Linear Voltage Regulator Basics

[tdbeng]

Hi All,

I'm using an LM317 (TO-220 package) to regulate a supply voltage of 7.5VDC from 12VDC. It's the basic LM317 circuit: 0.1uF cap on the input, 1.0cap on the output, forward diode from output to input, and two resistors on the adjust pin so Vout = 1.25*(1+r2/r1) + Iadj*r2. I'm using 100ohm for r1 and 500ohm for r2. The 7.5VDC is used to supply power to an ethernet hub circuit, and the max. current draw is 1A. The 12VDC input is coming from an unregulated power supply.

The problem is that the regulator is dissipating too much heat for my application, even w/ a large TO-220 heat sink.

What's the best way to reduce the heat from the regulator? Is it by reducing the input voltage (because the heat dissipation is in relation to the voltage drop)? I thought about putting a zener in seried with the input, but is this bad circuit design technique?

Thanks in advance!

 

[Lewish]

Well, putting a zener in the input might work OK, but all you are doing is moving the heat dissipated from one device to another one. This may solve your problem. I typically use just plain rectifier diodes, which generally have about 1.0volts drop at 1amp of current.
To truly reduce your heat output, a switch-mode power supply is the way to go. Heat output is the reason they were originally developed. A good design with adequate filtering can be just as quiet as a noisy linear. Just a thought.

 

[fil]

i agree with the last post - another thought is to use a switcher to regulate to 8.5v then run through the linear the power dissipation in the linear is down to 1W instead of 4.5W
use the national switcher program to design your switcher


have fun

 

[melone]

The linear will work VERY well in this application. First, you don't have enough front end capacitance. Try putting a couple hundred micro on the input. Second, put a couple of series diodes on front end like LEWISH suggested (start with one and go from there). Your diode will not get very hot, but it will drop your regulator power dissapation signifigantly. Like LEWISH said, 1V @ 1A (1W). Third, make sure you have a good heat sink. Use a 12W with thermal paste. I have used that regulator reliably and had great success. I would hesistate to put in a switcher, only because they are notorious for poor performance if not dialed in EXACTLY.

Good luck and keep us posted!

 

[tdbeng]

All,

Thanks for the help. I thought about a switcher, but I need to get something completed by the end of this week. I breadboarded the regulator with three 2.5A diodes in series, and this cut the input down around 9.5V. This noticeably took some of the heat from the 317. Also, I put the diodes a little further away from the 317. Previously, my main concern was the amount of heat being generated within such a localized area. Basically, the output voltage looks good, and I'm not melting the surrounding plastic! Thanks again.

 

[buzzp]

If you are going to put this on a PCB then be aware of the standard temp rating of the board at 105C. 130C is also available as standard from some board houses (they just dont tell you). Also, your resistor from adj to out should be 250 ohms as recommended by the app notes. Seems that something else is going on if you are melting plastic (what kind) with a TO-220 package with only 1A max flowing in the 3A package with a large heat sink. I guess the question is how hot is the regulator getting? Did you use the grease between the heat sink and package and how was it secured (this can have a significant effect on heat transfer).
Also, I don't necessarily agree with adding front end capacitance. This would depend on the source of the 12VDC. Adding capacitance will take out more low freq ripple, if it is present (depends on source) and won't help at all with your heating problem (don't think poster was implying it would either).
It sounds like the board layout may have contributed to the heating problems as well. Is the regulator laying flat on the board or standing up?