PDA

View Full Version : suggestions for period fitment of heat sink to ceramic 8087 40pin DIL



inotarobot
June 23rd, 2017, 03:41 PM
Hi all

I am looking for suggestions for period fitment of heat sink to ceramic 8087 40pin DIL or similar chips, that have the raised center,

Have seen them "super glued" on, but that material is not actually a good thermal conductor.

Generally what I have noticed it on ones already fitted is its only the tiny middle section that that heat sink is thermally attached to, the larger portion of the chip body, ie each side of the center raised portion is not actually touching the heat sink.

I don't feel this is correct to have maybe only 20% of top of chip in contact with the heat sink. In these instances I have worked heat sink paste into gap with a fine needle

One can use heat sink paste, and some form of bent clip from one end of chip to the other.

What glues or other methods have you used ?

Did you perform any thermal tests on the results of your mods ?

The old PCB's never had any heat sink fixing holes alongside these processor or co-processors, and its 100% unwise to try and drill board to add such holes. Often even earlier boards where multi layer, so one could cut a internal layer track.

Ok lets also add some replies for mounting heat sinks to the plastic versions of 8087 or like please. These typically are flat across the whole top surface.


Also when you 'glue' to top of chip, do you first take a pic or make note of the chip markings for future reference ? ie speed rating etc

Stone
June 23rd, 2017, 03:52 PM
My XT with its coprocessor has been running along just fine for ~ 30 years without any sink, fan or whatever. I'm certainly not about to alter its environment now! :-)

Chuck(G)
June 23rd, 2017, 06:03 PM
Most of the 40 pin DIL package is lead connections. The (heat-producing) die is actually quite small and located under the metal lid.

But if you must, you can take a standard cheap aluminum DIP heatsink (https://console5.com/store/dip40-heatsink-glue-on-thermal-epoxy-on-style-dip-40.html) and mill out an area for the lid, then attach the thing with thermal epoxy (http://www.arcticsilver.com/arctic_silver_thermal_adhesive.htm).

But I'm with Stone on this one--unless you're seriously abusing your NDP, you shouldn't need a heatsink.

mR_Slug
June 23rd, 2017, 06:28 PM
I don't know about period, but here are a few ideas:

I've never done it, but was once told you can mix epoxy and thermal paste together to hold on a heat sink. I can't remember what ratio was stated. With the height difference you could use some metal to shim it. Floppy disk shutters are good. Also aluminum foil can be folded a few times with a bit of compound on it, although it does start to become a mess after 3 layers.

Mechanics wire (or possibly cat5?) wrapped around the h/s and chip, twist the ends together until its tight enough.

A safety-pin of the right thickness could work in a similar way. You'd have to pull the latch with pliers, to get the pin into it.

The most reliable method i think, is to sandwich the chip and heat sink between two plates of steel slightly longer than the chip, and then use bolts either end to crank down the pressure. One of the plates could be the heatsink if you can drill through it. But this seems a bit over the top.

KC9UDX
June 24th, 2017, 01:10 AM
I think about this a lot. Most any undesirable way to modify something aside from using things that just didn't exist at the time, is a period way to do it, in my opinion. It would have been perfectly normaly for someone to super glue an inappropriate heat sink to that chip in a terribly inappropriate way, when the computer was new, or at any time in its existence.

I find that doing things properly and perfectly, and least-intrusively is just not "period correct" in all but the rarest cases.

Chuck(G)
June 24th, 2017, 08:46 AM
Huh? I'm not following this line of reasoning.

Long before the 8087, TRW was selling bipolar multiplier chips in DIP40 with glued-on heatsinks.

http://cdn.embedded.com/contenteetimes/images/01MDunn/BenchTalk/rememf2.jpg

inotarobot
June 24th, 2017, 04:02 PM
Huh? I'm not following this line of reasoning.

Long before the 8087, TRW was selling bipolar multiplier chips in DIP40 with glued-on heatsinks.

http://cdn.embedded.com/contenteetimes/images/01MDunn/BenchTalk/rememf2.jpg

Glued on yes. Likely with a a fine layer of thermally conductive glue.

I decided to use "Miss Google", and one of the early its was from https://www.masterbond.com/properties/thermally-conductive-epoxy-adhesives.


Common Applications Featuring Thermally Conductive Adhesives
Thermally conductive formulations are used for bonding, potting and encapsulation applications in a wide variety of industries. Some specific applications include:
Heat sink bonding
Potting/encapsulating sensors
BGA die heat spreader interface
Chip scale packages
Power semiconductors

inotarobot
June 24th, 2017, 04:18 PM
http://i.imgur.com/1TewOCf.jpg?1 this is a 8088 processor & 8087 coprocessor daughter board (aftermarket made ??), as found installed in my Sirus with the 8087 under the heat sink

side view of how heat sink was fitted

http://i.imgur.com/RTrdYIb.jpg?1

you can see the gap at each end.

