Guidlines/specifications for a good linear power supply? (PDFs, Links, Books)

[TheLaw]

Maybe some of you know my ongoing power supply ideas. It seems that I figure out one thing and then I'm stuck on another.

To ease these troubles, I was wondering if anyone know of some resources that explain how to put together a good power supply. Guidlines, information, etc.

I've been searching for a while now with limited success.

Maybe my search terms are off.

Any ideas are appreciated. Thanks.



PS This is what I've found so far: http://sound.westhost.com/power-supplies.htm

It's written up nicely, but I know most of it. I'm looking for how to optimize a power supply... How to bring it to the next level...

Thanks again.

 

[TheLaw]

Well now that I looked at some switching power supply stuff and it doesn't seem to complicated...That is for something simple. Like a 3A 5V supply.

Like 5 components? (Minimum)

Hmm....

 

[(*steve*)]

Lots more can go wrong with a switch-mode power supply, but having said that there are some chips that are not much harder to use than three-terminal regulators.

When I say "not much harder" I refer mostly to complexity. Layout is more critical.

A colleague of mine swears by the LM2576HVT-ADJ (you don't necessarily need the HV version, or the adjustable one)

If you choose one of these (and I mean *any* switchmode regulator IC), read the application notes very carefully as you need to choose the correct inductor (most important) and the correct diode (also most important) and the right capacitor(s) (another most important issue) and layout (most important) and also be aware of ripple levels at various load levels.

For the fixed voltage version of some of these chips, yes, 5 components is all you require.

edit: Here is the type of information you'll find in a datasheet. You need one (like this one) which tells you how to select all the components. Note that it even tells you that too low an ESR in the output capacitor can cause problems, and what a minimum value for this should be.

 

[TheLaw]

Lots more can go wrong with a switch-mode power supply, but having said that there are some chips that are not much harder to use than three-terminal regulators.

When I say "not much harder" I refer mostly to complexity. Layout is more critical.

A colleague of mine swears by the LM2576HVT-ADJ (you don't necessarily need the HV version, or the adjustable one)

If you choose one of these (and I mean *any* switchmode regulator IC), read the application notes very carefully as you need to choose the correct inductor (most important) and the correct diode (also most important) and the right capacitor(s) (another most important issue) and layout (most important) and also be aware of ripple levels at various load levels.

For the fixed voltage version of some of these chips, yes, 5 components is all you require.

edit: Here is the type of information you'll find in a datasheet. You need one (like this one) which tells you how to select all the components. Note that it even tells you that too low an ESR in the output capacitor can cause problems, and what a minimum value for this should be.

Yes the 2576-5 is precisely the one I was looking at. The schematic had a grand total of I think 5 components. Hmm...

Do all of the components have to be precisely what is in the schematic.

It was my accustomed belief that capacitor values never really mattered on power supplies as long as they aren't too low, or ridiculously high.

Diodes..Schottky only: Just make sure the ratings are appropriate....

Layout: As tight as possible I presume.

Inductors: Do they have to be exact?


Well I appreciate the information!

 

[(*steve*)]

Do all of the components have to be precisely what is in the schematic.

No, in fact they may need to be quite different! They may also be allowed to be quite different.

The component values will differ depending on the voltages (i.e. both of them) and output current.

The datasheet will answer this question (or at lease give you recommended values)

It was my accustomed belief that capacitor values never really mattered on power supplies as long as they aren't too low, or ridiculously high.

As a general rule, that's fine. It's just that "too low" and "ridiculously high" are not exactly quantative values.

The datasheet will answer this question.

Diodes..Schottky only: Just make sure the ratings are appropriate....

There are ratings for the diode that you need to consider.

wait for it...

The datasheet will answer these questions.

You need a fast diode too. Not all Schottky diodes will be fast enough.

Layout: As tight as possible I presume.

There are some leads which must be kept short, and some earthing must be done in a particular way.

But.

You guessed it...

The datasheet will answer this question.

Inductors: Do they have to be exact?

The circuit will operate with out of spec inductors quite well, but you may get one or more of the following:

1) instability
2) excess ripple
3) poor regulation
4) limited maximum current

In general, you will find that careful reading will result in...

The datasheet answering this question

Well I appreciate the information!

I guess you might note that I was suggesting a source of this information..

Datasheets are something you need to learn to read. After you've read a few and figured out how the writers of these things tend to put them together you will be able to find all sorts of answers to questions you have probably never even thought to ask before.

Have a look at the datasheet link I gave you and see if you can find the information. If you can't find it, ask me and I'll try to find it for you.

Maybe the datasheet can't answer all the questions. The one I pointed you to contained a full application note as well, many do not, and some can verge one the woeful. Indeed for some parts you may be forced to look at the datasheet from other manufacturers (and then hope they apply)

This post brought to you by the saying: "Give a man a fish and he will eat for a day. Teach a man to fish and he will eat for a lifetime."

 

[TheLaw]

Well I feel like I've asked so many questions and that there must be a simpler, less annoying way to get information...

Datasheets it is!

