Futurist's DRSSTC1

[futurist]

 Hi everyone,

I've been fascinated with Tesla coils since I saw one at the tehnical museum as a kid. Building one myself was the ultimate goal, but since electronics became just a hobby I never had enough courage to try to build one, until about half a year ago. Now after I solved most of the problems and got the coil running I advise anyone interested in the area to try to build one themselves. The project turned out to be great fun and valuable learning resource! I also have to mention that I wouldn't be able to build it without great community and writeups by Steve Ward, Mads Barnkob, Gao Guangyan and tips from others (Thanks to Hydron, Kizmo, zzz_julian_zzz and Uspring), so I hope this writeup will also be valuable resource for anyone wishing to build similar medium-sized DRSSTC

The coil itself is a design-hybrid of Steve Ward's DRSSTC1, Kaizer's DRSSTC I and loneoceans DRSSTC 2. The core of the inverter is CM200DY-24NF halfbridge IGBT brick driven by UD2.7C driver at around 65 kHz, with 700 A primary current. There were some design changes and problems I encountered, but to avoid confusion I'll write just the current specs. I recently changed IGBT brick to SKM400GB123DH6 and upped the OCD to 800A, but because I still didn't got around to test the coil properly I'll post those results later





Overview of the coil specs
- CM200DY-24NF 1200V 200A halfbridge IGBT inverter, set to a 700 A over-current protection
- 1200V 1uF EACO snubber capacitor
- 45 mm^2 copper bus bar (15x3mm) and 16 mm^2 welding cable for bridge-primary connections
- 2x EPCOS 4700 uF 400 V caps with KBPC5010 1000 V 50 A bridge rectifier in voltage doubler configuration, for 240 V AC in
- 340nF 16kV MMC 100 Arms
- Universal Driver rev. 2.7C
- Flat spiral primary 215 mm inner diameter, 8 mm copper tubing with 10 mm spacing, 9 turns tapped at ~6.8 turn
- Secondary 160 mm diameter on PVC drain pipe, ~2200 turns (~605 mm long), 0.25 mm double enamelled copper wire - 2x polyester resin coated
- Toroid 130 x 620 mm from flexible aluminum tube with 2x 400 mm diameter stainless steel serving plates
- Controlled over 10 m ST-ST 125/62.5 um optic cable
- Maximum 1.5 m sparks!

For the IGBT I decided to try the CM200DY-24NF which was the cheapest IGBT brick I could find, even cheaper than used ones (bought new for 62$ with shipping from aliexpress). At first I thought it may be fake, but it turned out to work greatly, even at 700Apk!

I decided to install the toroid a bit higher, to avoid constant ground strikes, it proved unnecessary because the streamers rarely hit ground with current settings. I hoped to get the same performance like fullbridge coil at 240VAC without voltage doubler, like Hydron's DRSSTC, but sparks are not that spectacular. Longer ones are certainly possible with this design - I think the coil lacks a little more coupling but I'm unable to change it until I rewind secondary

JavaTC output









Primary supports are constructed from laser cut 3x3 mm plywood glued together with PVA






For the middle part I left holes for screws, but they heated a lot so I removed them completely  and secured the supports using epoxy. Winding secondary was pretty straightforward, end plugs for drain pipe were created by pouring polyester resin. Secondary was then coated 2x with polyester resin. Didn't turned out great, but it's ok for first try




Winding primary was a little harder, it took some practice to learn how to bend copper tubing



Completed primary with strike rail, later I cut bent section of the strike rail

Whole coil








1:1024 current transformers are wound on medium-permeability 3E25 ferrite toroids (http://uk.farnell.com/ferroxcube/tn25-15-10-3e25/ferrite-core-toroid-3e25/dp/3056995)
Toroids are glued together with epoxy




Halfbridge before assembling

Also:
Gate resistors are 4.7 ohm 1W with 1N5819 Schottkys in parallel
Gates and emitters are shorted with 1.5KE33CA TVS
10kohm 20W resistor is connected across the output of the inverter




