5V switching IC

I've had good luck with the MC34063. $1.43 from digikey in quantity 1.

--
Regards,
  Bob Monsen

Each problem that I solved became a rule which served afterwards to solve 
other problems.
- Rene Descartes

Why change from the 7805 you can't get much cheaper

Dan

-- Dan Hollands

1120 S Creek Dr Webster NY 14580

A resistor and 5V zener?

You can save 25 or 30% by using a regulator that uses less Iq than a

7805. A TO-220 7805 hardly requires a heatsink at 25mA*19V = 0.475W dissipation.

But, if you insist, the 34063 that was suggested is cheap, but requires a relatively large inductor because of the low frequency and more external components than the 'simple switcher' type. Maybe you can use a cheap radial dipped spool type inductor with relatively high resistance. You could also use a switching diode such as a 1N4148 instead of the usual Schottky.

Best regards, Spehro Pefhany

--
"it\'s the network..."                          "The Journey is the reward"
speff@interlog.com             Info for manufacturers: http://www.trexon.com
Embedded software/hardware/analog  Info for designers:  http://www.speff.com

That is a neat circuit!

Realistically, the normal reason for replacing a linear regulator, is power/heat. Now at 20mA, with 19v dropped, you are only talking 0.38W. Switchers are always going to work out more expensive (remember you need to be looking at better capacitors, and quite a few other parts as well). Your requirement is unusual, because of the relatively high input voltage (most of the cheaper buck regulator modules, with high efficiency for low power, are aimed at battery operated systems, with perhaps 10v in max), and the units for higher voltages, tend to have significant standby current, and are aimed at higher power, making the savings rather small. For example, devices like the OnSemi 34063, have a quiescent consumption around 4mA. Even if the switcher then runs at 80% efficiency, you end up with an overall efficency for your low power, of only perhaps 50%-60%. I'd suggest going discrete!. If you look at:

This circuit costs the least of any design that I know of, with reasonable efficiency, and the parts cost will be less than the IC solution, especially once you have added the discrete parts to the latter. :-)

Best Wishes

As others have said, a switcher appears on face value to be overkill - is there any particular reason you want to use one? You could use a 78L05 quite happily.

Ken

Can you help me understand this better? Even with Roman's explanations, I don't have enough of a basic grasp to figure out what's doing what when. I would like to drop 48VDC unregulated to 12VDC and 5 or 3.3 VDC supplies. It seems straightforward enough, but the waveforms in SPICE are not very encouraging. The duty cycle on the main chopper can be quite short depending on load, and this has strange effects on the oscillator. I can continue changing the inductance and capacitance values randomly until it looks good, but that would like to understand the relationships better.

Thanks. Mike.

The regulator is a form of free running multivibrator. The switching decision is made by Q2, as its emitter voltage rises above or falls below its base voltage. Once a switching decision is started, the positive feedback via C2 provides the snap necessary for a clean and saturated switch to take place. The frequency is determined by the ripple voltage at the output of the LC filter, in combination with the time constant of C1 in parallel with C2 and the impedance of the zener in parallel with Rz. I think you would have more control of this time constant if you added some additional resistance between the zener and the C1, C2, base node. I think you want the capacitor time constant well under the on or off period, so that the output filter dominates the decision to switch.

As the input supply goes up, the inductance of L1 should go up at least proportionately, to maintain the ripple slope during the on time.

The Schottky diode does not have enough reverse voltage capability for a 48 volt source.

You could make one regulator to provide 12 volts, and make a second one to regulate that down to 5 or 3.3 volts. This would keep the on time reasonable for each (25% or more).

I don't see how much current you need at each of the output voltages.

There is a feedback loop, through Q2's emitter to the base of the PNP pass transistor (Q1). That is the source of the instability that causes it to oscillate. When Q1 is on, current is poured into the smoothing cap through the inductor, causing the output voltage to increase. Once it increases enough to shut off Q2, Q1 gets shut off too. The current through the inductor stays on for a bit, and thus the voltage continues to increase as its magnetic field collapses, but eventually stops, allowing the load to run on the smoothing cap. Thus, the output voltage eventually falls. When the output voltage finally goes Vbe below the fixed 5.6V threshold, Q2 starts up again, which again starts up Q1, restarting the cycle.

The thing that makes it oscillate is that the big inductor creates a delay between turning off Q1, and when the output voltage actually responds to the cutoff. The voltage continues to go up even after Q1 turns off, and then droops once the inductor's magnetic field is gone. Same is true of turn-on; turning it on results in current through the inductor, which gradually builds up to the point where it overcomes the load and starts charging the capacitor. So, the frequency will depend on the size of the inductor (how far the charge lags the voltage changes), the size of the smoothing cap (how long it takes for charge to change the voltage), and the current draw of the load (how quickly the inductor will overcome the load).

The cap C2 gives the turnon-turnoff a tiny bit of speed up, in that when Q1 turns on, it'll yank Q2 on just a bit more with positive feedback, thus making it return the favor to Q1. When Q1 turns off, it'll yank Q2 off just a bit more, speeding up the Q1 turnoff. This sharpens up the edges a bit, leading to a bit better efficiency.

