3
\$\begingroup\$

I've read some very good threads here on power and heat dissipation in VRs and believe I understand (most of) it. I have a 5V circuit operating a 12V relay (nominal operating current of 53mA), an ATtiny85, a couple BJTs, and a few passive components. I measure 54mA across the circuit. I didn't give much thought to VR-generated heat but did the math nonetheless.

TO-92 The input voltage is 12.8V so the power consumed is 7.8V * 0.054A = 421mW. With a TO-92 case and Rja(ambient) of 157.4C/W, I get a 66C temp rise and when coupled with an ambient temp of 25C results in 91C if I understand correctly. Although it's not approaching the max junction temp of 125C, it's quite hot. Other LDO VRs (e.g., 7805) generated even more heat. I could reduce the heat with a series resistor but that can create other problems according to other threads I've read.

TO-220 The TO-220 has a Rja of 23.3 yielding a total temp (including ambient air) of 35C. I have it mounted horizontally using the PCB as a heat sink. More PCB space required but much lower temp.

Have I calculated the temperatures correctly? The relay is not energized 100% of the time (closer to 75% worst case) but the circuit is in a small enclosed ABS box. Is it fair to say that a TO-92 would be a poor choice for this circuit? Anything I haven't considered?

Relay datasheet

TO-92 datasheet

TO-220 datasheet

\$\endgroup\$
5
  • \$\begingroup\$ I don't see how the relay current or duty cycle figures into this- it's not going through the regulator, right? So your ATtiny85 circuit is drawing 54mA plus the 53mA the relay draws? Total 108mA? That's a lot. \$\endgroup\$ Commented Mar 15, 2016 at 20:48
  • \$\begingroup\$ The heat generated by any linear regulator should be the same - they're just controlled power resistors after all. If you want a check on your numbers, breadboard it and see. Chances are an SOT223 regulator with moderate PCB heat sinking can handle your needs while taking up a lot less space than a TO220. Chances are also you should be able to get your current requirement to a tiny fraction of what you have now, especially in the time average - below 1 mA may well be possible by sleeping the ATtiny and not leaving voltage across resistors - mostly you'd see the regulator's quiescent current. \$\endgroup\$ Commented Mar 15, 2016 at 20:54
  • 1
    \$\begingroup\$ @Spehro Pefhany Correct. Can't believe I missed that. It is a 12V relay powered directly from the 12.8V supply, not the VR. When I was operating this on a breadboard, I completely forgot that the relay shouldn't be in the measurement. That leaves only the uC and BJTs. I'll have to redo the math now. \$\endgroup\$
    – unix
    Commented Mar 15, 2016 at 20:55
  • \$\begingroup\$ Did the regulator on the breadboard version get hot, or did you only conclude it would from the model? \$\endgroup\$ Commented Mar 15, 2016 at 20:56
  • \$\begingroup\$ @Chris It did not get hot. I was reading over the datasheets, was curious to do the math, and surprised by the resulting calculations. \$\endgroup\$
    – unix
    Commented Mar 15, 2016 at 20:59

1 Answer 1

2
\$\begingroup\$

Yes you have calculated the temp correctly, a simple units\dimensional analysis will tell you that.
Your options are: 1) Switch to a DC to DC 7805 drop in compatible regulator. (pricey) 2) Switch to a different package (more surface area)
3) Put a heatsink on it, this could be as simple as a few square cm's of aluminum to a full blown heat sink with fins on it.

\$\endgroup\$
2
  • 1
    \$\begingroup\$ Dimensional analysis will tell you if you're wrong, but it can't tell you that you're right. I have a nice wrench that measures energy, apparently.. (Newton-meters). \$\endgroup\$ Commented Mar 15, 2016 at 21:45
  • \$\begingroup\$ That's why I wrote unit\dimensional... \$\endgroup\$
    – Voltage Spike
    Commented Mar 15, 2016 at 22:19

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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