Switch Mode Power Supply Design / High Voltage / Forums

Forums

4hv.org :: Forums :: High Voltage

« Previous topic | Next topic »

Switch Mode Power Supply Design

1 2 3 next

Move Thread

LAN_403

Robert Francis

Fri Apr 04 2014, 05:35AM

Registered Member #2355 Joined: Thu Sept 10 2009, 09:12PM
Location:
Posts: 71

Hi guys,

I am looking to build a switch mode power supply and I was hoping I could get a little help.

Unlike normal SMPS I was hoping I could make mine AC/AC instead of the normal AC/DC or DC/DC but I haven't found anything on AC/AC SMPS.

I chose SMPS because of its ability to run on a high frequency and I want to feed the SMPS into a voltage multiplier which is why I want the SMPS to output AC.

I am unclear how a switch mode transformer works. I found some online but I got the impression they are not wound, and is the rectifier built into them?

I need the SMPS to reach 15kv and 250khz at about 500 watts. It would be powered by 120vac at 4.16 amps.

From what I read the AC from wall outlet goes into a rectifier, lets say full bridge with smoothing cap, from there it goes into some kind of PWM to chop up the DC into the frequency desired. From there it goes into the switch mode transformer where the voltage is increased due to difference between number of primary windings and secondary windings as well as making the chopped DC into an AC signal again.

With that I would just be outputting the signal and feed it into a voltage multiplier to be built at a later time.

I am a noob so I am sure I have missed things here.

Any comments or questions?

Thank you.

Ash Small

Fri Apr 04 2014, 07:00AM

Registered Member #3414 Joined: Sun Nov 14 2010, 05:05PM
Location: UK
Posts: 4245

Hi Robert,

Your description is reasonably accurate and you are on the right track.

While it may be possible to use a half-bridge design, I'm sure most here would suggest a full H-bridge design.

When you say 'they are not wound', that is incorrect. They do have a transformer, and the H-bridge effectively powers it with AC at the frequency of your choice. 250kHz 'may' be a bit ambitious if it's your first project along these lines. I imagine most here would suggest keeping the switching frequency to under 100kHz, The one I'm working on is designed for 25kHz, but anywhere within the range 20kHz to 100kHz should be reasonably straightforward.

You can google 'H bridge' for a general idea of the layout.

You should also find quite a few designs here that use a TL494 or similar controller, a gate drive transformer and four mosfets.

This should give you some idea of what to look for, google, etc.

I'm sure you'l have a few more questions, though.

EDIT: you will need some pretty hefty ferrite cores, to keep the 'volts per turn' fairly high, so that you don't have too many turns on the secondary, as insulation will be an issue here, as will stray inductance and capacitance.

Robert Francis

Sat Apr 05 2014, 05:48AM

Registered Member #2355 Joined: Thu Sept 10 2009, 09:12PM
Location:
Posts: 71

Hi Ash,

Thanks for the suggestions. I will do some more research on h-bridge design.

I will also have to read over several times the article on gate drive transformers in the wiki thanks for that heads up, I hadn't even heard of that part before.

I'm sure I will have more questions too, I crunched the math real quick on turns ratio for 15kv and 170 RMS in and it was about 90 turns on the secondary for just one turn on the primary and one turn doesn't sound like it is enough.

Thanks for the leads I will do some more digging and hopefully start to mock this thing up.

Ash Small

Sat Apr 05 2014, 11:36AM

Registered Member #3414 Joined: Sun Nov 14 2010, 05:05PM
Location: UK
Posts: 4245

There are a few tricks I can reccomend (these HV SMPS's are a bit of a 'black art'), You 'may' decide to use a 'voltage doubler' before the rectifier for the mains, as this will give ~340, so the turns ratio will only be ~1:45, although you will need twice as much ferrite in the core i you decide to do this.

You will probably eventually decide to use a few 'pancake secondaries', with the primary coils between them (I'll find a link for you later, and you may well decide to use either 4 or 8 ferrite 'C' cores, rather than 2. I'll find a link for this later too), then there is the problem of insulation, You will probably decide to immerse the actual transformer bit under oil, although there are other alternatives. We can provide some advice here later, as the design develops.