Sadly, for certain loads, I've found that the circuit will simply fail to oscillate, reverting to being an inaccurate linear regulator without current protection or temperature compensation. If you are trying to drop

48V to 3.3V with any kind of current, that is clearly a bad thing. Dissipation will go from Iload*3.3V/E to Iload*48...

--
Regards,
  Bob Monsen

A great discovery solves a great problem, but there is a grain of discovery 
in the solution of any problem.  Your problem may be modest, but if it 
challenges your curiosity and brings into play your inventive faculties, 
and if you solve it by your own means, you may experience the tension and 
enjoy the triumph of discovery.
- George Polya

...

Thanks, John. You and Bob are my new heroes.

The larger picture: I am driving two motors with a 36V 2A transformer. The

5v supply is for an MCU and interface circuits, about 150 mA. The 12v supply is used to help drive the MOSFETs, and to step down the voltage for the 5v regulator. Close regulation here is not strictly necessary. It has larger components -- BCP53 PNP, BCP56 NPN, ES1B diode, 7200 uH coil. PSPICE shows it switching well at about 200 mA load (I inexplicably misspelled 2.2 watts as 2.2 A in another message). The 5v regulator is as diagrammed in the original document.

My silence, for now, regarding your comments is due simply to trying to digest them. I respect the time and effort both you and Bob took to help. Thank you both.

Indeed! There is a prevalent illusion on this newsgroup that fewer parts-count is better, irrespective of performance.

The best switcher is built using some off-the-shelf chip OR rolling your own using, for example, an LM339, running at fixed frequency / variable duty cycle.

Load-dependent switching frequency structures suck ;-)

...Jim Thompson

--
|  James E.Thompson, P.E.                           |    mens     |
|  Analog Innovations, Inc.                         |     et      |
|  Analog/Mixed-Signal ASIC\'s and Discrete Systems  |    manus    |
|  Phoenix, Arizona            Voice:(480)460-2350  |             |
|  E-mail Address at Website     Fax:(480)460-2142  |  Brass Rat  |
|       http://www.analog-innovations.com           |    1962     |
             
I love to cook with wine.      Sometimes I even put it in the food.

Yes. There are a couple of comments here. The original poster, was asking about 10 to 20mA at 5v, from 24v. For this the little Black inverter is pretty ideal. However once the supply goes much higher than this, it becomes necessary to redesign the front end of the circuit. Also all switchers of this type, can potentially have problems with very large load variations. As load variations increase, chosing the time constant of the output components becomes increasingly critical. I too have seen some problems occasionally getting the circuit to start, usually where the circuit attached has a very 'soft' switch-on characteristic. The second poster, seems to have different input and output requirements. Given the need for multiple 'rails' here, It might well become a balancing act, between a multi stage version of this, having a version of this as a 'pre-feed' to a couple of linear regulators, or using a transformer with two secondaries. The best choice will depend on the loads involved, the accuracies required etc. etc..

Best Wishes

I certainly didn't go into much detail with my explanation. That was just the fly-over, to get you thinking about what the various parts do in the big scheme of things. I will be glad to discuss details when you arrive at the understanding to ask about them.

Roger Hamlett wrote: (snip)

The key idea here is, that just because a circuit is simple in parts count does not imply it is also simple to apply, optimally, to other applications.

This regulator reminds me of some of the designs that have been created by evolutionary processes. For the test application applied to the process, it works well, but, perhaps using some of the components in an unusual way. But altering the test situation requires re-evolving the circuit, and the result may end up radically different from the ancestor circuit.

The circuit is a hellacious p.o.s., and it doesn't work the way you describe, plus that comment about reducing the output filter capacitor to encourage oscillation is a nice try if your load does not mind 100% Vcc overshoot. So the circuit as shown is totally unreliable, it can blow out the load on turn-on, it can blow itself out by hanging in linear mode, it can blow itself out by destroying the NPN EB junction, nobody seems to have a handle on the critical time constants if any, the output voltage is largely an unknown with the zener ringing from an indeterminately large current spike, nobody seems to know what value of inductor is best, nobody seems to know how much filter capacitance to use, nobody has a handle on efficiency, and nobody seems to know what range of loading may be required, or not, for their blind random selection of components. That's quite the curiosity piece, Roger.

At no time have I 'described' anything about the way the circuit works. _perhaps you should learn how to read threads_, before posting. However the oscillator can work quite well, if used with care. Though it does take some thought about the whole circuit round it, rather than being treated as the typical 'black box' (though a different 'Black' here), that most regulators are treated as by many people. The time constants for the output stage, will depend massively on the circuit attached, and this is the biggest problem in the design, in that it cannot really be treated as a lump on it's own. It is a useable circuit, that can form the basis of thinking, for budget designs, which is what the thread started about. However the assumption is present that if it is to be used as the basis of something outside it's original work enviroment, it'll need to be redesigned by someboy who knows what they are doing....

Best Wishes

I'm reading the circuit description now. Boy, those sure are some horny transistors! ;-P

Cheers! Rich

At the very bottom of his page, he's got an adjustable dummy load using a TIP3055. In his ASCIImatic, he's got it upside down. :-)

Cheers! Rich