4
\$\begingroup\$

I have a project where my power supplies are giving me a bit of a headache. I have three voltage regulators that I am trying to run from a single bench power supply. There are two 3.3V regulators, and one 12V regulator. The current draw for the 12V regulator is up to 500 mA, and each 3.3V regulator peaks at 150 mA.

When I hook up about 14V to the inputs of these regulators, the 3.3V regulators start to cook. Doing the math, that's about 1.7W of power dissipation for each of the 3.3V regulators. They're in a TO-220 package pointing straight up, and they got too hot to touch. I was able to throw a random heatsink I found on the tab and cool things down, but the top of the heatsink was about 2.5 inches high - very flimsy and cumbersome.

I really don't want to sacrifice the single power supply input, and I also really don't want to go to an on-board switching supply for either the 12V or 3.3V supplies. What are some passive package+? solutions for dissipating up to 2.0W of heat that are physically secure and low profile (shorter than 1 inch)?

Edit: My reasons for not wanting a switching supply are:

  1. I don't want to deal with switching noise in the analog parts of my design
  2. I'm less comfortable with designing a switching regulator than using a linear regulator
  3. The PCB that I'm working on is big and expensive. I don't have the time/budget to "try out" a regulator. A linear regulator with enhanced heat dissipation seems a lot safer way to avoid making coasters than a switch mode regulator.
\$\endgroup\$
5
\$\begingroup\$

If you're stuck using linear regulators, you can put big power resistors in series with the inputs of the regulators to drop the voltage and share some of the heat dissipation. You're dissipating the same amount of power either way, but it might make it easier to fit on your board or whatever.

If you're using an LM7833, for instance, and supplying 150 mA, the datasheet says the dropout is 2 V, so the input voltage has to always be above 5.3 V. From a 14 V supply at 150 mA, this is a 2 W, 56 Ω resistor. The resistor just needs air circulation around it, not a heatsink, and then your regulator only needs a 500 mW heatsink instead. The highest power dissipated in the regulator will be at the current when resistor and regulator are both dissipating the same power, which in this case is about 95 mA.

\$\endgroup\$
  • 1
    \$\begingroup\$ +1, this was the solution i went to post. 56 ohms is a good choice, or you could stick a 47 ohm in to leave a little headroom in the "about 14V" supply. \$\endgroup\$ – JustJeff Aug 18 '10 at 23:24
5
\$\begingroup\$

voltage regulators come in two types, linear and switch mode. linear regulators work well when the voltage in is near the voltage out, but rather poorly if the voltage in is higher than the voltage out. What you need is a switch mode regulator. A few minutes on digikey suggests that a LM2574 will do the job. The down side of using such a regulator is that they require a couple of additional discrete components (a handfull of capacitors, a schottky diode and an inductor, in the case of the lm2574) to work, but they dissipate basically no heat.

\$\endgroup\$
  • 1
    \$\begingroup\$ I know about switching regulators, but I'm not comfortable with using them in a design. I'm also concerned about switching noise that such a part would introduce. I would much prefer to having a heat dissipation issue than have an issue with a switch mode regulator. \$\endgroup\$ – W5VO Aug 18 '10 at 12:07
  • \$\begingroup\$ You might look at better ways to even the power distribution then. Use multiple regulators, pre-regulate or find some other way to bring the input voltage down or use separate pass elements (probably shown in datasheet as "increasing current capacity" or such). Those should be easier to cool that single small hot chip. \$\endgroup\$ – XTL Aug 18 '10 at 15:14
4
\$\begingroup\$

As you say that you would rather have a heat dissipation/energy wastage problem than use a switch mode power supply you are left with only one solution - you must bolt your regulators onto a sufficiently large piece of metal. I used to make folded aluminium boxes for my projects and the usual solution was to bolt the regulators onto the case, with suitable insulation washers and mica sheets, and have flying leads from the PCB to the regulator devices. For higher dissipation I would add a heatsink externally mounted on the box.

Nowadays switch mode regulators are much better understood and characterised. There are any number of application notes and design guides that will help you to get a good, stable regulator cheaply and without the horrible power wastage that goes with a linear regulator.

Throwing away 3.5W to supply about 1W on the 3.3V rails may be OK for a hobby project but would not be acceptable for a commercial product.

\$\endgroup\$
1
\$\begingroup\$

You don't mention why you are against a switching supply solution... Is it complexity you're concerned about? There are dozens of very simple switchmode regulators out there. They are inexpensive and much more efficient than the linear regulators.

CUI and Recom both make 3-terminal switchmode regulators which are drop-in replacements for the old 78xx series.

If you're looking for something cheaper (you'll need to supply a diode and inductor though), look at National's LM2575 series of simple switchers, Linear has dozens of varieties, TI... There are tons of app notes to help figure out inductor sizing. There are very few reasons to need a linear regulator these days.

\$\endgroup\$
  • \$\begingroup\$ Couldn't find CUI's regulators, but the Recom part number is R-78XX-0.5 (R-783.3-0.5 in this case) - See bit.ly/dz0WGq, and TI makes a similar product, the PT78HT2XX (PT78HT233) - see ti.com/lit/gpn/pt78ht205 \$\endgroup\$ – Kevin Vermeer Aug 18 '10 at 18:00
  • \$\begingroup\$ The National LMZ12001 is also a SMPS with integrated diode and coil. More compact (SMD) than the 78xx replacements. Not cheap though. \$\endgroup\$ – stevenvh Jun 18 '11 at 13:09

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.