14
\$\begingroup\$

I'm using a 12V adapter and a 2S 7.4V Li-ion battery to power my electronics, and I would also like to power my MCU with it. To switch between the adapter and battery I'm using a BQ24133 from TI.

I will use an STM32L4 MCU and some other components that use 3.3V on a custom PCB. Everything together on 3V3 uses up to 150 mA when in full operation mode.

I'm searching for the best/cheapest solution.

1. What is the difference of using buck converter vs linear linear regulator to power an MCU?

2. Would a linear regulator (small packages) be a bad idea because it would heat up a lot because there is a big difference in voltage (12-3.3=8.7, 8.7*0.15=1.3W) ?

3. Would the frequency of switching, or output voltage ripple (noise) be a big influence on the normal operation of an MCU?

4. Conclusion, what is the best way to power it with the input voltages between 6V and 12V?

Thank you for you patience and your answers.

Thank you for all your answers. You all were very helpful. Up until now I used linear for my projects, but I think now I might go buck. If you wanna follow up the reason I asked this, and see what I am making, follow this link

\$\endgroup\$
2
  • 1
    \$\begingroup\$ For a project with similar configuration (only 9V adapter) I went with TLE42744DV33 in TO252-3. Since you have higher voltage differential and obviously not afraid of complex BMS I'd suggest using buck converter with enough filtering on the output and decoupling elsewhere. \$\endgroup\$
    – Maple
    Commented Jun 13, 2018 at 9:39
  • 1
    \$\begingroup\$ I don't have a thought that's full-fledged enough to be a standalone answer, but I'm surprised none of the good existing answers mentioned this. If you really want a linear regulator but can't handle the power dissipation in a single package, consider putting several in series. For instance, to get from your 12v down to 3.3v, you could use a 9v regulator followed by a 5v followed by a 3.3v. It takes up more physical space but divides the power dissipation among several ICs. \$\endgroup\$
    – dwizum
    Commented Jun 13, 2018 at 19:37

7 Answers 7

25
\$\begingroup\$

1) Buck converter is:

  • more expensive than a linear regulator
  • takes typically more room on the PCB
  • is typically more difficult to design (sometimes, just a bit more, sometimes way way more)
  • is more noisy (the amount of noise depends on a lot of factors, though)

But it is indeed much more power efficient, especially if there is a big difference in the input voltage vs output voltage, which is your case here. The buck will output almost the same power than it takes at the input (efficiencies are typically ~80-90%), whereas the linear regulator will take as much current as input as it needs to supply (which means that efficiency is Vout/Vin, something like ~27-44% in your case, which is very bad).

2) Yes, this is actually the only reason why linear regulator can be a bad choice: the efficency (and your computation of dissipated heat is fine). Now, having that much power dissipated leads to two big problems:

  • You will probably need a heatsink (check the datasheet of the linear regulator: at anything more than 1W, you need to check carefully even in TO-220 package. When using smaller packages, it is often not possible at all). So this negates the "more room on PCB" inconvenient of buck regulators.

  • If you run on batteries, it means much less runtime. Sometimes, you can't afford it (do the computation).

3) Most likely not, if you use standard integrated solutions to do the stepping down. Those are made to provide power to IC chips, and the datasheet/application notes of the step-down controller/regulator you'll choose should give you some information about the amount of noise you'll get. But for digital operation, supply noise isn't typically that much of a problem.

4) Given the huge difference in input/output voltages, the current you need, and the fact you'll partly run on batteries, it seems a logical choice to go for a buck. But you need to double-check all that yourself. Maybe in your case it is acceptable to have a huge TO-220 dissipating 1.3W in your enclosure, and the runtime you need isn't that high.

