Continue to Site

Welcome to our site!

Electro Tech is an online community (with over 170,000 members) who enjoy talking about and building electronic circuits, projects and gadgets. To participate you need to register. Registration is free. Click here to register now.

  • Welcome to our site! Electro Tech is an online community (with over 170,000 members) who enjoy talking about and building electronic circuits, projects and gadgets. To participate you need to register. Registration is free. Click here to register now.

LM317 heat dissipation (linear voltage reg)

Status
Not open for further replies.

Cobalt60

New Member
I want to know what I would do for cooling to run an LM317 at basically its maximum capacity. Just want ideas from some of you more experienced people out there who have probably already done this.

A typical LM317 has a max input of 40V, a low of a 1.25V output, and 1.5A of current, for 58W. My max input would be 36V, so in my situation, it would be more like 52W. And id like to not have to rely on fans.

This post will help me to design cooling setups for other ICs as well, so dont think Im going to be asking how to cool every IC or voltage reg I come across ; )

TIA

-Chris Placzek
 
Those power rating's are based on sufficient heatsinking AND ambient temperature, I wouldn't produce anymore than than a 1 off running that close to maximum power dissipation. Use a switching regulator as a primary drop down regulator into the linear one. It's from a design standpoint irresponsible to intentionally waste that much power if there are other options available, at least in my opinion.

Use Google's advanced search to search the forums here for 'heatsink rateing' and you should find more than one thread with all the math you need to value a heatsink for a given power dissipation. A switching regulator feeding a linear to remove the ripple is probably far more inexpensive, I'd imagine the IC and components for a switch mode supply are going to come in WAY under cost compared the hunk of aluminium or copper and fans you're going to need to keep 50 watts at bay, much smaller too for the enclosure if you're not disipating 50 watts in it. Whcih is just that much cheaper in materials.
 
Last edited:
Cool OK thanks. I want to say though, that I dont actually plan on running it at 50W quite possibly ever, just want the ability to do so should the rare occasion ever come up. And if that occasion did come up, I doubt it would be for any extended period of time. So I can sacrifice efficiency to gain simplicity. But your idea of using a switching regulator as a drop down is greatly appreciated and I will consider it.
 
Again, for heatsink consideration go to Google, and click "Advanced Search" and search for heatsink size (or variations like that) an just limit the domain to Electronic Circuits Projects Diagrams Schematics Electronics Circuit Project I know there was a very full discussion on this a few months ago. Sorry if I sounded like I was scolding =) I'm bad about that, and I hate linear regulators for efficiency => They do work though.
 
Last edited:
Youre helping me out greatly. I came here looking for help and within minutes you started providing me with just that. Thanks again.
 
Well generating 1.25V @ 1.5A using a 36V source voltage is very inefficient using a linear voltage regulator. I'm sure your aware of this as you are aware and grappling with how to best deal with the heat generated from that inefficiency.

Now if was to be used for say a bench variable DC power supply then there are things that can be done to improve the situation even if one wanted to continue to utilize linear regulator. One method is to utilize a tapped secondary transformer such that you could switch to ranges of variable voltages so that the voltage differential across the regulator is reduced.

In a very high power/current DC supply one can see that even the complexity of a motor powered servo controlling the position of a VARIAC transformer could be used to optimize the voltage drop across the linear regulator, although switching power supplies would most likely be more economical.

Anyway my main point is that the best way to deal with heat dissipation in linear regulators is to try and optimize the voltage differential across the regulator rather then trying to find the best way to remove excess heat from the regulator.

Wake up sir, this is the 21st century and we care about efficiency, LOL
Lefty
 
Last edited:
95% of the time it will use less than 6W. I simply want proper cooling should I decide to push the chip near to its design specs.
 
Yeah but lefty he does have a point. He said this wasn't under 'normal' conditions. So heatsinking for worst case is a good idea at least. Relay switched or servo controlled variac seems almost silly compared to using an SMP IC as a pre-regulator to get the voltage within the dropout range of the linear supply. Best of both worlds. You just need pick a linear IC that can handle higher frequency ripple. Every PC you'll ever use (more than likley) is operated under this principal, switch mode powering ripple/linear regulators.
 
Last edited:
No naturally convected heatsink will allow you to use the 317 at reliable junction temp, no matter what package type, at 50 watts dissipation. The thermal resistance from junction to case of the best package (TO-3) is 2 degC/w. If the junction to case resistance was zero, the heatsink would have to have a thermal resistance of 2.5 degC/W, worst case.