Update to this pic as I was handling the board, while reinstalling it, into the motherboard 40pin DIL socket, I ended putting ' very slight' pressure on the unsupported end of the heat sink, so as to get the board to seat into the motherboard socket.

Yes you guessed it. Heatsink came off. So the glue bond was not really good.

I wonder if in fact the glue used may have actually be thermally NON conductive and thus actually INHIBITING the natural air flow cooling of the chip
as originally designed by chip manufacturer.

Now I have this blog of glue on top of chip to carefully remove, to bring it back to original surface for air to flow and this cool.

I don't think the machine will get enough use to warrant getting special glue to put the heat sink back on.

Chuck(G)
June 24th, 2017, 05:31 PM
...now you can see why I recommended milling a relief in the heatsink for the chip lid.

But heck, you knew that good physical contact is everything. Look at the overclockers who lap their heatsinks (http://www.overclockersclub.com/guides/lapping/). Interestingly, rather than a mirror finish, a "matte" finish appears to be more desirable.

paul
June 24th, 2017, 06:06 PM
I wonder if in fact the glue used may have actually be thermally NON conductive and thus actually INHIBITING the natural air flow cooling of the chip
as originally designed by chip manufacturer.
Thermal resistance of the bond is dependant linearly on the thickness of the bonding/thermal material as well as its heat-conductive properties and area covered, so in theory even a dab of superglue would likely be very effective.

But, using a non-flexible material such as superglue would not allow differential thermal expansion of the two surfaces and likely fail quickly. Also, I've not 100% convinced that the "gold" cap has a good thermal connection to the die.

As you say, there is an academic question about whether the heatsink as shown in the photo is more effective than a bare chip, given that it prevents airflow to the top ceramic surface at the gaps. My guess is that it is no worst, given some radiative heat transfer.

For chips such as the 8087 increasing the airflow just slightly will be very effective at lowering the temperature. The 5150 doesn't get much airflow in the area where the CPU/FPU are located and I do recall measuring them once and finding them just within spec but put a small fan in anyway.

39288

konc
June 27th, 2017, 04:31 AM
Apart from the hard-to-remove glues there are also thermal conductive pads (3M has some great products). They are softer and "rubbery", I don't know how else to describe them. While they do a good job in keeping a heatsink in place, they can also be removed without the use of extreme force/tools etc and, at least the one I've tried, doesn't erase markings on the component. Just keep in mind that many listed for cheap on ebay are counterfeit and definitely not conductive.

lowen
June 27th, 2017, 05:07 AM
I'm getting ready to put one of these tape-type products to the test on a Z280. I know, not a 40-pin DIP, but the tape I got is the correct width for a 40-pin DIP. I bought a 25m roll of Sywon Professional 20mm-wide heatsink tape on Amazon for $15. If it works well, this is enough to last a really long time. You can get heatsinks with the double-stick thermal tape already applied.

The 'today' way using an adhesive rather than tape would be something similar to Arctic Silver's thermal adhesives (alumina ceramic or the silver).

You can mix your own more period-correct thermal adhesive. Get some zinc oxide powder and mix with regular epoxy, 1 part zinc oxide to 1 part epoxy (don't use a quick-setting epoxy for this; you want the time to mix it well). I can't tell you where I learned that, but suffice to say I worked at a place that used that mixture for several uses (along with the same proportions of a heavy silicone oil and zinc oxide; this place mixed that stuff up by the 55-gallon barrel). They also used some really heavy two-part silicones mixed with zinc oxide, but the adhesion of those isn't as good as the epoxy. Room-temperature-vulcanizing (RTV) silicones can work, but you might have fun with the acetic acid that they put off. This place used the epoxy mix for gluing thermocouples to devices under test. Since they used bulk thermocouple wire and welded it on-site, thermocouples were considered disposable.

Either way, zinc oxide is 80's-period-correct, and should have sufficient thermal conductivity for an 8087. If you were doing a Pentium or higher you would want the Arctic Silver, and a Pentium 4 would probably need the alumina version. The Sywon tape is marketed for GPU usage, but I've not dug deeply into its thermal conductivity.

Chuck(G)
June 27th, 2017, 08:11 AM
Here's a thought. Since you're not concerned about electrical conductivity, you might also consider metallic copper or aluminum powder, which would also impart a fair amount of thermal conductivity to epoxy. There are non-acid silicones, but I've never considered them as being particularly thermally conductive.

My standard heatsink compound, for what it's worth, is a big jar of Permatex anti-seize compound. Loaded with aluminum, hasn't let me down in decades.

KC9UDX
June 27th, 2017, 07:25 PM
If you did care about electrical conductivity, file your own beryllium copper dust; just hold your breath.