Solid Stage Marx Genarator Utilising Thyristors

It is hard to give ideal or exact values for R and C for the snubber circuit as it depends on a number of factors (including load, frequency and circuit element interaction and response, etc). Additionally you may have a problem due to the simple way you are driving the SCR gate (you may require additional R, L, C or diode elements depending on your SCR characteristics and how you are using it).

SCRs and TRIACS are really simple looking devices at first glance, but horribly more complicated than they appear (especially when they are used for special applications like this one).

There are some good documents available on the internet and SCR manufacturer websites that can give you a guide such as the following....

www.fairchildsemi.com Application Note AN-3008

http://www.onsemi.com DL137/D Thyristor Device Data

http://www.irf.com/indexnsw.html

etc

In the end you may just end up taking a 'best guess' and altering the values based on oscilloscope measurements and observations. If you are using a computer model beware, the response may differ from your real-world circuit.

It appears to me that you are in effect connecting two 2.5KV transformers "in series" so to speak, across your total 7KV string and so I would expect a breakdown. It might be clearer to you if you imagine each transformer as a 2.5KV capacitor (which is what the windings are from primary to secondary) and then you can see that, because all the primaries are common, your two end parts must sustain the entire voltage placed across the string between themselves. As with any series capacitor arrangement, without leakage resistors to equalize the effective voltage divider, one will most certainly take a larger portion and fail. Even were you to use resistors to equalize the voltage distributed across the windings, it is, and will become more in excess of the isolation rating, as you add more cells. You must find another configuration that circumvents this problem.

For this cirquit to work you have to make sure all the thyristors fire synchronized to the ns. Easiest way to achieve this is to feed their gate transformers with rectangular pulses at desired time and not sinusoidal voltage from propubly unmatched transformers that triggers them pretty much randomly. If you must stick to this scheme diacs and resistors on thyristor gates can help, but still the last thyristor to fire will suffer avalanche currents and RIP causing chain of failures since the others are connected in series. (This is possibly what's already happening as you increase voltage.) S.M.

Your drawing shows the gate pulse transformers in parallel (primary). The last one see basically the full voltage that you are producing between primary and secondary and laminations. You might need to place them in series even if this will introduce delays, and mount them on an insulating surface.

Don't forget that the 2500V rating for the transformers is probably at 50/60Hz. Things are different at high dv/dt. Isolation material behaves more like a series of capacitors and allows large partial discharge to ionize gas bubbles that eventually destroy the insulation properties of the material.

Thyristors are relatively slow semiconductors. They don't like large di/dt and are likely to fail prematurely in this circuit. MOSFET would be a better choice in this circuit even if they operate at lower voltage. IGBT could also be used.

Fiber optic firing might be a better choice for insulation. ABB makes this type of switches that way. Look them up. Check this link http://www.appliedenergetics.com/

Play safe. These are dangerous experiments.