thermal baggie things

Oops. It looks similar, has a similar name, but you're correct. It's not a plastic blade.

The real plastic razor blades are made by: The blades look like this: and cost about $9/100.

I've bought them at Ace Hardware and Home Despot stores, but can't seem to find them on their respective web piles.

For your amusement, here's a test comparing various thermal compounds. The winner is mint flavored tooth paste. For the Australians in the group, Vegemite was a close second and equal to Arctic Silver. Tap water worked the best, but was impractical to use:

I bet that plastic blade is good for removing those stupid stickers they put on notebook like "dual core" or whatever.

the

I just burned an hour searching with Google and looking at Google Images for something similar. I also checked iFixit for tablet construction. Nothing similar was found.

I repair laptops and have probably torn apart several hundred laptops. I have never seen such an arrangement. However, it may be in some newer laptops or on something I haven't repaired yet.

I find it odd that a vendor would use a thermal gel bag. Conducting heat through a PCB is not very efficient. There's an air gap between the CPU and the PCB creating a high thermal resistance path. The gel might be a good thermal conductor, but the plastic bag is not. There's also a risk of the bag leaking or bursting, which would both cause the CPU to overheat and make a mess. I just don't see it as a practical method of heat sinking.

So did I. When Googling for "silicone bag", I found lots of references to breast enlargement implants, which presented a momentary distraction. Rubbery and foamy were descriptions used for failed implants. Ummm... what were we discussing?

Thanks. I'll grind some numbers tonite or tomorrow. I'm busy throwing away about a third of the accumulated junk in the office.

But first, a wild guess(tm) and sanity check. Altera thermal resistance package specs: Which package/series are you using?

Pg 24. Looks like between 10-13 degC/watt to the BGA package in still air. If it's dissipating 1.5 watts, I would expect about a 20C temp rise above ambient or an operating temp of about 45C. Since you're actually seeing 60C operating temp, I would suspect that it's dissipating about 2.5 watts. Did you measure the 60C temp (under the BGA) with the covers in place? There could be some heat build up inside, which would explain the higher than expected temperature.

Yep, that works, although the labels usually fall off or peel without any additional assistance. Extra credit to HP for attaching large feature lists and advertising on every machine using non-removable glue.

If you've ever seen a "personalized" notebook, favored by the younger generation, you'll see why I like plastic scrapers. The various stickers and labels need to be removed. I use solvents, but to get to the glue, I have to scrape off most of the label first. The plastic razor blades are perfect for the purpose. However, they will scratch an ABS cases if used to aggressively. Breaking the sharp corners on the blade with sandpaper or a file does help. A hot air gun also helps by melting the glue, but in the end, the scraper is what does the dirty work. The only labels that present a problem are the government property tags, that seem to have a solvent in the glue that attacks the plastic. Those tend to leave a permanent mark when removed.

of

the

I know I've seen ads for them, but I just can't google any. It looks like the foamy stuff will work, but it would have been cool to just sticky down a little gell-filled pillow.

In my case, it is. The ADC has a power pad with big vias to the ground plane (layer 2, spreads out the heat) and a big copper patch on the bottom, layer 6.

The FPGA is a 256 ball Cyclone 3, with many balls via'd to the ground and power planes. That spreads the heat across the board surface.

Oh, one discovery that we made: the usual dogbone construction of BGA ball pad going to a via is thermally dominated by the narrow middle of the dogbone trace. It's much better to use a wide trace, namely make a racetrack oval of the pad and via. Even a small via running into a layer 2 ground plane has a pretty low theta, ballpark 20 K/W. The dogbone can be 10x that.

FR4 is still a much better heat conductor than air.

There's an air gap between

The similarity had ocurred to me. I hate it when women disfigure themselves, replacing perfectly good female flesh with globs of plastic.

I'm planning to go in today and clean up. We have two major customers coming in soon for company audits, all sorts of acronyms we never heard of, ISO9xxx, SPC, checklists running into hundreds of items. Yuk.

It's an EP3C5F256. Looks to me, from the IR images, that the die itself is a serious hot spot, with poor lateral heat spreading. So only the inner balls, fortunately ground, do much heat transfer. On future boards, we might add another patch of heat spreader copper on another inner layer.

I really ought to measure die temp. Maybe I'll wind up doing that, instead of cleaning my office.

A big part of my problem is that the PCB itself is rising maybe 20C above ambient, so the FPGA rise is on top of that. Hence the first fix, cooling off the board.

