Help Required for VIC-20 Colour-Block Screen

[norm8332]

I already wrote some time ago, that I consider heatsinks with the chips in the VIC-20 as rather useless. The thermal management is already taken care about by conducting excess heat away over the chip pins onto the ground planes. Witness the problems to desolder especially the VIC chip.

What kills the chips is not the heat, but possibly incomplete chip passivation (where creeping moisture lets them rot from the inside over the years), ESD onto ports, and defective PSUs.

... http://sleepingelephant.com/ipw-web/bul ... 0&start=12

Heating and cooling chips repetitively does cause their failure. It's well documented. See the link. There is much more research available on the web also. The best practice to extend chip life is to limit the actual temperature swing whether in use or even in storage.

http://ieeexplore.ieee.org/document/446 ... eload=true

[Mike]

You can cite as many documents as you want - what limits any usefulness of heat sinks in the VIC-20 is the rather small temperature difference between the chip operating temperature and the ambient temperature (regardless of convective or forced cooling).

When the package of a chip gets 40 °C warm, that's barely more than room temperature! At the best, we're talking about a temperature difference of 20 K. The VIC gets warmer - Eslapion once cited a figure of 60 °C. That's still well inside the absolute maximum ratings as stated in the data sheet, and actually shows, that the CBM engineers did a good job of getting the excess heat off the chip over the pins.

If there was no thermal management at all, the power dissipation of 1 W would heat up the chip die to unbearable temperatures within a short time: for a raw figure, the chip die is 4 mm x 4 mm x 0,25 mm in volume, i.e. V = 4 mm³. Take the heat capacity of silicon, c = ~700 J/(kgK) and density, d = ~2.3 g/cm³, and we arrive at the following figures:

mass of chip die: m = d x V = 2.3 g/cm³ x 4 mm³ x 0.001 cm³/mm³ ~= 10 x 0.001 g = 0.01 g

Energy stored per 1 K temperature increase: E = c x m x 1 K = 700 J/(kgK) x 0.01 g x 0.001 kg/g x 1 K = 0.007 J

Operating the chip for even for just a single second thus would increase its temperature by 140 K! Obviously, that doesn't happen: the excess heat first gets distributed into the chip package, then heats up the pins, then heats up the ground planes of the mainboard, and finally finds enough surface area (especially, the RF box) to be cooled away by convective air flow.

[norm8332]

All the studies are saying it is best to keep the chips within the most narrow temperature range as possible to limit the physical stress on the chip die. Some studies show that thermal cycling is the cause of up to 50% of chip failures over the long term. We are talking about 35+ year old chips here after all. That is a lot of cycles and it eventually will physically damage the chip die. The spec sheet concerning thermals is based on an "acceptable lifetime". We are well beyond that with the Vic-20.

I don't disagree with what you are saying, I know the heat is dissipated through other means.

In my experience with many Vic-20s in my possession, in about 30% the VIC chip exceeds 50°C. A properly sized and attached heat sink can lower that by 10°C, therefore lowering the amount of thermal expansion by a considerable amount.

I'm basing my opinion on the fact that these chips are already old and the desire to wring every single day of life out of them. If people don't care about that and will just purchase another VIC possibly sooner, then that's OK too.

[TheMightyMadman]

The VIC heatsink seems to fit well! (I took it back off afterwards to make sure the thermal paste spread properly )

Also, the case treatment is definitely working.

[Kakemoms]

When I decapped a VIC 6560 I found that it had a gilded plate beneath the actual chip. This was connected to the ground pin directly (no wire bond) and would have given more heat conductance out of the chip than from the other pins. All other pins were wire bonded (I didn't measure the wire thickness, but my guess is about 25mils). These would also conduct heat out of the chip, but probably less than the ground pin. I didn't examine them very well, but they might have been pure gold which is a very good heat conductor. The plastic casing of the IC would not be conducting heat very well. For that reason, I doubt that connecting a heatsink to the top of a 6560 would be very effective. If you have a 6560/6561 with a ceramic package, this may not apply.

3M has a thermal conductive chipbond glue which would allow connection of the heat sink to the pins. Whether it makes a difference or not I have no idea. If you want to test it, a TC probe to the ground pin might at least tell you if you are cooling it or not.

As for what is killing old IC's, it is a myriad of effects related to carrier diffusion, oxidation, electron migration, thermal and material stresses. And of course the more catastrophic varieties like static electricity, bad PSUs, coffee spills and overheating.

