LM317 Heat SInk Needed?

[bluebird]

I'm trying to save space on the pcb I'm working on. I was looking at SMD versions of the LM317 and figured the through hole version laying down flat on the circuit board would probably be able to dissipate more heat than any of the SMD versions. Maybe an aluminum standoff instead of a nut on the screw that ties the LM317 down.

I'm only powering a couple op amps and a few relays. I'm dissipating about 1.2 watts with the LM317, full load 100ma, 10 or so volts in/out differential. 
Does anyone see why I couldn't just lay the IC down on a pad of copper the size of the LM317 its self? 

 

[ruffrecords]

Laying it on a pad the size of the T)220 part will o little to dissipate the heat. Better to have it standing upright so air can flow round it. The theta junction to ambient is just under 40 degrees C per watt so your 1.2 watts will raise the juncture about 50 degrees above ambient; so it shold not reach more than 80 degrees C which is comfortably within its rating.

Cheers

Ian

 

[bluebird]

The theta junction to ambient is just under 40 degrees C per watt

Thanks Ian! I'm looking at the data sheet and I can't seem to find that info? On the data sheet it says the maximum Tj is 125 degrees. Is that assuming 45 degrees ambient (+80 = 125)?

http://www.ti.com/lit/ds/symlink/lm317.pdf

 

[Matador]

The information is in table 6.4 (Thermal Information), as the parameter Theta-JA (37.9 degrees C/W which Ian quotes). 

 

[bluebird]

Doh! Thanks!

 

[bluebird]

Just did a little more reading about heat sinks and found this blurb on the DIYstompbox forum.  It was a response to a guy asking about using a 7806 to regulate a tube heater and whether or not he had to use a heat sink.

On Semi also makes a 7806, and they spec theirs at 150C operating. The datasheet http://www.datasheetcatalog.org/datasheet2/9/0p4t1g1lw0spoo5ukh2s1uxchzky.pdf also lists two thermal resistances. One is the resistance from junction to ambient (that is, just the case sticking up in the air) and another of junction to case. The junction-to-ambient (JTA) is 65C/W, meaning if you make the chip generate one watt, the junction will rise 65C higher than the surrounding air. If the air is 25C, then the junction is at 90C.

Put another way, with 25C air, the most power you can dissipate internally is (150C-25C)/65 c/w = 1.923W.

However, if you bolt that thing to a heat sink with a thermal resistance of 0.5 c/w and either lap the metal surfaces smooth or use heat sink goo to get the heat transferred over the gap between the two down to 0.7C/w, then the total resistance junction to ambient is the junction to case resistance of the package (5c/w) plus the interface plus the heat sink, or a total of 6.2c/w, and the allowable power is (150-25)/6.2 = 20W.

For all power devices, how much you can dissipate is determined by
- the max junction temperature (from the datasheet)
- the junction-to-case thermal resistance (from the datasheet)
- the case-to-sink thermal resistance (how well can you mate the semiconductor case to the heat sink surface to let heat flow across?)
- the heat sink thermal resistance to ambient air (or water, whatever)
- the temperature of the surroundings that you're going to reject the heat to.

So the transistors thermal resistance to ambient rating (C/Watt) more or less takes on the heat sinks thermal resistance to ambient rating (taking into account there is a little case to sink thermal resistance as well).
And its probably good to add some more ambient temperature (above room 25C) to the equation if your working in a box with tubes.  Maybe 50C or 60C.

just in case anyone else is interested...

 

[abbey road d enfer]

I'm trying to save space on the pcb I'm working on. I was looking at SMD versions of the LM317 and figured the through hole version laying down flat on the circuit board would probably be able to dissipate more heat than any of the SMD versions. Maybe an aluminum standoff instead of a nut on the screw that ties the LM317 down.

I'm only powering a couple op amps and a few relays. I'm dissipating about 1.2 watts with the LM317, full load 100ma, 10 or so volts in/out differential. 
Does anyone see why I couldn't just lay the IC down on a pad of copper the size of the LM317 its self?

What's the thinking behind doing without a heatsink? Cost? Space? Mechanical issue? I've built a number of small items where I used regs without heatsinks, but there, I wouldn't take the risk; it's always difficult to evaluate correctly the internal temperature and airflow issues. In many units, I've used TO220 U-shaped heatsinks that mount flat on the PCB with one screw. They are very unexpensive, don't take too much space and provide an extremely reliable mounting (no stress on leads).

 

[Newmarket]

I tend to not use SMT regs for anything with significant heat dissipation. They look good initially but the thermal figures often relate to having on, say, a 25 sq mm Cu area - which often defeats the advantage of SMT.
It would be good to have an all SMT board but that's seldom the case as I nearly always have Thru Hole connectors and some caps etc.

I say stick a heatsink on it whichever way you mount it - those U shaped ones mentioned by abbey road' are good or there are ones that simply clip onto the device vertically. It may well be fine without heatsinking but I like to run things on the cool side and it's always difficult to know what the ambient will be.

 

[JohnRoberts]

The modern 3 terminal regulators use smart protection so they are hard to kill but running semiconductors hot reduces their MTBF....  I've seen old products without HS where the regulators got so hot their solder connections overheated (turned dull) and fell apart.

