Linear Vs Switching Regulation.

[Overclocked]

I have this big transformer capable of putting out 12Vrms@14Amps (168W) (It also has other windings, but I really dont need them). I was going to use a Linear regulator (LM317) Mounted on a big heatsink. Now Im starting to think there has to be a more efficient way of regulating Voltage. I believe I asked about this before (well Not THIS question) but I was told that if linear regs were to be used, I would only be able to use 85% of that current (and along with a lot of heat).

For example would it be a wiser choice to use a switching Semiconductor such as this one (Im not really going to use it, Im just using it for example sake)
**broken link removed**

1)Would there be any benefit in using a switching regulator vs a linear reg? 2)Would I still need Huge Caps to transform the pulsating DC from the rectifier to DC? I noticed on the inputs of the Regulator it says "Unregulated 7 to 40V"
3) Would I be able to use all 14Amps?
4) Since it is switching, would there be less heat output?

 

[Papabravo]

You're really in an area where answers to all your questions depend on too many other variables. Linear regulators are simple and straightforward. You seem to have a grasp of the issues.

At those currents choosing inductor cores that won't saturate, wire to wind the inductors that can hnadle twice the peak output current, MOSFET switches and Fast Recovery Rectifiers that can handle the load are really really big challenges. The guys who know how to do that stuff get paid really really big bucks. If you going to teach yourself about switching regulators start with something more modest.

You can still use your honkin transformer but at way less than it's maximum capability until you have a design or two under your belt.

I stronly recomment that you download a free copy of LTSpice/SwitcherCadIII from Linear Technology. Besides cranking out reference designs from a few inputs you can simulate them and ask what if questions, without the risk of toasting the parts or injuring yourself.

 

[Overclocked]

You're really in an area where answers to all your questions depend on too many other variables. Linear regulators are simple and straightforward. You seem to have a grasp of the issues.

At those currents choosing inductor cores that won't saturate, wire to wind the inductors that can hnadle twice the peak output current, MOSFET switches and Fast Recovery Rectifiers that can handle the load are really really big challenges. The guys who know how to do that stuff get paid really really big bucks. If you going to teach yourself about switching regulators start with something more modest.

You can still use your honkin transformer but at way less than it's maximum capability until you have a design or two under your belt.

I stronly recomment that you download a free copy of LTSpice/SwitcherCadIII from Linear Technology. Besides cranking out reference designs from a few inputs you can simulate them and ask what if questions, without the risk of toasting the parts or injuring yourself.

Ive always wanted to make my Own SMPS, but I dont have the knowledge of how they work. Im hoping that college will answer that question, and just not introduce Linear Regs (like in High school). Ive been to the library a couple times, and they have a VERY good resource on practically everything electronic (amongst other subjects...), better than anything On the net I have seen.

However, since I dont know how to design my own, I was going to just look up a Switching regulator thats been built into a IC.

Such as the program youve suggested, Im in it now and just experimenting (along with reading data sheets and such..)

 

[philba]

what are you going to use this PS for?

 

[Hero999]

Your filter capacitor doesn't have to be that big providing the minimunm ripple voltage is higher than the dropout voltage.

Normaly the formula for calculating filter capacitor size vs ripple is easy for linear regulators but it's much harder for switching regulators because the current actually increases when the voltage drops.

[latex]C = \frac{I_{OUT}}{2FV_{RIPPLE}}[/latex]

Perhaps you should double the value but I don't really know, I think you should experiment a bit.

Core saturation souldn't be a problem if you use the right core, if you don't like doing the calculations and winding your own then you can buy pre-made inductors that will work as expected providing they're used within the current and frequency ratings.

I thought most MOSFETs were pretty fast, I've heard about people using the IRF540 in high powered RF amplifiers above 5MHz at reasonable efficiencies, I don't see how typical switching frequencies between 50kHz to 500kHz can be that much of a problem. You can buy some specialised ICs that make all this stuff pretty easy.

By the way 168W is pretty small for a transformer.

 

[Papabravo]

Core saturation souldn't be a problem if you use the right core, if you don't like doing the calculations and winding your own then you can buy pre-made inductors that will work as expected providing they're used within the current and frequency ratings.

MY point was that finding a core and winding an inductor to handle 28 Amps(2 x Ipeak) is a bit of a challenge. Finding a suitable one off the shelf is probably difficult as well. I could be wrong however, I do medium power SMPS with some regularity, but I've never done one that beefy.

 

[Hero999]

Neither have I but I've seen some very large SMPs and they're nothing special, the inductors and switching transistors are just bigger. You can use a higher frequency to keep the core size down but that will reduce the efficiency due to the switching losses.

High power cores aren't that hard to find, a ferrite core from an old computer SMPs is perfect for currents this high. Even Craplins used to sell ferrite transformer cores suitable for use at over 500W but that was many years ago.

 

[Overclocked]

But wait, Arent the Switching ICs rated for a certain current? In the data sheet they even specify What type of Inductor to use. How Come Doughnut shaped (Yes, I know theres a proper name for them, but I cant spell )inductors are used instead of resistor shaped inductors?

I know for a fact that For the current I need, those types of switching regs dont exsist. The most I have seen a switching reg handle is 4 Amps. I figured I could parallel them all to get more current.

 

[Papabravo]

No. What you need to do is use an SMPS IC with an external switch. Read the application notes carefully to see which ones are designed for that function. I don't know of any example where SMPS are used in parallel.

