Opening switches simultaneously How to get rid of jitter?

[condorino]

Hi all,

I'm in a big trouble finding a switch that discharges a capacitor with 20kV on it.
My circuit requires to do this very qwickly (tens of nanoseconds), but I cannot use a single switch because the only existing one that can allow this voltage is a Thyratron, and I don't want to use it.

So what I want to do is to put in series a bunch of switches (like FAST HIGH VOLTAGE SOLID-STATE SWITCHES) and to open them at once.
The point is that the opening action takes place not "exactly" at the same time for different switches (jitter). This discrepancy in time can cause a single Switcher (maybe the last one that starts conducting) to have the entire 20kV voltage between its electrodes, thus breaking it.

I was thinking of a circuit to be in parallel with the series of switches, but didn't find anything working.

Can somebody help me?

Thank you all,
C.

 

[Carl Pugh]

Some methods of discharging capacitors are thyratrons, ignitrons, triggered spark gaps, and there was some research on placing the capacitor across a semiconductor and triggering it with a laser. There was also a SCR that would trigger automatically when a certain voltage was reached.
If you would explain exactly what you are trying to do, someone may be able to help you.

 

[jim hardy]

"" and to open them at once.." to discharge? Did you mean 'close them all at once..' ?


I'l be surprised if you find a semiconductor device thet'll do it, but i am years obsolete...


look into 'krytron' which is a cousin to the thyratron. You might get one for experiments from an old flashtube style copy machine.

old jim

 

[yungman]

You cannot use a series of lower voltage switches because the last one turn on will get the 20KV in full! And also you control circuit has to float and can take the switching speed. THis mean it has to be fiber optic or pulse transformer switching.I designed a precision switching circuit that discharge and charge about 200pF at 5KV with rise time in nano sec. It was a big project using at least 7 to 8 MOSFET on each side in series to pull up to 5KV and down to 0V that settle to within 1 or 2 volt in a few hundred nano sec. The circuit is quite complicated because each FET can take 1KV and I had to protect each and every single one with transorbs(TVS). You cannot use the total 1KV capability because TVS is not accurate and you have to guaranty the max voltage cannot reach 1KV no matter what. The drive is by a transformer design out of a PCB with ferrite glued on it. Then you have the driver circuit and all. I don't know your application, maybe you don't need this kind of precision if you just discharge it.

With this kind of circuit, for 20KV, it is going to be big. And as it get bigger, it is harder to get the speed you want because at this speed, propagation of the wire comes into play also.

I know at the time, there is a German company sell this kind of fast switchers, the name is Behlky? It was very expensive, that's the reason I design a substitute.

 

[condorino]

Thank you all!

I've looked for Krytron, Ignitron, ****tron. They all have the same problem as the Thyratron: very short life (10^7 commutations).

@yungman: yes, a series of switches is useless!... You have obtained a 5kV commutation with 1kV switches, now I have a question for you: which is the relationship (if any) between the number of (subswitches) and the total commutation time of the circuit? Is this number the major contribution to the increasing rise time? (yes, actually two questions).


Thank you again.

C.

 

[hisham.i]

Suppose that we put switches in series, and we put a large inductor in parallel to the load, thus when a switch is opened there is still a current in the load which decreases the voltage to a level that the switch can handle, and within this time all switches are opened.

 

[yungman]

Thank you all!

I've looked for Krytron, Ignitron, ****tron. They all have the same problem as the Thyratron: very short life (10^7 commutations).

@yungman: yes, a series of switches is useless!... You have obtained a 5kV commutation with 1kV switches, now I have a question for you: which is the relationship (if any) between the number of (subswitches) and the total commutation time of the circuit? Is this number the major contribution to the increasing rise time? (yes, actually two questions).


Thank you again.

C.

It is the driving circuit that cause part of the problem. I designed pulse transformers onto the PCB with primary and secondary winding as traces on PCB to get the high isolation voltage. Then I glue the ferrite core onto the board. I drive the transformers with DS0026?? MOSFET driver. The propagation delay of the DS0052 and the transformer path is main cause of the delay uncertainty between all the switches. I chose DS0026 at the time because it is bi-polar technology that has much much less drift than all the MOS technology. MOS ICs tend to drift a lot more because of the Vgs vary with temp. Maybe there are new ones that works better now, I designed this in 1996!

Also the Vgs on is different on each of the HV MOSFET.

If you just want to pull down to discharge, you might get away with using 1.5KV HV NPN transistors. At the time, I had to pull up from 0 to 5KV also, I need both direction and there was no 1.5KV PNP! It should be a lot easier to drive BJT compare to MOSFET. And also you might be able to find much higher voltage BJT than MOSFET so you need less stages.

 

[Carl Pugh]

Thryatorns don't have short lifetimes.
Google thryatron life.

 

[yungman]

One thing, I don't know your requirement, you might not need the requirement I had to deal with. In my case, not only I had to switch fast, I had to hold steady as if it is DC, I had to hold a 0V or 5KV for indefinite period and provide sink or source up to 10mA DC current. I had to continuously pulse the transformers to keep it in either direction. If you don't need to hold DC and sink current continuously to keep at 0V, you might be able to find much easier circuit to do the job. My design was in essence a DC switch. Also, in 95, I only can find 1KV FET and 1.5KV BJT, you might be able to find much better devices now.