Then there is the question of frequency. This 'must' be less than the resonant frequenc of the 'leakage inductance' and the 'stray capacitance', if I remember correctly.

I can provide some links as you deal with these issues. You'll probably have a few other questions as well. Good luck with the project.

Robert Francis

Sun Apr 06 2014, 05:50AM

Registered Member #2355 Joined: Thu Sept 10 2009, 09:12PM
Location:
Posts: 71

For the voltage doubler, would you recommend just picking up a step up transformer that inputs 110/120 and outputs 220/240?

Or do you mean buying an internal step-up transformer?

I was originally going to go with a low pf capacitor in the voltage multiplier attachment but if I use a nf capacitor instead I found I could get away with a considerably lower frequency for the SMPS, 60khz. Is this much more reasonable?

I still need the 15kv output for the SMPS though.

Would I go to a company like this to get the switch mode transformer made, basically give them my specs and request a quote?

Thanks

Ash Small

Sun Apr 06 2014, 11:52AM

Registered Member #3414 Joined: Sun Nov 14 2010, 05:05PM
Location: UK
Posts: 4245

You can build a voltage doubler for AC (mains supply) using a capacitor and a diode. Microwave ovens use these. It will also provide a DC output equal to twice the AC input voltage. It's basically the first stage of a voltage multiplier. I'll find a link later if you need one. This will be by far the simplest way to achieve 340V.

There are several reasons why I wouldn't use a company like the one you linked to, especially if you only want one. First, they don't like small orders, second, they don't usually wind stuff for this voltage (15kV), and third, it's not diffucult to do it yourself, if it's designed properly. (We'll get to the actual design later, once we've worked out all the requirements).

First question is 'what will this be used for? What output current do you require?'

I have to go out soon, but I should have more time to go into more detail later on, or tomorrow. Also, when it comes to actually designing the transformer, others here will have some ideas. I'll also find some links later, when I have more time, to other similar projects that other forum members have built in the past.

Ash Small

Mon Apr 07 2014, 03:56PM

Registered Member #3414 Joined: Sun Nov 14 2010, 05:05PM
Location: UK
Posts: 4245

Here's a link to the Wikipedia page on voltage doublers:

There are several different types, and even a quadroupler. Choosing one of the designs that has less ripple on the output would mean that any additional smoothing capacitor would be smaller, or that the output would be smoother still.

Depending on the application, it 'may' be advisable to include a 'mains isolation transformer' as well. This might also be advantageous if you want to increase the voltage even further at this stage (before the SMPS). There are pro's and con's depending on the application, etc.

As I said in my previous post, we need to know the desired output power (in watts) before we can start number crunching, etc.

Ash Small

Tue Apr 08 2014, 12:40PM

Registered Member #3414 Joined: Sun Nov 14 2010, 05:05PM
Location: UK
Posts: 4245

Ok, maybe you're not sure how much power you need?

I'll try and give an example based on the information you've provided so far.

From memory, 18cm^2 of ferrite is good for ~30 volts per turn at ~25kHz, for most ferrites.

If you are using 340V, that will mean around 10 primary turns. In this example we'll use 5 'pancake secondaries', each secondary sandwiched between two primary turns. This gives a total of ~450 secondary turns, spread over 5 pancake secondaries, each with ~90 turns. This arrangement will be good for ~5kW. This will give ~300mA @ 15kV, maximum.

Increasing the frequency means we can use a higher volts per turn OR use less ferrite. using less ferrite generally means less overall power and increasing frequency generally means more 'losses' and less efficiency. Also we must remember that the frequency has to be less than the resonant frequency of the leakage inductance and stray capacitance. The leakage inductance and stray capacitance are difficult to estimate accurately, so we try to design it so that these are minimal, and then measure them.

We need to know more about what you intend to use this for before we can get much further. I hope this gives some idea of what's involved. It's a fairly simple process, but we do need more information.

Can someone check my figures, as all the above is 'off the top of my head', thanks.

Robert Francis

Wed Apr 09 2014, 02:45AM

Registered Member #2355 Joined: Thu Sept 10 2009, 09:12PM
Location:
Posts: 71

Hey Ash,

I was thinking around 500 watts of power. I definitely don't need 5kW of power, even if I did I'd have no way of powering that from a wall socket. I realize you are overbuilding the transformer though.

