Simple SMPS / General Science and Electronics / Forums

Forums

4hv.org :: Forums :: General Science and Electronics

« Previous topic | Next topic »

Simple SMPS

1 2 next

Move Thread

LAN_403

iJim

Fri Jan 29 2010, 10:56AM

Registered Member #2662 Joined: Fri Jan 29 2010, 10:14AM
Location:
Posts: 36

Hi all,

First time poster, long time reader!

Iâ€™ve built numerous tesla coils, an induction heater, a few RF amplifiersâ€¦ etc, but I always find myself relying on either an expensive bench top PSU or bulky Variac. Iâ€™ve tried numerous times over the past few years to develop a simple DC-DC smps that provides current limiting, a variable DC output, and doesnâ€™t randomly go unstable resulting in destruction of the convertor and usually the expensive load!

Ideally I would like to minimise the parts count so that the design can easily be incorporated on my future projects. To this end I was thinking either a buck or boost convertor, meaning I donâ€™t have to make a transformer, I can simply buy a suitable inductor. Obviously isolation is an issue, but I can live without it for now, especially as many of my projects have an isolation barrier elsewhere.

Initially a power level up to 200W should be fine. An output voltage 0-50V would be great (Iâ€™m not even too bothered about 0V, 12-50V would suffice). It would be nice if the input voltage could be rectified mains, so 340V, but I guess this isnâ€™t going to be possible if a boost topology is used. Finally, as it's just for hobby use the output ripple is not really an issue.

So far I have produced two prototypes, one buck convertor with a 60V input, one boost convertor with 15V input. Neither work, both employ a UC3843BN I-mode PWM chip. I tested both designs using my signal generator to apply a variable duty cycle 50kHz square wave to the MOSFET gate (via an ir2303 in buck convertor case) operation is exactly as expected, hence Iâ€™m confident itâ€™s an issue with the UC3843.

A little research online suggests that the UC3843 is not suitable for buck convertors as its difficult to measure the true inductor current. I also spotted a post saying that the UC3843 cannot be used to give a variable output voltageâ€¦ I donâ€™t understand this, surely by altering the voltage applied to the Vfb pin alters the output duty cycle, it certainly does in the open loop test circuit.

UC3843 datasheet:

If anyone can give me any advice about my current approach or an alternate approach thatâ€™s better Iâ€™d be very grateful!?!

Apologies for the length of the post!

Steve Conner

Fri Jan 29 2010, 11:03AM

Registered Member #30 Joined: Fri Feb 03 2006, 10:52AM
Location: Glasgow, Scotland
Posts: 6706

No need to apologise for the length, you should have apologised for the lack of background research instead.

Use the search function, lots and lots of people here have built and modified SMPS. I prefer to modify existing ones, I've done two, and they can all be modded for variable output voltage, no matter what the controller chip. Some guy got variable 0-48V from a PC power supply once.

Also read all of these links:

Some of the UC384x series have duty cycle limited to 50%. This makes them unsuitable for driving buck converters, unless you never want more than 50% of the input voltage. However, peak current feedback is a good thing, it makes the power supply better performing and more stable than something based on a TL494.

iJim

Fri Jan 29 2010, 11:29AM

Registered Member #2662 Joined: Fri Jan 29 2010, 10:14AM
Location:
Posts: 36

Hi Steve,

Thanks for the reply and links, I must admit my search of previous posts was a little half-hearted, there are lots of posts about switch mode power supplies, but I didn't really see anything that ticked all the boxes... mainly a simple device, definitely no transformer, and a limited component count, I want to achieve this so I can squeeze the circuit on to many of my existing PCB designs thus giving me a variable input voltage without the need for a bench PSU. Hacking a cheap PC supply would be a great way, but its rather a bulky solution. I've made linear psu's before, they do the job, but the heatsinking is a nightmare, especially at low ouput voltages and high current!

I will try draw/post a schematic of what i've done so far, its possibly just a simple mistake that I keep making which prevents things from working. Do you have any experience with the UC384x series and could tell me if what I am attempting (ie variable output voltage) is fundamentally flawed and i'm wasting my time?

