Below shows the start of construction for the MMC. All the holes were worked out and drilled before hand (I used CAD software: Rhino version 3). Whilst this approach does not make the coil run any better, it looks a lot tidier, and importantly it ensures there is always sufficient spacing between components.

High voltage capacitors that can survive the grueling conditions of a Tesla coil are hard to find, and therefore inevitably expensive if you do manage to locate some.
Making your own rolled poly capacitor is another option, but this rarely tends to be successful for very long.
Nikola Tesla (1856 - 1943) used salt-water capacitors which consisted of wine bottles with a brine solution in them. This option will work and it can also be very cheap if your short of funds. However they are not very efficient. Glass has very high losses when it comes to RF frequencies, so your coil's power output will be down. It is also difficult to estimate the value of capacitance that you have made, important when it comes to tuning, unless you have a suitable meter capable of measuring the figure.


The option that a lot of people use is to create an MMC or 'Multi-Mini Capacitors'. These consist of quiet a few lower voltage rated capacitors joined in a series/parallel combination. This means that 10 individual capacitors each rated at 2000v, could now handle 20,000v being applied across their series string of 10. The drawback is that capacitors behave the opposite to resistors in respect of their overall value when connected in series.
This means that once you have enough capacitors in a string to meet the voltage requirement, you then need to keep adding identical strings in parallel to bring the capacitance value back up to what you desire. This is illustrated below.



The specs of the Cornell-Dubillier 942C20P15K-F are:-

Size	0.150uF
Voltage DC rating	2000v
Voltage AC rating	500v
Part Number	942C20P15K
Length inches	1.161 (29.5mm)
Diameter inches	1.811 (46.0mm)
Lead diameter inches	0.047 (1.2mm)
dV/dt	2879
I peak	432
Irms 70 deg C & 100 KHz	13.5

At one stage I wanted a capacitor in the region of 0.04286 uF, this value was decided on because of the NST transformer's value that I was using - two 10,000v @ 50 m/a units. The individual capacitors that I chose to use were the very robust and popular 942C20P15K-F made by Cornel Dubilier. These are rated at 2000v each and have a value of 0.15uF. These are 'metal foil' capacitors which are preferred over the 'metal film' type. The difference is to do with the connection onto the end of the plates, and it has been found the latter type can suffer from failure, unless the peak current is kept within the capacitor's specifications.

I found that four strings of 14 in series gave 0.04286uF (42.86nF) and would handle 28,000 volts. As my NST was 10,000v RMS (14,142 volts peak) this value was adequate.

Update February 14th 2010
The MMC is now used on a 200bps SRSG attached to a 'PIG' it is currently 72.2nF (measured) consisting of 9 strings of 19 in series. This gives a voltage rating of 38Kv


An Over-volted Capacitor



A clumsy accident occurred in which a piece of scrap wire dropped across the MMC, resulting in this one single cap (a Cornell-Dubillier 942C20P15K) ending up being connected across a 12Kv pig! The centre immediately blew out and was found to be completely carbonised.
The lesson learnt is not to put 12Kv across a 2Kv capacitor, and more importantly, also make sure no pieces of scrap wire are left unaccounted for.



MMC Software

There are several programs that will work these values out for you. One of the best for MMC calculations is 'TeslaMap'. It runs on Vista, XP 32 & 64 bit systems. In its menu goto 'View' >> 'MMC Calculator'.

A very important thing to remember is to fit bleed resistors across each capacitor. I used 10 Meg 0.5 Watt axial thick film resistors.

(Farnell Order Code: 129-2582).

The purpose of these is to discharge the capacitor after the circuit has been switched off. Capacitor banks can be fatal if care is not taken. Just because the circuit was switched off some time ago (hours, or in extreme cases even days!) does not mean to say that the MMC does not still retain a charge.
As the NST's secondary winding and the primary coil are wired across the MMC it should discharge naturally, but if a disconnection fault had occurred and stopped the coil, you may well just decide to pack the coil away for the day with lethally charged capacitors waiting to 'bite' you when the situation next permits.

The bleed resistors should also be spaced away from the body of the cap to avoid any flashover, likewise the capacitors themselves should be suitably spaced. Not only to avoid flashover between adjacent units, but also to allow cooling air to circulate.