SMPS Design high Wattage

[craigmillard]

Hi Guys,

I just came across this SMPS design and would like to give it ago but have no idea how to workout the type of transformer used/ specs etc...
The designer says its good for 125W
I Have emailed him for a clue but the site hasnt been updated since 2007 so not hoping to much..
http://www.audiohobbyist.com/diyparts/parts/kit/carpsu.htm
http://www.audiohobbyist.com/diyparts/parts/kit/psu/psuinstruction.htm (schem in PDF at bottom)

Any idea on how i can pick a transformer/ make one?

I know its based off a ETD32/34 core...

Cheers
Craig

[amptramp]

GAH!  Don't follow this schematic!

You will not get 125 watts through R15, the 2K2 in series with the output.  Are you sure it wasn't 1.25 watts?

Most modern inverters have a current limiter with sensing in series with the power FET's, either as current shutdown or in current-mode control.  You still have the problem that if R15 is reduced, you get a huge current flowing from the supply through the diodes because you have no inductance in series with the output diode bridge.  An inductance in series with the bridge requires either current-mode control or an inductor which is actually an autotransformer with the bridge going to a centre tap, the output at one end and the input going to a reverse-biased diode to ground.  Without that, any imnalance will cause the core to bias in one direction until it saturates.  Current-mode control or the tapped inductor provides a control voltage that resets the core.  (I invented the tapped inductor technique in 1976 at Spar Aerospace for an inverter that sat in the Shuttle bay and provided power to a series of Hughes satellites.  Spar wouldn't allow me the two hours to write up a patent disclosire, but TRW saw the design and patented it themselves!  Since it was a classified project, it would be extremely difficult to determine iif anyone was violating the patent or defend against TRW if we used it.)

[Lurco]

R15 is 2R2 !

[craigmillard]

Yup R15 is 2R2

Iv been looking into these SMPS's a bit more and came across this really good guide..
http://tahmidmc.blogspot.co.uk/2013/01/using-sg3525-pwm-controller-explanation.html

Looks like the SG3525 is a good easy to design around chip.. This guy has also a good guide on transformer winding!

http://tahmidmc.blogspot.co.uk/2012/12/ferrite-transformer-turns-calculation.html

What do you think of the design in the first link?

The guide has helped me a lot but im still not sure how you pick your frequency to switch at and why..? the above guide is 50KHz

[amptramp]

My mistake - I thought it was 2K2.  But the other objections are real - you have no short circuit protection, and in this design is the only way to reset the core is via C7.  If it is a high-voltage supply, the 2R2 resistor transforms by the square of the turns ratio into a much smaller resistance in series with the FET's on the primary side.  If you are transforming 12 volts into 120 volts, the turns ratio may be 10:1 or larger and the reflected resistance of the 2R2 resistor will be 20 milliohms or less.  There will be huge current pulses when the FET's switch on and you only have 2000 µF of electrolytic capacitors and nothing working at higher frequencies to contain the high-frequency hash.

There are more recent controllers using current-mode control and better circuit topologies that eliminate two of the problems here - core magnetization, lack of short circuit protection and uncontrolled pulse currents.

[craigmillard]

Cheers amptramp,

Your other pointers are why i started looking around.. the SG3525 linked above i think fixes the issues you have pointed out.. The article was written in Jan this year! Bit newer than the 2004 original one, ha!
Here is the schematic:


Can you explain why a certain frequency is chosen, say 50Khz?

[amptramp]

There is some latitude in choosing frequency, but it involves the tradeoff between efficiency and size.  If you operate at high frequency, the inductors and capacitors will be smaller and the conducted and radiated emissions from the converter will start at a higher frequency, so smaller components can be used in the line filters and the reaction time of the feedback circuit is faster allowing for more rapid response to line and load variations.  But there is an inefficiency caused when the power switching FET's go from low to high and high to low resistance.  There is a fixed loss for each switching event and if you operate at higher frequencies, you have corresponding higher losses.  Some circuit topologies such as resonant converters tend to alleviate the efficiency problem by making switching events occur at either zero voltage or zero current and since this is sine wave conversion, the upper harmonics of the interference generated by the converter are reduced - but the conversion frequency varies and this makes it difficult to design and test line filters for EMI (electromagnetic interference).

There are all kinds of circuit topologies with various properties - one that is seldom used is the current-fed push-pull.  There is no converter IC that handles it.  The nature of the feedback response is that it is very slow due to the right-half plane zero in the s-plane feedback response, meaning the initial effect of the feedback is to reduce the ability of the circuit to track line and load variations.  Sounds like a bad circuit?  For most purposes, it is.  But there is one place where it is ideal - if you have a sudden nuclear event whose gamma emisson causes all semiconductor devices to turn on momentarily, there is no damage and the circuit keeps operating.  A brief clarification of right-half plane zero feedback (otherwise called investment in financial circles):  if you are flying a helicopter and you are losing altitude, you can pull up on the collective control to raise the angle of attack of the rotor blades.  But raising the angle of attack also slows the rotor down, causing more loss of lift and this can become an unstable and unfortunate event very quickly.  To solve this problem, you always have to have a high speed on the rotor and usually, the operating range is only the top 10% of allowed rotor speed.  In contrast, flyback and most normal puish-pull converters have cycle-by-cycle response where all corrections can take place in one cycle.

[amptramp]

Looking at the circuit you have provided, there is still no current limiting or protection for the power switching FET's.  This is used to provide current shutdown in the event of the core getting magnetized (and this circuit has less protection that the first one you linked) since you have no reset voltage other than the difference in diode conduction voltages - not much and nowhere near as much as the small capacitor at the output of the diodes gave in the first circuit.  But more importantly, the most modern converters use current-mode feedback because it allows rapid feedback correction for line and load within one alternation of the push-pull output.

An analogy would be the cruise control on a car.  It normally looks for changes in speed to initiate a response correction.  Now suppose you added an attitude sensor to a cruise control.  Now you could add throttle as soon as you saw you were going uphill rather than waiting for the speed to drop and trying to maintain constant speed from speed measurement alone.  Similarly, it could back off the throttle if it saw it was going downhill before the speed increased.  A current-mode converter determines what is going to happen to the voltage by determining how much current it has to feed the load before having to react to a changing output voltage. Response time is a major hassle with converters - if you are building a converter for an audio power output, you don't want audio frequencies beating with switching response times to provide spurious outputs (like some of the ICL7660 switched-capacitor power supplies).