I need 3,9V and 5V from 12V. Boost or Buck?

Hello!
I'm building a device that is connected to 12V car battery. I use SIM900 module, which has it's own power supply(29302WU) to give it 3,9V. I also need 5V to the Atmega.
Which is the most sensible option:

  1. Boost the regulated 3,9V to 5V
  2. Buck the 12V to 5V again
  3. Other option?

I'd like to do this with discrete components(I can find both modules for 1$) and for as low cost as possible. Low power consumption is also key thing.

12 down to 5.

So I'm basically rectifying regulating from 12V to 3,9V then again 12V to 5V. Isn't that power wasting?

Your original post seems to say one device has its own regulation source, and you only needed a source for the Arduino. Nothing wrong with running two in parallel. Is easier too because the one source doesn't need twice the current flowing thru it, so smaller parts can be used, things run cooler, makes for more durability.
Make sure to connect all Gnds together.

Or run the Arduino at 3.9V. It can run at 16 MHz at 3.9V and still be fully in spec.

No. "Power" is most usually a construct of both voltage and current. (Expressed in watts or volt-amps.)

If you convert down from 12v to 5v, you drop 7v, but only consume as much current as the 5v load requires. ....Plus any inefficiency of the converter, and let's temporarily ignore the fact that car power systems aren't 12.000v, they're anywhere from 10v to 16v at any given time. Now, in addition, you'll have to convert from 12v to 3.9v, but only providing the current required by the GSM module. So you're not wasting much "power", since each converter is only supplying the current (1/2 the "power" equation) that its respective load requires.

You can buck from 12v to 5v, and 5v to 3.9v, if you have a regulator that can operate in such a tight margin... But then your 12-to-5v converter has to provide the current for both loads.

Converting from 12v to 3.9v, and then tapping the 3.9v to boost back to 5v is a Rube Goldberg solution that draws all the current from the 3.9v rail. In general, that's not the best way to approach this problem.

BTW, "rectifying" is the wrong term here. You're not rectifying anything since there's no AC source. (Unless you use particular switching topologies, but that's a different conversation entirely.)

CrossRoads:
Your original post seems to say one device has its own regulation source, and you only needed a source for the Arduino.

Yes that's right. I might be using SIM900 on it's own sometime(without integrated 3,9V regulator), what approach would you recommend then, if I had all in one PCB?
Could I get away with one 5V regulator, and maybe use voltage divider for the 3,9V?

Can I really run Atmega328 with 3,9V? Do I have to change fuse settings or anything? That would be definitely a ideal solution!
I just ordered Pro mini, that is guaranteed to run 3,3V.

SirNickity, thanks for the insight. I've a constant output regulator that can output 5.0V within 9-16V input. I sometimes mix up regulation and rectifying :slight_smile:

Thanks for fast responses, I feel way more confident with the choices when I hear a second opinion.

Could I get away with one 5V regulator, and maybe use voltage divider for the 3,9V?

No, not a "voltage divider", but a Low Drop Out (LDO) regulator, one that needs Vin = Vout + 1V (or less).
How much current at 3.9V did you need? Up to 150mA?

Voltage dividers (of the pair-of-resistors variety) will not work because the voltage will vary by current drawn. The more current is drawn, the more out of whack that divider becomes, and GSM modules are known for transient high-current demands.

There needs to be a feedback loop that alters the relative resistance of the divider based on load, which is essentially what a linear regulator does.

AVRs are well-suited to run at 3.9v. When you buy a complete Arduino or clone, the rest of the stuff on the board has to also run at the supplied voltage, which is not a given, so that's something you'll need to verify before using a pre-built dev board. If you build your own PCB, you are in control of that, of course.

CrossRoads:

Could I get away with one 5V regulator, and maybe use voltage divider for the 3,9V?

No, not a "voltage divider", but a Low Drop Out (LDO) regulator, one that needs Vin = Vout + 1V (or less).
How much current at 3.9V did you need? Up to 150mA?
http://www.digikey.com/product-search/en?FV=fff40027%2Cfff80182%2Cc0039b%2C10a40002&k=ldo&mnonly=0&newproducts=0&ColumnSort=1000011&page=1&stock=1&pbfree=0&rohs=0&quantity=&ptm=0&fid=0&pageSize=25

Oh okay, thanks for the info.
It can peak up to 2A and the module I have uses LDO 29302WU, so I'd probably go with that same in my own circuit, if I have to use 2 regulators. I just thought that I might get away with only 1 regulator. Just my nature trying to minimize the component count, although I'd only make max. of 5 units.

SirNickity:
AVRs are well-suited to run at 3.9v. When you buy a complete Arduino or clone, the rest of the stuff on the board has to also run at the supplied voltage, which is not a given, so that's something you'll need to verify before using a pre-built dev board. If you build your own PCB, you are in control of that, of course.

I will use barebone Atmega328.

I just tested the 3.9V with my other project I got on the table and I was able to go down to 3.85V until some I2C data got messed up. But this project had 20x4 LCD screen, I2C devices and everything. 3.85V! I always thought that AtMegas were 4,5 to 5,5V!
I think I can really use the 3,9V, but if not, then I just use another regulator.

Thank you guys!

They'll work a lot lower than that even, but you may have to tweak things. For e.g.: Set the brown-out protection fuse to allow a safe margin below your minimum Vcc; Make sure any pull-up/pull-down resistors are adequate but can be overridden with valid signals; Verify current-limiting resistors are low enough to pass sufficient current; Check that your oscillator settings are compatible with your oscillator at that voltage; Make sure the datasheet certifies the clock speed you intend to run at the minimum Vcc you will allow.

According to Figure 29-1, the part will run and be fully compliant at 16 MHz down to about 3.8V.
I've done the math before, can't find the chart I had pasted that into tho.