new problem

[Garyf]

Since my last post, I've replaced my primary with copper tube,  gradually increased bus voltage, best arc after tuning was about 16 inches before bridge blew at 420V. 
Decided to see if smaller coil would give better performance with TO-247s, ordered 3.5" X 16" coil from ebay, blew IGBT at low power.  Checked drive signal against the interrupter signal (w/signal generator input) and got this:

Just so I'm clear, the drive signal start and end should have complete pulses, not truncated?  Drive signal should show delayed start and stop to align with zero?  This may explain my growing bag of dead silicon.
The driver is an early Steve Ward variant without phase lead adjustment; I've started assembling Gao's driver but meanwhile I'd like to fix this.

[Mads Barnkob]

Which early Steve Ward DRSSTC driver is this?

Because he have written warnings on hit website that the early drivers should not be used due to being unstable/unpredictable or not taken certain scenarios into account. He advised on only using his UD1.3 or UD2.1 drivers, everything else is only left there for historic purposes.

[Garyf]

It's Tom Blitch's screamer design, similar to Steve's ~ 2008.  I think originally it worked OK
https://docs.wixstatic.com/ugd/9c6dc3_495554cc885a431c901e7bfd0cc4d8e0.pdf
Seems like it's just gating.  Before I try to troubleshoot, I'd like to confirm that something is wrong.....?

[Max]

Good evening,


The beginning spike is normal (for all UDs without startup oscillator* and probably for any similar driver). When your driver gets an interrupter signal, your primary current is zero (thus no feedback) at first, so your driver must switch a first time. After this, it can and has to switch based on the feedback.
Now as you use it with artificial feedback, the interrupter signal will activate the driver anywhere in that signal and it will take a random time to the first zero crossing.
 
On the other side that last pulse is not normal at all and absolutely deadly for your switches. Apparently it does not wait for the next zero crossing to turn off, but does it immediately. That indicates a problem with the flip flops, which assure in normal operation, that turning off can only happen in zero crossings.
 
The consequence of the behavior of your driver is fatal. At the end of the on time, primary current is maximal. Now if your driver switches at a random point of the period, you have very good chances to switch a considerable amount op current, causing incredibly switching losses and spikes. Bye bye IGBTs...
 
*Actually this does not solve the problem, but delays it by the number of periods your startup oscillator is working. So you‘ll get a strange spike in the middle.


Kind regards,
Max

[Garyf]

Thanks for confirming!  This explains some things.
Looking at the various driver circuits, most have a delay filter on the way to the FF that looks like this:

Checking the signal after the filter, I see that at lower pulse widths, I'm not triggering the Schmitt.

 The time constant for this on my board is 50 us, Gao's is 100 us, and the Ward 1.3 is 400us.
I'm using the OneTesla interrupter designed for the TS that is so minimal it doesn't even have this delay filter.
I guess I should have looked closer when matching interrupter to driver.

[Garyf]

I went back and re-read Steve Ward's design guide:  "Also note that the flip flop only does its synchronized shut down when PRE is high.  So this leads to choosing what value we want to use for the RC.  I typically size it so that t=RC=1.5P"

My time constant was too long for my 270 khz coil.  I cut it from 50 us to 5 us and now it works OK, but the message I took is that the value should be adjusted for each coil.   If I hadn't looked closely at the signals, I'd still be toasting IGBTs.

[Mads Barnkob]

I was wondering why I never ran into this problem, but I think it is because of coincidence of the resonant frequency of coils I build and when I used what drivers.

It is pretty tricky when Steve uses the DRSSTC1 driver in this design guide ( https://www.stevehv.4hv.org/drsstc_design.htm ), but the same mechanism is in UD1.3/UD2.1 without descriptions of it.

As far I can see:
R2 / C3 in UD1.3 http://stevehv.4hv.org/newdriver08/DRSSTC_pndriver1_3b.pdf has a time constant of 440 uS
R13 / C10 in UD2.1 http://www.stevehv.4hv.org/leadcomp/UD2_1revB/UD2_1revbschem.pdf has a time constant of 100 uS

I used UD1.3 in my DRSSTC1 ( http://kaizerpowerelectronics.dk/tesla-coils/kaizer-drsstc-i/ ) resonating at 70 kHz, so this never had problems
I used UD1.3 in my DRSSTC2 ( http://kaizerpowerelectronics.dk/tesla-coils/kaizer-drsstc-ii/ ) resonating at 330 kHz, this blew up a TO-247 half-bridge, then I upgrade it to 60N60 minibricks and no problems after that.
I used UD2.1 in my DRSSTC3 ( http://kaizerpowerelectronics.dk/tesla-coils/kaizer-drsstc-iii/ ) resonating at 37 kHz, so this never had problems either

[Garyf]

Upon further review:  looking back over various driver designs, the original steve ward:

He used this for several coils, then in 2008, rev 1.1b, he added a diode:

He originally recommended the time constant to be 1.5 times the resonant period.  I found it worked OK if the period was between 1 and 4. 
With the addition of the diode, it's much more forgiving, but still needs to be greater than 1 period.