If you go for a buck, here is what I can suggest:

  • One solution is to go for a complete module. You then don't need to design anything. Check mouser/digikey, they have DC-DC converters that you can just solder to your PCB like a linear regulator. If you're an aliexpress/ebay kind of guy, you'll probably find plenty of cheap things there also.
  • You can design your own (scary, I know... wow, an inductor!). What I can suggest in this case is to have a look at the tools provided by the various manufacturers (e.g. TI webench, but Linear technology has one too, ...). You just supply them with your requirements (input/output voltage, current, ...) and it spits out a lot of possible designs (you'll see some of them are actually very simple) with different chips from their catalog. All the passive component values are already calculated for you, and they even suggest you part numbers for the inductor, etc... So you just have to check the recommended PCB layout from the documentation to design your board, buy the parts, solder, and it should work.
\$\endgroup\$
14
  • \$\begingroup\$ This is a very good answer! I especially like the part about how SMPS converters often need more components -> it's a more complex design. \$\endgroup\$ Commented Jun 13, 2018 at 9:29
  • \$\begingroup\$ +1 @MortenJensen I absolutely agree with you. \$\endgroup\$
    – Long Pham
    Commented Jun 13, 2018 at 9:32
  • 8
    \$\begingroup\$ I think it's worth mentioning that bucks suck at ultra-low power. So if the MCU sleeps most of the time and only awakes for brief bursts, linear could actually waste less power overall. Dual regulation is also an option, but OFC the most difficult of them all. \$\endgroup\$
    – Agent_L
    Commented Jun 13, 2018 at 9:42
  • 6
    \$\begingroup\$ @Agent_L What you say is indeed important to consider, and it is right for average buck chips. However, with the advent of mobile, you can now easily find buck chips that outperform linear regulators even at light loads. If you lookup "micropower buck", you'll find more things than "micropower linear regulator", and looking at the datasheets, you'll see they indeed have lower quisescent current (bucks can operate in modes where it is sleeping most of the time, something a linear reg can't do). However, I agree this is not true for the usual buck, you have to specifically look for those. \$\endgroup\$
    – dim
    Commented Jun 13, 2018 at 9:58
  • 1
    \$\begingroup\$ @Joe Thanks. But don't worry, it's certainly just because you're fooled by my extensive use of the "obnoxious in-your-face formatting and lists" technique. You'll soon master this, too. \$\endgroup\$
    – dim
    Commented Jun 13, 2018 at 12:12
8
\$\begingroup\$

1) SMPS is more effective in converting energy, but is more noisy because of the switching. The linear regulator wastes power proportional to the difference in input and output voltage, but operates with low noise.

2) Depends on whether or not you can dissipate 1.3W - only the designer (you) can know that. 1.3W can be a lot of power for a small IC, so you might need a heat-sink.

3) Different switching frequencies make noise on different frequency-bands. Only the designer (you) can know if that will be a problem. You should follow a reference design for the specific MCU to ensure the input voltage ripple is acceptably low.

4) Depends on the how the trade-offs are weighted for the specific application. One cannot be objectively better than the other. It is almost always a trade-off in engineering.

\$\endgroup\$
6
\$\begingroup\$
  1. What is the difference in using buck converter vs linear linear regulator

Very minimalistic explanation:

SMPS

A SMPS (switch mode power supply, e.g. Buck) basically compares the output voltage to a given reference. If the output voltage is above the reference, the regulator basically cuts connection between input and output. If the output voltage is below the reference, input and output are connected. Output capacitance and inductance are used to store energy on the output side and smoothen the output voltage.

benefits: Efficiency and therefore power dissipation (--> heat) because the switches are either closed (no current -> no power dissipation) or open (lowest resistance state -> minimal power dissipation).

downsides: additional parts (usually a voltage devider, inductance, capacitance and maybe a ferrite bead for noise suppresion) and increased price (device itself and additional parts).

Linear

Unlike an SMPS a linear regulator doesnt use a transistor as a switch (on/off) but in linear mode (any state between on and off is allowed as well). This leads to increased power dissipation, as you can imagine the transistor as a regulated resistor that is being adjusted for a voltage drop of Vin-Vout.

benefits: cheap; easy; less/no noise due to no switching, might need only a capacitance downsides: efficiency, especially on high load;

  1. Would linear regulator (small packages) be a bad idea because it would heat up a lot because there is a big difference in voltage (12-3.3=8.7, 8.7*0.15=1.3W) ?

I would answer this with yes. If you have a look here and consider values like the ones on chapter 6.4 in e.g. this datasheet you will see that the thermal resistance easily goes over 100°C/W (meaning: a temperatue rise of 100 °C for 1W power dissipation). I think having this in a small case wont work, even with a (small, because small package) heat sink and lots of copper area on your PCB determined for cooling (so you won't be able to benefit from the small package at all).

