# High Voltage Switching - Cascode Circuit

Discussion in 'The Projects Forum' started by Terroman, Aug 12, 2016.

1. ### Terroman Thread Starter New Member

Aug 12, 2016
12
0
Hello all,

I have a particular power supply, capable of a linear output from 0 to 6,000 V depending on a 0 to 5 V input control. I'd like to switch the output of this power supply on and off repeatedly in order to feed the pulse train into a voltage multiplier to obtain tens of kiloVolts (purpose is for electric field experiments). The power supply is to stay on all the time and thus, circuitry needs to be added to its output. The maximum current output is 3.3mA @ 6,000V (20W).

First and foremost, I'd like to ask how this switching can be accomplished.

I have vastly searched for a solution and came up with a stacked transistor circuit implementation (20 NPN BJTs of a max VCE of 400V). I have biased the transistors in order to get 10uA for each base. The circuit seems to work on a simulator; however, since I am controlling the most-bottom NPN BJT, there is a small delay for each subsequent transistor to switch on and off. This is causing a voltage spike accross the VCE of the bottom transistors.

I will gladly post the circuit upon request, but for now I do not want to bias your opinion: how can I switch 6,000 V at a fixed frequency please (even 300Hz would be sufficient)?

Thanks in advance for your help!

Jun 4, 2014
1,893
375
3. ### Terroman Thread Starter New Member

Aug 12, 2016
12
0
Thanks for the link. It seems very complicated and I cannot understand the MOV connections. I was thinking of stacking solid-state relays, each capable of handling 400V between positive and negative outputs.

From your experience. is there anything I should keep in mind when using this topology instead of the NPN stack please?

Thank you!

4. ### AnalogKid Distinguished Member

Aug 1, 2013
4,518
1,247
Solid state relays are relatively slow-responding devices. What are the frequency, duty cycle, rise and fall times, etc. of the high voltage output? Also, are you limited to high-side or low-side only switching?

ak

5. ### Terroman Thread Starter New Member

Aug 12, 2016
12
0
I'm considering VOR2142A8 as a solid state relay. It has a turn on time of 0.13ms and an off-time of 0.05ms. Basically I can set it to 750Hz maximum. I'd prefer the relays to be low-side switching (i.e. stack relays from the ground up) but I'm open to any other topology which might work.

It is my intention to connect the output of the pulse train to a Johnson-Walton voltage multiplier circuit with 1nF capacitors. Considering the way the circuit operates, I believe that 750Hz should be enough as capacitors (High-Voltage rated 1nF) won't be discharged because of the way the diodes are connected.

If there's a delay in one relay, there's a possibility of frying the whole circuit. Is there an issue to switch on or off these 20 solid-state relays simultaneously please?

6. ### AlbertHall Well-Known Member

Jun 4, 2014
1,893
375
The MOVs are there to protect the MOSFETs from excess voltage.

7. ### Terroman Thread Starter New Member

Aug 12, 2016
12
0
What I cannot understand about MOV is their connections in that circuit in particular. Are the collectors of the MOSFETs connected to Ground?

I'm also considering using a similar approach for the relays: varistors put across the output of each relay.

8. ### AlbertHall Well-Known Member

Jun 4, 2014
1,893
375
No. Each MOSFET has an MOV between source and drain. If the voltage across a MOSFET is too high then the MOV conducts to protect the transistor.

9. ### crutschow Expert

Mar 14, 2008
12,991
3,227
The MOVs are connected directly across each MOSFET's drain to source to protect them from any momentary overvoltage during switching.
MOSFETs don't have collectors, only BJT's have them.
No terminals of the MOSFETs are connected to Ground.

10. ### Terroman Thread Starter New Member

Aug 12, 2016
12
0
Oh I see! I thought that the MOV were those circles connected to ground! Incidentally, what do those symbols mean?

An issue which I see in this configuration is the stress on the uppermost MOSFET, where there is a voltage difference of the full DC supply between the source and the gate. Won't 8 kV across that transistor cause it to breakdown?

11. ### AlbertHall Well-Known Member

Jun 4, 2014
1,893
375
Those squiggly things are transformers (drawn with the primary and secondary overlapped). They take the drive from the logic and feed it to the gate/source of the MOSFETs while isolating the high voltage.

12. ### Terroman Thread Starter New Member

Aug 12, 2016
12
0
Ah thanks AlbertHall! So basically they are 1:1 isolation transformers, correct?

The most important question at this point is about the uppermost transistor:
Will a voltage of 8kV between the drain and the gate cause overstress and break the oxide in between please?

13. ### AlbertHall Well-Known Member

Jun 4, 2014
1,893
375
That's what the MOVs are for. They restrict the maximum drain-source voltage to avoid overstressing the MOSFETs.

14. ### Terroman Thread Starter New Member

Aug 12, 2016
12
0
Ah! Because Vds = Vgs + Vgd correct? If we're limiting Vds, there'd be less voltage between Vgd and Vgs, right?

15. ### Terroman Thread Starter New Member

Aug 12, 2016
12
0
Can I use zener diodes instead of MOV though?

16. ### AlbertHall Well-Known Member

Jun 4, 2014
1,893
375
The published design uses MOVs. The article says:
"This circuit works reliably and has proven to be an excellent, cost-effective way of re-referencing our beam energy."
so I wouldn't suggest such a change.
Can you even get 960V zener diodes?

17. ### tcmtech Well-Known Member

Nov 4, 2013
2,034
1,631
To be honest for the complexity and costs and likely poor reliability involved you would be better off going with a common NST (Neon Sign Transformer) followed by a voltage multiplier circuit.

With minimal components a common 15 KV 30 ma rated NST can easily run a voltage multiplier that goes up over 100KV @ 1 - 2 mA and still be variable with nothing more than a basic Variac transformer on its input side.

Also, those transformers come in s solid state versions now that put out high-frequency AC (~40 - 60KHz is common) rather than line frequency AC as well.