- Blog entry posted in 'Building a Dual-Resonant Solid State Tesla Coil', October 18, 2014.
- The bottom layer of the PCB
- A 3-D rendering (done in Eagle3D) of the bottom of the board (The IGBT models are backwards though--the backs of the transistors should be facing down
- The top layer of the PCB
- A 3-D rendering (done in Eagle3D) of the top of the board.
- Less current-handling capabilities
- More difficult to mount high-power transistors
- More lead inductance, which could cause transistors to switch several microseconds too late and cause critical issues
- Practically impossible to match impedances between corresponding halves of the full-bridge, again causing timing problems
- Uglier--less professional. This isn't important for Tesla coils, but I prefer my boards to be neat. You'll see why when I show you my controller board in one of the next entries
Hello everyone, welcome back to my blog!
In my last entry I described the construction of the primary coil and strike ring which completed the primary resonant circuit. In this entry we'll continue to work out way backwards and look at the circuitry that drives the primary tank circuit.
In order to generate the alternating current required for the Primary coil and MMC to oscillate, we'll need to use a "bridge". You may recall from Part 5 in this series that I will be using an H-bridge configuration for my Tesla coil. This is done using four transistors that, when switched in a certain way, alternate the direction of current flow through the connected device. It is the same circuit as used in many motor drivers to change the motor direction. In our case, it alternates the direction of current flowing through the primary tank circuit.
The H-bridge was one of the first parts I thought about when starting the design for this Tesla coil. It is arguably the most important part in the design. While all parts of the Tesla coil are critical for its operation, improper construction of the bridge could cause issues ranging from poor output to severe damage of the device and surrounding property. When I first started thinking about my Tesla coil, I was going to build a very small one that used MOSFETs in the bridge. However, I quickly changed my mind and decided to use IGBTs instead. When switched on, MOSFETs act as a resistor, and thus the power dissipation (wasted power) increases exponentially with current based on Ohm's law: P = (I^2) * R. IGBTs, on the other hand, act as a diode when switched on, meaning they only have a single voltage drop regardless of the current. Therefore, the power dissipation only increases linearly based on Ohm's law: P = I * V. When dealing with the high currents seen by Tesla coils (often several hundred amps), you'll find that IGBTs are far more efficient than MOSFETs, and are a far better choice. They stay cooler and are less likely to fail after a short period of time.
After deciding to use IGBTs, I picked up a bunch of cheap ones from ebay -- 20N60A4Ds. These IGBTs are rated for 20 amps and 600 volts. The 20 amps is the continuous rating, so the pulse rating is significantly higher (the datasheet says 280A, depending on the pulse width and frequency). However, when I decided to build a larger coil I determined that 280A was not going to be anywhere near enough. I did a lot of my planning for around 500A or greater (pulsed at <10% duty). After some more research I eventually picked up some IGBT modules that are a favorite among the Tesla Coiling community -- CM300s. These modules are often already constructed in a half-bridge configuration (one half-bridge per module). They are rated for 300A and 600-1200 volts (depending on the model). Unfortunately, they're a fair bit slower than I had hoped for--They're really only supposed to be used for <100kHz operation. If you recall, the resonant frequency of my coil is going to be around 120kHz, so it would be pushing the limits of the modules. Using phase lead I could probably make them work, but it would be risky. For that reason I finally settled on using 60N60 IGBTs in a TO247 package. While these will get significantly hotter than the CM300s due to their small thermal dissipation area, they will be able to run much faster, which is very important in Tesla coil applications.
Another thing I feel I should point out is that each of the IGBTs really MUST have an anti-parallel diode in order to prevent back-emf (caused when the IGBTs are switched off) from damaging the transistors. If you build a SSTC, make sure the IGBTs you use have anti-parallel diodes, whether built-in or discrete.
While 60N60s are rated for a higher current than the 20N60s, I was still a bit concerned that they wouldn't be able to handle the extremely high currents involved in Tesla coils. I decided that I'd better double-up the bridges. Using two paralleled bridges instead of one will allow them to share the current, thus allowing them as a whole to handle about twice the current that only one of them would. I realize I've written quite a bit in this post without offering any illustrations, so the following are some snapshots of the PCBs I designed for the bridges.
88825 88826 88827 88828
Above we have (left-to-right):
In these designs I have kept the connections as short as possible and the polygon fills as wide as possible, to allow for better current handling and grounding. The schematic I used is here:
88829
Please excuse the sloppiness of the schematic (the labels and all). This schematic was never intended to be released publicly--I only created it so that I could design the PCB.
I chose to make PCBs so that I could have wide traces on the high-current paths, as well as have complete control over the trace impedances. I wouldn't trust hand-wired boards for several reasons:
Once the designs were complete, I sent them to my favorite manufacturer, who made 10 of them for me. The final version turned out quite well, and I am pleased with it:
88830 88831 88832 88833
In order to mount the boards to the large heat sink I needed two main things: Extremely thin electrical insulation and thermal compound. I was able to buy a 1 mil thick Kapton sheet to use for the electrical insulation. For thermal compound I went to RadioShack and bought some CPU thermal paste, used in computers to mount the heatsink onto the CPU. The combination of the two will prevent the back pads of the IGBTs from shorting to the heat sink, but will still allow for good thermal contact, ensuring that as much heat is transferred to the heat sink as possible. Keeping the bridge cool is sometimes one of the most difficult things to do in DRSSTC applications, so this part is critical.
The next step was to add a large filter capacitor, since we're running the bridge off of rectified mains. I bought a 3300uF RIFA capacitor from ebay for under $10 (very fortunate) which should do the trick. This capacitor will need to be mounted as close to the boards as possible, to make sure that the DC getting to the bridge is as smooth as possible. The rectifier I am using is a large 3-phase rectifier module, though I'm only using two of the connections. This will give me full-wave rectification on the output.
At this point I think I should point out that I am only using brass bolts and nuts for the connections to the PCB. Steel tends to heat up and is extremely inefficient at high frequencies, so it is very important that you avoid using steel whenever possible. Brass, aluminum, or copper are ideal for buses and connections between parts of your Tesla coil. Most of my buses are made from 1" x 1/8" aluminum.
The heat sink will be lifted off the shelf using standoffs, and an 80mm PC cooling fan (or two) will be mounted underneath. Good airflow through the heat sink fins is very important in order to keep it as cool as possible.
That's about it for this entry. I was originally planning to write a very long, detailed entry on the bridge, but I decided to cut it down to the bare bones. If you have any comments or questions, feel free to post them as comments to the blog, or start a thread on the forum.
My next entry will take a look at the GDTs ("Gate Drive Transformers") and the controller.
I know I left a lot out, so I'd be happy to answer any specific questions you might have.Comments