I'm actually considering buying a book on switching power supplies. I really want to get into power delivery but I'm still treading water here.

Thanks.

 

[TheLaw]

I have to comment on how friggin right you were about the datasheets. Whenever I used to go through them, I'd skip through 90% of it and think that it was all unimportant. But now as I read carefully, there's a ton of information.

Thanks for the splash of cold water.

 

[(*steve*)]

One of the very first books I bought was "CMOS Integrated Circuits" by National (National Semiconductor). It is dated Mar 1975, and it cost me what I thought was a small fortune at the time as I recall.

As an indication how old it is, it lists several, now long obsolete 74C chips as "Advance Information". Indeed several of those chips did not appear on the market until some time after I bought this volume.

I am reasonably certain that I have kept no other books I bought at that time in my life.

These days it is much easier to collect datasheets from the web. I make it a practice to collect datasheets (wherever possible) for any part I purchase, and certainly for any part that I consider in a design. And that goes all the way down to resistors.

Capacitor datasheets can be interesting because they'll tell you about variation with voltage, temperature, and frequency -- things you might not suspect will alter the capacitance. You'll learn what things like X7R and NPO mean for you. Importantly, you might learn where you might actually need to look at the datasheets before selecting a component.

Resistor datasheets are somewhat less interesting, however they tell you about stuff that many people don't realise -- that resistors have voltage ratings. For some value resistors in any series, the limit on dissipation is not based on thermal characteristics, but electrical!

 

[TheLaw]

They are quite useful to say the least. But now I have a real question. One that I could not find a real answer to even after scouring the datasheets...

Maybe it is do to my lack of knowledge in the field or maybe not.

The datasheet for the switching reg that I am looking at: http://www.onsemi.com/pub_link/Collateral/LM2575-D.PDF

If you scroll to page 22, Figure 33 shows an example circuit for an adjustable power supply. Note it has an "Optional Output Ripple Filter".

On all of the datasheets from about 5 different manufacturers of the same chip, only the adjustable voltage version included the ripple filter.

However, I would like to use the fixed version. (Figure 35 on page 23). Is there a reason why it is only showed in the adjustable version? I would like to have low ripple values. As low as possible...

Can I just plop in the extra inductor and capacitor in the fixed version?

Thanks.

I appreciate it.

 

[(*steve*)]

I can see no reason why you can't add it.

But remember it is a filter, and like all filters it has effects other than the one you want.

For example, although it will help remove the high frequency ripple, it will also diminish the high frequency response of the regulator. So whilst you will have less ripple, you will have a regulator with a poorer transient response to changes in load current.

And here is a problem with a general purpose power supply. It has to drive an unknown load. Power supply design is typically (perhaps ideally) done hand-in-hand with the design of the circuit to be powered.

edit: The bottom of the left column on page 18 is the written explanation of the additional filter. Since it doesn't say it's only for adjustable outputs, you can presume it's not (i.e. you can use it with anything). Note that it is probably most useful with an adjustable output since the inductor and capacitor are a compromise -- when designing for a fixed output you can more easily control the amount of ripple you will generate.

Note that this datasheet doesn't go into the mathematical details, it simply provides a set of tables and/or graphs. Other manufacturers and/or other chips may allow you more freedom (at the expense of more work and higher risk). See here for an example of designing a simple DC-DC converter.

 

[TheLaw]

I can see no reason why you can't add it.

But remember it is a filter, and like all filters it has effects other than the one you want.

For example, although it will help remove the high frequency ripple, it will also diminish the high frequency response of the regulator. So whilst you will have less ripple, you will have a regulator with a poorer transient response to changes in load current.

And here is a problem with a general purpose power supply. It has to drive an unknown load. Power supply design is typically (perhaps ideally) done hand-in-hand with the design of the circuit to be powered.

Ah I see your point quite clearly...

Okay let's try to find an application for this...

 

[TheLaw]

So after quarreling with the thought of death by electrocution (by way of mains current), I decided to do something a little less potentially hazardous:
http://www.ladyada.net/make/mintyboost/process.html

Essentially the idea is to use a buck boost type topology to be able to charge 5V USB devices. So, it is pretty straightforward, BUT, when I build an electronics project, I like to put more focus on performance/quality than on being economic. Of course, I won't be throwing in a $25 tantalum capacitor, but rather above average components.

So regarding the component selection. I've read the datasheet thoroughly, and for the main smoothing/input cap, it seems that the value does not have to be exact. It says anywhere from 10uF to 100uF is acceptable. I was curious as to what would happen if say I used a 220uF cap in its place. It should theoretically reduce ripple, but would it have any negative effects?

Switching power supplies are more picky about components than linear, I have determined. But does the value of that capacitor matter?

In addition to that, what effect does the value of the output capacitor have? The more farads the merrier? Or not?

Inductors, I have learned, are generally not to be played with. It seems that their values must be very precise. So that remains the same.

I just got a bulk shipment of components in from Mouser and I'm hoping to actually get a project going after quite some time. (I've spent more time trying to pick out a project than actually building them!)

Merry Christmas/Happy Holidays.

Thanks a lot.