# Building a DRSSTC Pt. 14 - Building GDTs & Ctlr

Blog entry posted in 'Building a Dual-Resonant Solid State Tesla Coil', December 18, 2014.

Hello everyone, welcome back to the DRSSTC build!

I'm sorry it's been so long since my last post. Between working on various projects at work and my own personal schedule, I haven't had much free time lately to post an entry. Hopefully this one will make up for it though!

Today I'll be discussing the construction of the gate drive transformers (GDTs) and the controller.

For those of you who don't know, GDTs are used to drive the gates of the IGBTs in the bridge that we looked at in the previous entry. They are necessary, or at least very useful, for several reasons. Arguably the most important reason for GDTs in DRSSTC applications is isolation. GDTs are often 1:1 transformers that protect the sensitive electronics on the controller from the high voltages that appear on the bridge. Some designs use 1:1.5 or 1:2 ratios, depending on the specifications of the IGBTs, but the ones I used are 1:1. Another reason why GDTs are necessary is because in an ordinary H-bridge you would need a high-side and a low-side driver. IGBTs are switched based on their Gate-Emitter voltage (that is the voltage across their Gate terminal and their Emitter terminal). However, you'll notice that not all of the transistors' emitters are referenced to ground, so it would fluctuate depending on the voltage across the load. Having this inconsistency means you wouldn't be able to control it with, say, 15 volts, since at a given time the voltage on the emitter could be 170, 340, or even higher (depending on your bus voltage). 15 volts on the gate and 170 volts on the emitter would not allow it to switch on. This is another reason why isolation is required. A third reason the GDTs are necessary is due to the gate capacitance of the IGBTs. If you look at how field-effect transistors work, you'll see that there is a thin insulating film (usually made of silicon-oxide) between the gate contact and the doped substrate. You may recognize that this creates a capacitor - Two conductive materials separated by an insulator. An IGBT is effectively a MOSFET switching a BJT, so the device will have a gate capacitance just like any other MOSFET. When running at the high frequencies of the DRSSTC resonant circuit, charging the gate capacitance (in order to turn on the IGBT) within a single cycle would require a very large amount of current. Looking at the datasheet for my IGBTs (found here), the Gate-Emitter charge required to switch on the device is 22nC:

89786

Based on the formula:

$I = \frac{Q}{t}$

where I is the current in Amps, Q is the charge (in Coulombs), and t is the time in seconds, In order to charge the gate capacitance in, say, 50nS, we would need:

$\frac{22\cdot 10^{-9}C}{50\cdot 10^{-9}s} = 440mA$

440mA to switch on the transistor. Most logic-level devices can only source a small fraction of that! Also, some transistors have significantly higher gate-emitter charges (some are even in the μC range!), meaning it could even take several amps or even tens of amps to switch on the devices. This is why we use GDTs.

When designing a gate-drive transformer the first thing you'll want to determine is how many turns you'll need. Too few and your core will saturate, giving you very poor performance. Let's start with the following formula for flux density:

$B = \frac{V\cdot t}{N\cdot A_{e}}$

B is the flux density in Teslas, V is the voltage at which you'll be driving the GDT in volts, t is the on-time of the signal in seconds, N is the number of turns, and Ae is the cross-sectional area of the core in square meters. Saturation occurs when the flux density reaches 0.2T, so we'll re-arrange the formula and to solve for 'N', since we're looking for the minimum number of turns that we need:

$N = \frac{V\cdot t}{B\cdot A_{e}}$

Let's plug in our values and see what we get. My controller drives the GDTs with a 15 volt signal, so that is our V. To get t we must first find the period of our signal. If you'll recall back when we were designing the resonant circuits, we determined that my operating frequency will be 122kHz. Since period is 1/frequency, that gives us a period of 8μs. However, we're not quite done yet. Since t is the on-time, we then have to multiply that value by the duty cycle as well. In this case I'm going to assume 10%, so 8μs * 0.1 = 0.8us. B we determined earlier will be set to 0.2T to signify saturation. Finally, we can generally get Ae from the datasheet. The Ae for the toroid cores I ordered from Digikey is stated to be $6.466\cdot 10^{-5} m^{2}$ . Plugging that all into the formula we get:

$\frac{15v\cdot (0.8\cdot 10^{-6})s}{0.2T\cdot (6.466\cdot 10^{-5})m^{2}} = 0.9279$

I'll round that up to 1 turn. Less than 1 turn will saturate our core, making it unusable for the Tesla coil. Now, this calculation really doesn't do us all that much good, as we're obviously going to want more than one turn on the GDT. However, it's an important thing to check to ensure you're at least on the right track.

