Stefan's Tesla-Pages

Vitamini((C) by STK): my Mini Coil System (wound on an 1" Vitamin C - tube)

This is one of my current projects, therefore this page is heavily
under construction !


Table of contents:

Specs of the mini coil system
circuit plan (final configuration)
Photos (26.07.2002 an 15.08.02)
Update May'99 (new gap)
some thoughts on a true micro coil system

The specs of my mini coil system are:

Control board
No extra controlboard. I use a commercial line filter (1A), a fuse and a switch, just 'plug and play'.

Filter board
No HV-filterboard. I think I don't need one at this power level. But I use a safety gap over the plastic caps.

Two 50Hz/5kV/20mA OBITs in series, measured values 18mA@9.54kV. Go to my neon page for more information about xfmrs. The optimum cap size for these xfmrs is C = I/(2.Pi.f.U) = 5.7nF (for 18mA@10kV). I applied 10µF of power factor correction caps (400VAC rated) and got 790mA on the primary side. With 230V line voltage, this works out to max. 182W input power (you can't perfectly compensate non-sinusodial waveforms with plain capacitors). BTW, when I disconnected the tank circuit caps, the input power decreased to 94W (perhaps the BPS decreased also due to the missing 50Hz-resonant charging). The gaps seemed to be happy and fired well though. These OBITs get HOT after 3-5 minutes of runtime!

Series gap
For the beginning, I used the old 7 gaps made of 8 heavy steel rings (modified set collars) on a wooden rod.
I'll make a new small flat RQ-gap (15mm dia copper pipes, 3cm long) in the future. Perhaps with a higher number of individual gaps, some coilers say that this seems to be important for low power coils. Other experienced coilers say that in this power range 2 gaps are enough - so lets try and see. I tried 6 gaps versus 9 gaps (with parts of the commercial 'valve gap' I describe on my sparkgap page, total gap width 6.9mm each) and Vitamini seems to like the 6-gap-version more (it runs smoother with it). Additionally, if the gap between two pipes is to small, corrosion will make it even more narrow over the time and therefore decrease the breakdown voltage by a significant amount.

First test runs with 3 small mineral water bottles (salt water cap) with a capacitance of 3.35nF totally (click here to go to my capacitor page), this was half of the optimium capacitance (cap matched to the xfmr).
Current configuration (I think it will be the final one) is a new approach, namely a MMC (MultiMiniCap). I tried to connect several commercial MKP-caps (ERO, selfhealing aluminized PP, series 1836 data sheet available at which leads you to 62nF/1kVDC in series/parallel. I used 2 strings of 20 caps each for my 10kVeff-supply. The caps were very cheap (only 0.12 EUR). The dielectric power loss in this cap (calculated for 3x32 caps, 1MHz, 5.8nF, 10kVeff, 100BPS) is approximately (no measurements up to now!) only 0.23W (loss factor of PP is 2e-4) assuming 20 cycles per beat and 100BPS. That is only about 0.4% of the power going into the cap (Ec x BPS = BPS.1/2.C.U^2 = 61W for 100BPS, 6.2nF, 14kVpeak)!
Wiring scheme:
         C1       C2       C3        C4
      O----||---o---||---o---||---o---||----o row 1
      o----||---o---||---o---||---o---||----o row 2
      o----||---o---||---o---||---o---||----o row 3
      o----||---o---||---o---||---o---||----o row 4
      O----||---o---||---o---||---o---||----o row 5

The gap (2mm wide) between the rows has the advantage that you can inspect the sidewalls of the caps in case some of them blew. Size is approx. 17x17cm2. Height of the caps is approx. 2.7cm. I use an 1.8MegOhm bleeding resistor across each cap. Therefore the bleeding current is 0.56mA or with other words about 0.9% losses (using 100BPS x Ec). Total loss is about 1.3% instead of 20% using the saltwatercaps. The old SW test cap (very thin glas) was 20x20x14 cm3. The losses in the glass were approx. 12W. Ceramik doorknob caps will be smaller (look at Terry Fritz' webpage) but much more expensive.

