Simple Switchmode Buck (Step Down) Regulator

© October 2021, Rod Elliott

Main Index Projects Index

PCBs may be made available for this project based on demand.  Given its simplicity, a board is unlikely.

For most projects, linear regulators are preferred, and for very good reasons.  They have low noise, and don't introduce high frequency switching noise onto the supply.  However, they aren't efficient, and can dissipate significant power.  Of course, this depends on the input voltage and load current, but even 100mA can mean that heatsinks have to be included so the regulators don't overheat.  The major benefit of a switchmode power supply (SMPS) is that it's far more efficient, and that means less heat.  In an ideal case, a switchmode converter reducing 24V to 12V only draws 1A from the 24V supply to deliver 2A at 12V.  Of course, this assumes 100% efficiency, which is never achieved.  Efficiency will usually be closer to 80%, depending on the design choices made.  A linear regulator doing the same job (24V to 12V) has a maximum efficiency of 50%.

Most SMPS you can buy use SMDs, and if you need to change something it can be either very difficult or even impossible.  While they are usually fairly cheap, you learn nothing by just buying a ready-made SMPS, and if it fails there's not much you can do other than buy another.  DIY not only lets you make changes as needed, but you'll learn things as you go, and that's never a bad thing.

The best option where you have a relatively high voltage to start with is to use a switchmode buck converter to reduce the initial voltage to something 'sensible', then use a linear regulator to set the final voltage.  This is more complex, but you can usually dispense with heatsinks (at least for low-current designs), and you don't need to use linear regulators designed for high input voltages.  With this project, you could reduce (for example) a +35V supply to 8V, then use a 7805 to get a clean +5V supply with minimal dissipation.  The output current with the circuit as shown is up to 200mA, but an inductor designed for a higher DC current can be used to get more if you need it.

The project itself is pretty much exactly as shown in the datasheet for the recommended IC - an LM2596T-ADJ.  This is a step down (buck) converter in a 5-pin TO-220 package (the datasheet I have claims it's a 7-pin IC, but that's wrong, and the pinout only shows five!), with a maximum input voltage of 40V DC.  In many cases, it could be used to reduce the main supply of a power amplifier from (say) +35V to +12V, suitable for running various peripheral circuitry.  This could be a Project 33 speaker protection circuit, eliminating the need for a dropping resistor for the relay and providing the circuit with a stable operating voltage.

There isn't anything unusual about the circuit - the schematic is almost identical to that shown in the datasheet, but I found during testing that the feedback capacitor isn't generally necessary, provided the output capacitance is high enough.  The datasheet makes the whole process appear very daunting (to the point where many will never even try the circuit), but it's actually fairly simple if a few sensible rules are followed.

Figure 1 - 4V To 16V Variable Output Converter Circuit

Naturally, the input voltage (including any ripple) must be greater than the output voltage, and the IC needs a minimum of 2V input-output differential before it can regulate.  With normal operation, I'd aim somewhat higher - typically with the input at least 5V greater than the output.  The adjustable version has a nominal reference voltage of 1.23V, so the voltage divider used for the feedback determines the output voltage.  The basic calculation is as follows ...

V_Ref = 1.23V
R_FB = R1 × ( V_Out / V_Ref -1 )

R1 should be between 240 ohms and 1.5k, with lower values within this range providing lower noise at the feedback pin (Pin 4).  For example, if you need an output voltage of 12V and R1 is 1k as shown, the total feedback resistance (R2 in series with VR1) will be ...

R_FB = 1k × ( 12V / 1.23V -1 ) = 8.756k

Conversely, you can determine the output voltage if you know the feedback resistances ...

V_Out = V_Ref × ( R_FB / R1 + 1 )

Depending on your specific requirements, you may need to use a higher (or lower) pot value, and/ or change the value of R1.  Note that the connections shown with a heavy line are fairly critical, and need to be low resistance.  This isn't as difficult as it may seem, because the ground trace can form pretty much a straight line from Pin 3 (which is also the tab of the TO-220 package).  Solder a copper wire along the length of the ground track (after components are installed of course).

Using a trimpot (VR1) is the simplest way to set the required output voltage, otherwise you'll likely need an unobtainable resistor value.  A small error is usually of no consequence.  The voltage can be set exactly, but that is rarely necessary for most circuits.  This is especially true since most external supplies have a reasonably wide tolerance anyway, so a nominal 12V supply may deliver somewhere between 11.8V and 12.2V, with some being worse.  If you need a 5V supply, it needs to be fairly close as some parts will fail with an over-voltage.

Figure 2 - Prototype Variable Output Converter Circuit

The prototype is shown above, built on Veroboard.  If you believe the datasheet, this can't possibly work well because Veroboard is not designed for high frequency circuits.  Despite this, the circuit works perfectly, but as shown in the photo, the additional output filtering is not included.  The extra 2.7Ω resistor and second 220µF cap make a big difference to the output noise, but of course the resistor will affect the regulation (-270mV at 100mA).  A switchmode supply is one place where it can be useful to add 100nF ceramic caps in parallel with the electrolytic caps, because like all switching supplies, there will be some very high frequency output noise.  You can also add ferrite beads to the DC output reduce the switching noise a little more (especially any very high frequency components).