As a rule of thumb I usually use a linear regulator if i need either very low currents (just a few mA at max), very small voltage drop (1..2V) and/or super clean supply voltage for an ADC or other analog parts. Means in most cases i prefer to use SMPS. These require more parts usually (more caps, resistors, inductance) so its a more expensive and 'complicated' solution.

  1. Would the frequency of switching, or output voltage ripple (noise) be a big influance on the normal operation of a MCU?

If you design a SMPS based on the devices data sheet there are usually calculations given for the expected ripple noise. These are usually within 1% of the output Voltage what is no problem for digital systems. I have created an excel sheet ot help dimension caps etc, but I dont know how to add an attachment here ...

Also you would probably want to add a 10..100nF cap to each supply input of the MCU and keep the traces from Cap to MCU short to minimize ripple seen by the power pins.

  1. Conclusion, what is the best way to power it with the input voltages between 6V and 12V?

As you need a big voltage step, more than a few mA and didnt mention any special requirements regarding noise (for analog stuff) I'd go with a SMPS.

\$\endgroup\$
3
\$\begingroup\$

There is no best way ! Everything is a bargain.

Generally, switch-mode power suplies have better efficiency than linear power suplies. However, they are much noisier than their counter part. This may be critical for precision circuitry.

Using linear regulator as post-regulation for switch-mode power suplies is good, as this satisfy 2 factors : efficiency, low noise. But, again, everything is a bargain ! This introduces more BOM cost and more board space !

  1. What is the difference ...

They differ from their working principle. Please use the Google !

  1. Would linear regulator ...

Maybe, it depends on your design.

  1. ...

Usually no if power supply pins have been decoupled. This might be a problem with analog stuffs (ADC, DAC,...)

  1. ...

I cannot answer this.

\$\endgroup\$
1
\$\begingroup\$

Buck converters are more noisy and expensive due to switching and external components like the inductor (you usually can't integrate this in the IC, but other externals may be integrated for smaller currents). Noise usually isn't a problem for digital circuits (which generate their own noise in the supply rail), but may be too much for analog. Depending on the amount of power you need, a SMPS may also be smaller since the high efficiency will mean less dissipated power (the inductor may be smaller than the heatsink).

Linear converters are usually cheaper and for lower powers it may also be smaller if few external components are used, but may require a heatsink for larger powers.

There is also the option of using a resistor and Zener, but this usually isn't even considered because the Zener will consume power even if the MCU won't (for example during sleep/standby), but it may be a viable option if you current draw is relatively constant.

The selection of the power supply is a tradeoff: you have to balance your budget, the size and the noise. Since you're possibly dropping from 12 V to 3.3 V, your thermal requirements will usually dominate, which will usually indicate a buck converter. However, if your application heavily uses the ADC, unless you can have an external voltage reference, it may be advantageous to use a linear converter, even for bigger sizes. Then, if your budget allows, you can also use both: you may use a buck converter to drop from 12/7.4V to 5 or 4 V, and then use a linear to go to 3.3 V. This will allow a smaller drop in the linear regulator, possibly avoiding thermal issues.

\$\endgroup\$
1
\$\begingroup\$

SMPS will provide you a more expensive solution to MCU internal power.
Linear Regulators instead repercutes in a less noisy DACs and Switching ones in more stable computation.
Switching power supply will compensate fast MCU switching computations. Instead you try linear regulators may provide equitable DAC quality but at the expense of higher chances your MCU will hangup.

Regarding a conclusion: linear regulators are often plugged with MCU which provide a reset on hang watchdog.
Switching is way more expensive but the demands are relevant to each application.
You may already prefer buying a SMPS modules which are easily adaptable to any needs.

\$\endgroup\$
2
  • 2
    \$\begingroup\$ Please fix your grammar and formatting. Are you sure about the first sentence, what I can barely understand ? \$\endgroup\$
    – Long Pham
    Commented Jun 13, 2018 at 14:57
  • 1
    \$\begingroup\$ I believe the MCU just need some decoupling caps close to it. \$\endgroup\$
    – Long Pham
    Commented Jun 13, 2018 at 15:05
0
\$\begingroup\$

Linear regulator using when we need a low noise (for example mic amplifier that feeds from the same power source as MCU).But there is an alternative - firsly switching DC/DC to 4V secondly down to 3.3V through LDO linear stabilizer.

\$\endgroup\$

Your Answer

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

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