Low current (~100ma) buck voltage converter simple, low cost options (48V to 12V

[dominicM]

I have 48V AC-DC power supply that I use to drive high power LED drivers. I also want to use the same 48V input to power Arduino Pro Mini.

I considered linear voltage regulators but with LM317 the efficiency would be 2.77%!   

I also considered voltage divider but wasn't able to get it to work. With high resistor values the voltage dropped drastically when load was applied. With lower resistor values they would dissipate around 3W with one configuration for example.

Lastly switching power regulators aren't ideal because the cost is quite high for input voltages over 40V such as over 2 euros for a single texas instruments LM25019. Then there's also additional components like inductors that are out of my scope of knowledge... The total cost for all components would be about 7 euros which is massive considering cheap ebay AC-DC power supplies are only 2-3euros, surely DC-DC conversion should be simpler than AC-DC?

Are there any other alternatives? What would you do in this case?

P.S. Sorry in advance if this is should be posted in beginner section, it didn't seem entirely beginner level.

[mazurov]

If input voltage variation is small (as it should be while driving LEDs) you don't need the whole DC-DC converter. Build a buck power stage, make it synchronous if you want efficiency, then add a square wave generator with settable duty cycle. I made one recently to drive 12V LED heatsink fans from 48V, works very well. It is not going to be the cheapest solution though, 2x60V MOSFETS and buck MOSFET driver will set you back ~$US8-10. Even auto grade monolithic controllers capable of delivering your current are still cheaper.

[dominicM]

Yes, the input variation should be minimal, though it does not matter as the 48V input is fed to constant current buck driver which in turn drives the led.

The cost is the same as the switching supply so it's not a great option. It seems odd that is it so expensive for such a low current solution. Is there anything else that could work? Maybe with lower required power output (~50mA)? Efficiency is important up to a point. 20% efficiency would be just fine but 2.77% is just horrible...

[mij59]

Hi,

First you need to calculate the total required power for the 5V supply, at this point your just guessing.
Maybe you need other voltages to complete the project.
You could consider a separate power supply.
The LM317 is a adjustable voltage regulator, a voltage regulator with a fixed output would also do.
The standard version of the LM317 has a max. voltage rating of 37V, so you would need the LM317HV version.
Efficiency is not so important in this case, a 5V linear regulator with an output current of 50mA would dissipate 2,15W.

[Whales]

For the LEDs: you could try assuming the 'load' would be stable and go for:
 - 555 timer providing an adjustable duty-cycle square wave (using a potentiometer)
 - control a MOSFET switching the 48V using this signal
 - smooth output via a capacitor

Extra complexities:
 - a 12V or similar linear regulator will be needed to power the timer/driving of the MOSFET (less than a few mA will be drawn)
 - LED electrical characteristics change as they heat up, so you will probably want to adjust your duty cycle to suit the LEDs when they are 'hot'. 

For the Arduino: a linear regulator might still be fine.  Calculate the max power draw & work out how many watts will be lost across the regulator.

[mariush]

Ok, so look at it another way.

You need 5v @ 100mA ... that's 5*0.1 = 0.5 watts.  For safety, let's say you need 0.15A and therefore the power required to produce 5v@0.15A is 0.75w.  A good switching regulator will typically have an efficiency above 80%, especially with so high input voltage / output voltage difference but even with 75% efficiency, you'd need about 1 watt of power from the input.

Switching regulators that are capable of working directly from 48v are few and expensive, so a good idea would be to reduce this voltage so that it will be within a range that's supported by most switching regulators.  That would be 3v --- 38-40v.
Let's go with 35v, just to be on the safe side (in case your 48v DC power supply is not quite 48v).

With a 35v input voltage, your switching regulator will pull about 1w / 35v = 0.02857 A  or about 28.57 mA.  Let's round it up to 30mA. 

