Have a look at https://www.jameco.com/Jameco/workshop/TechTip/555-timer-tutorial.html

https://core-electronics.com.au/tutorials/our-siglent-sdg1032x-arbitrary-function-generator-review.html

https://www.electronics-notes.com/articles/test-methods/signal-generators/function-generator.php

Since a 20 kHz square wave has a total period of 50 us, it's difficult to get ON times of 100-1,000 us. Did you mean 100-1,000 ns? Driving MOSFETS in series is pretty tricky; drive timing is all important for simultaneous turn ON and OFF, and various parasitic capacitive effects can be a real problem. Your best bet may be to use strings of TVS diodes across each transistor to clamp the drain-source voltage below their breakdown voltage. The TVS capacitive loading effects can be reduced by placing one or more SiC Schottky diodes in series with the TVS diode strings, to block much of the discharging of their clamping junction voltages. The most straightforward way to generate 8 kV pulses would be to use a half bridge of CREE/Wolfspeed 10 kV MOSFETS (if you can get them) on an 8 kV dc source.

Since you specifically ask for a block wave, I would assume you can just use a HV thyristor to switch a 8kVDC source on and off, with a low voltage 20HZ oscillator (probably best shaped as a block wave itself) to control the gate.

Rogier Brussee

The exact problem is finding a Mosfet or IGBT that blocks this amount of voltage. I would appreciate if you could suggest me some high voltage (8kV) products.

Bruce Carsten

Thanks for your reply. Actually, the frequency is 20HZ, not 20KHZ. Just as you said, I noticed stacking the MOSFET is very tricky. I went ahead and asked about the half-bridge Wolfspeed 10KV MOSFET, I hope i can afford the price of them.

Bruce Carsten Do you have any suggestions for an HV transformer? that supports this low frequency with possible lowest deformation of square pulse signal?

If you wanted just a 20 Hz square wave, even a SPDT vacuum relay would work for a while, but not likely down to 100 us pulse output.

A few questions that need answering to consider for use of a pulse transformer are:

1) Is this a pure unipolar output you want? That requires a net dc output component, which transformers cannot provide directly. However, it is possible to use an output or 'coupling' capacitor with a dc restoration diode to 'ground'. This could be made to work if the pulse duration does not have to be changed too rapidly.

2) What output "loads" are to be driven? Do you require just a voltage output, with no significant loading?

3) What deviations from an ideal pulse output can be tolerated? Rise and fall times, overshoot and ringing, droop, etc. These are all important to pulse transformer design.

A first pulse transformer design consideration is the output pulse volt-second product, which sets the minimum product of winding turns, core area and flux swing. Assuming an allowed flux swing of 3.2 Tesla with an Si Fe core, and 8 kV for 1000 us (1 ms) requires an area-turn product of 25,000 cm^2. A 1 cm^2 core area requires at least a 25,000 turn HV winding, and; a 10 cm^2 requires 2,500 turns minimum.

The next issue is output ringing, largely due to pri-sec leakage inductance and secondary winding capacitance, which also limits voltage rise and fall time, a bandwidth limitation. Unfortunately, bandwidth naturally decreases with secondary turns. Leakage inductance increases with the square of the secondary turns, which would be OK if the self capacitance decreased as the inverse square of the tuns, but as a rule of thumb the capacitance typically only decreases as the inverse square root of the turns. Thus the L-C product increases, lowering bandwidth and increasing rise and fall times. Capacitance can be lowered over that of conventional windings with sectionalized windings, such as used in some RF chokes, CFL transformers and the like. This ALSO increases the breakdown voltage!

A ringing output can be externally damped with an R-C snubber, with a snubber capapacitance about 3 time the winding capacitance. A more sophisticated method is to use winding resistance (with fine wire windings) to create a critically damped series L-C-R circuit. The distributed series resistance of the winding works better due to the distributed self capacitance of the winding.

Finally, output pulse droop (with a high impedance output load) is largely generated by increasing magnetizing current during the pulse and the primary winding resistance (and transistor and source resistance), so these need to be considered.

If a really clean pulse output is required, I'd consider clamping a transformers output to ground and the desired maximum voltage. This might be best achieved with a stacked array of secondaries, each with their own lower voltage clamping diodes and TVS diodes.

Bruce

I only know that High Voltage thyristors are routinely used to convert AC to DC and vice versa over hundreds of MW power lines, so I only know they exist and are known to work in roughly your frequency range. This is the first Google hit, so that is about how much expertise there is behind it.

https://www.infineon.com/dgdl/Infineon-540A_8000V_LTT_Thyristor-PB-v04_00-EN.pdf?fileId=db3a304336ca04c90136de6d3acf23ac

Bruce Carsten Thank you so much sir for providing precious detailed information. much appreciate it.

Rogier Brussee Thank you sir for your feedback. Appreciate it.

As so often happens, a requirement can appear simple on the face of it, and yet get progressively more complicated the deeper you delve. This is one of those cases. Producing an 8 V pulse is trivial, and 80 V (or even 800 V) is not getting that much harder.

By the time you get to 8 kV, however, you've entered a different world. Even with 50 years experience, I'd have to give the problem some serious consideration. If this were a consulting job, I'd ask a lot of questions about just exactly what was required and, often more importantly, just what was NOT required.

In other words, what assumed or "taken for granted" aspects, features, capabilities, etc. are not necessary. Thinking this way, i.e. "what can we give up to get what we want" is much harder that it would seem, but can sometimes help, occasionally a lot.

If you can't "Google the answer", you have to think it through.

when you connect semiconductor devices in series for high voltage, they can fail for several reasons. For example, different leakage current in non-conductive state; the device with lower leakage will have higher voltage across it, and that voltage might exceed the maximum rating. That can be fixed by placing resistors across each semiconductor device - the value of that resistor should provide leakage current higher than semiconductors, equalizing the voltage across the devices. Another reason - the gate drivers have slightly different delays, so the power devices get into conductive state at different time. Need to use very fast drivers with enough output current to ensure minimal time difference between your mosfets or thyristors. Another yet reason - getting your semiconductor devices outside of safe operating area. Could be even thermal transients; need to look through the datasheet very carefully.

Maybe you can use Behlke push-pull switches which can handle pulses from 300 ns - sec and the rise time of 10 -100 ns with a relatively simple driver circuit. You can find more details from the following link:

http://www.behlke.com/separations/separation_c8.htm

Shahriar Mirpour Thanks Shahriar for your comments. The only problem is money, my friend. It is way out of our budget to afford.