DRSSTC design questions + project progress

[Vaclav]

Hi I’m new here,

I want to build a DRSSTC for my state-certified technician final project.
I have from now on roundabout 38 weeks.
I’ve already started.

In the past I've already tried to build a DRSSTC. I got about 10cm arc from this coil:

Primary: 5 Turns 6mm²
Secondary:
110mm Diameter
370mm length
0,1mm enameled copper wire diameter (I think this is far to low)
100x280mm topload
72kHz resonant secondary resonant frequency
fixed coupling

I think it's better when I make a complete new coil.
Is this coil good?
final.txt


I already built:
H-Bridge
Interrupter (Steve Ward)
Control circuit
Gate driver (I don’t use GDT)
Galvanic isolated power supplies for the gate driver

I’ve tested all and it works as it should. (At least I think so)
I don't have tested the feedbackfunktion I just used my funtiongenerator.

I made some H-Bridge tests with a 75W Alight bulb.

Voltage at bulb:


When the interrupter stops the H-Bridge the last falling edge is not straight, it’s more like an e-function.
I don't have a picture of this because I killed yesterday one of my IGBTs....
I Switched the 5V off while the H-Bridge was oscillating.
Both IGBT switched on I think, because of a wrong signal and the short-circuit current killed them -.-
So I made this highly professional paint picture:


Is this curve ok?? or really bad??? Idk...

Which ferrite core is good for the feedback is this one here ok (I have some of them here from my first coil):
https://www.reichelt.de/ferritring-59770027-ft-140-77-p7921.html?

I've added the circuit diagram
I uploaded some pics as a zip file to my dropbox (can't attach them here cause they are to big):
https://www.dropbox.com/s/jcgcy6su8wmc3cz/Tesla.zip?dl=0

Greatings from Germany

Vaclav

[profdc9]

Does your controller have undervoltage protection and is it set up correctly?

If the gate drivers on the controller are nor forced off as the voltage falls when power is removed then you may get shoot through blowing the IGBTs.

Dan

Hi I’m new here,

I want to build a DRSSTC for my state-certified technician final project.
I have from now on roundabout 38 weeks.
I’ve already started.

In the past I've already tried to build a DRSSTC. I got about 10cm arc from this coil:

Primary: 5 Turns 6mm²
Secondary:
110mm Diameter
370mm length
0,1mm enameled copper wire diameter (I think this is far to low)
100x280mm topload
72kHz resonant secondary resonant frequency
fixed coupling

I think it's better when I make a complete new coil.
Is this coil good?
final.txt


I already built:
H-Bridge
Interrupter (Steve Ward)
Control circuit
Gate driver (I don’t use GDT)
Galvanic isolated power supplies for the gate driver

I’ve tested all and it works as it should. (At least I think so)
I don't have tested the feedbackfunktion I just used my funtiongenerator.

I made some H-Bridge tests with a 75W Alight bulb.

Voltage at bulb:


When the interrupter stops the H-Bridge the last falling edge is not straight, it’s more like an e-function.
I don't have a picture of this because I killed yesterday one of my IGBTs....
I Switched the 5V off while the H-Bridge was oscillating.
Both IGBT switched on I think, because of a wrong signal and the short-circuit current killed them -.-
So I made this highly professional paint picture:


Is this curve ok?? or really bad??? Idk...

Which ferrite core is good for the feedback is this one here ok (I have some of them here from my first coil):
https://www.reichelt.de/ferritring-59770027-ft-140-77-p7921.html?

I've added the circuit diagram
I uploaded some pics as a zip file to my dropbox (can't attach them here cause they are to big):
https://www.dropbox.com/s/jcgcy6su8wmc3cz/Tesla.zip?dl=0

Greatings from Germany

Vaclav

[Vaclav]

Does your controller have undervoltage protection and is it set up correctly?

I don't have any undervoltage protection.

I have 2 ideas:

1. Switch the IGBTs instantly off if there is undervoltage.
Worst case: I switch them off while peak current.


2. I switch the IGBTs off when there's 0 current.
Worstcase: needs longer (worst 1 complete cycle time at 82kHz --> 12,2µS) maybe Voltage drops in this time to far and I can't switch them of in time.
Solution would be to add more capacity on my power supplies.