If you want to protect your VIC-20 chips, I would start with aquiring a new PSU. Actually, I am building a PSU for myself by hacking a new old-stock Mascot PSU. The PSU had never been used, had cost about $200 when new, but already (its about 10 years old) it was showing 500mV p-p spikes on the 5V output(!). The cuprit seems to be bad capacitors of which there are about 9 inside. I have ordered new Rubycon high temp capacitors which will hopefully get it back to withing 50mV p-p (original spec).

So, my point is that old PSUs tend to get inaccurate. Add the Commodore PSU's normal "breakdown" mode which is to output high voltage on the 5Volt, and you have a real chip killer.

[TheMightyMadman]

Thank you all for making this into a 'heat-sinks versus no heat-sinks' thread...

The plastic casing of the IC would not be conducting heat very well. For that reason, I doubt that connecting a heatsink to the top of a 6560 would be very effective. If you have a 6560/6561 with a ceramic package, this may not apply.

3M has a thermal conductive chipbond glue which would allow connection of the heat sink to the pins. Whether it makes a difference or not I have no idea. If you want to test it, a TC probe to the ground pin might at least tell you if you are cooling it or not.

You can tell from how hot the heat sink gets that the thermal conductivity through the plastic is sufficient.

From what I have seen it can be more efficient to attach the heatsink to the pins (see this, among others), but that depends on the plastic of the casing and the material that the 'lead frame' (the pin harness of the IC) is made out of. The plastic of the casing typically has a thermal resistance of 50C/W whereas the lead frame typically has 38-47C/W, so in most cases it won't make a huge difference, particularly in chips with a low power consumption (such as all of the chips in a VIC-20 or C64). The bonds with the heat-sink would also have a smaller surface area if taken via the contacts, and it would look terrible

If you want to protect your VIC-20 chips, I would start with aquiring a new PSU.

I agree. The first thing I did was buy a modern switching PSU, and it works really well with my C64s and VIC-20

[norm8332]

Thank you all for making this into a 'heat-sinks versus no heat-sinks' thread...

It's a no-brainier, heat sinks help in a measurable amount. One side has actual scientific studies as well as first hand testing and in my case decades of experience. The other has a theory and are referring to spec sheets for a 5 year design life.

And Commodore can do no wrong, after all they designed these computers and it's chips to last forever....Not!

All this is about applying a $1.00 heat sink to a hot chip. If people don't want to do it, it's their computer and I couldn't case less but lets make sure the correct info is known.

[Floopy]

I have always wondered if which makes a better heat sink, copper or aluminum?

In the past I have always used aluminum because of it's inability to retain heat. That way it quickly gets rid of the heat, but copper absorbs it and keeps it longer.

Some tables I got off of the internet.

Substance | Specific Heat Capacity
| at 25oC in J/goC
------------------------------------------
Copper | 0.900
------------------------------------------
Aluminum |0.385

This just shows that for every one joule of energy applied to X grams, the temperature will increase X much.

Anybody have insight on this?

[norm8332]

On the high dissipation heat sinks they use a copper plug to make contact with the chip and aluminum for the majority. With the VIC it wouldn't matter much because of the low power. I personally use aluminum.

[Mike]

One side has actual scientific studies as well as first hand testing and in my case decades of experience. The other has a theory and are referring to spec sheets

It doesn't hurt if you name the 'sides'!

for a 5 year design life.

Where are you taking that figure from? I'm just asking.

Have you spend at least some thoughts, that the papers you refer to apply to nowadays chips? The chips today have a much larger thermal design power, if you operate them without the specified heatsink, and they have no internal throttling measures - they can literally explode and punch holes into your computer!

Compared to this, the chips in the early 80ies had structures that can be made out with just a magnifying glass, so to speak. It takes much more for any conceivable ageing effects to show. When the chips aren't operated, there's nothing of electromigration, etc. be had.

And let's face it: what operating time does the typical VIC-20 get nowadays? Do you switch on each of your VIC-20 daily, for more than 4 hours a day - as was maybe typical in those days? No? I won't want to speak for others, but the uptime of my VIC-20 is maybe once every 2 months, for 2 hours in each session. Mostly, to check what I've programmed lately in emulation indeed works on real h/w. I don't think the total operating time since I got my VIC-20 out of the closet in 2007 is ever going to surpass the time it spent switched on between 1983 and 1986.

