Three "simple" questions here:

1-I know the diferencies between linear and switching DC-DC converters/regulators. Are there any applications where the disadvantages (specially noise and ripple) of the buck and boost converter that you can buy on Aliexpress or such webs are really a problem?

2- If so, how clean is the output of a linear regulator in comparison with a buck/boost converter of that kind?

3-Would it be possible to step up first the voltage and then use a LM317 to regulate it down a bit and having a very clean output?

I have a 12V supply and I need 30 to closest to 0V I can go. So that idea came to my mind to combine both boost regulator and a LM317 to get the benefits of both and I think it would be the cheapest solution and with the lowest ripple/noise at the output (if my thoughts are correct).

EDIT1:Of course the boost will be manually regulated to have an input voltage of 2-3V above the output of LM317. EDIT2: The point is to make it as simple as possible so if the LM317 could reduce the noise sufficiently for no filtering in the output to be required. Also, if a commonly available boost/buck converter will be enough for everything I will go for that option. I don't know what uses are the SMPS not recommended for, because everything a could read is "not suitable for not sensitive applications" but it never specifies what are those.

  • 1
    \$\begingroup\$ Never use the LM317 for anything. The app notes and example circuits and datasheets are GREAT for learning about a lot of feedback concepts and clever circuit ideas. But the LM317 is hopelessly outdated. The biggest problem is the very large dropout voltage. This is still a good question if you change "LM317" to linear regulator. By the way, cheap DC-DC modules usually have very high quiescent current which can be a major problem in battery powered applications. Just keep that in mind. \$\endgroup\$
    – user57037
    Jun 12, 2017 at 0:03
  • \$\begingroup\$ Also, if you use a linear regulator to go down from 30V to 5V or something, the heat dissipation will be a major concern unless the current is very low. You might want to implement an adjustable boost rather than a fixed 30V boost. \$\endgroup\$
    – user57037
    Jun 12, 2017 at 0:05
  • \$\begingroup\$ it would be much simpler and more efficient to just consume the output of the buck/boost; if you have to ask about noise-sensitive applications, you're not working on noise-sensitive applications. \$\endgroup\$
    – dandavis
    Jun 12, 2017 at 7:24
  • \$\begingroup\$ I'm not working on noise sensitive applications at the moment but I want a PSU capable of that and if I can make it with a few components better than spending 100+€ on it. That's why I asked what the switching converters are not suitable to and what are those noise sensitive applications or at least the most common of them. \$\endgroup\$
    – DarKnight
    Jun 12, 2017 at 9:28
  • \$\begingroup\$ In my experience, step-up converters are far more noisy than step-down converters. So if you can avoid step-up, that's the first thing to do. After that, it mostly boils down to switch regulator layout, which is a big topic of its own. It is important to realize that the LDO doesn't necessarily filter out any of the high-frequency noise generated from the switch regulator, it might as well just let the noise past. The main advantage of the LDO is that it doesn't add much noise of its own. \$\endgroup\$
    – Lundin
    Jun 12, 2017 at 11:52

3 Answers 3


Switcher noise is a subtle thing, we shouldn't generalize.

Some loads don't give a damn about the noise (for example, a cpu) while others will care. Frequency of noise is also very important (most analog stuff like opamps and LDOs have decreasing rejection at higher freq due to capacitive coupling and decreasing feedback loop gain). So you should know your load.

Sometimes its not the output noise of the switcher that bites, but the noise it injects into its input supply (if same supply is used for sensitive circuits).

Or radiated noise, or noise injected into GND if the layout sucks, for example by violating the rule "You shall not let the square wave high di/dt current loop into the ground plane!" For example consider the EMU0202USB soundcard. There is a 3V3 switcher in it, it is located about 5cm from the cpu it powers. The cpu has tons of caps, but the switcher has none on its output! I guess they thought there were enough caps around the cpu. As a result, the 1MHz inductor ripple current runs in the ground plane over 5cm... also it has boost switchers to power the opamp, and the layout is not good. This turns the local GND into a mess, and the headphone jack is in the middle of the mess... it outputs tons of HF noise.

Different switchers have widely different output noise characteristics, and the noise they inject on their input is also very different. For example:

  • Buck has an inductor on the output

Thus its output current is a continuous sawtooth. The output inductor forms a LC filter with the output cap, reducing ripple. Remaining ripple will depend on frequency, inductor's value and parasitic capacitance, and output capacitors value, ESL, and ESR... and of course... layout!

