LM317's on positive AND negative rails....? Power supply design query

[iainpunk]

i freaking love the look of Manhattan style, especially if upright component mounting is applied. i've build a fuzz that style, but soldering onto the boards wasn't fun.

cheers

[Bunkey]

Do like.

Tbh there seems to be something great in every genre;
I was about ready to dismiss the traditional turret style amp until I came across the 6 Degrees FX pedals.
BC Audio's P2P amps look really nice too - that's kind of what I had in mind from the outset.

I suppose it's more a reflection of the effort expended than the particular approach applied.

- and I'd like to say straight lines but...





Got some nice protoboard on the way anyhow so I'll see how that works in the smaller amp and take it from there; might have to remember the upright component mounting idea though 

[iainpunk]

that immage looks a bit horrible, i generally don't like midair component connections, except for some nicely done (parallel) splices, that can look neat and fine, but just crossing wires and soldering the x hurts my eyes


for anti-parallel diodes, i use this thechnique, but i leave a longer bit of straight overlap in between the wraps, where the diodes recide.


cheers,

[Bunkey]

It's like the components have developed sentience and combined into some sort of organic mass. I'd thoroughly expect to come back to it one day and find it had grown a tone control overnight. Just oscillates away in the corner of a room somewhere powered by its own inductance...

*shudder*

What's the rule of thumb for heatsinking then?
Does the C/W of the sink used replace the JA C/W of the component it's attached to - So if the LM380 has a max temp of 150deg with a C/W of 50deg and dissipates 4W, it will rise 200deg in use and fail - but if I attach a sink with a C/W of say 15deg then that value takes over and the LM380 will rise 60deg instead? Roughly speaking (and ignoring ambient temps)

[Bunkey]

Actually it would be the JC resistance of the device plus the resistance of the heatsink wouldn't it.

So if the device itself has a JC of 37C/W then dissipating 4W is going to put it right on the edge regardless of how good the sink is - and that I guess is the determining factor in how much power these IC's can actually produce outright.

See you get told this stuff and it makes sense but it doesn't really click until it clicks 

[Bunkey]

So I got a pack of these heatsinks from RS because they're the same size as the chip and I figured it'd turn them into cute spikey hedgehogs whilst providing added dissipation.
https://uk.rs-online.com/web/p/heatsinks/5040716/

The JA of the LM380 is 107C/W where the JC is 37C/W, so the standard case is dissipating heat to 70C/W in the 8-pin package.

These tiny sinks are rated at 75C/W

...So really they're doing naff all for heat dissipation and serve only as a funky hat for my chips 

[Bunkey]

...There's something else going on with the calculation though isn't there because the LM317 for example has negligable JC resistance and a JA of 23.5C/W, which is being dissipated through the built in sink on the back. If I attach a heatsink to that which has a rating of 25C/W - it's still going to help the regulator dissipate heat better than it would do without it - that's just common sense... or is it? Are these winged heatsinks designed specifically to bolt to a TO-220 regulator actually insulating it if their SA is higher than the devices standard CA rating? 

Somebody please chip in before I'm sectioned for talking to myself here...

[iainpunk]

its always nice if the person you are trying to help answers his own questions.

or is it? Are these winged heatsinks designed specifically to bolt to a TO-220 regulator actually insulating it if their SA is higher than the devices standard CA rating?

no, the IC itself dissipates the same amount from the front and back in free air, you are only adding the sink to one side of the package, changing that side to 25. that adds 23.5/2 to 25, giving you 36.75

cheers

[Bunkey]

or is it? Are these winged heatsinks designed specifically to bolt to a TO-220 regulator actually insulating it if their SA is higher than the devices standard CA rating?

no, the IC itself dissipates the same amount from the front and back in free air, you are only adding the sink to one side of the package, changing that side to 25. that adds 23.5/2 to 25, giving you 36.75

But in that case 36.75 would be more thermal resistance and therefore the chip would run hotter.
I know what you're saying but the numbers should go the other way, subtracting from the overall JA of 23.5 by some factor related to the SA rating of the sink, if the sink is to make thermal dissipation more effective.

I appreciate the 23.5JA is derived through dissipation from all areas of the chip but to consider all the different values and their contribution to the overall dissipation is turning the thing into a full blown mathematical excercise and I find that process incredibly tedious...
I figured since the JC of the back is so small then most of the heat is being dissipated there anyway and dissipation through other areas could be reasonably ignored when just trying to work out a practical solution.



In this case I guess the 'bottom' is the built-in metal sink on the back of the chip.


But yeah the new JA should be less than 23.5 if the affixed sink is reducing heat; more than 23.5 would be insulating it.

Any thoughts? I appreciate your help

[iainpunk]

But in that case 36.75 would be more thermal resistance and therefore the chip would run hotter.
I know what you're saying but the numbers should go the other way, subtracting from the overall JA of 23.5 by some factor related to the SA rating of the sink, if the sink is to make thermal dissipation more effective.

this has been a really long time since i learned that in school, i guess my calculations should be in reverse, and i shouldn't post under the influence
the sink's resistance parallel to the package, not in series. oops.
that gives around 12.6, not 36.75,

cheers

[Bunkey]

Well I was after a rule of thumb for heatsinking - I suppose a half-baked approach to thermal dynamics will suffice instead.



Thanks for your time man

[iainpunk]

half-baked approach to thermal dynamics

no yesterday evening, i was fully baked. you seem to understand most of it quite well tho, so a bit of confidence goes a long way here

cheers