power supplies... why not PWM instead of transformer?

[ozerik]

OK, i have been thinking about this for a long time and still haven't gotten a proper response.

Let me preface: I'm pretty smart, but have no electronics or engineering classes under my belt - so don't hesitate to get into deep stuff, but please explain terms and stuff.

modern switching power supplies somehow increase the Hz of ac current, use transformers to get it close to the desired voltages, rectify, filter, and regulate the current, right? those things hooked to heatsinks in my power supply are mosfet voltage regulators, right? so that's How It's Done, with transformers and transistors (correct me if i'm wrong).

modern radio control speed controllers use PWM (pulse width modulation, right?) circuits to feed the motor little bursts of full-voltage, narrower bursts for less power, wider bursts for more.

OK, here's my question: would it work to pulse width modulate the AC wave to take the place of the transformer in the power supply circuit? it seems to me that light dimmers (the kind in your bathroom or halogen torchier light) use some kind of PWM technology to modify the shape of the AC wave and only feed the light a certain amount of voltage. would it be possible to rectify, filter, and regulate THAT current? I know the filtering would have to be much more thorough since the modified AC wave would be far spikier than a transformed (Hz-multiplied) wave, but it seems to me that transformers are less efficient and more expensive than the cheap-looking circuitry in the light dimmer i disassembled once. not to mention that the dimmer is adjustable and transformers are not(?).

so... why does no one build a power supply like this?

[Tracert]

My first take on the subject looks as though you would have one really bad side effect to PWM.

NOISE.

The amount of filtration and capacitance this type of signal would require in order to clean that type of signal would (most likely) cost much more then the standard power supply you see today.

The light dimmer that you dissassembled does not need to be clean because a light does not really care about a clean signal. The electronics in a PC is an entirely different story.

Clean Powersupply signal means happy pc components.

[larrymoencurly]

Weren't some early switching power supplies phase controlled like a dimmer? But since they switched 60-120 Hz instead of 60 KHz, the transformer (needed for safety isolation) and output filter capacitors had to be just as big as with a linear supply. ATX PSUs PWM high frequency AC.

[Paxmax]

The common PC powersupply is of a design called "Flyback"
Simplified, a PC powersupply, does it this way:

Raw AC is first filtered so it won't spew to much noise back to the line and it's powercord(which would act as an antenna). It's not neccessary, but needed to pass the athorotives legislation for EMC emission.

Then it's rectified and stored temporarily in high voltage caps. They smooth out ripple somewhat and provide headroom for glitches in the AC mains. Again this is needed by legislation and for proper working of the next part, the switching unit. If you don't use caps the voltage would at some points dip towards 0V and the switching unit won't be able to push any current.

So, the switching unit feeds the transformer. Yes, we still need a transformer to provide galvanic isolation between the highly deadly +200 -> +400V DC, otherwize you'd be toasted even if you touch a supposedly "safe" +5V line. Galvanic isolation means getting rid of a pure low resistive(metal) path between the mains and the +5(or whatever). This is done inside the transformer which consists primarily of three parts: A winding which is isolated totally from the Core which only transfers the magnetic induction field to second (also isolated) winding.

Basicly you don't need a transformer but only a coil to do the switching down voltage trick, that method is called "buck" switching. The opposing method of raising the voltage with a coil is called "boost" switching.Todays topic is however "flyback" switching.

The flyback transformer is optimized for much higher frequencys then ordinary 50/60Hz. As maybe said before in this thread, switching frequency used today is usually somewhere between 60KHz and 500KHz. The higher switching frequency, the smaller transformer needed. However at the higher frequencys you gets all sorts of more problems.

The transformer has a primary side with alot of "turns" and a secondary with fewer turns to get lower voltage(but higher current). Since the AC mains is recified we can't just hook it up directly, a transformer only works it's charm with AC.

The input voltage per winding is directly proportionate to the output voltage per winding. So if you had 400 volts on the primary with 400 turns you'd get 1V per turn.... by having 10 turns on the secondary you get a 10V pulse out. That transformer has a ratio of 400/10.

Since it doesn't operate on DC, we simulate AC by letting the switching unit pulse current throuh it. A transformer is somewhat slow to react on changes even though the DC resistance of the winding might be 10 Ohms!. The current will rise slowly(compared to how a resitor would react) even though 400V DC is applied on it's windings.

This part is often misunderstood about the way flyback switching works: You don't "use" the energy(or output voltage) you get when you apply the voltage, you just charge it up and "store" it in the transformer temporarily.
Much like you do when you lift your bicycles frontwheel and spin it by hand. Faster and faster. You are effectivly inputing and storing energy into your wheel
When the switch turns off however, the transformers secondary side instantly changes polarity and unloads, delivering a energy kick which is primarily stored in a cap on the secondary side.
That switcher turning off can be compared to you suddenly pressing the brakes really hard on that spinning front wheel! The energy is moved from the spinning wheel to the bikes frame and it jerks up wildly. You have transformed 2 pounds of rotating mass into 20 pounds of bikeframe, moving it an inch or so off the ground
After that cap there is a sensor circuit sensing if the voltage has reached nominal or not. If it hasn't, it tells the switcher to keep them pulses coming! Each pulse raising the voltage a little bit. When proper voltage is reached the switch unit stops and rests... waiting for the output voltage to drop. Usually there is also a filter coil after the cap which blocks the 60-500KHz pulses and then another cap aswell. That cap serves as a reserve for changing in loads, providing more time for the switch cycle to engage again when the voltage suddenly drops.