The donut shaped inductors are called toroids

 

[Hero999]

There's normally an option to add a MOSFET on the output. You can also buy SMPs controllers which actually require a MOSFET to work they have the advantage of being more flexible than the sort with a built in MOSFET.

Toroidal inductors are often used because the feild leakage is negligible.

 

[Overclocked]

No. What you need to do is use an SMPS IC with an external switch. Read the application notes carefully to see which ones are designed for that function. I don't know of any example where SMPS are used in parallel.

The donut shaped inductors are called toroids

Do you mean something like this?
**broken link removed**

 

[Overclocked]

To help me out a bit more, I bought a book.
http://search.barnesandnoble.com/booksearch/isbninquiry.asp?EAN=9780750673297&x=40212612

I had a look inside via Amazon, and it does look pretty intresting..

 

[hjames]

Scarily enough, these wattages are in the range of recent PC motherboards - they'll convert the 12V from the power supply into a 1V supply for the processors at close to 100Amps. Of course they tend to be multiphase monsters...

 

[Overclocked]

Scarily enough, these wattages are in the range of recent PC motherboards - they'll convert the 12V from the power supply into a 1V supply for the processors at close to 100Amps. Of course they tend to be multiphase monsters...

Err.. 1.5V Depending on your Proc, and its closer to 75Amps. Thats why theres such a high current on the 3.3V and 5V Rails. Thankfully, because of Cool and Quiet tech (by AMD- I think theres another by Intel) and other tech, the Proc Lowers how much power it draws by adjusting the clock cycle.

 

[hjames]

I think the current king-of-the-hill in terms of power requirements is the Intel quadcore:
ftp://download.intel.com/design/processor/designex/31559401.pdf
page 18: 1.1V, 125A, 130W design power.

I think some of the older Pentium D's were pretty close to this as well.

They stopped drawing power from the 5V lines about when the pentium 4's showed up. The last motherboard I had that had that configuration was a AMD dual socket one - the main failure mode of that board was for the 5V pin on the power connector to melt (2x 50W@5V through 2-3 pins just isn't healthy).

In any case, there's another possible architecture: the Aglient power supply on my desk is a linear power supply with a "pre-regulator" sitting on the AC side of the transformer. Essentially a triac based circuit coarsely regulates the voltage on the filter capacitor a bit above the regulated output, and a beefy pass transistor does the final regulation.

edit: Hmm, I think I'm misremebering the operating manual for some other power supply:
**broken link removed**
The operating manual for the E361xA power supply has some schematics around pg 34, It uses some SCR's to switch in secondary windings as needed.

 

[Hero999]

the Aglient power supply on my desk is a linear power supply with a "pre-regulator" sitting on the AC side of the transformer. Essentially a triac based circuit coarsely regulates the voltage on the filter capacitor a bit above the regulated output, and a beefy pass transistor does the final regulation.

This is called a hybrid regulator and is very common in compact high power LAB power supplies. I doubt it uses a triac switching power supply (thouse normally aren't isolated and are only used for very low power levels, it probably uses a normal switching regulator simular to that used in a PC PSU followed by a linear regulator with a very low dropout voltage.

 

[Overclocked]

I think the current king-of-the-hill in terms of power requirements is the Intel quadcore:
ftp://download.intel.com/design/processor/designex/31559401.pdf
page 18: 1.1V, 125A, 130W design power.

I think some of the older Pentium D's were pretty close to this as well.

They stopped drawing power from the 5V lines about when the pentium 4's showed up. The last motherboard I had that had that configuration was a AMD dual socket one - the main failure mode of that board was for the 5V pin on the power connector to melt (2x 50W@5V through 2-3 pins just isn't healthy).

The fact a Proc uses a lot of power really isnt a good thing. Im hoping this Next Quarter (Q1, 2007) that procs wont need 500W+ To use SLI or Xfire (or Even Run the dam thing for that fact..).

 

[Overclocked]

Alright, Ive been reading around for a bit, but now Im stuck. Ive designed my PSU to be 15V, 10Amps. Ive calculated pretty much everything, but now Im stuck on which Controller to get. I Used 90VAC to 130VAC as My Input Voltage (I want it to plug into the wall..). I have everything calculated around that, however I am finding that most controllers have a Max Input of 40V DC. Ive seen the circuit they use in the book (see attached Circuit)

On one circuit they have 900uH for the Inductors and 150uF for the output caps (same circuit, just those 2 things are different). The book doesnt explain what the circuit does. Does it Regulate the Input voltage for the Controller? Or is it a safety thing?

BTW, Ive decided to use a buck converter, I wouldnt want to use a Flyback converter since I would have to wind my own transformer. I want to try to keep things Simple.

ADD: Those Inductors Have cores On them even though my drawing doesnt show it.

 

[Nigel Goodwin]

It's just a mains rectifier circuit - that's all - with UK 230V/240V mains you would get around 340V on the output. The inductor is just part of the mains suppressor circuit - along with the capacitors. It reduces intereference from the following switchmode circuit - and also reduces external interference entering the PSU.

 

[Hero999]

most controllers have a Max Input of 40V DC.

The controller is normally powered from a simple zenner and resistor regulator but if you need more current then build a very small power supply using a small 1.2VA mains transformer, rectifier and smoother capacitor.