In both cases I have to add a diode in series to my power supply.
For the Undervoltage detection I would use one relay for each supply because I need galvanic isolation.

Vaclav

[Vaclav]

I think it's better when I make a complete new coil.
Is this coil good?
final.txt

[profdc9]

A relay might work, but it is slow and requires a lot of coil current to operate.

Perhaps use an optoisolator such as the 4N28/4N35 to pull the signal to ground if there is undervoltage.

[Vaclav]

A relay might work, but it is slow and requires a lot of coil current to operate.

Perhaps use an optoisolator such as the 4N28/4N35 to pull the signal to ground if there is undervoltage.

I will use the 2. idear because the caps are big enouth.

I have ~1A Driver current.
At least 220µF on the driver.
84KHz^-1 = 12µs --> 15µs for safety.

I think the effective current is quite constant.
So I use this equation.

U = I/C * t --> (1A/220µF) * 15µs  ~ 68mV

68mV  voltage drop is realy good.

This is the schematic for the undervoltage detection.

I could use op amps but there is not enough space in the case.
Detecting 0V is ok for me.

Can some one have a look on my Coil design?
I wanna buy some parts for it this week

Vaclav

[Mads Barnkob]

Which ferrite core is good for the feedback is this one here ok (I have some of them here from my first coil):
https://www.reichelt.de/ferritring-59770027-ft-140-77-p7921.html?

It might work, despite it is only a 2000 permeability core, it is advised to use around 5000 or higher for CTs in a DRSSTC.

Can some one have a look on my Coil design?
I wanna buy some parts for it this week

Vaclav

You are within all the best-practise limits of designing a good run of the mill DRSSTC, I see no short comings in your JavaTC design.

[Vaclav]

It might work, despite it is only a 2000 permeability core, it is advised to use around 5000 or higher for CTs in a DRSSTC.

Just to be sure this one here, T38 material would be much better?
http://www.farnell.com/datasheets/2168097.pdf

[Mads Barnkob]

I have used N30 material plenty of times, T65 and T37 is also good choices, too high and the 50% frequency roll-off might get into your resonant frequency region, check the datasheets.

[Vaclav]

When the interrupter stops the H-Bridge the last falling edge is not straight, it’s more like an e-function.
I don't have a picture of this because I killed yesterday one of my IGBTs....
I Switched the 5V off while the H-Bridge was oscillating.
Both IGBT switched on I think, because of a wrong signal and the short-circuit current killed them -.-
So I made this highly professional paint picture:


Is this curve ok?? or really bad??? Idk...

The new IGBT is here.
This is the actual voltage curve with a 75W light bulb as load.
I don't know is that good or bad?

[Mads Barnkob]

The switching transient spikes are within reason, very low energy and short lived, could be better, depends what DC bus voltage you did this test on?

The trailing edge with its RC characteristic also puzzles me, I doubt it would be able to damage a IGBT brick, there is simply not enough energy in it for that, they are after all rated for high short circuit currents for 1ms.

It could be busbar inductance and snubber caps RC time constant? Try to calculate it from your values. Or change the snubber caps to higher or lower values and see if the slope changes.

You have built a very neat looking bridge, clean layout and no mess of wires, so I really doubt that is the source of your problem.

On a little side-note, how satisfied are you with the EEVBlog diff. probe? If you want to, please do write a review in the electronics subforum.

[Vaclav]

The switching transient spikes are within reason, very low energy and short lived, could be better, depends what DC bus voltage you did this test on?

Test was with 230V DC

I used this connectors for my snubber caps and my TVS diodes.
https://de.farnell.com/amp-te-connectivity/1-5380758-1/mini-spring-buchse-tht-20-16awg/dp/2310196?ost=2310196&ddkey=https%3Ade-DE%2FElement14_Germany%2Fsearch
https://de.farnell.com/amp-te-connectivity/5050864-1/terminal-discrete-socket-6-53mm/dp/2452556?ost=2452556&ddkey=https%3Ade-DE%2FElement14_Germany%2Fsearch
Maybe the resistance is to high?
I will make a test without the snubbers tomorrow.

On a little side-note, how satisfied are you with the EEVBlog diff. probe? If you want to, please do write a review in the electronics subforum.

I can't say much about it its my only diff. probe and this was literally my first measurement with it.