GDT is checked on oscilloscope using 100 kHz squarewave, toroid core is N87 material




Laser cut 3mm plywood capacitor supports




MMC capacitors with vishay HV 10Mohm balancing resistors
Capacitors are 0.68uF rated for 2000VDC, Irms = 25 A, per datasheet 8 caps in series and 4 strings in parallel gives 0.34 uF at 16 kV DC (5.6 kV AC) with Irms = 100 A
Originally I wanted to use 12x 942C20P15K-F capacitors for 0.45 uF at 4000 V DC. Such low impedance primary circuit isn't suitable for my design, Steve Ward used that MMC with his high-coupling helical primary coil, in my primary circuit cap ratings would be heavily exceed and they would fail soon




Cad drawings for MMC enclosure




Cutting MMC enclosure, thanks to a friend who kindly provided me with laser cutter




MMC enclosure completed






Whole bridge assembled + MMC
0.1uF 2000VDC film capacitor is connected from negative rail to ground, and Graetz bridge is connected in voltage doubler configuration




Closeup of the electronics
UD and Bus capacitor precharger/discharger and fan speed controller




Bus capacitor precharger/discharger and fan speed controller is made using ATMEGA328P and two 30A relays. Precharge resistor is 150ohm 10W, and discharge resistor is 1kohm 10W. Originally I had two LM35DZ temperature sensors for MMC and halfbridge to control fan speed, which were later removed because I had problems with EMI jamming MCU, and had to add protection for reset pin, multiple ceramic capacitors and common mode choke to deal with it. For now it seems to work fine, but in future I plan to make more reliable and universal fan speed controller and bus precharger using STM32, with protected inputs/outputs.




Scope shot of first power-on with adjusted phase lead


Playing The final countdown

The interrupter was assembled by my friend - Arduino Nano 328A with modified oneTesla interrupter code for playing MIDI directly from USB + 10 m ST-ST 125/62.5 um optic cable.
The midi sound is good and only problem are occasional "pop" sounds, they are caused by the way second note is handled; if second note plays when first is playing, the second note is ORed together but limited to some max. pulse width
Also expect stuck notes at >160us on-time with some midi files, I don't know the reason

It took me a whole afternoon to get the MIDI interrupter working, so anyone using old code from oneTesla;
to run directly from USB I did the following (gnu/linux)

in oneTesla source change MIDI boud rate to 38400
install http://www.varal.org/ttymidi/

ttymidi -s /dev/ttyUSBxxx -b 38400 &

Using aconnect connect Midi Through port to ttymidi port (to get output list: aconnect -o)
aconnect xx:xx xx:xx

To play (Midi Through port)
aplaymidi midi_file.mid -p xx:xx

Or using something like Midieditor

On windows use this as Midi-serial converter
http://projectgus.github.io/hairless-midiserial/

and this for the loopback interface
http://www.tobias-erichsen.de/software/loopmidi.html

Any feedback or questions are welcome!
Happy coiling!

Edit: Wrote description for each photo

[Mads Barnkob]

You have implemented some nice solutions and also got some good tools available, I am envy at having a laser cutter at disposal

The laminated primary supports was a good solution, I like that, I mounted my own with nylon bolts through from the underside of the primary coil base

Do you have any additional heat sink than the slab of copper? I like it though, its nontraditional and a DRSSTC really only needs fast thermal transfer and not a huge heat sink for a average large power dissipation. Good choice.

I will add your coil to the list of proven IGBT switching speeds at my website

Did you experiment with higher coupling between primary/secondary, something like 0.15 maybe. The recommended coupling from JavaTC is based on SGTC. You could saw a hole where the secondary sits and lower it that way?

Are you sure that it is tuned to best performance? When running off a regular interrupter and not MIDI, how long on-time do you use for the longest sparks? JavaTC has the maximum time for energy transfer calculated to 50 uS, I am not sure if its at 200% or 300% of this figure that should indicate that its not tuned properly.