Most DRSSTC coilers wind their GDTs with 8-15 turns, though it often involves a bit of trial-and-error, as well as some waveform testing using a signal generator and an oscilloscope. I re-wound my GDTs several times before I got an acceptable waveform. Something else to watch out for is having mismatched inductances between the secondaries. If your secondaries aren't the exact same length, you may get some switching delay that could cause your coil to perform poorly, and even damage the IGBTs. Steve Ward (well-known for his extensive work in DRSSTC building and design) has recommended simply using CAT5 ethernet cable left in its sheath. The conductors inside are individually insulated and already twisted in pairs, reducing leakage inductance significantly. This also allows tight, even winding of the primaries and secondaries. His method is to solder the ends of all the white wires together and use those as the primaries, and use each colored conductor as its own secondary. Here are the two GDTs I wound, one for each bridge:

89787

I took several waveform measurements with these GDTs -- Some with a damping resistor, some with the GDTs connected to a dummy (capacitive) load, and some with both a damping resistor and a dummy load. Here is the most recent waveform I got, feeding a 20v pk square wave from my function generator into the primary:

89808

Notice there is some slight overshoot and a very small amount of ringing, but it is almost negligible. The output waveform is still very neat and clean, and should be able to switch the IGBTs without a problem.

I didn't really mention this before, but keeping the windings very tight is critical. Higher leakage inductance will likely cause more oscillations with the IGBT gate capacitances, giving you a lot of ringing on the output. This ringing can sometimes be damped using a resistor, but too much will give you a poor output waveform.

For an extremely useful reference while building and testing your GDTs, I suggest you refer to Richie Burnett's gate-drive transformer page, where he shows different waveforms, explains the causes, and offers suggestions for how to fix them.

Finally it's time to move on to the controller. The controller consists of the driver chips, the flip-flop, the feedback circuitry, and the over-current protection. So far mine is just lashed together on veroboard, and I am not terribly pleased with it. There are a lot of long paths which create unnecessary inductance. I will eventually be swapping this out with a PCB, but as my coil sits this is what the controller looks like:

89788

I have made my first attempt at the PCB controller design but am not yet happy with it. I would like to lay out the components a bit better, widen some of the power traces, and optimize it to have the lowest possible inductance, but here is the current design (I apologize in advance for the sloppy labeling):

89789

That's about it for this entry. Next entry we'll look at the interrupter, which sends signals to the controller telling it how and when to switch the bridge. There will probably be a fair amount of coding that would go into that--code which I have yet to write--so the next entry is probably a ways off. I figured I should go ahead and post this one anyway, though, since it's been quite a while since my last one.

If you have any comments or questions, please feel free to post them as comments to this blog, start a thread on the forum, or send me a PM. I welcome any and all feedback!

killivolt, December 18, 2014
Nice; I'm reading early in the Morning at the moment. I didn't know you could wrap RJ-45 or network cable like that? Good read, Matt. Thank you. kv Merry Christmas:)
DerStrom8, December 19, 2014
Hiya kv, Network cable is a little tough to wrap tightly, but it works beautifully in DRSSTC GDTs. The number of conductors is perfect, and the fact that they are already twisted in pairs greatly reduces the leakage inductance. I just realized I forgot to post the waveforms, so I will be updating this blog entry shortly :)
DerStrom8, December 19, 2014
Entry has been updated.
sabo34, February 02, 2016
I'm having problems with the gate drive signal when I use the interruptor. A plane 80khz square wave goes through the gate drive transformer fine, but if I use the interruptor to send ~100 microsecond bursts, the signal rings really badly. I've tried resistance in series and even a couple snubber circuits, but I can't get a clean sharp turn off. Any thoughts?
DerStrom8, February 02, 2016
It sounds to me like your leakage inductance is too high. How did you wind your GDTs? If the wire is loose around the toroid, wind it tighter. Make sure you keep your connections to the GDT short. Make sure the wires are well-insulated from one another. Not knowing how you wound your GDTs makes answering your question a bit difficult. If the above suggestions don't work, you may have to redesign and re-wind your transformers. It may be worth starting a thread on the forum, and posting screen captures of your waveform, as well as photos of your GDTs.