Need some custom pedal power supply help

[PRR]

> these heat sinks on order:

DigiKey Search links do not work for anybody else. Copy-paste the part-number. (Even then, be careful. This week a client wasted 25 minutes of my time with an almost-right part number one digit wrong.)

I only see one "38" sink at Mouser, and it is non-stock. It is also under a half square inch per side, far short of my 2"x2" rough-guide.

> rated 38o C/W ... vs 28o C/W .... Is less or more better?

What R.G. said only less.

The number is a temperature for One Watt of heat put in it.(*)

38 is hotter than 28.

"Your" sinks will run hotter than R.G.'s sinks.
____________________________

Modern Silicon/Epoxy parts can run VERY hot and live. I have seen a power transistor working fine even though its heat had melted the solder on its legs. (But for how long? And how much leakage? In this case leakage was tolerated, in other circuits it might go crazy.)

With old age and full trash-can, I like to hold down to 50 deg C. (This is in part because audio power amps thermal-cycle on every boom of the music; steady heat does less damage.)

On the face of it, 3.055 Watts at 38deg/W in the sink, plus 5deg/W inside a LM317, is over 114 degrees C *rise* above ambient. Assume ambient is 30 deg C (room-temp plus transformer heat), you are mighty darn close to LM317's number of 150 deg C at the junction.
___________

(*) Actually, convection can be non-linear and radiation is very non-linear. Heatsink makers often run the sink VERY hot, they convect better, radiation becomes non-negligible, and the number looks better. Assume these numbers may be very best-case. Round up (bigger sink, lower deg C/W) generously.

An aside: if the sink is flat (not warped or dinged), at the 38deg C/W level sink-goo is IMHO not essential. It can reduce interface from 2 to 0.5 deg C/W at best; in a 38+5 deg C/W budget a couple C/W difference looms very small. At this point, put $3 into bigger sinks, not a tube of goo.

[tubegeek]

Question about heat sink goo:

In a pinch, would lithium grease (white grease such as I stock for my bike chain and you undoubtedly stock for your chainsaw, PRR) be any good?

[PRR]

Lithium grease for a chainsaw?? Must not have chainsaws in your neck of the non-woods.

Classic thermal paste is a lot of Zinc Oxide (old fashioned sunburn cream). Cheap and conducts heat much better than air or straight oil. Aluminum Oxide (ditto) is also done. Silver is also touted. This gunk can be 80% of the paste. Which suggests that any Silver is garnish, not meat.

I don't think there is much Lithium in Lithium Grease. AFAICT, lithium stearate and lithium 12-hydroxystearate have one (two?) Li on the end of a very long chain, like C18H35LiO2. They are just tags on the end of a long mostly-C&H soap used to bind the oil.

Inasmuch as oil is more thermally conductive than air, it may "help".

But as R.G. says, this is Ohms Law. Remember your series circuit. If you have heavy flow in a path with a 2 Ohm and a 38 Ohm connection, "optimizing" (greasing) the 2-Ohm down to 1 Ohm even zero Ohm isn't going to make any real difference. The 38 Ohm wire will still get hot. And I suspect that at 3 Watt flow, a 38 C/W sink will get hot (in terms of happy Silicon). Whatever time/money put in grease will be much better spent on the sink.

[R.G.]

As Paul says, both this is Ohm's Law (albeit twisted into usefulness for lies, d@mned lies, and advertising) and anything is better than nothing.

Air is one of the better insulators. Goose down is "warm" because the micro-barbels of the down trap air and don't let it circulate much, so it approximates still air. Air-blown foams and other entrapments all the way up to aerogels just hold the air still so it can't convect.

If I had to heat sink something and had no goo, I'd use any grease at all. Anything is better than air.

Actually, as I think I mentioned, the rubber thermal pads are really handy. They're worse than goop+mica, but then as Paul said, if you have a 38C/W sink, why squint 1-2C/W on the interface.

One lateral-thinking way to fix this is to lap the semiconductor to the sink surface with fine abrasives. This flattens both surfaces microscopically and improves heat flow by flattening the micro-mountains of rough surfaces. But that's a far-out labor of love.

If you're going to do many sinks, even many over years, get goo or use rubber thermal pads. I'm still working on my original can of gook. Even better, find friend who are dum... er, enthusiastic enough  to want goo of their own and split up one order 40 ways. Or order a $4 syringe from Frys.