Overall, great result that also looks good

[futurist]

Thanks for the feedback 

Laser cutter is really neat thing, this was my first time using it and I was amazed how it simplifies things.. and it's partially responsible for the coupling
I called my friend and asked when he had some free time on cutter and he said - tomorrow, so I was awake at 4a.m. drawing new mmc enclosure
Next day I after I cut everything I glued primary supports and messed up with inner diameter (for second time!).. originally the idea was to have 0.16 coupling when secondary is placed directly on stand
Then I could lower the coupling by adding plywood discs under the secondary, however because of too large inner primary diameter coupling turned to be too low anyway

I already planned to rewind the secondary starting from lower point, this is only temporary secondary because the tube warped a bit when I made end caps and I had to repair some overlapping turns with cyanoacrylate
Funny though, I probably won't be able to fry it anyway because of the low coupling



Secondary construction


About the cooling
I noticed that only few inner turns of the primary get hot, other components are barely warm, MMC is stone-cold
Now I see it isn't clearly visible on pictures, but under 1.5kg copper bar there is old pentium cooler I salvaged from university





Only potential problem could be blowing hot air towards bus capacitors because I couldn't mount cooler sideways, it currently doesn't happen

I know isn't tuned well, 150us is needed for max. spark length
I'll do a more thorough tuning now when I don't have to fear about blowing up igbt brick, it can tolerate 600A for 1ms per datasheet

Edit: added secondary drawing

[Mads Barnkob]

Great idea with a CPU cooler on that huge copper piece, that will definitely works better than most large heat sinks, for a DRSSTC

When you redo your secondary coil, you should get a brass bolt to conduct the current to the topload, instead of a very lossy steel bolt. Another thing, is the steel bolt going all the way through the end cap? If it is, you risk internal flash over to the inside of the secondary coil and that could potentially punch through the plastic and burn up your coil from the inside. Scroll down to the bottom of the secondary coil design article to see pictures of this: http://kaizerpowerelectronics.dk/tesla-coils/drsstc-design-guide/secondary-coil/

[futurist]

I epoxyed brass nut at the base of the secondary for grounding, it's just glued on the tube and doesn't go inside
Endcaps are made like this (from inside out) plywood disc > polyester resin > plywood disc with stainless steel tapped rod > polyester resin
Only upper tapped rod is connected to the windings, I'm not worrying about electrical connection to the toroid because copper wire is 0.25 mm and tapped rod is 6mm diameter

Recently I replaced CM200 brick with SKM400GB123DH6 and upped OCD to ~820 Apk, also I wound new secondary starting from lower point (coupling is ~0.16 now)




Footprint of the Semikron brick is few mm off so I had to adjust copper rails a bit


Yesterday I finally tested the coil, it performs much better and I'm getting constant ground strikes (~150 cm) at ~600 Apk primary current




I measured primary current waveform while tuning the coil using homemade CT and 10 m coaxial cable (thanks to Hydron for tips)


Primary current is still notching at higher than 120 us on-time so I believe spark length can be improved further, especially with startup oscillator by making the coil oscillate at upper pole
I recently finished assembling UD+ driver and I'm planning to test it on the coil

Current transformer used for primary current monitoring

[Mads Barnkob]

You did a good job on the end caps on the secondary. Lots and lots of resin!

The IGBT upgrade really did a job on it, you might even need to raise the whole coil up on a box or build a taller base for it, if you get constant ground strikes then it is too close to ground.

I have not tried any drivers above UD2.1, so where you might be able to get more performance with more careful tuning, the UD+ might just do it for you instead? Isn't that the whole point of the startup oscillation? Or does it need just as careful fine tuning the primary tap to also hit the upper pole?

The DIY current monitor is something everyone should make and own, if not investing in a old Pearson or similar commercial current probe. These home made current transformers are actually on par with a similar Pearson up to around 1 MHz.