I have stated my point - there are much more pressing reasons why the chips in a VIC-20 can fail. One first ought address these (and in the case of "moisture creep+bad die passivation", there's not much one can do about this), and when these measures have been exhausted, then it's sensible to put heatsinks on the hottest chips.

...

That being said, I have no interest to turn this into an extended argument. We two have different judgements about one detail what's best to prolong service life of our friendly computer. Nothing worth to pick a quarrel about, really.

[norm8332]

I never quarrel online, just a friendly debate

The 5 year figure is an industry standard assumption. There are thermal studies for 80's era low dissipation devices if you or anyone is interested that can be provided.

The problem really isn't electromigration, it's the expansion and contraction differences between the plastic chip package and the silicon chip die. Similar to when metal is bent back and forth to failure, the differing physical stresses can cause cracking of the die that slowly migrates to failure or die delamination from the base. The tiny wires that are embedded in the plastic substrate and connect to the die can become intermittent or disconnected from the die at the weld point.

Any reduction in the temperature swing of the chip will result in lower stresses thereby reducing the damage. Chips can even be damage by being stored for years in an environment where there are wide temperature swings such as an outdoor shed or attic.

We basically agree that much more should be done to protect the VIC-20 such as power supply etc. but heat sinks shouldn't be completely ignored, especially at the minimal cost. I have about 5 VICs in my collection where the VIC-1 chip exceeds 50°C as stated above.

I've also said my part. All of my VIC-20s have a heat sink on at least the VIC-1 chip and I feel better for it.

[ral-clan]

I installed both heat sinks and a small fan in my VIC-20.
http://www.sleepingelephant.com/ipw-web ... php?t=7117

[Kakemoms]

Heating and cooling stresses a chip over time, so heat sinking can increase the lifespan of a chip by reducing the maximum temperature and increasing the time it takes to heat up and cool down. It won't repair a chip, but it will prolong the life of a healthy chip.

That is not what decays such a chip over time. I have worked with lifetime testing of components, and if you take early catastrophic failure out of the equation, its basically a function of temperature and time.

Thank you all for making this into a 'heat-sinks versus no heat-sinks' thread...

No need to thank yourself

You can tell from how hot the heat sink gets that the thermal conductivity through the plastic is sufficient.

No you can't. Thermal conductance is the important factor here, and for plastic, which has a low thermal conductance, it means that the internal temperature of the chip is much higher than the outside temperature. Had the casing been of a highly conducting material (like silver), the outside temperature and inside chip temperature would have been almost the same.

[TheMightyMadman]

No need to thank yourself

For starting a thread requesting hardware help, or for posting a picture of a heatsink? I kindly requested that the thread didn't end up like this. The more we go on, the less useful this thread is to others.

Thermal conductance is the important factor here, and for plastic, which has a low thermal conductance, it means that the internal temperature of the chip is much higher than the outside temperature. Had the casing been of a highly conducting material (like silver), the outside temperature and inside chip temperature would have been almost the same.

As you say, the thermal conductance of the die, pins and casing will determine how much a heatsink will reduce the temperature of the chip internals by, as the thermal energy must transfer between the die and the heatsink through them. We can't replace the plastic casing with solid silver, so we have to make-do. Surely what matters is that a significant proportion of the energy is successfully transferred to the heatsink, which can dissipate it into the surrounding air more effectively than the plastic case because of its increased thermal conductance and surface area, thereby reducing the operating temperature of the chip? This would be more effective if the casing had a greater thermal conductance, but you can't write off its effects as insignificant because of this.

Based on the previously referenced data, the pins of an IC from the 1980s are likely to have a greater thermal conductance than the plastic casing by between 4% and 24%, meaning they will be somewhat closer to the temperature of the die. So (as for your original point) it stands to reason that attaching a heatsink via the pins may be more effective than attaching it to the case, but for 4-24% would you bother? Assuming that the reduced contact area doesn't balance out the increased thermal conductance, which could make it pointless altogether.

From what I've read the thermal conductance of the die, pins, and case should be designed to match one another as closely as possible, to minimise differences in the expansion of the materials which can cause separation and damage contacts. Do we know whether MOS took this into account when designing their chips? Because this would mean that the heatsink wouldn't be any more effective even if it were directly on the die.

[TheMightyMadman]

I installed both heat sinks and a small fan in my VIC-20.
http://www.sleepingelephant.com/ipw-web ... php?t=7117

That's really neat! The pictures don't seem to be working for me though, I'm afraid?