I removed the top cover and replaced it with Safeway cling wrap, the super thin Saran-wrap clone. It's thermally similar to the cover (low theta but blocks convection) and it's transparent to the IR imager.

stuck a

about 1 C,

under the

one

could be

away

but I

using an

Personally, i think your best bet is a quite small fan fully inside the case. All it has to do is stir up the air inside, makes a surprising difference.

?-)

be

an

Unfortunately, that's not very practical here, especially as a retrofit. And fans aren't reliable. Vibration could be an issue too, as some of these boxes get mounted near some very touchy stiff. Passive cooling would be best.

I've used hex threaded stand-offs, brought up to be a bit above the pcb and then bolted TO-220's to 'em.

George H.

ed text -

With a clearance hole in the PCB, or alongside? Cute.

Leads to another idea, an insulated version of that standoff, with a big bottom foot with peel-off sticky.

Want to join my "fan" club?

I just love adiabatic nightmares. The hot air has to go somewhere. If you just insert a fan into a closed box, it will help with the hot spots, but will also heat up areas that are currently relatively cool. If you just want to do that, maybe a piezo fan is the answer. About 2 cfm (0.9 liters/sec).

With a closed box, all the heat is removed by radiation from the case. (Incidentally, painting the extrusion black will help a tiny bit with radiation efficiency). If the outside ambient air is colder than the average temperature inside the box, then adding intake and exhaust holes will greatly improve cooling.

At low power dissipations, fans also add heat. A small 30mm * 7mm fan burns about 60mA at 12V or 0.72 watts. With the unit currently drawing an alleged 6 watts, adding the fan is a 6% increase in heat load. That's not going to cause any problems. However a larger 40mm

• 10mm fan will need 130ma or 1.5 watts, which is a 25% increase in heat load. That might be a problem.

Foundit in the paint department. P/N 1380971 $1.79 with 5 double edge blades. Cheap. I bought two spares.

A major part of electronic design is thermal design. Especially fast stuff... current-mode opamps, ECL, FPGAs clocked at hundreds of MHZ, or part of FPGAs clocked at several GHz.

The hot air has to go somewhere. If

Or by convection. In my situation, the box is bolted down to a structure, so most of the cooling is via conduction. Probably the sliding top cover gets mostly convection cooling.

Filling the space below the board with the Bergquist 0.25" thick pad stuff looks like the best fix. Chip temps drop by about 15C, on average. Installation is easy: just drop it in between the spacers and replace the board.

I suppose you don't have much in the way of thermal problems if all you use is 4000-series CMOS and 555s.

Have you ever designed a board with a current-mode opamp? Or an FPGA?

Ever done a multilayer PCB design with serious thermal management issues?

This pulse generator was water cooled:

??

re

't

I'm curious why don't did you bolt the pcb to the bottom instead of using the slots on the sides? saves the drilling operation

We use a similar box sliding the pcb in, the edge launch bulkhead SMA connectors with a nut on the outside of the end plates holds it in place probably helps a bit with cooling too

-Lasse

It could be

laterally away

does, but I

using an

And

boxes

Maybe you could consider heat pipes. That is passive and beats most goo tech and it gets the heat from the hot chips to the case.

?-)

Is that good enough? I never saw you indicate what chip temps you wanted. You mentioned the chip die temperature of 85°C, but I haven't seen your calcs to show what the case temp is for the target die temp. Have you figured this out yet or are you just going to play it by ear?

LOL! I was in some overclocking groups some time back and people were both fanatic about cooling and irrational. Even though the goal of the goop is to resist the flow of heat as little as possible, they treated it like it was somehow a magic thing and could lower chip temperatures by many degrees. They never understood that even if you use a perfect heat transfer between the two surfaces, you can't lower the chip temperature any more than the number of degrees dropped across the interface with whatever you were currently using. So going from magic fluff A to magic fluff B could only make a very small difference no matter how good magic fluff B was. It is more a function of how bad magic fluff A is and we all know magic fluff A is pretty good stuff...

One of the things I was never able to figure out was if a faster flow of coolant was significantly better than a slower flow. Once you are moving the cooling fluid fast enough it is not in equilibrium with the heat sink, does moving if faster help. The context was water cooling where it would also not be in equilibrium at the radiator... as a first order assumption. My take was moving the coolant faster just meant it was more of a constant temp with smaller temperature rises at the heat sink and less drop at the radiator. I think the improvement of heat transfer would not be significant.

I'm working at the other end of the thermal scale these days, ultra low power devices that scavenge power from things like heat or light sources and electric fields.