Technical data of the ERO-caps (KP1836, 62nF, 1000Vdc):
description high dV/dt rates, high current handling capabilities
application High voltage, very high current and high pulse operations,
deflection circuits in TV sets (fly-back tuning). Protection
circuits in SMPSs. Snubber and SCR commutating circuits.
rated voltage 1000VDC
permissible AC voltages (rms) up to 60Hz 330VAC
test voltage 2000VDC for 2s
self inductance ~6nH
dielectric polypropylene film
electrodes aluminium foil
construction extended aluminium foil, internal series connection,
double-sided metallized polyester carrier film
maximum pulse rise time 1500V/µs
dissipation factor (tand) at 100kHz 1e-3
permissible AC voltage versus frequency (cw)
- 1.5e4 Hz: 330 VACrms
1e5 Hz: 100 VACrms
2e5 Hz: 60 VACrms
3e5 Hz: 50 VACrms
5e5 Hz: 30 VACrms
1e6 Hz: 20 VACrms

Now lets calculate how hard I drive them:
I use 2 chains of 20 caps in series. My frequency is 800kHz.I use 10kVeff, tank circuit data is ~6.2nF/~9uH which works out to 14kV/370A peak values (Z=38Ohms). Calculation of the actual rise time: 1/4 period of 800kHz is 0.3125µs. In this time, the voltage rises from 0V to Vpeak, for the rise time the steep beginning is the point to view at. Here the rise time is Pi/2 higher than the average, resulting in 14kV*Pi/2 per 0.3125µs over 20 caps in series. Result is 3.5kV/µs per cap instead of the rated 1.5kV/µs and therefore factor 2.4 above the specs.
Now lets look at the AC-voltage. Duty cycle is low, so the derating due to thermal problems is negligible. 20 Caps in series with permissible 330Vrms (due to corona build up) are 6.6kVrms total. I apply 10kVrms, this is factor 1.5 above the specs.
Now on to the DC-voltage. 20 Caps in series with permissible 1000Vdc are 20kVdc total. I apply 10kVrms, this is 14.1kVpeak and therefore only factor 0.7 of the rated voltage (that means I'm in the specs here).
Total inductance in my series/parallel connected MMC is 1/2*20*6nH which is 60nH and therefore 1/150 of my primaries inductance and therefore also negligible.
All in all, I drive them too hard, yes. But considering the fact that a typical hobby coilers system will run not more than 1 hour every weekend for max. 10 years, the caps should survive only a total runtime of 500 hours and not the 10XX hours, the manufacturer gives the specs for.
One thing to consider though: the MMC consists of many (n*m) caps, therefore the lifetime of the MMC (before the FIRST cap will fail) is approx. 1/(m*n) the liftetime of one cap alone.
Read more about MMCs on the GTL-Website GTL (sorry, this is only in German!).

For increased cap lifetime I added about 38Ohms in series to the safety gap for reducing the stress on the caps (rise time) down to the value of standard TC operation (with other words, I limit the current to a value below 370Apeak this way).

12 turns saucer shaped (height 1", width 2"), made from different types of wire approx. 1.5mm dia (what I had at hand...) Total inductance should be up to 17µH. Tap point depends on the size of the top load of the secondary. Well - that was the theory. I now add as much topload as I have available in this size, that is 2 to 3 small toroids on each secondary. Using the test cap (only half of the optimum primary capacitance), the tap point is on the outermost turn, with the MKP-caps I tap at turn 9 now.

Secondary (actual parameters of 'Vitamini', the first secondary coil of the twin system)
1.14" PP form ('Vitamin C' - tube), winding length 5.14" (1:4.66, 0.6cm bottom margin, 13.05cm winding length, 0.2cm top margin), around 850 turns (at 98% filling factor), magnet wire dia 0.15mm (AWG34), inductance 4.18mH, parasitic capacitance around 2.23pF, natural frequency 1.65MHz. I used a 'Vitamin C'-tube as it is sealed on one end by design and has a very neat cap on it. This cap is now fixed on the chassis and holds another tube with a slightly smaller diameter than the secondary. The secondary can slide up and down on this tube inside and gives the possibility to adjust coupling.