The inductor I used for the prototype is a 100µH, 1.75A DC unit bought on ebay, and while it's only tiny it works well for low current (generally less than 200mA) operation.  Normally, I don't recommend buying parts from ebay, but both the IC and the inductor turned out to be fine.  Virtually any inductor with similar specifications can be used if you can't get the same type where you live.  In reality, it's the inductor that's really at the heart of any switchmode circuit, and this is one of the main reasons that I haven't produced many other SMPS designs.  You may also notice that I didn't use a Schottky diode, but used a fast-recovery type.  This reduces the efficiency slightly, but that doesn't matter much when a circuit is used for low-power applications.

If you need more current, use a lower value inductor with a higher DC rating.  The converter ideally operates in CCM, so the inductor has a net DC component.  If the inductor core saturates, the IC will attempt to draw very high (and destructive) current.  The datasheet recommends against using open-core inductors, but I found no issues.  Note that the diode will carry a current of about 1.3 times the output current, so make sure that the one you choose is up to the task.

Figure 3 - Converter Circuit - 15V At 110mA Output

The trace shown was taken using a 2.7Ω resistor and 220µF filter network at the output.  The noise is greatly reduced, and would be reduced further if the filter were located further from the converter.  Some stray flux from the inductor will always be present, and distance is your friend (see below).  This was just a quick test, and I also verified that the preset output voltage is rock-steady with changes to the input voltage.  For the 15V test, I varied the input voltage from 20V to 40V (actually 45V for a moment as I was looking at the wrong meter).  It's also very good (without the 2.7Ω resistor of course) when the output current is changed from zero to 55mA and then 110mA.  This also holds good for other output voltages (I tested at 5V, 12V and 15V).

While you'd expect that adding a 100nF cap in parallel with the 220µF output cap would help, it made exactly zero difference.  I expected this, but unlike any number of pundits elsewhere, I know that the self-inductance of an electrolytic capacitor is negligible, and have proven many times that polyester, polypropylene or snake-oil capacitors in parallel make no difference until you're looking at MHz frequencies.  If it makes you feel better then use a low-value bypass by all means, just don't claim that you can hear the difference.

I also took the same measurement with around 150mm of wire between the SMPS and the filter with load, and the noise was reduced to 1.6mV at 55mA and 2.2mV at 110mA.  Just this distance was enough to ensure that no stray flux from the inductor could influence the output.  And no, the 100nF bypass cap still made no difference whatsoever.

Naturally enough, I wanted to see how much current I could get with 12V out and 24V input.  480mA is so far past the design goals that it's being silly, but the converter handled it.  The IC became rather warm (and did so rather quickly), but the voltage was rock-steady at 12V, and the ripple was still reasonably low.  It's perfectly alright for the sort of application for which the circuit was designed.  Note that the frequency is actually around 55kHz, and is not 174kHz as shown by the scope.

This project is designed so that you can use an existing supply (of no more than 40V) and derive an output voltage that suits additional circuitry that can't handle the higher voltage.  There's nothing special about it, and it just shows that a simple SMPS can be built on Veroboard (and is very small).  My prototype measures only 44 × 18mm, and could easily have been made smaller.  Even on Veroboard, its size could be reduced simply by using PCB pins or fixed wiring instead of the connection loops seen in the photo.

This is a useful circuit, and it often can be used exactly as shown in the photo, without any extra filtering.  There is some output noise (it's a switchmode supply after all), but that won't affect many circuits at all, especially if they aren't in the audio path.  As with any SMPS, ideally it will be in a screened enclosure, or kept well away from audio circuitry.  The operation frequency may be well above the audio band, but it can still create intermodulation artifacts if the noise picked up by audio circuits.

The missing link is an equivalent negative regulator.  While the IC can (in theory) provide a negative output, the implementation described in the datasheet is not (IMO) suitable for producing a negative supply with the typical voltages found in power amplifiers.  While the lack of a negative supply is likely to be a problem for audio circuits, it doesn't matter with most control systems.  These may include the P33 speaker protection circuit, or 5V circuitry used to operate a motorised volume control.

The datasheet goes to great pains to closely examine every possibility for input and output voltage, and includes warnings if you use tantalum capacitors.  I don't recommend tantalum caps for anything if they can be avoided, and the use of 'ordinary' aluminium electrolytic caps means that stability issues are unlikely.  During the testing phase, I subjected the circuit to considerable abuse, and it performed as expected under all conditions.  It's not perfect of course.  There are some conditions where the output may become slightly unstable (but with only about 100mV variation), but that's of no consequence if it's followed by a linear regulator.  Any issues can be solved of course.  (The datasheet is 45 pages long!)

Unfortunately, at the time of writing, these ICs are a victim of the current 'silicon drought', and are only available from my preferred supplier on back order (March 2022).  You can get ICs that claim to be the 'real thing' from China (of course), but they could be either fakes or factory rejects.  You take your chances with anything bought through on-line auction sites.

LM2596 Simple Switcher^® Datasheet

Main Index Projects Index