Are you sure it wouldn't be easier to just have a company build the transformer, give them the specs, 60khz 15kv 500watts and the input voltage? I don't mind paying a little more, even $200 - $300 for the transformer wouldn't bother me, especially since I have a chance of needing more of them in the future.

For the input voltage, would it be best to use a cockcroft walton generator style symmetrical cascade rectifier for maybe 1-2 stages in order to raise the voltage enough to reduce the number of secondary turns needed? If so, the calculator I used showed it to need about 850 uF which is a lot higher than the capacitors I have seen at the 400-500 volt level.

I want to try and build a small modular voltage multiplier with removable stages that plug in like a video game cartridge to sell to colleges that don't have the budgets for a unit like those seen in wiki articles. I figured if I can get it working and it is fairly reliable I could also sell replacement stages and silicone insulating gel for when capacitors or diodes burn out.

I was going to using high dielectric strength epoxy and silicone gel to insulate the components and I have a small vacuum chamber to degas the epoxy.

Ash Small

Wed Apr 09 2014, 11:18AM

Registered Member #3414 Joined: Sun Nov 14 2010, 05:05PM
Location: UK
Posts: 4245

You've hit on a few of the points there, firstly, the the capacitors in the doubler stage need to be large for any real power.

Getting these wound maybe an option in the future, but I suggest that at the 'prototype stage' it's best to experiment a bit yourself. The cost that these companies have is mainly in setting up/programming the winding machines. Once thay are set up they can produce thousands cheaply, but it's much cheaper to wind 'one off's' by hand, or with a simple 'winding machine' that is pretty easy to build.

If we extrapolate from your '500 watt' figure, we get ~30mA output from the SMPS, to be fed into the multiplier. This figure is before losses, so it will in fact be lower. Multipliers are lossy as well, especially if they have lots of stages. 15 stages is about the maximum you can get to before the losses become huge. This would give a theoretical current of ~2mA, but in fact it would be less, maybe as low as one milliamp, at somewhere around 200kV. Even with less stages, you could lose most, or all, of your power in corona, and effectively have no output at all (I can't say for definite, but at these voltages there are losses due to corona, etc). Just having three or four stages shouldn't be too much of a challenge, though, and you could have ~5-10mA output, I guess. (you are more likely to have at least a few mA output)

If you are planning to sell these to schools, etc, then the safety aspect will be important, 20 to 50mA is considerel lethal for kids, so ~30mA is possibly, but unlikely to be lethal (maybe someone else can add something here?). A 'mains isolation transformer 'may' be required as well, although this could possibly be used to increase the voltage prior to the SMPS. Also, 'motor run capacitors' are what I'd use for any doubler, although as you point out the price could still be prohibitive. Electrolytics could be used for smoothing after a rectifier, and bipolar electrolytics 'may' be suitable for a doubler at 50-60Hz, I've not tried it yet, but it does sound feasible.

The next point is, suppose we just decrease the amount of ferrite in the core and up the number of windings. With a core one tenth the size I suggested above, you need ten times the turns I suggested above to maintain the 'volts per turn' thing, and ~4,500HV windings on a small core is pretty much impossible, if you don't want the insulation to fail. Increasing the frequency will reduce the number of windings, but increasing the number of windings also increases 'stray capacitance', and this SMPS 'must' be run at a frequency lower than the resonant frequency of the leakage inductance and stray capacitance.

Maybe you're starting to get an idea of some of the issues involved now?

It's generally accepted that the 'pancake secondaries sandwiched between primary turns' gives the lowest leakage inductance, and 'the fewer the windings and the greater the spacing between them' generally decreases the stray capacitance. It's not quite that simple, though.

As I hinted at above, the method to use here is to choose a core size that will be suitable for the power at the frequency you have in mind, then wind the windings on a bobbin, then measure the leakage inductance and stray capacitance, and check that the resonant frequency of the leakage inductance and stray capacitance is above your intended operating frequency.

This post is long enough already, so I'll stop for now.

1 2 3 next

Moderator(s): Chris Russell, Noelle, Alex, Tesladownunder, Dave Marshall, Dave Billington, Bjørn, Steve Conner, Wolfram, Kizmo, Mads Barnkob