The twin is nearly identical to the first one described above. Top load is slightly different arranged, but has approximately the same capacitance for the virtual ground point being located at the base of the coils (where they are connected).

For the test runs with the sw-cap, I used a toroid with a mean diameter of 10cm, and an 8cm one topped with a half sphere. Goto the MKP-cap testing (final results) for a description of the final setup with bigger toroids.

Nothing special as the input power is only 180W. If driven in twin mode, no ground is required anymore! For some experiments like arcs to grounded objects, a kind of safety gap from the real RF ground to the virtual ground point has to be inserted! Despite of carefully balancing the twin system with different sized top loads, the virtual ground point is still on a high potential (resulting in a high gap width of this vgnd safety gap). I think the reason for this behaviour is that the arc is starting not exactly at the same time from the two toroids, leading to a time dependend shift of the virtual ground point towards one of the secondaries. The reason could be a temporarily increased top capacitance of the secondary from which the arc starts first, therefore the vgnd will shift towards this secondary. A reality check with a very short arc (by placing the secondaries in a short distance) showed that this behaviour is independent of spark length - I couldn't bring the vgnd safety gap electrodes narrower when decreasing the distance between the toroids without sparking in the vgnd safety gap.

A wooden housing, consisting of 3 levels:
- bottom level is 1+5.2cm high and contains the line filter, fuse, xfmrs and gap,
- middle level contains the test glass bottle sw-caps / MKP-caps,
- top level is 0.4+7cm high and contains the primary coil and the movable ground rail for the second secondary (the 'twin').

max. sparks
Test runs with sw-cap:
First test run (with sw-test-cap!): 7.5cm long arc from wire on toroid T2 to ground wire. Seems that the 4.3pF toroid is to small (many breakouts at same time). The distance between the strike rail and the toroid is about 14cm. Lots of corona (a really blue cone) from strike rail to secondary. Tapping between turn 9, 10 and 11 brought no difference in output length. Even between turn 8 and 12 only a very small difference was detectable. Conclusion: the system seems to be extremely overcoupled and the secondary therefore has to be raised some cm upwards.

Second test run (with sw-test-cap!): much more top load (T2 + T1 + T3 + 8cm-ball), 11.5cm arcs to grounded wire. Tap point at turn 11. Secondary was rised so that the lowest turn is now 1cm above the first primary turn. Breakout also from top turn, so a small toroid should be placed between top turn and big toroid. The corona cone is still present.

Third test run (with sw-test-cap!): For the third run, I made a small toroid which can slide up and down on the secondary above the winding. The strike rail is now replaced by a smoother one with thicker wire diameter. Breakout from the top turns and the corona cone are supressed this way. The secondary now can also slide up and down on a smaller tube inside it, so the coupling is adjustable. Again, 11.5cm max. arc length (as always: more topload - longer sparks).

First test run in twin mode (with sw-test-cap!): 11cm arc length between the secondaries (no ground wire attached!). With other words: I doubled the voltage but couldn't achieve longer sparks. As suggested by many other advanced coilers, I can now say by own experience that it is the power that makes the spark length, not the voltage.

Final results:
Results with final MKP-cap:
First test runs with matched MKP-multicap-cap (again in twin mode): 20cm arc length (16.02.99) between the secondaries (no ground wire attached!). The bigger tank cap (as opposed to the test runs with the SW-cap) allowed me to use bigger toploads here: main toroids are 17cm x 6cm and 19cm x 4.5cm on top of the two small toroids mentioned above. There is a lot of energy inside the small system now, I got sparks all over the secondaries up and down - even on the twin secondary which is only driven by the base current!! Frequency of this setup was about 715kHz (turn 9, which is about 8µH(???) and 6.2nF in the tank circuit) or, when calculated via the data of the secondaries (4.18mH, parasitic capacitance around 2.23pF, toroid capacitance about 8pF), 770kHz. Considering the fact of the toroid capacitance being not known very well and 5% tolerance of the tank caps, this is a good match of the two results. With the new MKP-Caps, the arcs are much longer and louder now. I place the twin now always in a distance of 16cm (at the end of the extending bus bar), this will give a beautiful display of a heavy, continuous and very long arc.
An impressive demonstration would be a 25W light bulb inserted into the base wire. It will light up as driven with about 5W (guess). Ok, only 1/36th of the input power is not very spektacular. But this is the average power! Duty cycle is only about 0.0025 and therefore the peak power dissipated in the filament will be approx. 2kW. No wonder that the filament was dancing around heavily inside the bulb :-)