BTW the UC3843 is the 100% duty cycle version.

Cheers.

GeordieBoy

Fri Jan 29 2010, 02:48PM

Registered Member #1232 Joined: Wed Jan 16 2008, 10:53PM
Location: Doon tha Toon!
Posts: 881

For 0-50V @ 4A (200W max) from rectified 230V mains (340VDC max) I would choose one of the following two topologies:

1. Half-bridge converter
2. Two-transistor forward converter (also sometimes called assymetric half-bridge.)

Both of these will include galvanic isolation in the form of a ferrite power transformer as an inherent part of the design.

For the half-bridge you want a dual-ended controller like the TL-494 (although one of the newer "Unitrode" UCxxxx range is better) with the D-type pulse-steering flip-flop built-in.

And for the forward converter you would want a single-ended controller like the UC3843. But you want the version with the 50% duty cycle limit built in that Steve mentioned to ensure magnetic reset of the transformer core.

Current-mode controllers don't directly control the output voltage, they usually control the average inductor current (the one that you said was difficult to sense!) This current drives the load and charges the output reservoir capacitor, so you can indirectly control the output voltage by manipulating the average inductor current. If you fill the output reservoir capacitors faster than the load is emptying them of charge, then the output voltage will increase, or conversely if you decrease the inductor current below the output current, the voltage will decrease. The outer voltage-mode loop essentially implements this control.

Beleive me, a PC power supply is not a bulky solution or expensive solution. There has probably been more design effort put into minimising the size, weight and cost of these than any other power electronics system!

If you absolutely know what you are doing then design a new SMPSU from the ground up, and if you get close to the power density of something on the market these days, then you did a good job. If you are not 100% confident of what you are doing then I'd take something like a PC supply and start modifying it. Things like built-in over-voltage protection and s/c protection will probably catch a lot of your mistakes without blowing something up.

Finally, if you've made something before and it wasn't stable then read up on feedback systems, control theory, stability criteria etc. There is no easy way around this stuff, if you don't compensate the controller properly, then at best it won't regulate very well, and at worst it will go haywire when you least want it to!

-Richie,

Steve Conner

Sat Jan 30 2010, 09:39AM

Registered Member #30 Joined: Fri Feb 03 2006, 10:52AM
Location: Glasgow, Scotland
Posts: 6706

"Full disclosure"... I've never designed anything around the UC384x, although I've repaired and modified SMPS that use it.

The average inductor current in a buck converter is difficult to sense because the inductor is floating. One end is connected to the switch, and the other end is the output voltage, so you need a floating current sensor. Low-voltage controller chips like the LTC1628 have a built-in differential sense amp, so they just use a sense resistor in series with the inductor, at the output capacitor end.

But with a 50V output voltage, and the UC384x series that have a single-ended current sense input referred to ground, it gets more complicated.

GeordieBoy once told me about a trick using two CTs that are allowed to "reset" on opposite half-cycles, and the outputs of them are combined to produce a DC sense voltage. But this stops working at extreme duty cycles.

When all's said and done, how much money do you have and how much time do you want to spend? I picked up two Xantrex 1kW switching power supplies second-hand for the equivalent of about $150 each. A 150V, 7A one and a 600V, 1.2A. They are proper bench power supplies with digital voltage and current readouts, and 10-turn knobs. I probably couldn't have even bought all of the parts to make them for what I paid for them!

Though again, in the interest of full disclosure, they're an old voltage-feedback design that whistles a merry tune at light loads, and suffers from horrific overshoot on load dumps.

iJim

Sat Jan 30 2010, 12:56PM

Registered Member #2662 Joined: Fri Jan 29 2010, 10:14AM
Location:
Posts: 36

Hi guys,

Thanks for the input!