Buck's input current is, however, a fast-edged square wave, so it's going to inject lots of HF into its input supply, and this noise may find its way to other parts of the circuit.

Say you got +15V for your opamps. And you use a buck to get +3.3 for your micro. This buck will draw HF square wave current from +15V and screw it up. In this case, a filter is needed... on the input of the buck, not on the output!

  • Boost does the opposite, inductor is on the input.

It draws a sawtooth current from the input, and outputs a square wave current. Its output is thus inherently noisier than a buck.

Ripple and spikes

Also these two types of noise must be distinguished.

A 100kHz switcher will have 100kHz ripple on the output for example. As said above, buck will have less HF harmonics in the ripple waveform due to output inductor providing a free LC filter. Opamps and other stuff with falling PSRR at HF like this.

It is actually easier to filter out HF ripple (like 500kHz) than LF ripple (like the 25kHz switchers of times long past), because the inductor and caps are smaller/cheaper, so you can use more effective filters in the same space. Also when the inductor is small you can get a shielded one for cheap.

And... the other noise type is the spikes. This is the combination of a MOSFET switching large current in nanoseconds, thus very high di/dt, thus nasty L.di/dt spikes everywhere there is inductance. These switching spikes have a verrrry wide bandwidth (like GHz) and a tendency to couple into anything and leak everywhere if the layout sucks.

Additionally, the spikes are modulated by output current and the mode the switcher is using. If it is in powersave mode you get trains of spikes, then as it shuts down, it goes silent. Your slow opamps will rectify and detect these modulated HF spikes, your fast opamps will attempt to process them as signal...

Some switchers use slow/soft-switching on purpose to reduce the spikes. Also, ferrite beads work great.

Now, to answer your question:

You need 12V input, 0-30V out, and you don't specify the current, nor the target efficiency, so I can't answer.

Anyway, if you need efficiency, use a buck-boost like this one. I won't recommend the external FET version for higher currents since it needs a 4 layer board and this doesnt fit your 100€ budget. I'm not sure it can be coerced to go down to 0V, but I guess it should, as output voltage can be controlled by injecting a known current into the feedback node.

You can also use a pure boost, followed by a LDO. However at low output voltages, efficiency will suck. If this is powered by a 12V battery, then you will care about efficiency, so you have a problem.

If this is mains-powered, then ditch the useless 12V requirement, get a 32V or more DC supply, and use a simple buck to make your 0-30V output!

  1. Switch-Mode converters are not suitable for precision instruments like multimeters or precision current sources.
  2. According to the datasheet on page 5, the LM317 output noise is "0.003 % V-out".
  3. You could use a switchmode converter before a linear regulator, however, the average noise will still be higher than a single linear regulator. Page 7, figure 8, in the datasheet gives more details about ripple rejection.
  • \$\begingroup\$ Any other application where they are not suitable for? I can imagine audio, data transmission... but not anything else (not even sure if those are, actually). I don't know what projects will come to my mind in the future so I would like to have a bit of information in this subject. \$\endgroup\$
    – DarKnight
    Jun 12, 2017 at 9:50

To expand on #AH L s #3. Yes you can do it. If you heavily filter the switchmode converter there is no obvious reason (in principle) you couldn't get good results.


In the first place, switchmode converters typically operate at high frequencies, like 50 to 500 kHz. At these frequencies the intrinsic rejection ratio of a 317 is not the greatest, although it's much better at 10 kHz than at 500 kHz. See the data sheet.

Secondly, regardless of the intrinsic behavior of the 317, suppressing high-frequency noise is something of an art, and particularly at the higher end it can get tricky. That's not to say it's impossible, just that some experience is a good idea.

  • \$\begingroup\$ I know I can put there some heavy filtering with RLC filters or some opamp low pass/ high pass filters but I wanted to keep it as simple as possible so if one chip would do everything on itself I don't have to brake my mind designing some filters. First of all because I don't have a scope. \$\endgroup\$
    – DarKnight
    Jun 12, 2017 at 9:33

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge that you have read and understand our privacy policy and code of conduct.

Not the answer you're looking for? Browse other questions tagged or ask your own question.