Max. Heatsink Thermal Resistance = (Max Junction temp - AmbientTemp) / Power = 2.5 deg C/W

Subtract 2 from 2.5 and you end up requiring a heatsink of .5W/degC thermal resistance or less. And that's if you are running the 317 at its absolute maximum junction temp of 150 deg C. The package will not allow the heat to escape to its case fast enough to dissipate the heat to a reliable junction temperature no matter how big the external heat sink. It would be almost like trying force over one amp through a 1 ohm resistor at 1 volt, where the resistor is the junction to case thermal resistance. An infinite heat sink would give you a 130 degC junction temp, which is too close to the 150 degC absolute max.
 
The datasheet for the LM317 shows that when the voltage across it exceeds only 15 then it reduces its output current to protect itself.
With 40V across it its max output current might be as low as 150mA.
The current-limit graph shows the typical max current for different voltages and temperatures.

The thermal resistance for the TO-220 "T" package is 4 degrees c/W and its max allowed chip temp is 125 degrees C.
So if the ambient is 30 degrees C then a heatsink with a thermal resistance of 1 degree C/w will cause the chip to be near max allowed temp with a power dissipation of only 19W.

Is sombody talking about the LM317 in the TO-3 metal case? Its typical thermal resistance is 2.3 degrees C/W but could be as high as 3 degrees C/W.
With a 30 degrees C ambient and a pefect heatsink then at 3 degress C/W it will be at its max allowed temp with a dissipation of only 31.7W.






















How did you calculate 50W?
With a perfect heatsink (impossible) with a 30 degrees ambient then the chip will be at its max allowed temp with a power dissipation of only 23.75W.

If the ambient is at freezing and the LM317 had a perfect heatsink then the chip will be at its max allowed temp with a power dissipation of only 31.25W.
 
OK I guess I should have looked more into the specs of the 317. So youre saying the 317 is really designed more for a max of like 23W? Thats fine, so then now my design goal would be a heatsink to dissipate that 23W.

I came up with 50W assuming a 36V input, 1.25V output, and 1.5A (more like 52W).

Again, typical uses would involve more like 5W (12V in, 5V out, less than 1A). I simply want to be able to use the chip to whatever its max allowed would be.

Thanks again for the help.
 
I came up with 50W assuming a 36V input, 1.25V output, and 1.5A (more like 52W).
The datasheet shows how the LM317 reduces its output current when it has over 15V across it. A graph shows a "typical" one (some are much worse):
1) With up to 13V then its max current is 2.2A.
2) With 36V then its max current is reduced to 480mA.
3) With 40V then its max current is 300mA but is guaranteed to be 150mA or more.
 
If you want to get the full typical current limit out of the LM317 with a TO-3 package, without the junction exceeding a maximum of 125 degC, at a 30 degC ambient, I figure you will need a heat sink with a thermal resistance of .8 degC/W or less. That's using a junction to case thermal resistance of 2 degC/W for that package, as given in National's datasheet, and a 2.2A × 15V = 33W device dissipation. Some devices may allow you up to 3.4A × 15V= 50W for a brief period, until the junction gets too hot, but that is the max current limit, not the typical of 2.2A.

BTW, I see a procedure for calculating a heat sink is given in National's datasheet, which is pretty much the procedure I used.
 
Last edited:
OK so I am all set as far as the heatsink goes, but while I have you here, might as well discuss what I would do if I used a switched mode step down converter.

If I am not mistaken, there would be basically 2 parts I would need; a PWM and a smoothing capacitor. So what would I do for a PWM? And what would the specs be of an appropriate smoothing capacitor?

Thanks again

-Chris P
 
"Use a switching regulator as a primary drop down regulator into the linear one."

Im having trouble comprehending how I would do this. On my previous statement I was remembering info on how to smooth a rectified AC voltage, not a switched DC voltage, sorry. So I really have no idea on how to implement a switching regulator before the linear one. Thanks for any help.

-Chris P
 
"Use a switching regulator as a primary drop down regulator into the linear one."

Im having trouble comprehending how I would do this. On my previous statement I was remembering info on how to smooth a rectified AC voltage, not a switched DC voltage, sorry. So I really have no idea on how to implement a switching regulator before the linear one. Thanks for any help.

-Chris P

You feed primary power into the switching regulator to efficiently lower the voltage by a lot, then you feed the output of the switching regulator into the linear one to get an accurate and quiet output voltage. THat's all it is.

THe "smoothing" in a switched DC converter is a part of the converter in the form of output filters like capacitors and inductors. It's like if you have square wave with a peak of 10V but the duty cycle is only 50%, then the average voltage is 5V. You filter it to get rid of the peaks and end up with a 5V voltage. IT is much more efficne than a linear regulator because power in (ideally) equals power out. Converting a higher voltage to a lower voltage requires less input current than output current (because power in=power out and the energy for the higher output current was converted from the excess input voltage). For the same reasons, converting a lower voltage into a higher voltage requires more input current than output current (the energy in the excess input current is used converted to be the energy in the higher output voltage).

