it is indeed likely the weak part - 4kV sounds like a low rating for a coil that needs brick igbts, and will limit your max current.

[Teravolt]

lead solder will not flow on aluminuim

[Mads Barnkob]

Good to see you are taking on building a large DRSSTC, atleast that is the feeling I got when you are using SKM400 bricks.

There is a few plot holes in your plan. Driving SKM400 at 150 kHz is that something you calculated first? ( http://kaizerpowerelectronics.dk/tesla-coils/drsstc-design-guide/igbts/ ) Sounds like its going to take some gate driving power and will dissipate quite some switching losses.

Your MMC is grossly underrated, you simply need to build something better and bigger. At 150 kHz that little poor thing will only handle 24Arms and even at 400Apeak you will be hitting 4500Vpeak and then you are not far from destruction of it. I used the Wima FKP as a substitute in MMC calculator ( http://kaizerpowerelectronics.dk/calculators/mmc-calculator/ ), but that also shows a 33 degree Celcius temperature rise at 200 uS/BPS, you are going to scorched that little cardboard and caps

I am not sure, but I think the UD3 was built to be used for QCW aswell, so I would expect it to work into some 500 kHz region.

[Laci]

Thanks for the suggestions!We talked about switching speeds and losses here: https://highvoltageforum.net/index.php?topic=549.msg3415#msg3415

The MMC is poor because they had no more in the local store.The plan was to build a 4 in series,8 in parallel MMC

I plan to use freewheeling for long pulses if the losses won't be much.Philip probably used 2kV CDE caps,2 in series for 4kV and 4 in parallel just like my MMC.

What about the primary?I can buy copper tube from a local store from 0.6 cm diameter.The thicker tube stays cooler than the thinner?

[Netzpfuscher]

That's right, 500kHz is the maximum for the UD3.

Perfboard is not ideal for a high current gatedriver. I would prefer SMD mit big ceramic decoupling to keep the switching noise out of the driver.

I drive my SKM400 with a 13 to 10 GDT with 24V on the GDT driver so you get 18V at the gate.

[Teravolt]

the universal driver can do 500khz but the SKM400 probly can not what frequency is your coil? your snubber will not work if you are soldering to aluminuim. at least bridge the cap connection to the screws with copper at the end of your assembly

[Laci]

That's right, 500kHz is the maximum for the UD3.

Perfboard is not ideal for a high current gatedriver. I would prefer SMD mit big ceramic decoupling to keep the switching noise out of the driver.

I drive my SKM400 with a 13 to 10 GDT with 24V on the GDT driver so you get 18V at the gate.

Thanks for the suggestions!I'll try to etch a PCB and use 10uF THT X7R capacitors,two in parallel for each N-P MOSFET pair(IRFZ44NPbF and IRF4905PbF).If I succeed to modify an ATX PSU,then I can change the gate drive voltage.
I also received a CY8CKIT-059 board and checked your UD3 firmware,but some files are missing.Do you have a simpler firmware something like the UD2.7 or I should make one on my own?

the universal driver can do 500khz but the SKM400 probly can not what frequency is your coil? your snubber will not work if you are soldering to aluminuim. at least bridge the cap connection to the screws with copper at the end of your assembly

The SKM400 bricks definitely can't do 500kHz,but hopefully they can 150kHz.I plan to enhance the MMC connections or even make a completely new one if this fails.The snubbers are good,1uF and mounted directly to the bricks with the screws.

[Netzpfuscher]

There are no files missing, I think psoc creator didn't find the bootloader files. This is a common issue, go to the schematic of the UD3 QFN project, double click the bootloader component and set the path to the correct booloader binary "\common\binary\UARTldr". Then it should compile.

I have no smaller firmware, because I'm not interested in a step back ^^

If it still not compile send me the log file.

[profdc9]

I made a board for the CYC8KIT-059 based UD3

https://github.com/profdc9/DRSSTC-PCB-Pack/tree/master/Psoc5-power

so you don't have to etch a board.

