ormally I use linear voltage regulators like the 7805 or LM317. These are simple to use and work great but can dissipate quite some heat if powered from a higher voltage like a 12V board accu and delivering 0.5 to 1.0A at 5V. Cooling is then definately required. Heat also means battery power is waisted. An alternative is to use a switching regulator. Voti provides affordable versions of the (not so modern anymore) LM2575T-5.0. This is a fixed 5V version. The documentation from National Semiconductor is pretty detailed. I will not repeat is here. Instead this page describes my experiments with the LM2575.

Last updated: March 7, 2004

Circuits tested

First I tried the textbook examples from the NS documentation. This was done with components I had at hand. This covered a number of components, including a small inductor of 20mH, a 2 x 1.1mH toroid coil, a bare toroid core, 1n4148 diodes, later a schottky diode and several capacitor values.

The standard buck regulator circuit delivers a +5V output. This is a rather straightforward circuit and will operate with a wide spread of component values. Optimal values should be selected however to get minimal ripple, minimal noise to the power source, maximum efficiency etc. Note that the quiesent current is about 10 to 14mA.

The buck-boost regulator circuit delivers -5V output. I hoped this would a nice way to generate a negative voltage for e.g. an LCD-display. However this circuit requires component values that are similar to the buck regulator at high load. Probably because the LM2575 as buck-boost regulator tries to bootstrap its own operating power. Together the need for subtantial bigger output capacitors I abandoned the idea of using the LM2575 in buck-boost for an negative LCD voltage. There is an alternative further down on this page.

The inductor

The inductor needs to be selected corresponding the load current. With an input of 12V and a low load (tried <10mA) the LM2575 works with a small 20mH/60ohm inductor (also available at Voti). For higher loads lower inductor values are needed (order of magnitude 330uH). The inductor value also depends on the input voltage. The documentation shows graphs that help to determine the right value. Just make sure that you find one that van handle the load. Common 'resistor package' style inductors have too high series resistance and are not usable. Also make sure that the core does not saturate at the operating current.
I used a toroid core inductor because these produce the least EMI because the fields are contained within the core. Besides a 1.1mH inductor I did wind my own on a bare core of which I had no specifications (frequency range, Al). This inductor worked equally well. If you know the Al value of a core you can determine the needed turns using:
L = Al * n².
Make sure how Al is specified (e.g. in uH per turns, uH per 10 turns or uH per 100 turns).

The diode

Initially I did not have acces to a schottky diode. This is the preferred type of diode since it is fast and has a low forward voltage drop, thus increasing overall efficiency. Instead I started with a couple of 1n4148 diodes in parallel. This works fine since it is a fast diode. The forward drop can be 0.9V at 100mA. If you have diodes with similar characteristics (from the same batch?) it is ok to parallel them. The maximum current is 200mA for a 1n4148, I used 2 in parallel to test a 200mA load. Note that the diode spec should be at least 1.2 times the output load. If you want to withstand continues short circuit, the diode must be able to handle the limit of the LM2575 which is over 1A. Then use a bigger diode. E.g. the SB350 (3A) or SB530 (5A) family. The latter should be normally available at Conrad.

The capacitors

The documentation states these should have a low ESR rating. It is difficult to find the actual ESR value specified at companies like Conrad. At least take ones that are referred to as low ESR. A higher voltage capacitor then strickly needed is said to have lower ESR values. It also seems to increase the life-time of the capacitor.
I tried lowering the input capacitor to below 100uF and found that gradually the noise it generates increases on the input. Beware if you power more (and sensitive) circuits on the same input power source.
The output capacitor influences how much ripple is present on the output. This also depends on the load. If needed you could filter it as described in the documentation. For digital circuits it might not be necessary. On my breadboard setup, which is not ideal with regard to the common ground point mentioned in the documentation, I measured 100-200mV peaks at a 0.5A load. These reduced to < 20mV with the L/C filter in place.