In a linear regulator input current (ideally) the same as output current and since the input voltage is higher than the output voltage, more energy enters the regulator than outputted. WHere does this energy go? Well the linear regulator takes the extra energy in the higher voltage and turns it into heat.

Like if you had a source of 50V but you needed 5V, and the linear regulator needs the input voltage to be at least 1V higher than the input (called the dropout voltage), you would use the switching regulator to take 50V down to 6.5V (an extra 0.5V for head room) and use the linear regulator to take 6.5V to 5V, which burns off much less excess voltage as heat than if you just took 50V down to 5V linearily.

Use an IC though made for the thing unless you want to learn a whole lot more about switching regulators. THey have some pretty complex control algorithms and such. Depending on your requirements for the output voltage, you could probably just get away with only a switching regulator. No linear regulator needed.

Linear regulators:
Very fast transient response, cheap, simple, very low noise but very inefficient.

Switching Regulators: Slower transient response, more complex, more parts and more expensive, and more noise but very efficient.
 
Last edited:
Try a switch mode IC, they require a few external components and stages. It may not seem related to your original question but if you can variably control the approximate voltage the linear regulator gets you don't waste the heat in the first place so it never overheats. You only ever supply the linear IC with enough voltage to regulate out the ripple.
 
The point of learning how to and using a switching regulator before a linear one is that you can avoid the heat generated in the linear stage in the first place by supplying it with a voltage just high enough for it to regulate without wasting a huge amount of power. It generates less heat and is therefore more efficient, it can simplify case design as well because you don't have as many clearance issues.
 
First, thanks again for the help, and for any more you provide me with : )

"THe "smoothing" in a switched DC converter is a part of the converter in the form of output filters like capacitors and inductors."

So is the smoothing external to the switching regulator or is it a part of it?

"Try a switch mode IC, they require a few external components and stages. It may not seem related to your original question but if you can variably control the approximate voltage the linear regulator gets you don't waste the heat in the first place"

But how difficult is it to variably control the output of a switching reg?

So basically it would go something like this (in my situation):

smoothed variable DC -> switching reg -> smoothing capacitor -> linear reg

So ideally for me, the switching reg would be variable. It would also be just super if somehow they adjusted to each other so that there was one voltage control, that automatically adjusted the switched regs output to be slightly higher than the linear regs.
 
First, thanks again for the help, and for any more you provide me with : )

"THe "smoothing" in a switched DC converter is a part of the converter in the form of output filters like capacitors and inductors."

So is the smoothing external to the switching regulator or is it a part of it?
It is a part of the regulator, but not the IC because capacitors and inductors are almost always too big to fit onto an IC. Plus they are the most crucial parts to customize in a switching converter. You can get modules where everything is integrated into one package, but they are rare and expensive.

"Try a switch mode IC, they require a few external components and stages. It may not seem related to your original question but if you can variably control the approximate voltage the linear regulator gets you don't waste the heat in the first place"

But how difficult is it to variably control the output of a switching reg?
Dead easy! Switching regulator ICs almost never have a fixed output voltage. THey use a resistive divider that steps down the output voltage to a level tolerable to the IC. This scaled down voltage is then fed back to the regulator so it knows what the output voltage currently is in order to regulate it. The ratio of the resistive divider sets the output voltage of the regulator (relative to the internal voltage reference of the switching converter IC). Just use a potentiomter instead of a fixed resistor divider and you instantly get variable voltage. Really, it's no different than what you do with the LM317 to set it's output voltage, whether it be fixed or variable. By no different, I mean exactly the same thing.

So basically it would go something like this (in my situation):
smoothed variable DC -> switching reg -> smoothing capacitor -> linear reg

Smoothing capacitor is lumped in with the components that are needed to get a switching regulator to work. But that is just semantics. Some linear regulators don't need one to operate, but they aren't switching square waves so they don't need to filter them out. THey work better with them, however.

I am not sure what you mean by smoothed variable DC. But the source voltage should be DC and it's better if it is smooth than not- makes the regulator have to work less to produce a good output. They can only react so quickly and supress input noise by so much after all.


So ideally for me, the switching reg would be variable. It would also be just super if somehow they adjusted to each other so that there was one voltage control, that automatically adjusted the switched regs output to be slightly higher than the linear regs.
Depending on the noise requirements of your output voltage, chances are you can get away with just a switching regulator set to be variable. But if not, it's going to be tricky to get the two potentiometers controlled by a single knob and to use it to get the output voltages not to scale proportionally to each other, but to maintain a fixed voltage difference between the two outputs. Sure it's possible, but the extra circuitry required to do that are probably outweighs the benefits. See if you can't just use a switching regulator alone.

This is filler because a post can't just be a quote.
 
Last edited:
Status
Not open for further replies.

Latest threads

New Articles From Microcontroller Tips

Back
Top