So you need to find something that drops the 48v to 32v and let's 30mA go through. Well, V = I*R  so  (48-35) = 0.03* R  therefore  R = 433.33 ohm.  Smaller value means less voltage drop but more power wasted in resistor. Larger value means higher voltage drop and higher current limit.
433 ohm is not a standard value, so let's go either:
* 1 x 390 ohm = 390 ohm
* 330 ohm + 100 ohm in series = 430 ohm
* 270 ohm + 150 ohm in series = 420 ohm
* 2 x 820 ohm in parallel = 410 ohm

For 390 ohm,  you'd have (48-35) / 390 = 13/390 = 0.033 A  and the power dissipated in the resistor will be P = I*I*R = 0.43w so in theory a 0.5w resistor will be enough but in practice a 1w rated resistor will be safer.
With higher value resistors, the power dissipated will be lower, but not by much, it will still be safer to use a 1w rated resistor.

So now you can choose any switching regulator that works with around 35v at input... even a cheap MC34063 switchin regulator works with up to 40v, so it would be fine: http://www.digikey.com/product-detail/en/MC34063AP/296-17766-5-ND/717432  (note that some only work with up to 36v or 38v, depends on manufacturer)
You can use online calculators ( http://www.nomad.ee/micros/mc34063a/ or http://dics.voicecontrol.ro/tutorials/mc34063/ for example) or the one on Sourceforge ( http://sourceforge.net/projects/mc34063uc/ )  to figure out the inductors and diodes and their specs. Everything should cost you less than 2$.

Now keep in mind that the switching regulator will not always use 30mA. If you don't use 5v  @ 0.1A,  the switching regulator will not use 30mA.  The switching regulator doesn't act like a LED so you can't leave it at that.

48v + ----- 390 ohm --------- regulator --- 48v -

At 30mA , the regulator appears as a 35v / 0.03 = 1166 ohm resistance.   You have 30.85mA going through the 390 ohm resistor, 12v drop and 371 mW dissipated.  The 1166 ohm "resistance" will see 35.97v on it and waste about 1.11w.
But at 10mA, the regulator appears like a  resistance of about 4700 ohm .. 390 ohm resistor will see 9.45mA flowing through it and drop only 3.68v, and the regulator will see 44.32v  which is more than the 40v maximum so that's not good.

(going to continue this post, editing it now)

[cosmos]

Ok, so look at it another way.

You need 5v @ 100mA ... that's 5*0.1 = 0.5 watts.  For safety, let's say you need 0.15A and therefore the power required to produce 5v@0.15A is 0.75w.  A good switching regulator will typically have an efficiency above 80%, especially with so high input voltage / output voltage difference but even with 75% efficiency, you'd need about 1 watt of power from the input.

Switching regulators that are capable of working directly from 48v are few and expensive, so a good idea would be to reduce this voltage so that it will be within a range that's supported by most switching regulators.  That would be 3v --- 38-40v.
Let's go with 35v, just to be on the safe side (in case your 48v DC power supply is not quite 48v).

---- snip

I like your approach, but just want to caution that using resistors to bring the voltage under 40V means the solution depends on the load being there, no load means 48V again and risk that poof goes the switching regulator.
Might want top look at using Zener(s) instead? 
With a ~10V Zener you should be able to go closer to the max rating of the switcher, hence less losses in the preregulation too.

[mariush]

cosmos, I agree, and I was  editing the post as you wrote the message. I'll edit this message and add more info and a possible solution.

Now keep in mind that the switching regulator will not always use 30mA, it's not a LED to suck as much power as the 390ohm resistor will allow it to suck.
If you don't use 5v  @ 0.1A,  the switching regulator will not use 30mA.

48v + ----- 390 ohm --------- regulator --- 48v -

If your device uses 100mA at 5v (5v*0.1A = 0.5w), the regulator will use about 0.66w to produce this if it has that 75% efficiency.

So your regulator will behave like a resistor of about 2600 ohm in series with that 390 ohm regulator:

390ohm : 16.05mA flow, 6.26v drop , 100mW wasted
regulator: 16.05mA flow, 41.74v drop , 670mW wasted 

Well that's not good, we wanted 35v.

At 10mA on 5v (5v * 0.01A = 0.05w), the regulator will use about 0.066w to produce this with 75% efficiency, so your regulator will behave like a 34 kohm resistor:

390ohm :  1.4mA flow, 0.54v drop, 0.76mW wasted
regulator: 1.4mA flow, 47.46v drop , 66mW wasted

Much worse. Now your regulator is close to 48v.

