Tesla Transformer with Printed Coils Again, This Time Complex with MIDI
Having visited the electronics exhibition in Germany last year, the author drew attention to two exhibits, the combination of which would have given him an excellent result. This is the Tesla spiral transformer, built by Daniel Eindhoven and presented at the Elektor magazine stand, as well as new gallium nitride keys, positioned by Texas Instruments as intended for high-frequency pulse converters.
And here is the result of work that lasts from November last year: a Tesla transformer with printed coils, giving discharges up to 50 mm long and working from a 3-amp Type-C connector in a laptop (it is clear that through a long cord). From there, he not only feeds, but also receives MIDI data from, for example, LMMS or Albeon.
The device has everything open, both hardware and software, namely:
USB MIDI class driver: CC BY 4.0
STM32 CubeF0: three-point BSD license
Other software, hardware, circuits: GPLv3
Everything is here .
Repeating it is difficult. For example, you have to solder QFN with a very small step. The author plans someday to start selling assembled sets. But if you make three transformers at once, you get an excellent orchestra:
Composition of the device:
Main board with installed SMD components
Upper board with a printing coil
Kapton tape />
Bottom board with a printing coil
Four 8 mm plastic washers
Four 20 mm nylon screws M3
Four connecting nuts 40 mm long, on one side internal thread M3, on the other hand, the same external.
Four nuts. M3.
Three foil capacitors at 47 nF.
XT60 connector with a counterpart. Socket
for microcircuit with spring contacts (it will have to be disassembled, see below)
The resonant frequency of the device is relatively high. It is difficult to build a full bridge on ordinary silicon MOS transistors at such a frequency, mainly due to losses on charging the gate capacitance and switching on transistors of the required size.
It is possible to receive discharges in air at such frequencies using class E. But then the transistors will work in linear mode with large power losses, as here .
Daniel Eindhoven solved this problem with conventional bipolar transistors operating with emitter followers. It turned out fast switching with low delays. But they will require a huge driver, the amplitude of the signal at the output of which is equal to the amplitude of the signal at the output of the terminal stage. The repeater after all amplifies only in power, and not in voltage. Daniel took a driver chip for this on MOS transistors, but it can not have more than 32 V at the output. And bipolar transistors in the terminal stage are not very efficient.
Keys on gallium nitride and silicon carbide require special gate drivers, for example, with a range of output voltages from -6 to +10 V. This protects the gates from bursts when disconnected. The driver should be placed as close to the key as possible.
Texas Instruments released a device with two built-in keys on gallium nitride, included half-bridge, containing all the necessary drivers. For about half a year they have been selling at $ 9 apiece.
From here, the
Lays here. The
whole circuit can be divided into five functional blocks:
, feedback signal receiving
circuit, Idling circuit,
Full bridge with gallium nitride keys. Board
Of all the board versions, the author decided to show three:
- 1.0, November 22, 2018 - no-load, improperly designed power scheme
- 1.3, December 22, 2018 - with no load, the first discharges in the air, gallium nitride keys failed after 1 - 20 minutes
- April 1.5, 12, 2019 (just Cosmonautics Day) - works well and reliably, but there is something to improve. In README.me it is said that in this version, when the discharge interacts with a finger or a screwdriver, the keys still fail, but if this is not done, they work indefinitely. That's only 15 watts - decent power, and it is better not to put a finger in any case because of the danger of thermal burns.
To achieve the result, 6 versions of circuits and boards were required. Basically, it happened because of the specific requirements for circuits on such keys. Some versions would not have to be developed if the author immediately listened to the recommendations in the datasheet. Typically, the designer interprets the maximum permissible parameter values as “recommended”, but this is not the case with gallium nitride and silicon carbide keys. If TI requires you to place the components “as close to each other as possible” on pages 10, 13, 14, 15 and 16, and not just once — only 8 times — that really means “as close as possible”. “Close” means not “at a short distance”, but “close to the face”. The smallest spurious inductance is obtained by placing the components on different sides of the board and connecting as many holes with through metallization as possible.
Not following these recommendations, the author ruined ten LMG5200 modules because he thought he knew better. And now we turn to the control scheme, there it is simpler:
Did you notice anything unusual? As if DN and DP on the USB interface are shorted to each other? It’s just a symbol from USBLC6-2SC6, but in fact USBLC6-4SC6 is used, the author will correct this in the next version of the circuit.
Everything is normal here. The usual STM32F072, it is cheap, contains the ARM M0 core, and works without quartz, clocked from USB. For one dollar, you get a microcontroller with USB and timers, requiring a minimum of external components. The author chose a chip in the QFN package, because it was easier to solder than QFP.
There is also a DFU loader, which means that a programmer is not needed to upload the firmware. The first time you have to solder a jumper on the back of the board, and it will turn out to fill in the firmware, for example, through dfu-util. Then it can be removed, and before each flashing, transfer the device to the appropriate mode with the button.
And another scheme is the boost converter, which allows power supply of the device’s power circuits from Type-C:
... But the author was initially refused by JLCPCB, because they could not guarantee the absence of short circuits. He promised to take fees without a guarantee, and then they took up the order:
Worried there in vain. Of the 70 circuit boards with short circuits, only one turned out to be. Check it out is simple. Ring the secondary windings of several boards and compare the resistance (the author got about 180 ohms). If one board has slightly less, then there are short circuits.
Ready-made assembly of three boards:
When connecting boards to each other with wires, it is important to ensure that the rations are smooth, without sharp protrusions. Do not spare the Kapton tape, because during the debugging process, the following happened too:
From here, the
boards with windings are separated by 8 mm thick plastic washers and nylon screws. And the bottom board is connected to them by conductive connecting nuts with a length of 40 mm.
The idea of installing the electrode in the spring socket from the socket for the microcircuit is taken from here .