From 28V DC to 3.3V DC

[cv007]

Powering a 3.3V mcu from 28V DC, probably using about 5mA or less normally but will plan on 20mA in case something changes. The mcu probably will see little use and normally runs for maybe 20-30 seconds, but could occasionally be a few minutes. Environment probably can be up to 100F, and no air movement (enclosed case). This will also be using only through-hole components.

Even with the low current heat seems to always be a concern when dropping that much voltage. I am looking at a AP7381-33V-A (TO-92) which has 165C/W for its ambient thermal resistance. If I am dissipating ~450mW (24.7v x 20ma) at an ambient of 37C it is less than the 150C limit (165*.45+37C =~110C). Still seems too hot.

My main question- would it be a good idea to spread some heat to another device, such as a 15V 1W zener diode? Let the zener take a good portion of the heat and let the voltage regulator be more 'comfortable'?

[thephil]

I'm not an expert on thermal design but I'd probably use a buck converter for that kind of voltage step-down. 

[cv007]

I guess I should have mentioned I would rather not do dc-dc in this case. Trying to keep components simple and I think heat can ultimately be managed here, and if needed the next step would be to use something like a TO-252 package to get the heat into the pcb.

[Terry Bites]

I've often used the LV3842. You only need a few extenal components to get a fully functional regulator.
It will operate at 100 real degrees without complaining. It will shut down at 170'!
Lets say its 75% efficient at 20mA 28V in and 3.3V out, the power loss is only ~30mW.
If you are really worried about ambient temperature check for "automotive grade" parts.
eg LMR51606. I would lean towards this ic in your application.
TI seem to have the widest range of HV input DC-DC converters.
If your 28V source is automotive some input over voltage protection is a must, a TVS would be a good idea.

A bit of advice. Dont buy cheap DC-DC converter modules from amazon, ebay, aliexpress etc.
They cant be trusted.

 

[pcprogrammer]

How about using a couple of 78xx in series. Have the first one drop 10V and the next one drop another 10V to end with a final one dropping only little less than 5V.

[EggertEnjoyer123]

If you calculate the current right, you can also use a power resistor which is probably going to be cheaper.
A 1k resistor can drop 20V at 20mA, and it would dissipate most of the power.

[phil from seattle]

I used to be where you are - not wanting to deal with a switch mode converter.  But, I decided to give it a try and found it to be pretty easy.  TI has design tools that will pretty much do everything for you. Some of the ICs are in SOIC packages - easy to solder. The big advantage is SW mode converters run pretty cool. And the costs are not terrible. JLCPCB carries a lot of the TI parts for well below western disti prices, btw.

As to a 28V to 3.3V linear - at 20 mA, that works out about 1/2W so you will want a fairly big package. You'd be pushing SOTs and SOICs. And any kind of case will push your ambient temp up making thermal management more of an issue. Also, if you decide to expand your circuit, you may have to revisit your power supply. Yeah, you could do a cascade of say 28 to 12V and 12V to 3.3V but seems pretty amateurish.

[Benta]

We're talking about 500 mW worst case here - not a big deal. Pretty much any 3.3 V voltage regulator will tolerate that. No reason to overengineer. Pick a linear regulator of your choice, today most likely a LDO, they are more common. They mostly have thermal shutdown, so a short-circuited output is not a big deal.
As an example TI LP2985. SOT-23, give it a bit of copper on the PCB pads for cooling and it'll be fine.

[Psi]

When you only need 20mA there are lots of simple DCDC options with integrated fets and sometimes diode too
All you need is an SMT inductor in like 0805/1206 package and two caps.

[PCB.Wiz]

This will also be using only through-hole components.

and if needed the next step would be to use something like a TO-252 package to get the heat into the pcb.

Those are mutually exclusive, as TO-252 is not thru hole, but going SMD will give you a lot more options.
There are very few Thru Hole > 30V low current regulators. 

eg the TO-252 SE8633J1-HF at lcsc, is 60V 1.8uA 3v3 16c/150

[dobsonr741]

No need to increase component count with a buck. Look at buck modules, like the DD4012SA. Widely available from Ali. If you are not into that kind, you can find more reputable 3 pin buck modules.

[mariush]

There are regulator packages with better thermals than TO-92's 165C/w

D2PAK goes down to around 63C/w .... 28v in, 3.3v out , 20mA ... that's half a watt wasted, so the temperature rise should be manageable

See for example LD1086 / LD1085  or *1086 as they're made by other manufacturers ... only catch is the absolute maximum voltage of 30v could be an issue.


Switching regulators... that's the correct solution.

If you want to keep it through hole, the *34063 and *33063 switching regulators are available in DIP package, they exist all over and they're super easy to implement, there's online calculators for the parts and even software on sourceforge that does the calculations.

[cv007]

I have done a couple dc-dc and have had no problems with them, but am sticking to linear in this case for several reasons.  The more common the part, the better.

I will try some combos to see what the 'burned finger' test shows (if I can continuously touch parts without getting excited, then I'm probably good). I thought maybe this was a common scenario where maybe spreading heat among a few components was a common solution.

I do have a bunch of LM317M's in TO-252 I can also play with and see how they act. Ti has an online power designer where the LM317 does show up as an option with my inputs.