So what can you do... Well, if you can make sure the regulator will always have a minimum load, for example by connecting a power on led to it to waste about 10mA, you can then add another resistor in parallel with the regulator to make the overall "virtual resistance" change.

Let's go with a 1800 ohm resistor in parallel with the regulator.

Now at the minimum load of 10mA, your regulator uses 66.6mW the 1.8k resistor takes some of the load:

390 ohm : 32.34mA flow, 9.1v drop , 212mW wasted
1800 ohm  : 21.61mA flow, 38.9v drop, 840mW wasted
regulator : 1.73mA flow, 38.9v drop , 67mW wasted



So the regulator is happy with 39v on the input and outputs 5mA at 75% efficiency and overall you're wasting 1,119mW to do it, or an overall efficiency of 4.45% - hey, it's bad, but still better than a linear regulator.

With 50mA on 5v, things change a bit. With the same 75% efficiency, the regulator will need 5v * 0.05 * 100/75 = 0.333W or 333mW to produce that, so we're looking at:

390 ohm : 29.4 mA flow, 11.47v drop, 337.63 mW wasted
1800 ohm : 20.29mA flow, 36.53v drop , 741.16 mW wasted
regulator : 9.13mA flow, 36.53v drop, 333.52 mW wasted

You now have a reasonable 36.5v on the regulator and you used 1411mW to output 250mW so about 17.17% efficiency



With 100mA on 5v, it's even better. Now you have 666mW wasted in the regulator and about 33v on regulator, but you're only using about 1850mW to do it (in the picture below, 675mW in regulator, 1862mW overall, so 675*100/1862 = 36.25% efficiency).




So both resistors will have to be 1w rated. Of course, you can play with the values in that circuit simulator and get some compromise. But just two resistors and a minimum load (a led using 10mA for example) can be enough to allow you to use any cheap 20 cents switching regulator (excluding inductors and diodes which add to the cost)

[SeanB]

There are a lot of telecoms regulators designed to run a MCU off phone line voltage, either isolated or not, and they are pretty rugged and often have a cheap development board available for free inside corded phones with a digital display.

[Alex Eisenhut]

Are there any other alternatives? What would you do in this case?

P.S. Sorry in advance if this is should be posted in beginner section, it didn't seem entirely beginner level.

Find your budget and shop for buck modules on eBay.

Or look on eBay first and set your budget...

http://www.ebay.com/itm/DC-DC-LM2596HV-Buck-Converter-5V-60V-to-1-25V-26V-Power-Module-48V-to-3V-5V-12V-/301216970372?pt=LH_DefaultDomain_0&hash=item4621ee3684


http://www.ebay.com/itm/DC-TO-DC-Buck-Converter-Step-Down-Module-18-48v-to-5V-1-5A-12V-1-5A-dual-out-/261422288256?pt=LH_DefaultDomain_0&hash=item3cddfb8d80

Naturally I'd try them out on a dummy load first...

And make sure you never exceed the maximum input voltage under any and all conditions.

[dominicM]

Hi,

First you need to calculate the total required power for the 5V supply, at this point your just guessing.
Maybe you need other voltages to complete the project.
You could consider a separate power supply.
The LM317 is a adjustable voltage regulator, a voltage regulator with a fixed output would also do.
The standard version of the LM317 has a max. voltage rating of 37V, so you would need the LM317HV version.
Efficiency is not so important in this case, a 5V linear regulator with an output current of 50mA would dissipate 2,15W.

First I need to clarify that it does not need to be 5V. The assumption you made is that since it's arduino it's 5V but Arduino Pro Mini has built in regulator that can take anywhere from 3.5V to 12V. Since one sensor needs at least 5V my new range is 5V to 12V. The arduino itself if 3.3V version and this is my primary voltage for most sensors/devices.

The absolute maximum current draw by arduino is 150mA, this is limited by the internal voltage regulator among other things. Nominal (measured) amperage is ~63mA with minimal change except possibly at start-up/shutdown or another unknown (to me) event.

LM317 maximum voltage is not 37V, in fact it has no specified max voltage. It has maximum voltage differential of 40V which means it would work with 48V input and 8V (or higher) output: 48-8=40V

The problem with low efficiency is that  I would like to keep my device on 24/7. 2.15W wasted heat + power used by device would add up quite quickly over a month 24/7 especially considering I plan on having one of these in every room.

