2

I could only find one question semi-relatable to mine on SE, but I have a more demanding application in mind. I have designed a 32 stage 40 VAC -> 1000 VDC Cockroft-Walton voltage multiplier which I plan to drive with the 2.6 VAC clock from my MCU. To get my SMD capacitor sizes submillimeter, I have to run the clock at over 3 MHz...

My plan was to use a LT8410 to boost my battery's 3 VDC to 40 VDC and switch that w/ the clock through an optocoupled solid state relay:

circuit diagram

I went with this topology bc I read somewhere that SSRs "have fast switching speeds which can range from 1 to 100 nanoseconds" but the smallest \$t_{\text{ON}}\$ (Turn-ON time) I've found was 500 ns & I need less than 50 ns to obtain an acceptable square wave at 3 MHz.

So is it even possible for an SSR to switch this fast or should I have went with a different topology to deliver the 1000 V?

Landon
  • 167
  • 1
  • 7
  • If your circuits all share the same ground, why do you need the ssr? – BeB00 Jun 19 '20 at 04:54
  • By which I mean, why cant you use a mosfet or other switching device? The main advantage of the SSR is isolation, but it looks like you dont need that? – BeB00 Jun 19 '20 at 06:25
  • http://tinyurl.com/ycp99ajc – BeB00 Jun 19 '20 at 06:39
  • @BeB00 I wanted to avoid unexpected spikes in current being drawn from the pins on the MCU as the MAX is quite delicate. Although the SSR isn't providing full isolation at least it is routing dynamic current draw away from the MCU. If I can't find a fast enough SSR I will probably use 2 CMOS inverters to make a switch. – Landon Jun 19 '20 at 07:26
  • 1
    *have fast switching speeds which can range from 1 to 100 nanoseconds* My opinion: whoever wrote that article **has no clue**. SSRs are generally used for mains switching which is 50 or 60 Hz, so there is no need for ns response times. What could be the case is that the transition from off to on of the SSR is in the 1 to 100 ns range. That doesn't mean it can be switched on/off in a few ns. TLDR: an SSR isn't going to work in this application. – Bimpelrekkie Jun 19 '20 at 08:15
  • @Landon That's not how this works. In the configuration that you're using it, the SSR will provide no benefit, and will just slow things down and mean that you cant do 3MHz. Mosfets will completely isolate your mcu from the high voltages and currents (you will still have to properly decouple your power lines etc). You also don't need to buy CMOS inverters - the circuit I showed you will work just fine (it probably needs some modifications to help it a bit, but that would also apply with inverters) – BeB00 Jun 19 '20 at 08:21
  • 1
    I also think that your system isn't what I would choose. I would use a **transformer** to boost 3 V directly up to 500 V and then double that or go directly from 3 V to 1000 V. A CW multiplier can work but in my view not with the amount of power that you need and so many stages in series. The losses in each stage will prevent you from reaching the 1000 V. I also suggest that you study other high voltage generators and see what is commonly used. "Designing your own" only works if you have enough experience of what works and what doesn't. Also: some solutions work in theory but not in practice. – Bimpelrekkie Jun 19 '20 at 08:22
  • Also I agree with @Bimpelrekkie on the CW topology not necessarily being the best choice. You should investigate other transformer based topologies (flyback is a popular one) – BeB00 Jun 19 '20 at 08:24
  • @Bimpelrekkie All of my components have to fit on a cm^2 board to meet project requirements. It also has to be very low power since we are running off of TDK solid-state batteries. The smallest smd transformer i found was 1:10 & drew 4 A from the 3 V supply. I simulated my CW multiplier in Multisim & got 960 V out; BAT62 diodes have very low voltage drop. CW is most common for my application (electrostatic actuator driving) bc it is a near zero power load. – Landon Jun 19 '20 at 08:37
  • 3
    XY problem. The op is trying to shoe-horn something in that is not fit for purpose. – Andy aka Jun 19 '20 at 08:38
  • 1
    [3 volts to around 10 kV](https://electronics.stackexchange.com/questions/66072/charging-a-circuit-to-around-10kv-from-3v-two-aa-batteries-for-discharge-over/72739#72739) – Andy aka Jun 19 '20 at 08:42
  • Thank you Andy, that's a really cool find but I'm trying to avoid having to use a transformer. – Landon Jun 19 '20 at 08:50
  • 1
    *It also has to be very low power* You're trying to make 1 kV into 10 Mohm, that means 100 mW is dissipated in the 10 Mohm resistor. 100 mW is not what I call "very low power". My guess is also that those "solid state batteries" cannot deliver the amount of power needed. TLDR: You're focusing too much on **implementation** while forgetting about the **architecture** and **system design**. In my opinion, your system doesn't make sense and is **impossible to implement**. But feel free to prove me wrong :-) – Bimpelrekkie Jun 19 '20 at 09:21
  • 2
    *All of my components have to fit on a cm^2 board to meet project requirements* Not the first time that "a customer" asks for **the impossible**. That's why **you** need **proof** of what can and cannot be done. That's where an **architecture study** comes in. A starting point is to investigate similar designs, what is used there? What size are they? Why is that so? A customer can easily write down anything that they want, small size, small cost etc. That doesn't mean that what they ask is **possible**. – Bimpelrekkie Jun 19 '20 at 09:26

1 Answers1

2

So is it even possible for an SSR to switch this fast or should I have went with a different topology to deliver the 1000 V?

The design of solid state relays often prevent this. In order to not consume much energy when off they disable the booster required for the high side switching.

You can do this yourself with a high side gate driver and N-channel mosfet. Eg: 1EDN7550.

Jeroen3
  • 21,976
  • 36
  • 73
  • I would probably just stick with an N channel driving a P channel, rather than a whole gate driver – BeB00 Jun 19 '20 at 17:28
  • The boost converter I'm using has a disconnect switch that I could use when not active. – Landon Jun 19 '20 at 19:34