Polyphonic Tesla Coil MIDI Interrupter (and SSTC side project)

uzzors2k, Thu Oct 22 2009, 08:48PM

Polyphonic Tesla Coil MIDI Interrupter

This is a little something I've been working on for the last month. It's a modular MIDI player, which can be used as a Tesla coil interrupter or synthesizer. For those who aren't familiar with MIDI, it's a music standard used largely by keyboards and drum pads, in which simple commands are sent serially, such as note on and off information, but not actual musical. That job is left to the synthesizer, which is simply told when a note is on or off, and must create the corresponding sounds (frequency, waveform, etc) itself in order to make audible music. So no matter what the original song is supposed to sound like, the MIDI signals simply tell us when a key is on or off (and a lot of other information, but nothing that's interesting for this project). This is perfect for Tesla coil modulation, as every SSTC or DRSSTC can be interrupted, and by interrupting the coil at the same frequency as the note being played, we can create music!



This project has evolved from a monophonic MIDI player using a single ATmega32, to a multi-channel, multi-note (polyphonic) MIDI player using several ATtiny2313s working together. Each ATtiny2313 can play two notes at once, using Timer1 and the two Compare Match modules. The AVRs are organized so they all receive the same input, but only start playing notes once an enable input (PD5) is set high. Once an AVR has used both it's compare match modules to play notes, ie is full, an output pin (PB0) goes high. This way they can cooperate, and in theory play and infinite number of notes at once. The MIDI channel that the AVRs listen on is set using the four most significant bits on PORTB. I've made a small example circuit where three ATtiny2313s are chained together, and can either play 6 notes on one channel, 4 notes on channel 1 and two on channel 3, or two notes on channels 1,2 and 3. Depending on the switch positions. The firmware was made using WinAVR. Firmware, schematic and a command-line program for determining the compare match values here can be found further down.

For anyone else who wishes to tread down the path of MIDI, I recommend you read this article by Paul Maddox. It was a reference I used throughout the entire process, and saved me a lot of headaches. (not to say I still didn't encounter enough of them.) Also, if you wish to mix several notes together you need to keep the duty cycle low (~5%). This is obvious once it's been pointed out (thanks Steve W), because you can only send one bit of data down the line, and mixing two square waves results in areas at V/2 - which would require 2 bits of data to describe. Keeping the duty cycle low reduces the chance of an overlap. And the more square waves you mix the more bits are required. It's easy to see graphically if my explanation is poor.



The firmware itself is interrupt-driven. MIDI bytes are processed as soon as they are received, and once a full MIDI packet has been received a note is either turned on or off. When a note is turned on the correct off-time and on-time values are looked up and stored for quick access, and interrupts are enabled for one of the compare modules. If both compare modules are now busy an output pin is set high to alert any other AVRs in the chain that it's full. Once a compare module generates an interrupt, an output pin is toggled and the time until the next interrupt is determined depending on whether it's in a high or low state (to get an asymmetric waveform). When a note off event occurs the AVR first checks that it's actually playing the note before stopping one of the compare modules. The main program loop is just used to indicate when notes are being played and check what channel to listen on.

So far I've done little more than test this on a quarter watt PC speaker, but it's plays the Maple Leaf Rag perfectly, so I can't wait to try it on a SSTC or DRSSTC. The next steps in this project are designing a PCB, boxing up the interrupter, and finally making a dedicated SSTC. Don't be afraid to test this interrupter before I do, it'll still be some weeks until I can rig up a TC to test it on.

Wed Nov 18 2009

Progress is still snailing along. I made a PCB for the MIDI interrupter but ordered the wrong box size and only received one of the two 12MHz cyrstals I ordered. Still, I had enough to test the full circuit over a toslink connection, which works great with a small speaker. With a temporary 8MHz crystal on the last ATtiny2313 I couldn't get 6-note polyphony though, only 4-note. I hope this is due to the crystal speed, and not something in the code... I'll publish the PCB design and upgraded firmware once I've pinpointed the problem.



I later dusted off an old secondary and half-bridge from another SSTC project and fired it up using several different drivers. In the end Steve Conner's MKI PLL was the only one that worked reliably, thanks again man. With a 2250µF filter capacitor power draw is almost exactly 1kW and the discharge is fairly silent, no mains hum at least. Heatsinks stay cool and it appears to be able to run continuously in CW mode.




I figured a musical test was in order, so I hooked everything up for a trial run. Results left a lot to be desired... Low frequency notes gave decent sparks, but higher notes gave smaller streamers until I could only hear them without my earmuffs on. Eventually (within 10 seconds) the output died altogether, even when playing low notes. I turned off the coil at this point and removed the keyboard and set the coil for CW. When I turned it on again, after charging up the filter capacitor, I was blinded by an exploding bridge of IRFP450s. The toslink module was unshielded, and I hadn't disabled the IRFP450's internal diodes. Now before I go buy some expensive diodes, was this the likely cause of failure? This coil's resonant frequency is 625kHz.

Sun Nov 29 2009

I've tested the PCB and final source code now which all work, so it's a go for anyone looking to try this! I put the assembly in an ugly project box, and made a little teaser video. An actual Tesla coil test will have to wait until I get the high speed diodes I ordered from China.

Teaser video. ]mvi_1188.avi[/file]
Project files. ]tc_midi_interrupter.zip[/file]

Sun Jan 10 2010

Holidays are over and it's back to work! I reassembled the Tesla coil and popped some brand new MUR1560 diodes into the full-bridge this time. So far they appear to have solved the problems I had earlier, and the coil can be interrupted without exploding. I have a habit of turning off the logic supply before the main power, which caused some headaches today. So despite constantly forgetting to power the coil down in the right order, thus loosing at least four full-bridges, I managed to get it functioning again.

At first I tried the MIDI interrupter carefully, in non-inverted mode (like a DRSSTC, so ON-times are short) this resulted in absolutely no output from the coil. I suspect the PLL isn't able to lock quickly enough, so that's one problem which needs fixing. So when that failed I switched the interrupter into inverted mode, so it runs CW until interrupted. And it worked!!! I've made a few videos flaunting my coil already.

Super Mario Bros. 2
Maple Leaf Rag

The video doesn't do the sound quality justice with all the background noise, in person the sound is quite clear, easily audible without ear muffs on.

Well, there's the evidence needed for this project! I been in contact with another coiler who has been using this interrupter with a DRSSTC, so it works with them too!