[dominicM]

Are there any other alternatives? What would you do in this case?

P.S. Sorry in advance if this is should be posted in beginner section, it didn't seem entirely beginner level.

Find your budget and shop for buck modules on eBay.

Or look on eBay first and set your budget...

http://www.ebay.com/itm/DC-DC-LM2596HV-Buck-Converter-5V-60V-to-1-25V-26V-Power-Module-48V-to-3V-5V-12V-/301216970372?pt=LH_DefaultDomain_0&hash=item4621ee3684


http://www.ebay.com/itm/DC-TO-DC-Buck-Converter-Step-Down-Module-18-48v-to-5V-1-5A-12V-1-5A-dual-out-/261422288256?pt=LH_DefaultDomain_0&hash=item3cddfb8d80

Naturally I'd try them out on a dummy load first...

And make sure you never exceed the maximum input voltage under any and all conditions.

As my options look increasingly limited, I might have to resort to ebay cheapos    Wanted to avoid this as I would need to have 2 individual PCB's with cables running from 1 to the other and back, very messy...

[dominicM]

For the LEDs: you could try assuming the 'load' would be stable and go for:
 - 555 timer providing an adjustable duty-cycle square wave (using a potentiometer)
 - control a MOSFET switching the 48V using this signal
 - smooth output via a capacitor

Extra complexities:
 - a 12V or similar linear regulator will be needed to power the timer/driving of the MOSFET (less than a few mA will be drawn)
 - LED electrical characteristics change as they heat up, so you will probably want to adjust your duty cycle to suit the LEDs when they are 'hot'. 

For the Arduino: a linear regulator might still be fine.  Calculate the max power draw & work out how many watts will be lost across the regulator.

Well, I am using LED drivers that take 48V as input and output constant current to the LED directly so I don't need anything else to drive the LED's. Potentiometer would be no good as well as everything will be controlled over wireless comms. I am not really sure what you mean by assuming the load?

[mariush]

The problem with low efficiency is that  I would like to keep my device on 24/7. 2.15W wasted heat + power used by device would add up quite quickly over a month 24/7 especially considering I plan on having one of these in every room.

1kwh = $0.10 - $0.30, even cheaper depending where you live.

2w * 24 * 31 =  1488w = 1.488 kWh or less than half a dollar a month.

Electricity costs is not a problem, heat is.  With 48v in and 12v out and 100mA, you're going to dissipate 36v * 0.10 = 3.6w - without heatsink, the regulator will heat up about 55-60c/w so you'll need a heatsink that's going to be large.
A switching regulator will not waste as much power as heat, but like I explained switching regulators that accept 40v+ are more expensive. It's cheaper to use a couple of 1w resistors to dissipate a couple of watts (in worst case) safely. At 60mA, if you check my post above, you'll have less than 1w wasted in the two resistors.

And by the way, the Arduino Pro Mini uses a MIC5205 as a voltage regulator: http://www.micrel.com/_PDF/mic5205.pdf
If you check the datasheet, at page 3 you'll see that the regulator works at up to 16v - they probably say 12v so that stupid kids won't damage the chip with 16v or maybe because the copper pad on which the regulator sits may not be large enough to keep the chip cool.

[dominicM]

The problem with low efficiency is that  I would like to keep my device on 24/7. 2.15W wasted heat + power used by device would add up quite quickly over a month 24/7 especially considering I plan on having one of these in every room.

1kwh = $0.10 - $0.30, even cheaper depending where you live.

2w * 24 * 31 =  1488w = 1.488 kWh or less than half a dollar a month.

Electricity costs is not a problem, heat is.  With 48v in and 12v out and 100mA, you're going to dissipate 36v * 0.10 = 3.6w - without heatsink, the regulator will heat up about 55-60c/w so you'll need a heatsink that's going to be large.
A switching regulator will not waste as much power as heat, but like I explained switching regulators that accept 40v+ are more expensive. It's cheaper to use a couple of 1w resistors to dissipate a couple of watts (in worst case) safely. At 60mA, if you check my post above, you'll have less than 1w wasted in the two resistors.