[Netzpfuscher]

I think there is problem with your driver. On the inverter output is no component which can cause this. Please measure your gate drive if there is also such a signal.

The P-Channel FETs in your driver are AC coupled and you have a pair of UCC27321 and UCC27322. If you disable the the drive signal at the ADUM1200 with a zero, this zero gets to the UCC pair. The upper part gets inverted and the N-Channel conducts which is fine. But the lower UCC is not inverted and with a zero on the input the P-Channel conducts, but after a short time the coupling capacitor is full and the P-Channel goes to non conducting and the IGBT gate gets a floating input on one side.

With such a uncontrolled decay of the drive signal it's possible that you get massive shot troughs which can kill the IGBT easily. 

You should change you gate drive that it is capable to transfer DC. One option is to toss the P/N-channel fets and use only the UCCs which should be strong enough for the small bricks. I have driven a CM300 directly with a UCC pair.

[Vaclav]

It could be busbar inductance and snubber caps RC time constant? Try to calculate it from your values. Or change the snubber caps to higher or lower values and see if the slope changes.

I made measurements with one and with no snubber cap.
The measurement didn't changed.
And I don't understand why this should make any difference.

I think there is problem with your driver. On the inverter output is no component which can cause this. Please measure your gate drive if there is also such a signal.

Yellow = Gate
Blue = Interrupter signal
Gate1:

Gate:2

Gate3:

Gate4:


I think the bad spikes are due to this:

In the final assembly I will use shielded cables.

The P-Channel FETs in your driver are AC coupled and you have a pair of UCC27321 and UCC27322. If you disable the the drive signal at the ADUM1200 with a zero, this zero gets to the UCC pair. The upper part gets inverted and the N-Channel conducts which is fine. But the lower UCC is not inverted and with a zero on the input the P-Channel conducts, but after a short time the coupling capacitor is full and the P-Channel goes to non conducting and the IGBT gate gets a floating input on one side.

With such a uncontrolled decay of the drive signal it's possible that you get massive shot troughs which can kill the IGBT easily. 

I think adding a resistor is not a bad idea.

You should change you gate drive that it is capable to transfer DC. One option is to toss the P/N-channel fets and use only the UCCs which should be strong enough for the small bricks. I have driven a CM300 directly with a UCC pair.

I realy want to keep my driver design.
Adding the resistor for safety is probably not a bad idea.

I noticed if there is no load at the bridge the bridge voltage stays high when the Interrupter turns it off.


When I add the 75W bulb I geht this:



I think I have some parasitic capacity at the bridge output.

Time from Hight to 0V is ~8µs --> Tau = 1,6µs
cold bulb resistance = ~59R
hot bulb resistance = ~770R (I measured the current at 230V AC and calculate it)
1,6µs/59R = ~27nF
1,6µs/770R = ~2nF

When I change the duty cycle of the interrupter.
The bulb temp change --> bulb resistance change--> ramp change.

I think there must be some capacity between 2nF-27nF.

[Mads Barnkob]

Try to add a 10K resistor across the inverter output, I use this on my large DRSSTC. It has to do with the UD2.1b driver with phase lead has issues with residual MMC charge and it can act up weird and trigger randomly.

Not that you have this jittering issue here, but if you got residual charge on your bridge when the power is off, bleed it away.

[Vaclav]

Try to add a 10K resistor across the inverter output, I use this on my large DRSSTC. It has to do with the UD2.1b driver with phase lead has issues with residual MMC charge and it can act up weird and trigger randomly.

Not that you have this jittering issue here, but if you got residual charge on your bridge when the power is off, bleed it away.

Do you know where the residual charge came from?

By the way I really want to say thank you for the nice help I get here.

Next step will be building a bigger winding rig.
The old one is just too small.

