Sizing PSU for a Project

[fagiano]

Hi, I'm trying to choose a power supply for my project. I want to use a "off the shelf" wall wart
but I'm wondering what rating to choose.

My project uses a peltier element that I want to power with around 1.6A plus I have a MCU a couple of thermal sensors and wifi transmitter. All togheter I don't think I pass 1.7A. How much should I oversize my psu, 2A? or I shold give it more headroom?

Also, given that my digital stuff is regulated at 3.3V can I drive the peltier straight from the PSU(+ power MOSFET) or I should also add another voltage regulator?

ciao
Alberto

[Electr0nicus]

First of all:
Do you use linear dropout regulators for your µC and WiFi stuff, or are you using a switch mode PSU approach?
WIFI modules consume very much power, and they also have very low supply voltages. So 200-300MA on the 3.3V rail for the Wifi module is very common.
The µC, f.e. a AVR, doesn't need very much power.  So let's assume 350mA on the 3.3V rail (WiFi + µC)

Then assume you have a 12V wall wart.

If you using a linear regulator from 12 to 3.3V, you will dissipate (12-3.3V)*350MA = 3.045W
That means, you will need heatsinking for f.e. a TO220 regulator. The positive thing about linear regulators is, that they are extremely easy to built. Just treat them some bypass caps, and they are fine. Also consider that the current flowing out of the linear regulator also has to flow in, thus when you have 3.3V 350mA on the output, 12V 350mA have to be applied on the input. So If your pelitier element uses 1.7A and Wifi + µC 350mA, you are already above 2A. So I would use a 2.5A or 3A wall wart.

Switching regulators, on the other hand, are very efficient, but complicated to built, dependend on your experience.

Driving the pelitier element directly with, f.e. a µC output pin, which has 3.3V, is totally fine. Just remember to select a logic level MOSFET, with a V_gs(Th) of 2.5V or lower.

[fagiano]

First thanks for your reply. For the digital part I'm pretty much copying the devboard I'm using for prototyping(the spec claims 250mA peak with WiFi) plus I've added a IO expander and a couple of
thermal sensors; 350mA is a good bet. All this is behing a voltage regulator LM1117. To switch the peltier I have a logic level power mosfet that I already tested with this setup.
Currently I have my breadboard prototype feeding the digital part behind the voltage regualtor and feeding 5 to 8V directly from my bench PSU to the peltier + mosfet.

When I meant running the peltier directly, I meat directly from the wall wart with no regulator in the middle.
I was wondering if would be ok to just split the output of a wall wart and connect one part straight to the peltier + mosfet and another part to the 3.3V regulator that goes to the digital stuff.
I measured the current consumption of the peltier and it seems ok to run it from 5 to 7V. It would draw 1 to 1.6A straight from the psu. Because the element can tolerate up to 15V and my power mosfet is rated up to 30A I thought there is no reason to regulate the voltage/current. Am I making any sense?

sorry for the silly questions but I'm a newbie, 6 months ago I didn't know what a capacitor was for.

thanks for your time

ciao
Alberto

[mariush]

7.5-9v @ 1A adapters will be easy to find - most routers, switches, scanners, faxes  etc are using these adapters so it would be easy to find some good adapters on eBay or goodwill or some overstock store.

Also super easy to find and cheap will be monitor power bricks, those are usually 12v @ 2-4A .. ebay is full of these, from cheap and dubious quality to original bricks from monitors that die of other reasons (such as kids punching the screen)

The first ones i mentioned are unregulated, they'll go up to about 11-13v with no load, or little load.  The monitor adapters are usually pretty close to 12v at most loads

Personally, I'd go with a MC34063 or something similar to step down the input voltage to about 3.6-3.8v then use a small linear regulator to get 3.3v.

For example, NCP4589 can do 300mA with 0.23v drop at maximum load : http://uk.farnell.com/on-semiconductor/ncp4589dsq33t1g/ldo-300ma-3-3v-tri-5sc70/dp/1885143

They're cheap, use two if needed, or if you want go for something a bit more expensive but which can do let's say up to 1A like MCP1826 with only 0.25v dropout as well:

http://uk.farnell.com/microchip/mcp1826s-3302e-db/ic-ldo-3-3v-1a-sot-223-3/dp/1578424
http://uk.farnell.com/microchip/mcp1826-3302e-at/ic-ldo-3-3v-1a-to-220-5/dp/1578413
http://uk.farnell.com/microchip/mcp1826s-3302e-ab/ic-ldo-3-3v-1a-to-220-3/dp/1578423

I've only pasted and searched for fixed 3.3v, you might find adjustable versions are even cheaper.

Anyway, the idea is that even with a switching regulator like MC34063 being 75-80% efficient then use a low drop regulator, you'll waste less than 3 watts as opposed to using a linear regulator directly. The LM1117 would require about 1-2v ABOVE 3.3 to give smooth output, these ones above only need about 0.25-0.3 volts.

