Main Site Documentation

Custom Power module needed


I need to build a custom Power supply module.
It needs to fit in a quite tiny housing and I want to build it on a breadboard. (I know that this is a bit converse)

I think I need a similar thing like on the Power Extender Module. (see picture)

I want to power a MedusaMini + N18 + RS485 + Temp & Humidity.
So I think 500mAmps output on the 5V DC/DC converter should be enough.
My external power supply is 12V but it should work with lower voltage as well (longer wires through the whole house).

I also want it to stay as cool as possible to minimize the effects on the Temp & Humidity module.

Could anyone give me some tips which parts with DIP housing to use?


I would use one of the Usb wall chargers. They are very small. They are getting cheaper and cheaper, Have 1000-2000mA and use standard USB cables. You can cut the Usb cable and put suitable connector for your board.


A USB wall charger has AC input and 5V output, right.
I need 9…12V DC input and 5V + 3.3V output and not more than 500mAmps.

And cheap wall chargers usually get warm or even hot. I can’t have that here.

But thanks for the tip anyway.


I would use the LM2596. You can get modules pretty inexpensively

You can use one for 5V and the other for 3v3 or just a linear regulator for 3v3, but it will heat up somewhat. Even the LM2596 will heat it some, if you really need 500mA. That is 2.5W output. At more than 80% efficiency you will need to put in 3W, with the other 1/2W turned into heat. If most of the draw is at 3v3(say 400mA) the linear regulator will need to turn another 0.7W into heat, so using two would probably be a good idea.


I use these little 78 replacements from Murata.

They never get hot, even when drawing 1.5 amps.

Or this one from Recom which is only 500mA which is what you asked for.


Thank you all for the tips.
In fact, the only module that needs 5V is the RS485, and it says the power consumption is 0 mAmps at 5V. (In fact I assume there will be a minimum current < 1 mAmp).
So I assume all of the power is needed at 3.3V.
If I use 2 separate DC/DC converters the 5V can be even more low in power.

I’ll dig into it over the week.

@ GHI: does the RS485 Module really need the 5V, or is it somehow possible to use it with 3.3V only.
How clean does the 5V need to be?, can I distribute the 5V line for RS485 separately over long cables?

Then I would supply my devices with 5V directly and only have to create the 3.3V volts from it.
How exact must the 5V be for the RS485 module. Would 4.5V work as well (assuming some voltage drop on the cable).


You can also use a 3.3V tranceiver like


the 5v is used in the GHI RS485 module, it’s directly powering the MAX13487. Check schematics on the catalog page.

No matter what you end up using, you will need to take into account voltage drop. The reason power-over-Ethernet uses higher voltages is that you get less voltage drop through the resistance of the conductor between source and destination. You then need to transform it (efficiently) down to your desired levels. If I was approaching this I’d be using at least 12v power in and transforming that down to the 5v/3v3 at the destination.


That was my orininal plan.
But I want to use DIP components and I want as less power as possible to be transformed into heat.
But when I look at farnell (or similar) I get so many DIP DC/DC converters with 12V input and 5V output at 500mAmps, and I can’t tell the difference.
So that’s my problem so I wanted to ask if someone with more experience could help me out.
Even if it’s just something like a “don’t use a converter with this or that in the sepc” or “this feature means high efficiency = low heat output”, …


a switching reg like @ Dave has pointed out is efficient and doesn’t generate much heat during the down-conversion. Linear regs are the nightmare for heat, so avoid them. Those Murata modules (or even a jelly bean LM2595 module from ebay in a pinch) are not bad at their job, and save mucking with this


Yes, use a switching regulator. They are not all the same, but most are relatively similar. The data sheets can be helpful in discovering the differences. You seem to be concerned about size, efficiency, price (I would guess) after the basics like input and output voltage and package are satisfied. The other thing is how many external components they require.

The efficiency at the input voltage and output voltage/current you need will determine how much energy will be turned into heat by the regulator. With a switching regulator this will likely be 5% to 20% of the output wattage (a linear one could be 250% or more). With a high efficiency regulator you will be able to minimize the temperature rise due to the regulator. However, if you really need 400-500ma at 3v3, that is almost 1.5W that will be dissipated and need to be removed from the enclosure. I have noticed that while the N18 display doesn’t get “hot”, it is several degrees above ambient.

I have read reviews of some of the newer smart thermostats that they read high. Many of them employ a simple offset adjustment, which allows the user to calibrate them. Unless you can figure out a way to get all that heat out (perhaps the fan, suggested earlier) or use a remote sensing device, you will need to figure out an acceptable “calibration” method to get the results you want.


Do you mean the fan I suggested in this post?

That is far too small and weady to cool the entire device, but it might just keep the sensor closer to room temp if insulated from the heat source a bit.

The offset calibration should be a fixed value so long as the average power consumption stays the same and excluding drafts and things. Remember that one way of measuring power is by measuring it’s heating effect. One problem is that you do introduce is a bit of lag due to the specific heat capacity of the materials.


The device used by GHI is an auto direction IC from Maxim. It is only available in 5V at present. I have not been able to locate any 3.3V transceivers with this capability. I use this same device and I’ve had to use 5V (luckily devices with USB ports tend to have this available) and use a level converter in between, as GHI has done.

The datasheet shows 4.75 to 5.25.

As the current drawn is very small, you are unlikely to see a great voltage drop unless your cable is really long. How long will it be?

If you can do this and the current and cable is not too long, just fit a local capacitor pair at each module to clean up the supply. :slight_smile:


I’m not totally sure about the cable length, but I guess it might be up to 10 meter.
And I wanted to use a standard 4x2 twisted pair cable like a CAT 5.
Since RS485 is a linear bus and I have a star architecture I wanted to use one pair for RS495 in, one for RS485 out and two pairs in parallel for power.

But since I don’t want to make too big experiments here (it has to be wife compatible, and by this it should instantly work) I guess I go by using two regulators, one 12 to 5V and one 12 to 3,3V. Or would it be better to use a 5 to 3,3V here?


@ Dave McLaughlin - ?


Nope, that’s just a 3.3V device which I can easily find. :slight_smile:

The MAX13487 has auto transmit enable without the need for an extra IO port to control the transmit and receive state.

Great for the likes of Android or Linux systems where you don’t have access to the transmit buffer register to know if it is cleared or not, before switching back to receive.


I did wonder :slight_smile: