1KW Variable SMPS University Project

Thread Starter

matt09

Joined Oct 26, 2012
2
Hi all,

I wish to build a variable lab style SMPS with output voltage from around 1V up to 100V and variable current from 0-10A. I have quite a lot of experience from doing electronics as a hobby for many years as well as studying the subject. I haven't tried anything like this before though. I am building this as part of a university project so this will actually happen and hopefully this thread will be open the next year or so as I progress.

First of all does anyone know of any existing designs which can be modified to achieve the output I want (I've checked I am allowed to do this). I haven't been able to find any so far?

Failing this I have already been looking into designing my own from scratch. I am struggling in choosing a topology, I know flyback is more suited to variable voltage but not suitable for high power and ideally I want to achieve 1KW or at least 500W. At the moment I am going with half bridge but could anyone tell me if this is good idea for variable voltage? I need to have isolation as it will be mains powered.

I have various ideas so far on how to achieve such a wide voltage range from an SMPS with pulse width limitations. So far I am looking at doing one of the following after reaching my lower voltage limit, presumably of 5-10V.

1. Have a second converter running in series off the output of the first which will rechop the waveform a second time and easily reach down to 0V

2. The same about but have two running in parallel.

3. Have a linear regulator stage running from the SMPS secondary which cuts in at a given voltage which will then go down to almost 0V

4. Lower the switching frequency once a given voltage is reached so that a higher mark-space ratio can be realised.

5. Feed the disable pin of a driver chip with a clock signal so that i get a duty cycle of the duty cycle, effectively missing out some pulses.

If anyone could advise which direction to go on with this element that would be great.

I am also struggling with choosing a base switching frequency as I want it to be as low as possible so that I can get the narrowest possible pulses, could anyone advise a starting point?

I also have no idea if any of the 5 suggestions will be needed? It has been suggested that it will not be possible to achieve a 1% duty cycle to give a down to nearly 0V at any reasonable frequency. From my initial calculations this seems likely as I would need a td + tr of around 100nS at a low 40KHz switching freq. There are many Mosfets around this speed or faster, but I'm assuming I would never be able to reach this and get proper pulses in reality?

I haven't looked into it yet but a head start on the best way to have a variable current limit from 0-10A would be great.

I'd really appreciate any help as these questions are really holding me up from making progress with calculations and have done for some time. I probably already have most of the components and a suitable core from being a hobbyist for so many years but I will get on to this later.

As the post is so long I have made questions bold to make it much more readable.

Many Thanks,

Matt.
 

R!f@@

Joined Apr 2, 2009
9,918
Man tht is some project. :D

Wish I could help you but designing SMPS is way out of my league.
I did made a 0V to 40V with 10A CC and CV linear one though..!

I don't think a single stage SMPS can go down to 0V. And to be variable is another question.

Like u said. With a SMPS of 120V DC at 10 amps out put is possible. And to it u can use a linear CC and CV circuitry.

And to limit power dissipation, u can adjust the out put of SMPS. Like say switchable out puts. Say at 20V steps. When the linear output goes down to 15V the SMPS output will switch down to 20V. This way the linear power dissipation is minimized. And if the out put voltage is increased the SMPS output can be tracked with it, giving a linear drop out margin.

Hey....tht is some Idea right. I think I want one of the type too for my bench.
:D
 
Last edited:

bountyhunter

Joined Sep 7, 2009
2,512
Hi all,

I wish to build a variable lab style SMPS with output voltage from around 1V up to 100V and variable current from 0-10A. ///

Failing this I have already been looking into designing my own from scratch. I am struggling in choosing a topology, I know flyback is more suited to variable voltage but not suitable for high power and ideally I want to achieve 1KW or at least 500W.///

If anyone could advise which direction to go on with this element that would be great.

I am also struggling with choosing a base switching frequency as I want it to be as low as possible so that I can get the narrowest possible pulses, could anyone advise a starting point?
The best topology would be a H-bridge converter using four FET's chopping an unregulated 300V derived from a FWB run from the 220V line.

Best switching frequency is dictated by switching losses and core losses in the magnetics. 50 kHz would be a good starting point for this design.

I would be remiss if I did not point out: this would be an extremely difficult project for an experienced power supply designer. I know, I designed the 1kW rack mount supplies sold by Power 10 in the late 80's.