Dan

That's right, 500kHz is the maximum for the UD3.

Perfboard is not ideal for a high current gatedriver. I would prefer SMD mit big ceramic decoupling to keep the switching noise out of the driver.

I drive my SKM400 with a 13 to 10 GDT with 24V on the GDT driver so you get 18V at the gate.

Thanks for the suggestions!I'll try to etch a PCB and use 10uF THT X7R capacitors,two in parallel for each N-P MOSFET pair(IRFZ44NPbF and IRF4905PbF).If I succeed to modify an ATX PSU,then I can change the gate drive voltage.
I also received a CY8CKIT-059 board and checked your UD3 firmware,but some files are missing.Do you have a simpler firmware something like the UD2.7 or I should make one on my own?

the universal driver can do 500khz but the SKM400 probly can not what frequency is your coil? your snubber will not work if you are soldering to aluminuim. at least bridge the cap connection to the screws with copper at the end of your assembly

The SKM400 bricks definitely can't do 500kHz,but hopefully they can 150kHz.I plan to enhance the MMC connections or even make a completely new one if this fails.The snubbers are good,1uF and mounted directly to the bricks with the screws.

[Netzpfuscher]

You must double check the pinout of the dev-board project in psoc creator and the pcb files. I shuffled the pins around to fit the Ethernet-Interface.

I work on a second ethernet option, which uses a ESP32. With this you get WIFI and wired ethernet and a small problem with noise is solved. The W5500 ethernet chip uses a unprotected SPI protocol (no checksums) which can lead in false data under some conditions. The ESP32 interface connect through the UART of the UD3 and uses a packet based protocol with a transport layer. every packet is protected with a CRC32 and a sequence number. If a packet gets corrupted the transport layer resends the packet until it was received correctly. It's like a simple version of TCP/IP.

This is the new interface board for now: https://www.olimex.com/Products/IoT/ESP32/ESP32-POE/open-source-hardware
I think I integrate it in the new PCB if I find a little bit time.

[Laci]

I made a board for the CYC8KIT-059 based UD3

https://github.com/profdc9/DRSSTC-PCB-Pack/tree/master/Psoc5-power

so you don't have to etch a board.

Dan

Looks great,although I'll make most of the driver on a perfboard.

There are no files missing, I think psoc creator didn't find the bootloader files. This is a common issue, go to the schematic of the UD3 QFN project, double click the bootloader component and set the path to the correct booloader binary "\common\binary\UARTldr". Then it should compile.

I have no smaller firmware, because I'm not interested in a step back ^^

If it still not compile send me the log file.

Thanks,it works!Not sure if the bootloader location was the problem or something else,but it's OK now.
I want to use a bluetooth 4.0 BLE module and android app for interrupter instead of fiber optic,which I bought for cheap and it's of course clone and can't pair my phone with it.Firmware update can solve this hopefully.
Another idea was the using of an ethernet module,which I received factory defective:none of the two ethernet LEDs turns on.
Sadly I don't know how to get started with the UD3,so it'll take for a long time to make a working board.

[Laci]

Some progress updates around here.The inverter is almost done,just the GDTs should be mounted somewhere(B64290L0082X830) and add the precharge circuit.Driving these large bricks seems to be possible!I will upload some scopeshots soon and hopefully estimate switching losses.

[Laci]

I'm getting close to the destination of this project.
The inverter only requires the precharge circuit now.It looks nice,doesn't it?


I still can't send gate waveforms because I don't exactly know how to run a test on the UD3 without feedback signal.In this picture you can see the four LEDs.LED2 blinks not constantly,while LED1 and LED4 is ON and LED3 is OFF.Is this how it's supposed to work?

