Avalanche Based High Voltage Pulser

[ghost896]

Hello everyone. I've been researching high voltage switcher (2500V enough for me) topologies with very low rise time and in many ways the proposed design is the one I've linked below. Although it is explained a little in the article, there are still things that are not clear. A desired pulse can be obtained with the Avalanche transistor stack, but I don't understand how the other half of it produces a differential pulse. I guess it triggers the other half with a delay, but I still couldn't understand the working logic. I would appreciate it if you could clarify the working logic.

https://mountainscholar.org/bitstream/handle/10217/67552/Woolston_colostate_0053N_10962.pdf?sequence=1&isAllowed=y

 

[FvM]

No differential pulse, an unipolar pulse with separately controlled rising and falling edge, applied as voltage difference to a floating load.

 

[ghost896]

No differential pulse, an unipolar pulse with separately controlled rising and falling edge, applied as voltage difference to a floating load.

I thought it was differential when it was 2 symmetrical structures, thank you for the correction sir. Well, I couldn't understand why 2 structures were used. It also triggers both sides with a nanosecond delay I guess.

I understand how a single stack works, but I don't understand why using 2 stacks. By 2 stacks I mean the circuit at the right and left ends of the load.

 

[Easy peasy]

as the stacks are in hard parallel at each end - it can only be for more current.

 

[ghost896]

Doesn't it divide the current using parallel?
When I look at Avalanche-based products, I see that the pulse width is 0.5 us, but in my topology research, the pulse width is around 10 ns. How did they increase the pulse width and how can I increase it? If I connect a capacitor parallel to the output, will it work?
(Example PCD-21 Output)

Best Regards

 

[Easy peasy]

Sigh, if you have 2A in each leg that adds to 4A, if you have max amps in one leg you have 2 x max amps in 2 legs in parallel ... the coax cable appears to be the cap storage for the pulse .... so a longer coax = a longer pulse, and presumably a bit higher current too ....

 

[ghost896]

Sigh, if you have 2A in each leg that adds to 4A, if you have max amps in one leg you have 2 x max amps in 2 legs in parallel ... the coax cable appears to be the cap storage for the pulse .... so a longer coax = a longer pulse, and presumably a bit higher current too ....

I guess it is not possible for me to use a long coaxial cable. After all, isn't the purpose of the cable to have a capacitive effect? Shouldn't I just throw a parallel capacitor instead of a wire, sir?

 

[PlanarMetamaterials]

I guess it is not possible for me to use a long coaxial cable. After all, isn't the purpose of the cable to have a capacitive effect? Shouldn't I just throw a parallel capacitor instead of a wire, sir?

In an avalanche pulser I previously designed, I just used capacitors. They worked great for my purposes, but I just wanted a very long fall time (~1us). The pulse shape was definitely "sawtooth" and not rectangular.

 

[FvM]

Transmission lines can be used to generate rectangular pulses with a pulse length double the electrical cable length, e.g. with avalanche transistor or mercury wetted reed relays as switch. The cable is storing energy, not acting as capacitor but as delay line.

The transmission lines in the post #1 design are not defining pulse length, just isolating the load from the pulser during active pulse time. Pulse duration is defined by delay between first and second trigger.

 

[ghost896]

Transmission lines can be used to generate rectangular pulses with a pulse length double the electrical cable length

1)There is a product called PCD-21 and it works in the same topology. It can handle high voltage 0.5 us. The length of the cable is 50 mm. Does your calculation match these values?

2) I designed a circuit like this schematic, but I did not pay attention to the output impedance. Can I increase the pulse width if I connect the PCD-21's cable to my circuit? Or would I have to do impedance matching?

3)If I have to, what's the quickest and easiest way to do it, sir?

--- Updated Nov 30, 2022 ---

In an avalanche pulser I previously designed, I just used capacitors. They worked great for my purposes, but I just wanted a very long fall time (~1us). The pulse shape was definitely "sawtooth" and not rectangular.

Actually, I don't want a full square wave, just 0.5us voltage around 2500V. Fall time can go as long as it wants, my frequency is too low.

 

[FvM]

Propagation delay of 50 mm cable is in a 200-300 ps range, it's almost irrelevant apart from a certain effect on pulse edge. 6 ns delay line in post #3 has 1.2m length if we assume 50 ohms coax.

 

[ghost896]

You are right, it is in post #3 but not in PCD-21. I'm really curious how they can keep the pulse width 0.5us.