Did you have any flash over issues with the higher coupling? If you had, it could also just be out of tune.

[futurist]

I tried to lower primary fres as much as possible to see the limit, and at 55 kHz and low BPS I got few running streamers

Startup oscillator won't make it easier to tune the coil, however it could theoretically help in getting bigger streamers from lower primary current

[Uspring]

The primary current waveform shows an interesting rise at the end of the burst. Current seems to increase almost as fast as at the beginning of the burst, where there is no arc loading. The only reason I can think of, is, that the secondary isn't loading the primary anymore at the end. Closer inspection of the waveform seems to indicate, that the frequency jumps up at there. This will bring the coil out of tune and thus decrease secondary loading. Out of tune always means, that the running frequency is far away from secondary resonance. Probably the coil is running at the upper pole there.

How can this come about? Coils with only primary feedback, i.e. no frequency steering with e.g. a startup oscillator, tend to lock on to the pole nearest the primary resonance frequency. If primary res frequency is lower than secondary res, that is the lower pole. For a higher primary res than secondary res it is the upper pole. My belief is, that during the burst, the secondary res drops due to arc loading below primary res, which in turn can make the coil jump to upper pole operation.

Lower pole operation with the secondary res above the lower pole can lead to notching. Since the arc load will decrease secondary res, the secondary res will move closer to the lower pole, pulling the secondary more into tune. The increased loading of the primary will then limit primary current.

In upper pole operation, the decrease of secondary res will move the coil away from the pole (=running frequency) and thus tend to detune the coil, causing the primary current limiting to be less effective.

[futurist]

That's interesting observation, makes sense when you think about it

One thing still bothers me

In upper pole operation, the decrease of secondary res will move the coil away from the pole (=running frequency) and thus tend to detune the coil, causing the primary current limiting to be less effective.

Why does then Steve use upper pole operation for QCW?

Today I did short tests with UD+, it needs more tuning but at same tap point works better than UD2 for some reason
Because current rise is very slow I get 2m streamers from 250-300us on-time


For that reason midi mode is not that spectacular (max. ontime 200us), waveform:


I have few options, increase or decrease primary fres, and/or startup oscillator frequency
Any suggestions?

[Uspring]

Why does then Steve use upper pole operation for QCW?

Here's an ascii art diagram (somehow adding graphics didn't work for me). The frequencies involved are the lower pole fl, the primary resonance fp, secondary res fs and the upper pole fu.

-----------fl----fp---------------fs----fu-----------> f

You can run the coil with zero current switching at the lower and upper pole. How well power is transferred from the primary to the secondary depends on the difference between running frequency and the secondary resonance fs. The lower, the better. The upper pole is much closer to the secondary res than the lower pole. Thus the upper pole operation is tuned better. Now add arc loading, which reduces secondary res:

---------fl------fp--fs------fu-----------------------> f

The difference between secondary res and upper pole increases a bit and the difference between secondary res and lower pole decreases a lot. Generally upper pole operation will yield a better tuning but requires an extra effort in electronics to make the coil run there.

The dynamics is quite different for both modes. In lower pole operation breakout will start late during the burst due to the bad tuning. Then the arc will bring the coil better into tune, which can lead to a sudden dump of the energy of the primary into the secondary causing rapid spark growth. After this, primary current will be limited to a moderate value due to the arc loading.

In upper pole operation, breakout will start earlier and there will be no sudden dump of energy into the secondary, since tuning doesn't get better due to arc loading and the primary hasn't stored as much energy prior to breakout. I can think of 2 reasons, why Steve Ward uses the upper pole: Firstly, tuning is better and secondly, it doesn't change much during the burst. This allows a better control of power delivered to the secondary. This is an advantage for QCW straight arc production, which requires a slow and smooth rampup of power.

If you're looking for the longest possible arcs, lower pole operation is probably preferable, since the energy dump adds (for a short time) to the power supplied by the inverter. This mode requires relatively long bursts, though. For limited burst length, e.g. high notes in a music playing coil, the upper pole operation is better, since breakout and spark growth begins earlier in the burst.