[PGPG]

If you have 28V at input you should expect overvoltages to say 40V or more. It is hard to find 3V3 LDO tolerating it.
In such case I'm using LM317 first and then LDO fixed at 3V3.
Have in mind that LM317 is not connected to GND so maximum voltage it can tolerate during pulses you don't count to GND.

[PCB.Wiz]

I do have a bunch of LM317M's in TO-252 I can also play with and see how they act. Ti has an online power designer where the LM317 does show up as an option with my inputs.

317s are cheap and cheerful, but they have high minimum iq requirement.

[Simmed]

3.9k bleeding into 3.3v zener
 

[cv007]

There is a LM2936 which is a fixed 3v3, and it has a 40v and HV version for 60V. Also has various package options.

317s are cheap and cheerful, but they have high minimum iq requirement.

Always a downside for each solution.

3.9k bleeding into 3.3v zener

A zener was my first thought, but the 20mA may or may not ever be seen so using it as a constant seems wasteful and ends up with heat that was not needed. That led to my next step in thinking where a 3V3 regulator would get some help in the heat department by dropping the voltage before the regulator, and end up with the advantages of a regulator. Maybe more choices for the regulator when its max voltage requirements drop (?)

[SteveThackery]

I have done a couple dc-dc and have had no problems with them, but am sticking to linear in this case for several reasons.  The more common the part, the better.

Could you just explain again why you must use linear voltage regulation, such as the 317?

It's just that you can get three-terminal switching regulators that are - effectively - a drop-in replacement for the ubiquitous three-terminal linear regulators. I've attached a picture of one (this is a 5V part, but it's just for illustration). If you want simplicity, you can't get much simpler.

Anyway, just a suggestion.

[PGPG]

317s are cheap and cheerful, but they have high minimum iq requirement.

What!?
Do we understand iq the same way?
LM317 has iq about 50uA (hope I remember well as I was measuring it in 90s) and it not varies with load current what you can take in count counting feedback resistors.
What 317 needs is high minimum load current of about 5mA that in typical 317 application is ensured by using small resistor values in feedback divider but you are not forced to use that way. I was using LM317 with even 2k7 resistor from output to feedback pin (about 0.45mA/0.5mA (50uA difference) in feedback resistors). Of course you loose some precision.
This approach is possible in two situations:
1. when your load in all conditions consume 5mA or more, what fulfills LM317 needs, or
2. when you don't care if LM317 regulates or not - if you use it to distribute power between two regulators you don't care how it is distributed until the current is little higher.

[PGPG]

Could you just explain again why you must use linear voltage regulation, such as the 317?

I don't know OPs rationale, but can explain my.
Long time ago (previous century) in RS232 125kHz RFID reader we used NS MAX232 clone to have RS232. Everything worked well until NS removed this IC from their offer.
We replaced it with ST clone and suddenly got a serie of readers (already flooded) with reading distance reduced from about 50mm to 3mm. Reading distance was all contained inside the casing .
The reason - first IC used about 170kHz to generate RS232 voltages and second used frequency very near 125kHz. Our decision - never more any DCDC inside RFID reader.
As some door-locks use 24V we decided that our readers should work being powered from 12V or 24V and you have a need to linearly reduce 28V down to 3V3.

[PCB.Wiz]

317s are cheap and cheerful, but they have high minimum iq requirement.

What!?
Do we understand iq the same way?
LM317 has iq about 50uA (hope I remember well as I was measuring it in 90s) and it not varies with load current what you can take in count counting feedback resistors.

That is Reference pin current.

What 317 needs is high minimum load current of about 5mA that in typical 317 application is ensured by using small resistor values in feedback divider but you are not forced to use that way. I was using

Yup, as bad as 10mA in the corners, typically ~3.5mA but you cannot design to that typical
That imposes a minimum resting current ( or quiescent current) to your system.

TI do have a variant that has a lower system quiescent current, but that's lower voltage too.

[cv007]

What 317 needs is high minimum load current of about 5mA that in typical 317 application is ensured by using small resistor values in feedback divider

So I assume that is why the output feedback resistor is typically shown as 240 ohms in the datasheets, as a way to guarantee the minimum load current under any load (and at all in-out voltages and in all temperature ranges). If you have your own minimum load, I assume you can subtract that and come up with better feedback resistor values.

[NiHaoMike]

If you don't need the voltage to be accurate, have you considered the classic zener + resistor + bipolar transistor circuit? The only current it will need in addition to the load current is the current the zener needs to drop a reasonably stable voltage.

[Benta]

If you don't need the voltage to be accurate, have you considered the classic zener + resistor + bipolar transistor circuit? The only current it will need in addition to the load current is the current the zener needs to drop a reasonably stable voltage.

Please explain why that would run cooler than a regulator IC. 

[PCB.Wiz]

What 317 needs is high minimum load current of about 5mA that in typical 317 application is ensured by using small resistor values in feedback divider

So I assume that is why the output feedback resistor is typically shown as 240 ohms in the datasheets, as a way to guarantee the minimum load current under any load (and at all in-out voltages and in all temperature ranges). If you have your own minimum load, I assume you can subtract that and come up with better feedback resistor values.

Yes, if you know a minimum system current (even during boot & power cycling)  you can subtract that.
The MIN 317 current varies by vendor, and by voltage. MAX values can be typically 5/10/12mA

Another solution for low current regulators, is a low current LM431/LM432 and a NPN transistor.