circuit plan (final configuration) of Vitamini:

Data has to be recalculated! I made some minor changes on the geometry of the primary and never checked the values against a second program, so it might be a bit off. Toroid capacitance is not determined very well because of the two toroid approach (which has not found its way into an easy mathematical expression up to now).

test setup with sw-test-cap third test run (with sw-test-cap)

final configuration (MKP-caps and twin setup) (26.07.2002)

Kurt's Black&White (40cm dia) and my Vitamini (2.9cm dia)

Vitamini has a big advantage over Black&White: you can store
her in a small card box (31x22x25cm3 ;-)) (15.08.2002)

Here you can see the toroid arrangement. The grey cylindrical thing below the twin is just a box.
The small anti corona ring on the secondary can slide up and down to find the best position.

A look inside. In the box are the main power switch, the line filter, the
compensation cap (now reduced to only half of the original value due to space
restrictions), the 12VDC power supply for the fan, the fan, the two HV xfmrs
and the spark gap.
In the middle layer, the caps with their bleeding resistors are located as well as
the safety gap with its current limiting resistors.
On top of all is the primary. (15.08.2002)

... some more details: You can see the fan in the left front corner and the outlet holes on the
backside directly behind the spark gap. every second gap is shorted with a small crokodile clamp
because the gaps are made a bit to wide. I shorted every second gap instead of just taping only
e.g. the left half of them to spread the heat over them all.   (15.08.2002)

Primary construction  and strike rail. You can see now that I removed the three outermost turns
(unused) to prevent the primary from further arcing to the strike rail.  (15.08.2002)

The parts of the secondary. On the top of the secondary, I screwed an empty film container
(ask in a photo shop) and srewed the top lid beneath the toroids. So I can simply snap the
toroids onto the secondary.(15.08.2002)

The parts of the twin (second secondary). Here the toroid arrangement can be slid onto the film
container. The original lid of the 'Vitamin C' - tube was mounted onto the cylindrical box, so I
can snap the secondary in place.(15.08.2002)

Yes, the housing is ugly (I used what I had at hand, perhaps I paint it somewhere in the future). After seeing Richard Hull's Tesla coil in an Attache case, I decided to try this with my Vitamini if time permitts (I'll skip the Variac and place the twin inside the case instead of it)...

Some electrical data around the current twin system (values are not fixed because I'm still 'playing'):

9.5kVeff, <182W, ~6.2nF/~9uH tank circuit => 14kV/370A peak values (Z=38Ohms), f~800kHz, top load ~ 7pF => max. output 400kV (from Ep=Es) , 20cm arcs, that equals 9.1W/cm.

Some thoughts on twin coils:

I for myself prefere to look at the base current going into the secondary here: In a standard 2-coil system, the secondary is grounded at its base. Here you have the base current Is.
(_________)=^-~´\,-~ I I I I I I I I IcI IoI \ IiI / \IlI/ +-----o Is _|_ - gnd

If you connect an identically twin secondary and disconnect the gnd and drive only the first secondary as before (by the primary), then the behaviour of the coil should NOT change:
  _________                _________
     I I                      ItI
     I I                      IwI
     I I                      IiI
     I I                      InI
     IcI                      IcI
     IoI                      IoI
   \ IiI /                    IiI
    \IlI/                     IlI
                  virtual gnd point
                 (not connected to any real gnd!!!)