[Sanguinicus]

Here's my Geofex Spyder. It's a triumph of efficient use of space.

Build report: http://www.diystompboxes.com/smfforum/index.php?topic=111457.msg1025466#msg1025466

Pictures: https://www.dropbox.com/sh/4rlghw3h1mxcuir/AAA6lTcGNdtvpwoDFLkQgCg1a?dl=0 Also pictures of the pedalboard I built.

Unfortunately there's no picture of the second veroboard stacked on the bottom one.

[Natman]

Thanks again everyone -here's what's nagging me a bit:

"Your" sinks will run hotter than R.G.'s sinks.

The unit of C/W only indicates the energy "trapped" by the heatsink per unit put in (and more is bad ultimately). 

However, isn't it GOOD to trap more enrgy, thus sucking it away from the regulator up to a point?

Or is it stritcly the running temp at saturation, which is bad since it's in intimate contact with said regulator? In my (very) short investigation in between tasks at work, it seems like the larger heatsinks have higher C/W values.   

I'd much rather see them use "power dissipation at temperature rise" since it's more intuitive for what the heatsink's actually doing. For some reason, few of the models list this parameter.

I recognize that the heatsink then has to eject the heat by dissipating to air, but this is a nearly fixed rate (natural convection will increase with hotter temps). Wouldn't copper be better than aluminum for most applications?

[PRR]

> dissipating to air, but this is a nearly fixed rate (natural convection will increase with hotter temps).

What you said. The "sink" can't "trap" power; it would blow-up or melt.

It spreads the power to a large amount of air. Which is cheap, and tends to wander-off, carrying the heat power with it.

The *rate* of convection goes up at least as fast as the temperature rise.

An air-cool engine has smaller fins for its power than a transistor amp, because an engine's cylinder and piston can run at much higher temperature than the delicate processes we use in transistors. The extra-hot air rises through the fins much faster.

> rather see them use "power dissipation at temperature rise"

At what temperature? Sometimes there's reason to hold to <50C to reduce silicon leakage, other times we know the conditions are heavy but un-fussy and we allow 100C rise. Such a chart would have to have many lines.

> Wouldn't copper be better than aluminum for most applications?

If it weren't so hard, steel would be better.

Copper may be best but Aluminum is SO much cheaper that you can do fat Al for less than medium Cu.

Copper is standard for in-chip heat spreaders. Heat density is high, size is small, cost is small. If you have a metal-tab TO220, file the tin on the tab, it may be copper under the plating. (However steel is also done, and steel was routine in TO-3, tho sometimes abetted by a small copper spreader.) 

Any metal carries heat a thousand times (+/-) better than thin air. "Most" of the "heatsink resistance" is not in the metal, it is in the layer of air which the heat must flow through to reach the "whole world" of air.

Yes, as a commercial proposition, the heatsink designer thins-out his metal enough to reduce cost with "not much" harm to the performance. If metal were cheap, we might aim for 2% in the metal 98% in the air (as we do for power wiring). If an expensive metal is specified, you might aim for much higher loss in the metal to reduce cost.

Steel and zinc are the cheapest metals. I have seen a steel sink doing its job just fine. This is especially true when a large structural member (chassis) is available in steel.

But Aluminum is "large" for its cost, which means large cross-sections. It is much cheaper to work into fancy shape than steel. Since the electrolytic process, Aluminum is real cheap. More per pound than steel, but not much more for a given volume (area times length), and that volume gives better conductivity.

Why not Silver? There was a racing motorcycle company with a hot head. It was decided to explore silver. They even found a silversmith who would buy-back all the scrap. But when they looked at the strength of Silver they saw it would not contain their large explosions. They had to throw quite a lot of Copper in the mix to get strength. (Which also meant that a scrap silver buy-back would have to charge an assay fee and refining mark-down.) In the end it was still weaker and did not carry off heat much better than Aluminum alloy, while being attractive to theft.

Heat-pipes conduct much more than any metal. Some laptops heatpipe from the CPU to a fin-array out on the edge of the enclosure.

But for a few Watts and no killer problem about space, several square inches of reasonably thick scrap-metal is the usual path.

[R.G.]

How hot is 1W?

