# Very small, very high ratio, boost converters

Discussion in 'Electronic Design' started by Michael, May 23, 2008.

1. ### MichaelGuest

Hi there - I need to build a very small, very high ratio boost
converter. It needs to go on a PCB that can be no greater than 1cm in
width and can be no greater than 1cm in height. Length is more
flexible. It will have a 5VDC input and needs to have a 1KV, 1ma
output. Regulation is not terribly important.

First I thought about just using a normal DC/DC boost converter and
using a HV fet on the output. However, it seems all HV FETs are
massive. The smallest relatively HV FET I could find was the STMicro
STN1NK80Z, which is an 800V (I could probably get away with just an
800V output - but 1KV is my goal) 250mA FET in a SOT-223 (7x6.5mm)
package. However, I noticed that RDSon is ~13 ohms. So since my boost
ratio is about 200, that'd mean I'd have about 200ma on average (if
I'm remembering my maths right) going through the FET, producing a
2.6V drop. So I don't think this FET is even going to be able to
handle my 1ma load. So I'll probably have to go up to a larger FET.
There are more FETs in the D-PAK (AKA TO-252) package available. The
Infineon SPD06N80C3T looks much better 0.9 RDSon. A quick calculation
shows my boost ratio ending up at about 208 once the RDSon is factored
in (assuming the resistance of the inductor is much smaller than
RDSon). So I'd have about 208ma going through the inductor. Now, for
choosing the inductor, I may be completely off in the way I am
calculating its value. This is the equation I came up with:

Einductor = .5LI^2 = output power * (1/ boost ratio) * (1/operating
frequency)

If that is right, if I used a 100uH Sumida CDFH6D28, I'd have to
operate it at 462KHz. I could put a couple of those guys in series, or
find a higher inductance inductor, but already 462KHz seems nearly
reasonable already. That circuit is seeming almost doable, though
driving that big HV FET at that speed will take a decent amount of
power.

However, I remember being told that boost converters with high boost
ratios do not work very well for some reason that escapes me. So then
I started thinking about transformers. However, I can't seem to find
any that are this small. Heck I can't seem to find any reasonable
sources for boost transformers (none of the big three - Digi-Key,
Mouser, and Newark) seem to have any.

So then I started thinking about PZTs (piezo transformers). PZTs are
typically very small with very high ratios - however again I can't
seem to find anybody that sells any.

With all of this in mind - anybody have any recommendations as to a
possible route for me to go? Are my calculations for the boost
converter right? I am not typically a power guy - so I could
definitely be messing some things up here.

Thanks!

-Michael

2. ### Wim LewisGuest

How about a boost to a lower voltage (100-200V?) followed by a Cockroft-
Walton multiplier?

3. ### mikeGuest

Stray capacitance is a killer in this kind of applicaiton.
Can you use that to advantage by resonating the whole structure.
Tap the drive way down on the inductor. Don't need hv fet any more.

Might also be interesting to look into the details of how they
make stun-guns work. 9V in 250KV out.

4. ### Tim WilliamsGuest

Obviously, one step isn't going to cut it. The most that's practical off a
single inductor is, oh, let's say 10:1. The duty cycle is already awful
(ca. 90%). You can get peaks in the 20-50x range, at nil output current
and insane loads to the driving transistor (since it has to handle all that
voltage AND current).

The obvious solution is a transformer, which ends up having the same size
as a single winding (an inductor), more or less. Put your ratio there and
you get much happier duty cycle, current and voltage capacity.

You could probably tinker with a camera photoflash charger. Those are
around 1cc. I don't know how much current they do, and 1.5 to 3V stepped
up to 200-300V is quite a bit lower than 5V going to 1kV, so you'd probably
need a new transformer at least. But sure, a self resonant sort of
inverter, just what you need.

Tim

5. ### Guest

Maybe a CCFL backlight power supply? They're usually small and output
600V-ish.

6. ### MooseFETGuest

You don't say how important but I'm going to guess 10% or so.

[....]
When the duty cycle is large, the upper harmonics are large. You need
an inductor that is not too lossy at all the frequencies that have
much energy in them.

The pass transistor has to handle a large current when on and a high
voltage when off. The same applies to the output diode.
You may need to abuse a transformer intended for audio if you want to
get it from them. I'm thinking of something like a MET-23. I don't
like this idea but I'll put it out there.

The turns ratio is sqrt(1600/3.2)= 22.4 which is a bit low for turning
5V into 1000, but lets push on with thinking about it.

The inductance on the 3.2V side will be more than 3.2 Ohms at 300Hz
so:

3.2 / (2 * pi * 300) = 1.7mH Lets use 3mH below.

We can assume that the inductance is more than this. You didn't say
if you were making many of these or just one but I will assume several
so we do need to work with whatever the inductance of any one. At
this point you'd need to but one and measure it to see if the rough
guesses are right.

I'm going to suggest a half bridge series resonant design to make the
secondary voltage hit about 250V or 330V so that only trippling or x4
is needed in the rectifier. Unfortunately, you will end up with quite
a large AC current in the 3.2 Ohm winding when you do this. This
really messes with the desire to keep things small because it pushes
you up into bigger transistors.

You also need a 0.1uF capacitor that can take the high current.

All in all, I'd rather get a transformer intended for the purpose.

Here we hit the "use a PIC" suggestion. A micro can drive the
switching devices and produce the needed timing. A voltage divider
off the 1KV can do the feedback. The micro can also tune the
switching rate onto resonance.

The transformer is rated at 65mW at 300Hz but if we run at, lets say
10KHz, we won't burn it out at a higher power level.

7. ### PaulGuest

....snip! there goes a lot of interesting design stuff!.....
Why build when you can buy? This sounds like an "off-the-shelf"
device, for example: http://www.emcohighvoltage.com . If you acan
build one cheaper and better, then you can put these companies out of
My point is, custom electronics is great for those applications
where nothing exists to satisfy cutomers, but if there are a host of
products already out there, why go through the misery of design,
testing, electrocution, and customer unhappiness?
I used to work in a research outfit, and it used to drive us
electronics guys nuts when labs would want to build stuff (wasting
weeks and months of time), when they could buy a far superior thing
from an electronics supplier. Why would you want to waste your
precious time on something that highly skilled people have already
perfected, and reduced the costs by economies of scale?