And by the way, the Arduino Pro Mini uses a MIC5205 as a voltage regulator: http://www.micrel.com/_PDF/mic5205.pdf
If you check the datasheet, at page 3 you'll see that the regulator works at up to 16v - they probably say 12v so that stupid kids won't damage the chip with 16v or maybe because the copper pad on which the regulator sits may not be large enough to keep the chip cool.

I am not in the US and the power cost could be as high as 0.35USD. Like you mention the heat dissipated is 3.6W + 100ma load. with 4 devices in the home (could be more) thats 3.7W * 24 * 31 = 2.752W * 0.35USD = ~3.4USD * 4  devices = 13.5USD. This is insanely high for something like this. It should be no more than 1USD a month (with some power saving tricks). The device is meant to reduce power consumption so it's not something I can live with long term.

Will have to look into voltage dividers more then as it's even simpler than LM317 and more efficient like you say.

It's a clone so it's probably a good idea to stay under the limits just in case. And I actually did blow one up with a badly regulated dc adapter Capacitor went, maybe it was just defective or it couldn't handle ~15.6 voltage spike at start-up...

[dominicM]

@mariush

That took me a while to read and understand

I think I will order MC34063. Not sure where you saw 20 cent regulators though, the cheapest I saw was 40 cents(euro)
It's still far from ideal but I don't think there's a much better solution. Will use your resistor suggestion and will try to source 36V PSU (few table-top are available in 36V) in the future or even 24V with lower Wattage led might work. Will have to try it out and see how if it works.
Also would using 12V instead of 5V make a significant difference or is it fairly close?

Thanks a lot for such a detailed answer!

[mariush]

I'm in Romania, so I added the 24% VAT to the price listed on the sites and converted to dollars using google:

21 cents each if you buy 25pcs :  http://www.tme.eu/en/details/mc33063ad/voltage-regulators-dc-dc-circuits/texas-instruments/#  (33063 is just a (better) version of 34063, both are made by lots of manufacturers)

DIP version is a bit more 'expensive' at $0.38 if you buy 10 pcs : http://www.tme.eu/en/details/ka34063/voltage-regulators-dc-dc-circuits/fairchild-semiconductor/#

RS-Components have them for 0.24 euro if you buy 40, a bit more if you buy smaller quantity: http://ro.rsdelivers.com/product/on-semiconductor/mc33063avdg/swregulator-15a-step-up-down-inv-soic8/7862401.aspx

Farnell has them for about 40-50 us cents each if you buy 10pcs or more.  You can get clones and no-name chinese versions of 34063 on ebay, probably 5-10 for $0.99. You can get car usb chargers that have 34063 in them for about 2-3$... they're cheap and simple to use, but have almost no protections so you can kill the chips easily.

[dominicM]

Ah, yes that's quite cheap, oddly enough though ebay has them for 2 euro+ for one plus shipping...

[dannyf]

A resistor divider will most surely not work.

you can have a 48v regulator. You can have two serial regulators , or a pre-regulator plus a regulator, ...

[dominicM]

Can you be more specific? Why wouldn't resistor divider not work?

How would "serial regulator" work? Can you give an example of what pre-regulator I could use?

[dominicM]

@mariush

I remembered that I attempted to use this regulator: http://www.mouser.com/catalog/specsheets/MPS_MP2456_r1.0.pdf

It's 50V and only 0.65EUR a@ Mouser

Looking back at the datasheet I am not sure why I gave up on it. What do you think about the device? Would it work and is there anything that might trip me up? It's not a popular component so there's very little info on it and the datasheet is not that great.

[mikerj]

For a very cheap and cheerful switching regulator with high input voltage capability you could do worse than the Roman Black switcher.

[dominicM]

An interesting solution but if I can I will try to stick with something more common. I am confused enough

[dannyf]

The output from a resistive divider is load variant. To minimize that, you have to run them at high current levels, which leads to inefficiency.

linear regulators can also be configured as switching regulators so check the datasheet for examples.

[dominicM]

Saying it woun't work is misleading when it wclearly  works just with low efficiency. Either way, efficiency is relatively good compared to linear regulator.

I examined the LM317 datasheet with nothing resembling a switching regulator configuration. It's not very useful to say you can do x without any explanation. In any case I don't see why you would have switching regulators if linear regulators can work as switching regulators, it just doesn't make sense. Please explain what you mean.