That's a trick question. power (i.e. watts) has no temperature. 1W makes different things heat up different amounts, and how much is purely a consequence of how easily the heat can get out. Mother Nature (in the form of the Three Laws of Thermodynamics) says that temperature is a side effect of heat transfer. Temperature goes up **without limit** until the heat being produced can get out of whaterver it's in. If that melts or vaporizes the thing making the heat, well, OK, so be it. 1W in a household iron won't even get it perceptibly warm. 1W in a grain of wheat bulb makes the tungsten filament heat to over 2000C.

In a normal human environment, all heat eventually gets out into the air (or water, or dirt, etc.). How hot the thing gets depends on how well it couples to the air around it. Air carries heat away by the few molecules hitting a surface getting heated by the touching, then bounding off faster than they came in. It's a very thin film effect because air insulates so well.

The thermal resistance of the air is pretty fixed on a molecule-by-molecule basis, so to get more heat out of something, you have to increase the number of air molecules hitting the thing and carrying off heat. One way is by blowing air over the thing. Fan cooling is really effective and getting things cooler in most cases. Another method is to leave the air to do its own wandering around, but make the surface it can hit be much bigger, by spreading the heat out over a bigger surface. More surface means more molecules can hit a hot surface and carry away more heat.

Heat sinks are heat spreaders. They spread out those watts so that each unit of surface area only has to get a tiny fraction of the watts out, so the overall heat flow can get out without a high rise in temperature to force the watts out. It's a lot like using many high value resistors to make a lower overall resistor. If you want a 1-ohm resistor to conduct 1A at 1V of drop across it, you can make this up by connecting a hundred 100-ohm resistors in parallel. Each one carries its fraction of the current, but the overall voltage stays down.

As a final gee-whiz, the best material for heat sinks/spreaders is not a metal at all, nor even an electrical conductor. It's cubic carbon crystals - diamond. They conduct heat better than any other solid. To get better than that, you have to use moving fluids or gasses.  Diamonds have been used for some particularly high end semiconductor stuff. But they only use it to spread the heat out of the chip to a metal, usually copper, but sometimes silver, as silver is about 20% better than copper, but many times as expensive. Aluminum is generally used next; it's about half as good as copper, but much cheaper again.

There is a recent shakeup in this lineup. It is possible, although they're still working on it, that boron arsenide is as good or better than diamond. Maybe. 

[Natman]

You guys are so much fun to nerd out with!
Brings me back to my university days, wondering why conductivity on some ceramics (diamond) was better than metals.
We had a prof. who deominstrated heat pipes and had EVERY student stumped as to what was going on physically.

On a side note, all my parts are shipped, but I forgot to include my xformer in the cost. This project will cost me just over $100 all in, not counting solder and wire. Still can't get anything commercial for that money. Stay tuned and thank again!   

[Natman]

Hey RG, I just realized that I'm not clear on 2 things:

I am going to use my center-tapped secondary for TWO supplies just like your dual supply at the bottom here:

1) How exactly do I rectify these? I had intended to use 2 full bridge rectifiers as if they were separate but sharing GND. 

2) It is clear that the outputs will be +9V and -9V (18V and 12V in my case). So recognizing that, can I just wire one jack "backwards" (or leave it tip negative) to get both +ve V or is there more to consider?

Much appreciated!

[Natman]

Hey all, I got my parts and started populating my board. Slapped the heatsinks on my 2 big regulators with goo last night. I even found a great idea for the enclosure. Pics to come... 

Here is an interesting discussion I found and was wondering how valid the small capacitors are for this kind of application? If there's truth to it, I might pair the 0.1 uF caps with the 1000uF and go with 100/0.1 uF on the output side... 

http://electronics.stackexchange.com/questions/21686/whats-the-purpose-of-two-capacitors-in-parallel

Lastly, I sketched a schematic in paint for how I think the 18V secondaries should be wired. Please correct me if I'm wrong!

[antonis]

IMHO you should place a 10 - 100 μF cap between regulator OUT & GND..
(if you desire a decimal percent of ripple voltage..)

And a power diode (like 1N4001 or similar) from OUT to IN would help during "abnormal" situations..
(like preventing damage to the regulator in case the input voltage gets interrupted..)

[Natman]

Cool, will add diodes. 
Is there any reason to use electrolytic (polarized) for the large caps and not for the smaller ones?
I have both on hand but seems to me like it would make no difference. 

[PRR]

> I sketched a schematic

That makes my eyes hurt.

There's no reason for two grounds, there is no need for *eight* diodes, and the CT connection is not clear. (Drawing is nice though!)

Go STEAL "study" several of the many-many-MANY bipolar supplies already published on the Web. Maybe even R.G.'s site.