Efficiency

The power efficiency of the cicuit depends on number of factors. A good inductor helps, but primarily the diode is important. With a pair of 1n4148 in parallel and a load of 200mA I measured 71% efficiency. With a proper Schottky diode this became 75% and went up to 79% increasing the load to about 400mA. These values are not extremely good for a switching regulator nowadays, but they do allow operating without a (big) heatsink for the regulator.
When the input voltage gets closer to the output, the efficiency increases a bit as well (I got up to 84%). But then again so it does for a normal linear regulator...

Buck regulator with additional negative voltage

The above circuit shows a way to generate a second voltage. The toroid coil at hand had 2 sets of windings. So I tried to see if I could get some power from it. This works great as long as there is at least some current drawn on the main output (a few mA is enough). The coil then acts as a sort of transformer. In the above circuit the out-pin of the LM2575 switches between about -0.5 and 11V. Since the circuit output is 5V the 'transformer' is driven by a about 5.5V (top) AC voltage. This is also available on the second winding and after rectification gives about 5V. The 5.6V zener was not needed in my case since it was meant for a graphics LCD that works between -5 and -8V. Actually the zener lowered the output voltage just a bit too much and I left it out.
Note that the voltage on the secondary winding depends on the input voltage. E.g. a car battery ranges from 11 to 14V. Then the negative voltage varies as well. Also the voltage was not perfectly symetrical. One polarity side was a bit higher then the other. If you want to be sure there is enough voltage available to get the zener conducting at all times, just add a couple of windings yourself until you covered the margin you need. Of course you can even use that method to create a second widing if there is none in the first place. When you draw as little current as I do thin wire should be fine.

The image above shows not a lot of components are needed. The biggest component I used is the coil. It would be nice to test some smaller versions, but I could not lay my hands on the right ones.

Other output voltages (>5V)

The LM2575 is available in several version with different output voltages. The documentation mentions only differences in resistor values in the feedback circuit internal to the LM2575. I tried to add a series resistor to the feeback pin to see if I could change the output voltage. This works like a charm. With a 10k potmeter in series I could control the output between 5V to over 10V using a 12V input. The formula for the output voltage is:
Vout = 1.23 * (Rf + 4100) /1000.
So Rf=0 gives 5V, Rf=3k3 gives about 9V and Rf=5k6 gives about 12V.
This shows the LM2575-5.0 can be used for voltages higher or equal to 5V. I expect this is all within the normal bounds of the LM2575-5.0 because the maximum input voltage is 40V. The switch output can handle this 0-40V anyway and the feedback pin remains at 5V as in a standard operation.
This tunable output should make it possible to use the LM2575-5.0 as a pre-regulator for a linear regulator like the LM317. With the right stereo potentiometer (10 or 22K) it should be possible to keep the LM317 input about 3.5V higher than its output over the whole range. This should result in a low noise power supply with low loss, even at low output voltages under maximum load (1A).

The result

The LM2575-5.0 is a simple regulator to use. The additional components around it (schottky, low ESR caps, toroid) need some attention for optimal operation. Perhaps the only non-common part is the inductor, which might require some searching in the market. You can make your own if you find a bare core with known Al.
Having a second set of windings can also give a (low current) second output voltage, provided there is some load attached to the output.

The output can be a bit noisy at higher loads (<200mV). Then a L/C filter as usually is found in switching power supplies will help (reducing ripple by a factor 10). Also you might want to consider a linear regulator as follower for even better results.

The output voltage can be higher than 5V with the use of a feedback series resistor. This should even be possible in a e.g. a 1.2 to 25V power supply with a LM317 follower circuit.

Since the efficiency is relatively high a small heatsink or none at all will do. The remaining loss (about 20%) will be dissipated by the LM2575, the diode and probably a fraction by the capacitors. So the 20% will not all go to the regulator itself.

Last updated: March 7, 2004