JR

 

[bluebird]

What's the thinking behind doing without a heatsink? Cost? Space? Mechanical issue? I've built a number of small items where I used regs without heatsinks, but there, I wouldn't take the risk; it's always difficult to evaluate correctly the internal temperature and airflow issues. In many units, I've used TO220 U-shaped heatsinks that mount flat on the PCB with one screw. They are very unexpensive, don't take too much space and provide an extremely reliable mounting (no stress on leads).

I am trying to make a table top unit so I want it to be as small as possible. There will only be one preamp tube but many voltages so many regulators. The heater regulator will already have a large vertical fin heat sink, but the +/- voltages, phantom, B+ all are light current draws so I'm just trying to get away with a small area for the power supply.
I have looked into dc/dc converters for this project but always chicken out and go back to what I know.
One other possibility is to have a slab of aluminum bolted to the chassis with tapped 4/40 holes and lay all the (insulated) regulators down side by side off the edge of the PCB. But then I have to have that piece made...

I think I like the laying down on a little heat sink tagged to the PCB idea. Thanks Abby.

 

[bluebird]

I've seen old products without HS where the regulators got so hot their solder connections overheated (turned dull) and fell apart.

JR

Yes and I always try to make the traces around regulator pins as large as possible, also around resistor leads what will be dissipating any amount of heat. I usually use double thickness copper for my pcb orders because its just not that much more money. Its a little harder to solder around ground planes sometimes but well worth the piece of mind for me.

 

[Newmarket]

I am trying to make a table top unit so I want it to be as small as possible. There will only be one preamp tube but many voltages so many regulators. The heater regulator will already have a large vertical fin heat sink, but the +/- voltages, phantom, B+ all are light current draws so I'm just trying to get away with a small area for the power supply.
I have looked into dc/dc converters for this project but always chicken out and go back to what I know.
One other possibility is to have a slab of aluminum bolted to the chassis with tapped 4/40 holes and lay all the (insulated) regulators down side by side off the edge of the PCB. But then I have to have that piece made...

I think I like the laying down on a little heat sink tagged to the PCB idea. Thanks Abby.

How many of these units are you planning to make - one off ? a handful ? ongoing production ?
Spacewise TO220 regs  take up quite a large area when mounted horizontally onn the pcb due to the leg bend radius and then the heatsink takes up considerably more. They do look good though and have a solid fixing. Depends on the dimension where you are most pushed for pushed for space.
It shouldn't be too difficult to craft the tapped aluminium piece if you are handy with a press drill and tapping set but I can see you might not want to do many that way.

 

[Newmarket]

Yes and I always try to make the traces around regulator pins as large as possible, also around resistor leads what will be dissipating any amount of heat. I usually use double thickness copper for my pcb orders because its just not that much more money. Its a little harder to solder around ground planes sometimes but well worth the piece of mind for me.

With resistors that are going to dissipate anything significant it's worth using small ceramic stand offs to keep them off the pcb and have some air around them.
Can't argue with 70um Copper but I have to say that my day job involves highish power medium voltage stuff (few hundred volts) - PFC / DC-DC / Pulsed Power etc. and I've never had to go with it. Having said that it's four layer stuff with large power planes although the inner layers are thinner than the top / bottom layers.
The difficulty of soldering the extra copper made me smile - I have a compulsion to increase the thermal spokes attaching pads to planes. The default / standard thicknesses always look too thin to me ! So I have increased  copper on four layers. Needs a fair bit of heat to solder - but desoldering and clearing the holes is the real problem. Have a powerful temp controlled iron on the bench though as well as the 'usual' temp controlled Weller.

 

[Monte McGuire]

The difficulty of soldering the extra copper made me smile - I have a compulsion to increase the thermal spokes attaching pads to planes. The default / standard thicknesses always look too thin to me ! So I have increased  copper on four layers. Needs a fair bit of heat to solder - but desoldering and clearing the holes is the real problem. Have a powerful temp controlled iron on the bench though as well as the 'usual' temp controlled Weller.

I'm preaching to the choir, but this is part and parcel of the trials of modern PCB layout, especially when it's done well! Short of full-on hot air and IR pre-heating, you can often make things work nicer by using a hot air gun to pre-heat the foil around the work you're planning to do, just to thermally pre-load the surrounding foil. Then, your iron will work a little faster, without having to thermally load the surrounding foil so greatly.

In my experience, PCBs and components largely get damaged from peak temperatures, where laminates de-polymerize, and where active components re-diffuse. So, if you can pre-heat the surroundings, use only a safe maximum temperature, and get the soldering work or re-work done quickly, then you will do no damage. If an attempt fails (and I know you know this) then the best idea is to back off, let the work cool, and try another approach. Lingering with high powered tools using dangerous temperatures is never good - the answer in those situations is almost always pre-heating and possibly more flux.

I want to also add that the 'old days' of components designed to use external heatsinks has largely passed for relatively low power devices, at least for designs using modern silicon.  High performance amplifiers sold in packages with die-attach pads absolutely cannot be connected to an external heat sink - the PCB has to be designed as the heat sink, or at least structured as part of the thermal mechanical system. The good part is that this is possible, and if your assembly is good, then this is one fewer point of failure - no more poorly greased or under torqued heatsinks, or over-torqued fasteners that tore the insulating sheet and failed the board. All that BS, as well as the heatsink, mounting hardware, and thermal insulator, are gone from the BOM, and good riddance.