[Mads Barnkob]

Uspring, thank you very much for the detailed, illustrated and easy to "understand" explanation on tuning and coil modes, it is not that easy to get your head around in the first place.

So for high impedance DRSSTCs it actually has to be detuned even more than a "regular" low impedance DRSSTC, maybe some day I will have a driver that can force my large coil to operate at upper pole and see what difference that makes when playing MIDI

[Uspring]

In principle, you could tune the primary almost arbitrarily low in upper pole operation, since the upper pole will stay close to the secondary resonance. So tuning would be good almost independent of the choice of primary res. There is a problem, though:
For strongly damped secondaries, the upper pole will vanish, leaving only one frequency near the primary res at which you can obtain zero current switching. The coil will then need to run there to avoid hard switching. If the secondary res is at this point close to the primary res, this is not a problem, since it stays well tuned, but it certainly puts a limit to the extent to which you can tune the primary low.

@futurist:
2m streamers from a coil with your electronics is a respectable result, but certainly you can try to improve it. A bit surprising is the the slow initial rampup of primary current. If your bridge switches between e.g. +-330V, then the primary cap should gain 660V for each switch or 1320V for a complete cycle. That translates to a current gain of 1320V / 2*Pi*f*L. Putting in values of 69kHz and 15.6uH, that amounts to almost 200A per cycle. The scope shot seems to indicate a value somewhat above 100A. This calculation presupposes no secondary loading, though. Initially it shouldn't be very large.

The wavy envelope suggest to me a not too far match between primary and secondary resonance. Generally there is some back and forth thrashing of energy between the primary and secondary. The amount of thrashing depends on the proximity of primary and secondary res.

[futurist]

Thanks for the explanations

Yesterday I played with lowerupper pole operation, things that I changed from previous run:
I lowered startup oscillator to 63 kHz (3 cycles) and further detuned coil to turn ~7.17 which increased spark output a lot for 200us on-time (previous tap point was at turn ~6.85)
Detuning caused short sparks for midi mode at higher notes due to low on times

Further detuning stops spark growth, I'll try to lower startup oscillator further and see if it makes any difference
At 300us streamers are not getting any longer and current waveform looks like this


Fres measurement at turn ~6.85 few days ago
Lower pole was 55.3 kHz and upper pole 65.5 kHz

By my experience so far I have a feeling that most people overexaggerate their streamer length
I never got continuous 2m streamers, just caught a few, and more frequent ones were 1.5-1.7m long

Note that my coil is just over 130cm tall which makes 1.5m streamers look incredible

[Uspring]

You got me confused.

Yesterday I played with lower pole operation, things that I changed from previous run:
I lowered startup oscillator to 63 kHz (3 cycles) and further detuned coil to turn ~7.17 which increased spark output a lot for 200us on-time (previous tap point was at turn ~6.85)

and

Lower pole was 55.3 kHz and upper pole 65.5 kHz

at tap 6.85. At tap 7.17 I'd expect the lower pole to be even lower than 55.3kHz. So it looks like you're running at the upper pole, since that is much nearer to your startup oscillator frequency. As far as I can judge from the scope shot, that seems to be indeed the case.

[futurist]

Sorry my mistake, upper pole operation

[Mads Barnkob]

By my experience so far I have a feeling that most people overexaggerate their streamer length
I never got continuous 2m streamers, just caught a few, and more frequent ones were 1.5-1.7m long

Note that my coil is just over 130cm tall which makes 1.5m streamers look incredible

Maximum spark length is usually those once in a while extra long sparks that can you find by scrolling through the frames of your video recordings, while the continuous streamer length is shorter.

I never recorded sparks longer than 1.5 meter on my DRSSTC1, but I think it can do better from tuning and more power, I was limited by 10A fuse back then, on 250VAC single phase.

Great work so far, looking forward to hear more about how you get the best performance from the UD+