Totally agree about computer PSUs, after years of refinement and optimisation they really are an engineering feat, 100â€™sW for essentially a few pounds! Iâ€™ve also got a pretty fancy bench PSU, all singing, all dancing, 1kW, it did cost Â£500+ though! Still I like to experiment and as switching supplies seem to be my nemesis I shall continue trying!

After reading your comments its probably worthwhile dropping the idea of a buck or boost conertor and go for a more complex design incorporating a transformer. I like the idea of a half-bridge, Iâ€™ve built several fixed output DC-DC convertors using the self oscillating ir2153d. Iâ€™ve seen designs for variable output voltage half bridges but canâ€™t get my head around how they work? How does one use the duty cycle to control the output voltage, doesnâ€™t it have to be fixed to 50% in a half bridge when youâ€™re driving a transformer?

Cheers

Steve Conner

Sat Jan 30 2010, 01:02PM

Registered Member #30 Joined: Fri Feb 03 2006, 10:52AM
Location: Glasgow, Scotland
Posts: 6706

like this!
Scopetrace

It's called "bipolar PWM". When you vary the duty cycle, the widths of the "upward" and "downward" pulses vary identically, so the average value is still zero, and hence it passes through a transformer fine. The above picture was taken from the output of the half-bridge in an ATX power supply.

The two-switch forward converter doesn't use bipolar PWM, though, just the ordinary unipolar kind. It's hard to visualise, I didn't get it until I attached a scope to one and looked at the waveforms. But it works and the transformer won't saturate unless the duty cycle exceeds 50%.

iJim

Sat Jan 30 2010, 01:54PM

Registered Member #2662 Joined: Fri Jan 29 2010, 10:14AM
Location:
Posts: 36

Ahh... it all becomes clear! Thanks! I guess I need a half-bridge driver with independant High and Low side inputs.

I'll have a play around and post my results.

Cheers.

Steve Conner

Sat Jan 30 2010, 02:00PM

Registered Member #30 Joined: Fri Feb 03 2006, 10:52AM
Location: Glasgow, Scotland
Posts: 6706

Well, yes, if you were to use a driver chip, it would need independent inputs.

But the thing to realise is that you can combine the gate drive signals for the two FETs in such a way that the resulting signal looks just like the half-bridge output, so it will also pass through a transformer. Hence no high-side driver chip is needed. It's very common to use a little three- or four-winding gate drive transformer (GDT) to connect a TL494 etc. up to a halfbridge. More info in the TL494 and UC3525 datasheets.

I believe the two-switch converter can also be driven by a GDT, at least, I've seen commercial SMPS that did it.

Peter02

Sat Jan 30 2010, 06:16PM

Registered Member #1488 Joined: Sat May 17 2008, 10:41AM
Location: Germany
Posts: 18

Hi,
when i needed a good powersupply with reasonabe power and stability i decided for a buck converter
because i had not that much experience with working directly from mains voltage
and wanted a conventional transformer to take care of the isolation as well so there would be less dangers for me to bother with at the beginning.
Having a 2.5KW toridal transformer with 60V Secondaries at hand i wanted to build a powerwise fitting buck backend for it.
I wanted the Converter to be sweet-tempered in a way so that i would not have to be afraid of shorting the output or other kinds of abuse.

So i searched internet for possibe control schemes and found this one

After reading it several times and comparing with other control schemes like peak current or voltage mode i decided to give it a try and it worked quite well from the beginning.
The hardest part still was getting the current sense to work, but there are a few good tips in the article on how to do this.
Now the supply i have build doesnt look nice but works very satisfying without any failures for years now although i sometimes use it for stick welding.
It delivers up to 140 Amps at lower output voltage. i did not include a power limiter so it sometimes happens that when drawing arcs while welding when the max voltage and current is set too high
the fuse triggers, but in normal operation this hardly should be a concern.
The controller is build around a tl494 + opamps, made to work as discribed in the article. Switching devices are 4 IXFX180N10 in parallel + Diodes (dont know which at the moment)

Going for a Half-Bridge or similar has lots of advantages and may be the better way but back then i just didnt dare to and i did not really care for the weight.

-Peter

1 2 next

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