You don't say at what voltage your peltier element requires 1.6 A so I don't know if it would be enough
or you could go with something a bit more expensive

[fagiano]

mmh I'll have to get one of this MC34063 and learn a bit more about this kind of regulators  .

the peltier I'm using can go up to 15V and suck up to 65W. However for my tests the sweet spot for me is to drive it with 6V to 7V. At these voltages it draws respectively 1.6A and 1.92A. Around 12V  it would draw probably around 4/5A that is way behond my needs. The cooling effectivness flattens out after 8V-2A as my heatsink is not able to dissipate enough heat.

thank you for your advice

Alberto

[mariush]

Dave has a video with everything you need to know when it comes to using that chip:



There's also a page which does all the calculations for you based on the formulas in the datasheet:

http://dics.voicecontrol.ro/tutorials/mc34063/

You don't need exact values for the parts most of the time, you can round them up a bit especially if you do the calculations for a current output a bit more than what you need, you'd only lose a bit of conversion efficiency.

This mc34063 is very cheap and simple to use, there are other chips a bit more expensive but also more efficient, but it doesn't seem it would be much of a benefit for it.

I understand that 6-7v would be the sweetspot for you, but do a bit of market research and see what's available at lower prices on the market.
You may find out that a 12v monitor power adapter or some 18-19v laptop charger would be cheaper than a standalone 7.5v 2A wall wart style of adapter. The cheap ones usually don't do more than 1A and the really cheap ones often can't even do 0.5A even though the label says more.

In a pinch, you could also use a regular computer power supply, and get it running simply by shorting the power good wire in the atx connector to ground. Cheap computer power supplies are also everywhere

[fagiano]

Thank you! I missed that episode, very educational. I wish Dave would do a bit more newbies videos. Anyway that gives me plenty to work with.
I think I'll give a try to a 12V wall wart; I have several in my computer parts box.
I did some investigation and for this project, seems that the LM2576-ADJ is a better fit, as it can handle 3A. I definitively do some experiments with the MC34063. Tomorrow, as I get the parts, I'll see if I can blow up something

thank you again!

ciao
Alberto

[McMonster]

I did some investigation and for this project, seems that the LM2576-ADJ is a better fit, as it can handle 3A.

You can use an external transistor with MC34063 to get more current.

[Flávio V]

The problem with a "better than mc34063" switcher here in portugal is that there isn't any better....(there are lm2576s but they are overpriced to near 4-6€)

I will still figure out what does't works in my lm2576 power supply.....

[mariush]

But you don't need 3A of power...

Remember, the idea is to only use this MC34063 to get a noisy 3.5-3.7v 3-400 mA from whatever input voltage you have, which you can then filter down to 3.3v using a cheap low dropout linear voltage regulator.

It's for the DIGITAL part of your circuit, which you said will only need 300-350mA at 3.3v

The MC34063 avoids the problem of having the linear regulator dissipating 3-4 watts into the air (you would need a good expensive heatsink to keep a linear regulator cool when it does 12v to 3.3v directly). As the linear regulator now only does 3.5-3.7v to 3.3v, the power wasted in the worst case is  (3.7-3.3)  x 0.3A = 0.12 watts, which can be easily dissipated by the regulator pins and traces on the pcb.

The peltier element is not something that requires an accurate input voltage, so you don't need a linear regulator to get those 2-3A of power to it. 

[fagiano]

I was thinking to use the LM2576-ADJ  to go down to 6/7V+, then pass that directly to the peltier and through the linear regulator at 3.3V for the MCU.
If I pass 12V straight to the peltier, it is going to draw a lot more current than I planned(60W) and I'd have to regulate current at thet point.

Alberto

[fagiano]

I just built a step-down converter with the LM2576-ADJ on a breadboard 12V to 7.2V(theoretically is actually 6.8ish). Anyway, it gets kinda hot("cannot keep finger on heatsink too long" hot) when I attach a 1.7 to 2.0A load. As it gets hot, the voltage drops and it starts to draw less current. Am I crazy for building such a circuit on a breadboard? Is it supposed to get that hot?

thank you for your time

ciao
Alberto

[mariush]

Well, read the datasheet, make sure you selected a proper inductor, output capacitor, schottky diode etc

http://www.ti.com/lit/ds/snvs107b/snvs107b.pdf

the datasheet has a whole set of pages with the formulas and recommendations .. page 14 and forward.

A breadboard is not really a good place to test switching regulators, they work best with large pads of copper on the ground and voltage traces
See also the section Heatsink/Thermal considerations in the datasheet.

It's natural for the IC to heat up a bit, after all it has a lot of stuff in it, error amplifiers, comparators, drivers, transistors.. but they may be kept cooler with the proper inductors and diodes.   

[SeanB]

Needs proper layout, definitely not a breadboard. Needs short, wide, low inductance tracks between the input capacitor, output inductor, diode and output capacitor at least, as well as a good low inductance ground plane. Birds nest it with a chip on a piece of plain PCB using short leads and you will have better performance. Read the Appnote from pages 13 to 20 and note where they show you need short tracks for certain components and why.

[fagiano]

Ok, fair enough. I did go though through the datasheet, I had a strong feeling that a breadboard was not the ideal test. I was just too curious to see something moving, and I couldn't wait until this weekend to lay out a board(I gotta go to work). At my newbie stage this stuff is very exciting .

Basically, I have built "Figure 2" of the datasheet. I bought the exact components suggested in the datasheet. The math for the output seems to work out fine. Just the heat was a bit scary. Actually, I had a 2.2K resistor instead of a 22K. Now the voltage and current are stable and the heat is not as terrible but still too high.

anyway, thank you soo much to everybody!! I'm looking forward to make a board!

ciao
Alberto