[Teravolt]

hi there Vaclave, I like your design however I am with Netzpfuscher. dump the mosfet driver fets and the bootstrapping unless you can make it work. the UCC's will drive each igbt directly for short periods. I am not completely clear on your circuit because it is kind of broken up and hard to get a clear picture. I am assuming that the ADuM1200 are for galvanic isolation and they are driving the UCC's? (That would be my choice.) I have tried this with the

https://www.nve.com/Downloads/il710.pdf

and using them to drive

https://www.digikey.com/product-detail/en/ixys-integrated-circuits-division/IXDD614CI/CLA356-ND/2424696

with isolated supplies

https://www.digikey.com/product-detail/en/murata-power-solutions-inc/MGJ6D051505SC/811-3229-ND/6202202

later I ended up using

https://www.infineon.com/dgdl/Infineon-1EDI60N12AF-DS-v02_00-EN.pdf?fileId=db3a3043427ac3e201428e5da08f372a

to me I see a lot of shoot through at your driver scope shots. make your leads to the gates as short as possible and add dead time at u3c and u3d

 http://www.modularcircuits.com/blog/articles/h-bridge-secrets/h-bridge-control/

I hope that I have not overstepped on your thread. I am interested in what you are doing do you have a topside picture of your driver box

[Vaclav]

hi there Vaclave, I like your design however I am with Netzpfuscher. dump the mosfet driver fets and the bootstrapping unless you can make it work.

To me it looks like it is working I don't can see any problem.

This is the schematic with the pullups
Driver.pdf

the UCC's will drive each igbt directly for short periods. I am not completely clear on your circuit because it is kind of broken up and hard to get a clear picture.

I know the schematic is confusing...

I am assuming that the ADuM1200 are for galvanic isolation and they are driving the UCC's? (That would be my choice.)

That's right.

to me I see a lot of shoot through at your driver scope shots. make your leads to the gates as short as possible and add dead time at u3c and u3d

How can you see shoot through??
Can u explain that to me??

In theory I can adjust the dead time with P8 and P9 on the control board.
But I don't see any difference on my inverter ouput voltage and the current consumption (according to my power supply) of the inverter don't change when I adjust the trimpots.
Control.pdf

The connection to the gate can't be shorter.

I hope that I have not overstepped on your thread. I am interested in what you are doing do you have a topside picture of your driver box

[Vaclav]

Short explanation of my construction:
I have on box with 2x 230V AC inputs. On input is only for the inverter so I can use my variac to change the inverter voltage. The other is for the control voltage power supply.
Inside the box: Resistor for inverter cap charging. Relay to short out this resistor. Rectifier for the inverter voltage. Some Fuses + connectors for 230V Output (to the control voltage power supply).

There is a self-build power supply with 2 Transformer (each 2*18VAC output):
The power supply outputs are:
-4x 18DC each for on driver (galvanically isolated)
-12V for the fans. I get the 12V DC from an AC/DC power supply. This is a workaround. At first I wanted to supply my drivers with 3x 18V because I thought the 2 driver to the negative rail are at the same potential... that was a mistake and I had to get one galvanically isolated supply more.
-5V for the control board (with a regulator from the 12V)

Control board:
-feedback with adjustable phase lead. Signal via current transformer.
-over temperature detection of the inverter via NTC
-overcurrent detection via a current transformer
-under voltage detection
-optical fiber input for the interrupter
-some type of interrupter signal monitoring (if the on time is to long the interrupter signal is forced low)
-dead time adjustment to avoid shot through (independent adjustment for positive and negative edge feedback signals)
-4x adum for isolation between low voltage and inverter
-4x outputs to the 4 drivers

2x complete identical driver boards (each for one brick / two IGBTs):
Each driver board hast 2 identical drivers on it.
Every driver has one 5V regulator for the HV side of the adum.

[Teravolt]

Hi Vaclav, the picture, gate 4, has a lot of noise I think remove the 2k's that you have added and add 2k from IGBT gate and emitter and see if that helps. to my knowledge not many people hear have tried to do direct drive and usually uses transformers. I have seen this before. also try adding incrementally resistance say 2k 2w temporarily across 1/2 of the bridge output to the center tap of mains multiplier capacitors. The idea is to add a mild resistive loading to your bridge output to the center tap of the capacitors to damp transient noise generated by emi. If you look at your blue gate signal the transients are synchronized with your transitions if the IGBTs. this also can be caused by not enough dead time. I have tried what your doing with good results but I wanted to reduce the complications of lots of circuitry so it can easily be reproduced I have outlined this in my thread

https://highvoltageforum.net/index.php?topic=632.0

Using direct drive I think IGBT transition times small enough to use bricks up to 100khz out side normal operating frequency, also phase manipulation or PWM is more possible