For someone with no power supply design experience this project would be effectively impossible, ie have such a massive and long learning curve as to not be viable.

First of all does anyone know of any existing designs which can be modified to achieve the output I want (I've checked I am allowed to do this). I haven't been able to find any so far?
I did not save any of the schematics from our designs, but they were included in the user manuals. You could try to get a manual from an old Power 10 (1kW) power supply.

good luck
 
Last edited:

Thread Starter

matt09

Joined Oct 26, 2012
2
Thanks for the responses, Bounty Hunter I am aiming high to begin with, if necessary the project could be stripped right down to just a 20-100V output and end there with no other controls, but I want to go far beyond that. Thanks for the links, I will add those to my library of stuff so far.

Thanks for the atx designs but I need to still make my own circuit board, I think that would definitely be cheating!

It has been suggested I could use two secondary windings and switch between them where the ratio is much lower to achieve the last 0-20 volts or so. Does this seem better than the switched frequency method also suggested (I'm not sure there are any driver chips with a variable frequency?)

Thanks
 

bountyhunter

Joined Sep 7, 2009
2,512
It has been suggested I could use two secondary windings and switch between them where the ratio is much lower to achieve the last 0-20 volts or so. Does this seem better than the switched frequency method also suggested (I'm not sure there are any driver chips with a variable frequency?)

Thanks
I don't think a switched freq design is viable for 1kW ballpark. I think you need fixed frequency PWM converter (H bridge).
 

bountyhunter

Joined Sep 7, 2009
2,512
It has been suggested I could use two secondary windings and switch between them where the ratio is much lower to achieve the last 0-20 volts or so.
You're not getting advice from a knowledgable source. That is a good technique on LINEAR designs with variable output to minimize power dissipation (used in most linear lab supplies). In a switcher, it really doesn't help.
 

catt87

Joined Nov 18, 2010
2
My current project I'm working on is also a variable switchmode power supply, except it's somewhere in the 300 watt area. Matt, for 1000 Watt, a full-bridge topology is what it takes. In my design I plan to use a half-bridge, with a balancing winding, parallel to the primary winding and with the same number of turns.

I plan to run my supply in current mode, and for that I chose the UC3825, but the UC3846 could also do the job. I think it's pointless to tell you that there is little material to go by on the internet, concerning adjustable SMPS, especially down 1 digit volts. Currently I'm stuck at the main power transformer design stage, because there is almost no literature that tells you exactly how to calculate the darn thing or exactly what are the effects of trying to get low voltages (<10V @5 Amps for example) out of it, so I think I'm gonna bite the bullet, calculate the transformer for the maximum voltage and current I want (that's 50 V / 5A) then start bread boarding the contraption and see how it behaves.

Concerning the adjustable frequency idea, it's not all that crazy. I thought of it as a method, and you could do it with any IC, as long as you see the "CT" pin on it :D Basically, instead of a capacitor, you could feed it your own ramp, of any frequency you want (well, ok, in reasonable margins, of course)

I don't know if you've come across this or not, but there are a few schematics for the Sorensen supplyes:
http://www.diysmps.com/forums/showthread.php?319-Lab-SMPS-help&highlight=sorensen

The control method isn't all that complicated, except there's a lot of circuit to comb thru to find the basic blocks that you need.

Also, if anyone has any pointers on how I should start designing the power transformer, I's really appreciate the help.
 

MarpK

Joined Jan 13, 2012
1
You're not getting advice from a knowledgable source. That is a good technique on LINEAR designs with variable output to minimize power dissipation (used in most linear lab supplies). In a switcher, it really doesn't help.
Can you please elaborate on this? I was thinking about multiple secondary windings as well. I figured that if the voltage to be converted is lower to start with, that the resolution would be greater. Like if I used an 8-bit ADC to sample the output voltage, my resolution with a 100 volt supply would be 100/(2^8) = 0.390625 volts, whereas the resolution with a 20 volt supply would be 20/(2^8) = 0.078125 volts. Wouldn't this allow the last 20 volts to be more precisely regulated? Or is there some further complication with the smoothing circuitry that I'm not thinking of?

It's funny, I was thinking about building a nice supply myself with a big 1.4KW microwave oven transformer. I thought it'd be an easy project. Hehe, I guess not...
 
Top