[Netzpfuscher]

The UD3 should produce output without feedback, without feedback you see the start oscillator.
I think the watchdog has triggered. You can reset the watchdog with "kill reset"

But for operation the UD3 needs a watchdog reset signal at least every 20ms. Your controller should send this "0xF0 0x0F 0x0F" (In Binary!) or you need to disable the watchdog.
Just set the "watchdog" configuration to 0. But if the connection dies the UD3 won't stop to fire the coil 

LED1 = Interrupter
LED2 = System fault
LED3 = Over current
LED4 = Receive data

[Laci]

LED1 is always on no matter what I do."tr start" should start the interrupter,so turn LED1 ON and "tr stop" should stop the interrupter,so turn LED1 OFF?
LED2 still lights up randomly and with the "eeprom save" command
LED3 is always off yet,without OCD
LED4 turns off when I connect to the bluetooth module so that should work

I extended the app with a terminal to use every command,but noticed problems with the serial communication colours,so that needs some improvements.

EDIT: LED3 and LED4 turns ON when I start the bootloader.What's the purpose of it?I tested the UD3 with many settings changed,still no output signal...

UPDATE: some waveforms with a simple pwm(130kHz,800us ontime).Yellow is the gate of one IGBT,blue is the inverter output with a resistive load.I tried to increase the gate resistance which removed the overshoot on the rising edge,but not on the falling edge.The current gate resistors are just 1 ohm.Rise and fall times are pretty much the same,around 450ns from 0% to 100%.Overall I'm quite pleased with the waveforms,didn't expect such a good result.We'll see how they perform with a proper load after getting the UD3 to work!






2nd UPDATE: I forgot about the UVLO,now the UD3 passed transient mode,hopefully other functions works too.I think I saw a maximum value of 100kHz for the autotune functions(which by the way seems to work too-outputs <10A without CT).Is the autotune really limited to 100kHz?

[Mads Barnkob]

Good progress on the coil and inverter, this really shows that if you pay enough attention to the brigde design and making it as compact as possible and with good snubbering, the resulting waveform show it was not in vein, good job.

Netzpfuscher: There is a lot of good trouble shooting and background information about UD3 in this thread, maybe make a FAQ in your own original post with a copy of these answers? Or just a link to different threads where it is dicussed?

[Netzpfuscher]

I think I write a FAQ in the github wiki, that everyone can benefit from this.

No, the autotuning function isn't limited to 100kHz, this is a bug. I correct this as soon as possible.  Edit: It's fixed: https://github.com/Netzpfuscher/UD3/commits/master

If you find more bugs please feel free to write an issue:
https://github.com/Netzpfuscher/UD3/issues


If transient mode works all system faults should be gone. Due to this thread I think it should be a big thing to add a alarm monitor where you can see which signals are not good.

[Laci]

I did a small test on 50V with experimental tuning on an old primary and already see some nice sparks.
The driver seems to work in transient mode.At the start frequency the interrupter is also working,might need some more smoothness and fix the unexpected behaviour of the app with fast clicks.
Primary current feedback is not working yet.I have a 1000:1 CT with 5.1 ohm burden resistor.What settings should be changed to get it working?When do I know if the UD3 senses the primary current?

An alarm monitor would be nice in the app too even as simple as four indicators like the LEDs.

Related to the inverter: I see not much overshoot on the output.In this case how should I adjust phase lead?
One thing I noticed while putting the driver into a plastic box is that the wire lenght and type from GDT drive to GDT primary is critical.Rise and fall times increases drastically if two thick twisted wires are used instead of multiple thin twisted wires from driver board to the frame of the box.

[Netzpfuscher]

Have you connected the CTout, ZCDA and ZCDB?
Have you set the "ct1_ratio 1000" and "ct1_burden 5.1"?

If you connect over putty and connect to the UD3 type: "status start" then you should see the current.

You test with transient mode and it runs? What do you mean with interrupter, the transient mode is the normal interrupter.

You know that the UD3 sees zero crossings if it switches more cycles than the set start_cycles. I should add a indicator ^^