[antonis]

Is there any reason to use electrolytic (polarized) for the large caps and not for the smaller ones?

Large caps handle low frequency ripple and mains noise (and major load changes..)
Small ones handle noise and fast transients..

Electrolytics have good ability to filter out low frequency ripple but they are not good at filtering higher frequency noise because they tend to have large internal inductance & relatively large internal series resistance (ESR).
(but they are much cheaper than non-polarized caps of the same capacitance - if you can find them at the range of mF..)

Small caps are non-polarized (mostly ceramic - I hate tantalum  ) with low ESR and low inductance giving them excellent high frequency response and noise filtering capabilities.
By themselves aren't enough to do the whole job as they cannot store enough energy to deal with the energy needed to filter out ripple changes and large load transients - so we have to deal with a paralel combination of both kinds..

P.S.
If you study MANY bipolar supplies (as suggested by Paul  ) you will find parallel combination of three or more caps (especially for output decoupling purposes at benchtop stabilized power supplies..)

[Natman]

OK is this better? (add diodes across regulator)

Quickly looking at the online bipolar supplies, I generally see them all rectifying the full secondary (36V in my case) and 78**, 79** regulators. Guess I'll need a -ve regulator huh?

I would rather use the parts I have on hand and treat the secondaries as -18/0/+18. I mainly want to maintain the full current rating if possible. Notice how I drew my jacks. One is pin +ve, the other pin -ve, which is fine since that's what they'll be feeding. 

[PRR]

> Quickly looking at the online bipolar supplies

Look slower.

Your drawing is still UN-like any other I've seen. Put the ONE ground in the *middle*. This will show that you do not need _8_ diodes (they duplicate each other); four will do.

http://www.circuitstoday.com/wp-content/uploads/2011/03/5-volt-dual-regulated-supply-for-f-to-v-converter.png

If you don't get it right, the lights go dim and then the cellar fuse blows. (How would I know?)

> and 78**, 79** regulators. Guess I'll need a -ve regulator huh?

Since the '79xx came down to same-price as the '78xx, it has made MUCH more sense to use CT PT and 78/79 pair. It saves a bridge.

If you "must" use only positive regulators, there is a way. So out of style that I am not seeing an example. But conceptually trivial. Build two regulated DC supplies. Don't ground anything yet. Then wire the "-" of one to the "+" of the other, and call that point "ground".

Is this really a BI-POLAR supply? Or are you building two supplies, opposite polarity, common ground? If the latter, and you have *separate* windings, you probably just build two supplies (using + regs if that's handy).

[Natman]

Ya, that's the thing, I'm not really looking for bipolar here. I am debating whether to try and separate the windings by finding the center tap and soldering 2 separate leads? It's risky surgergy but I would feel ALOT better about it.   

I bought my transformer because I saw it had 5 secondaries. Then I realized the two 18V's were really a center-tapped 36V. I thought fine, I can still make it work (I still think so).

Here's a concern:
I NEED my 12V to replace the wall wart rated for 500 mA. I don't actually know how much current it draws, but I don't want to cut it too close. The pedal has a tube at full voltage inside so I'm guessing there is some demand. If I rectify the 36V with a single brisge, my current rating will still be 500 mA but now it's for the 18V and 12V supplies combined and I'm definitely too low for the 12V. 

The more I think about it, the more I am tempted to wire two 9V taps in series to feed my 18V supply and leave the 12V separate, using just half of the center-tapped secondary. This sacrifices a 9V supply but overall it might be wiser.

Again I really appreciate your help guys!

[tubegeek]

Lithium grease for a chainsaw?? Must not have chainsaws in your neck of the non-woods.

Only the serial killers really use them around here.

Classic thermal paste is a lot of Zinc Oxide (old fashioned sunburn cream). Cheap and conducts heat much better than air or straight oil.

Thanks! That's easy enough to find, even around here in the non-woods...

[Natman]

Here's a long overdue update on my findings.

Even though it took forever fo rme to get organized, I measured the current on my Jouster (below). It peaked at 480 mA when both channels are on and that is a definite no-go with the 500 mA rated xformer so I bought a new one rated for 1 A. Bad news is a I lose one 9V secondary. Considering the 18 V and 12 V are the most important to me, it's a small sacrifice.

I am also going with the standard bipolar schematic. Pretty sure it will work. Just have to wait for the transformer to arrive...