Voltage regulator or DC/DC converter

Yes. You'll need a decent heatsink also with the 7805.
Suggest you use a little switching step down regulator such as

Higher current output parts are available also.

Thanks for confirming my suspicions.

I just came across this, which looks like it might be a useful substitute, and still quite reasonably priced.

The LM2575 is not a drop-in substitute -- it is a switching regulator controller, just one component of a switching regulator. You still need to add an inductor, diode, and capacitors around it, and route it properly on a PCB to keep the switching loop tight, etc.

The Pololu part suggested by CrossRoads IS a drop-in solution.

--
Beat707: MIDI drum machine / sequencer / groove-box for Arduino

RuggedCircuits:
The LM2575 is not a drop-in substitute -- it is a switching regulator controller, just one component of a switching regulator. You still need to add an inductor, diode, and capacitors around it, and route it properly on a PCB to keep the switching loop tight, etc.

Understood; I hadn't meant to suggest that it was a drop-in replacement. The data sheet makes it clear what extra components are required.

Are you in US or in Europe? Should add to your profile for better source suggestions.

Good call, I've done that now - I'm in Ireland.

The data sheet makes it clear what extra components are required.

Just be aware that building a DC / DC converter is not simply a matter of putting the parts round it. It requires a PCB with a very critical layout requirement.

Hmm, only 4 parts to support a fixed regulator per datasheet - that looks doable!

Well I have had trained engineers working for me that have failed to get the layout right first time and some even second and third time and they knew what they were doing.

That is pretty good incentive to just buy for a 1-time use.

Grumpy_Mike:
Just be aware that building a DC / DC converter is not simply a matter of putting the parts round it. It requires a PCB with a very critical layout requirement.

The datasheet suggests that a single ground point or ground plane be used - I'm using a ground plane anyway. It recommends keeping certain paths as short as possible - I have all the relevant components within a square of less than 1".

Are there other requirements beyond what the datasheet specifies? I could just buy something and kludge it into the circuit, but part of the reason for designing and building something like this is as a learning exercise.

The datasheet suggests that ......

Yes that is a good starting point and all my engineers read and understood the data sheet and the recommendations. In practice however component placement current paths are crucial, 1" square is not particularly tight for this sort of circuit. Also decoupling into and out of the regulator are important as are a number of other factors like the load capacitance and even the load current. Sometimes you get oscillations at a certain load current, anything above or below is fine.

The point I am making is not that it is imposable to make your own it is just that it is not easy to get it right first time. Better men than me have failed at that.

Grumpy_Mike:
The point I am making is not that it is imposable to make your own it is just that it is not easy to get it right first time. Better men than me have failed at that.

That's discouraging. The implication is that there's no reliable way to design this board and be confident that it will work, other than the empirical (and pricey and time-consuming) approach of building it and seeing if it fails, then trying to figure out why.

Or pay the premium for a pre-packaged DC-DC converter, of course.

oscarBravo:

Grumpy_Mike:
The point I am making is not that it is imposable to make your own it is just that it is not easy to get it right first time. Better men than me have failed at that.

That's discouraging. The implication is that there's no reliable way to design this board and be confident that it will work, other than the empirical (and pricey and time-consuming) approach of building it and seeing if it fails, then trying to figure out why.

Or pay the premium for a pre-packaged DC-DC converter, of course.

This is why electronics is an art as well as a science - well once you are in high frequency domain that is. There is no such thing as a resistor at high frequency, everything has stray capacitance and inductance. At RF frequencies a ceramic capacitor and a plastic film capacitor behave completely differently for instance. Even a simple trace on a PCB is a transmission line with impedance of the order of 100 ohms..

Actually it's still a science so long as you have the facilities for simulating Maxwell's equations for your board layout and components!

The implication is that there's no reliable way to design this board and be confident that it will work

Yes that is what I am saying. Most things you can hack together but switch mode stuff is the one thing that has bitten me consistently through the years. It is a specialist aspect in electronics and unless you want to make loads of them I would suggest a pre made board.

http://products.cui.com/CUI_V7805-500_Datasheet.pdf?fileID=2982

http://products.cui.com/CUI_V7805-1000_Datasheet.pdf?fileID=5694

http://www.recom-international.com/pdf/Innoline/R-78AAxx-0.5_SMD.pdf

If you're using Analog on your board be careful, these are more prone to produce noise and possibly EMI. Highly recommend running a separate LDO (Low DropOut) Linear Voltage Regulator cascaded from your DC-DC Conv to supply your Analog In, if not two, one for Analog, one for Digital.

SEPARATE YOUR ANALOG GROUND PLANE and have it only connect with the rest of the circuit at the power source.

There is voodoo afoot with DC-DC conv, your component placing has to be precise, your inductor just right, built in shielding with these drop ins is worth the trouble.

Let me tell you, sourcing inductors alone is easily worth it to save the hassle. I have found datasheets recommend obsolete inductors. At least you have the specs but still... Voodoo.

Oh, and don't forget to follow Atmel's app notes on putting an inductor or at least a ferrite bead on your AVCC and DVCC.

OscarBravo,

What kind of budget are you looking at for your dc to dc? Do you have limited space for it? I am assuming it needs around 500ma only?

There are lots of new chips that require a minimum of components and are very forgiving to design(as long as you follow good layout techniques as mentioned above). Linear and National are your best bets for this, Ive tried a handful of them and all gave out clean power on the first(mostly second) try. Even the cheap MC34603 works well with recommended application notes. High switching frequencies are the hardest to work with.Pay attention to the input and output caps, If it says it requires 2x 33uf tantalum or ceramic chip use only those.

Roy

Hard to beat this price + free shipment:

http://cgi.ebay.com/MC34063-Voltage-Adapter-Step-Down-I-O-5-24-2-5-12VDC-/250850211128?pt=LH_DefaultDomain_0&hash=item3a67d67538

I've had good luck with several modules from them, but haven't bought this one yet.

Lefty

I'm not badly stuck for space. Price isn't the biggest obstacle, but it's always good to keep costs down.

I'll need the same basic board design to handle 24V, 48V and -48V supplies, either directly plugged in or through PoE. I might also need a 12V version. Rather than re-design the entire circuit around the different PSU requirements, I've designed it with a "daughterboard" arrangement. The individual daughterboards take whatever supply voltage they're designed for, and output +5VDC to power the mainboard as well as an output scaled in the 0-5V range to represent the actual supply voltage (the whole point of this circuit is to monitor the parameters of the power supply).

The daughterboard is 1.2" by 0.9", so there's a little room to play with. The input and output supply pins are arranged at 0.1" spacing so I'll be able to mock up daughterboards on stripboard and test various components before committing to a PCB layout.

As for power consumption, I'll be driving an Atmega328 and a W5100 with a number of fairly low brightness SMT LEDs and (optionally) a 16x2 LCD without backlight. I think I'd rather have the headroom of a full amp available.

When you say -48v do you mean -48v or 48v positive ground(used in telecom systems)? I ran into the same problem where in I needed to design a power supply that can handle 12v, 24v and 48v positive ground and having the built-in ADC measure the voltage. I had grounding problems since I used a bridge rectifier to make the input to the supply always in the correct polarity. I eventually solved the problem without using an external ADC with isolated supply with a little loss in voltage measurement resolution.

I used the LM5007(LM5010 for 1A) from National, total parts cost of the PS is around $3. range is 8-75vin @500ma. Use their webench tool to design/modify the recommended circuit.

Roy