The current is not going into the (now virtual) gnd anymore but into the base of the twin secondary instead. Same current as before with only one secondary and therefore the same voltage appearing at the discharge terminal. With the phase shift of 180°, the potential difference between the two toroids is now two times Vo!! With the same energy consumption as before. I can't imagine why the base current should change when you replace the 'mirror plane' (real gnd) for a real coil. I have no scope until now, so I can't measure the base currents in the two configurations. Please mail me, if you have any thoughts on this!

Update May'99: the new gap (flat arrangement of copper pipes)

I'm currently making a new gap. Total gap width should be around 7mm for 9.5kV. Due to the limited space, I only can make 10 gaps. For adding some flexibility, I try to space them each 0.63mm+0.254mm=0.884mm. The strange values are based on some ceramic spacers I have at hand :-) I think I'll need somewhere around 6.9mm total gap width, this will be 8 gaps (the gaps will decrease a bit if the spacers are removed because I mount them with some tension on the base board). So I have 2 gaps more for compensating some corrosion which will occure over the time. For the beginning, I'll short these two gaps with two small crocodile clamps.

It is obvious that this mini coil system is still to big and heavy to convert it into a battery-powered system.
So perhaps I'll build a true Micro Coil System (µTC) with some of these smaller commercial MKP-caps sometimes...

YES, I do have some ideas of its values:

Stunguns usually run on 2 x 9V in parallel (alkaline batteries) because they are very powerhungry (approx 2.5A). With the voltage dropping down to 8V this results in 20W. A 9V battery of the brand 'Energizer' has 600mAh. If we leave 100mAh in the battery, we have 1Ah. That leads to 24 minutes max. total runtime. That's ok for a 20W batterie powered µTC which will be only a gimmik and will be designed for several short-time runs of max. 30 seconds each.

Considering the losses below 5W, the output would be in the range of 15W(cw).

Power = Ec x bps, with the data given above: 15W / (1/2 x 2.6n x 15000^2) = 51bps. I won't use a higher capacitance because this would lead to a physically bigger cap. So it will get a max spark length of 0.85 x 7cm = 6cm. The reduced bps (51 instead of 100) reduces the output by a factor of 0.85 according to a fit of my own measurements on my 2"-system. An 50bps-TC with 15W equals an 100bps-TC with 30W when you reduce the spark length by this factor of 0.85. The named 7cm are the value of this 'virtual' 30W/100bps-TC (from the 'bad value'-curve in the efficiency diagram on my technological background page).

This all leads to approx. 6cm spark length for this batterie powered µTC!

If I would use a 7x higher capacitance (18nF), the spark length should increase to 14cm!!! Guess what I'm doing now? Yes, you got it - I'm going to look for smaller caps (1000VDC => 20x 330nF, 630V => 32x 560nF or 38x 680nF).....

(Oops - perhaps I should try to get some data on the stungun first :-)

Fuer die 1"-Microcoil sollte ich ein paar WIMAs uebriglassen und als EMMC konfigurieren:
4.75kV / 18mA mit 120% resosize macht 14nF.
(Update 10'2004: better use up to 160% = 18nF for a static gap (320% for SRSG))

Zwei WIMAs in Serie ergibt 16.5nF => passt
ACpeak ist 7.1kV, ok bei 12kVDC.
AC-faktor ist 5, das ist doppelt so hoch wie im 4"-MMC, also guter Test.
Kurt's WIMA-MMC liefert "fully safe design"

drei WIMAs in Serie ergibt 11nF => passt auch noch (fast genau reso=12nF)
ACpeak ist 7.1kV, ok bei 18kVDC.
AC-faktor ist 3.4, das ist 1.5 fach so hoch wie im 4"-MMC, also auch noch ein guter Test.

Statt der WIMAs (33nF/6kV) mit den KP1836 von ERO:
7 caps ist "fully safe" nach Terry's MMC-calc bei 22C/W (siehe Auslegung des 2"-MMCs, update March'2002) => 9nF
6 caps ist "nearly safe" nach Terry's MMC-calc bei 22C/W => 10.3nF
5 caps wäre 12.4nF
4 caps wäre 15.5nF