There is about 10 times more to tell about the flyback design, most of it consists of all problems that will arise in a switchmode PS. Thats the basics, I know a pic would help, but I'm too dead tired at this moment. Maybe later.. no promises I'm sick'n'ill.

Okey, class dismissed, or, were there any questions?

[Since87]

Very good Paxmax.

In my experience, not many EE's understand that a flyback transformer needs to be thought of as an inductor moreso than a "transformer".

[bluce ree]

I'm an EE who's worked on 5 switching (CLASS D) audio amps and 2 high current switching power supplies (alll fixed voltages.)

PWM does mean pulse-width-modulated.

The application you descirbed as, switching the AC without a step down or isloation transformer is called an "Offline switching power supply." Obviously some 'splaining is necessary:

AC cord------>Bridge Rectifier--------->FilterCaps----->common-mode choke---->regulator circuitry.

you should review the following topics when researching power supplies:


CLASS D or Switch Mode Power supplies: Forward type, Current mode control, Zero voltage switching, HALF BRIDGE DRIVERS, FULL BRIDGE DRIVERS, H-BRIDGE DRIVERS.



1) Offline switching power supplies may have voltages that exceed 384 Volts right after the Bridge rectifier. Remember to use Fuses. Remember to heatsink just about everything, even the bridge rectifier. A cooling fan isn't a bad idea either.

I get really nervous when people ask me power supply questions, because YOU MUST CALCULATE EVERYTHING AND CHECK YOUR WORK CAREFULLY or you may die, and die very quickly, or start an electrical fire. I STRONGLY SUGGEST YOU SPEND ABOUT 1 month researching Offline Switchers, about 1 month calculating everything, then the rest of the time getting parts and building it section, by section. Offline switchers are extremely dangerous because the only thing between your circuitry and the power company is your breaker or fuse box.

When in doubt, calculate it. then run a circuit simulation, then build it. It takes a long time, but when you're dealing with power electronics mistakes can cause fires or be fatal. In the development of all the switch mode CLASS D products I've worked on I've seen each one explode so many times, SOOOO MANNY TIMES.

If you have specific questions, please ask me.

[Crazy Jayhawk]

Transformers aren't really that inefficient. A transformer with 90% efficiency isn't terribly unusual.

[oz2cpu]

Almost all the new ATX PC psu types are made using push-pull, not flyback.
in a flyback converter the "transformator" IS a coil.
in a push-pull converter it is a transformator :-)
Remember what is the goal for the design of a PC PSU;
low cost, low cost and more low cost !
no one specify power loss.
today it is possible to design a PC PSU will almost no loss at all,
it could be made at about 2 cubic inch size and no need for a fan.
But this design will need to use syncronyous retifiers
on the primary side and on all outputs, now all diode loss is almost gone, now use better low loss material for all inductors and transformators, but that is also much more expensive..

[larrymoencurly]

I have a 300W Powmax (Leadman) with only a single MOSFET on the high voltage side. Does that indicate it's a flyback inverter with a transformer?

[bluce ree]

1 MOSFET on the high voltage side, hmm it MAY be a flyback (it sounds like it,) but you may want to take a closer look--it may infact be a PFC circuit, and not the actual power switching circuit (PFC = Power Factor Correction, it's a technique used to increase the efficiency of a power supply.)

Transformers are very inefficient. no offense, but I am NEARLY POSITIVE that 90% efficient transformers DO NOT EXIST for POWER SUPPLIES. Perhaps for communications, like a telephone 600ohm 1:1 transformer, you MAY get 90% depending on the circuit.

Also, high power transformers WILL HEAT UP, greatly reducing the amount of current it can drive.

[larrymoencurly]

I'm pretty sure that this isn't a PFC component because this Powmax was probably the second-cheapest model they made, and it's the only transistor on the high voltage side that's mounted on a heatsink. There's another high voltage transistor, but it sits alone, and I think it's for the standby +5V supply.

[The Overclocker]

Originally posted by bluce ree
Transformers are very inefficient. no offense, but I am NEARLY POSITIVE that 90% efficient transformers DO NOT EXIST for POWER SUPPLIES. Perhaps for communications, like a telephone 600ohm 1:1 transformer, you MAY get 90% depending on the circuit.

Also, high power transformers WILL HEAT UP, greatly reducing the amount of current it can drive.

(scratches head) um...no

i have used many types of transfomers. all of them offer efficancy around 90% - in a computer powersupply, we are talking at about 20Watts of heat given out - not much at all for something that size.

i have some very large transfomers - we are talking about 1000VA in my amplifiers - after a long period of use, they get walm - nothing more

(now the MOSFETS are something different - at the moment i am using four heatsinks with a passive rating of 0.2c/w)

[bluce ree]

...depending on the power supply circuit.

[macawgumbo]

instead of using a pvm psu, how about somebody integrate a ups into a pc case. It would be much simpler and no need for a surge protector or battery backup. Us a mini modified version of an APC battery pack that gives you 60 minutes uptime at 20% load on a 550W psu? Why wouldn't this be the most efficient? I hear no noise coming from my ups?? Plus, it doesn't need much cooling.

[Carnil]

A battery pack that can give 60 minutes of uptime for a desktop computer will either be a lead-acid battery costing around $100 and weighing 40-50 pounds, or NiMH/lithim ion/whatever battery that weighs 2-3 pounds and costs $500.

[larrymoencurly]

I once saw an ancient computer magazine that showed a PC Power & Cooling PSU with a built in UPS. I think the power rating was about 300W.