28

Looking at this page which refers to the initial release of the Z80 microprocessor, it prices the ceramic one at $65 versus the plastic version at $59.

I know that at least with the 6502, the earliest run was ceramic and now considered quite desirable since they changed over to plastic-only at some point in the mid-late 70's. But besides having a rarity factor, is there any advantage to having a ceramic chip instead of a plastic one? I presume ceramic has better heat transfer properties, but these things didn't run that hot to begin with.

5 Answers 5

24

One factor is indeed better heat dissipation properties of the ceramic housing. In those times both Z80 and 6502 were NMOS and had considerable heat release, so better heat transfer properties of ceramic housing could help achieve higher ambient working temperatures. Second factor is temperature range. In most cases, temperature range of the chip is defined not by silicon chip itself, but by its housing. Solid plastic housing could literally tear out connecting wires off the chip during extreme temperature changes, unlike the ceramic housing, where the silicon chip is only glued (or soldered) by its bottom part.

6
  • Welcome to Retrocomputing, and thanks for the answer. Did you mean "could help achieve lower ambient working temperatures"?
    – wizzwizz4
    Commented Apr 25, 2017 at 18:17
  • 1
    When I was working on on-board satellite computers in the late 80ies, the only CPU accepted by the ESA as on-board CPU was a 1MHz Z80 in ceramic package (Even if a Z80H was long available). Also, the metal lid that covers the die actually IMHO acts as additional radiation protection.
    – tofro
    Commented Apr 25, 2017 at 18:24
  • 16
    @wizzwizz4 No. CPUs work great in lower ambient temperatures. The challenge is keeping them working in higher ambient temperatures, and ceramic packages help with that immensely. You'll typically find that a ceramic package supports a much higher peak ambient temperature than a plastic one, other things being equal. (There are ceramic package CPUs that are rated to operate at 200C! That is not a typo.) Commented Apr 25, 2017 at 19:05
  • 1
    @tofro That's probably in part because the ceramic 1 MHz Z80 was already space-rated, but the Z80H wasn't. Compare e.g. space.stackexchange.com/q/4922/415.
    – user
    Commented Apr 25, 2017 at 21:42
  • Very low ambient temperatures can make plastic rather brittle ... in the end, you don't want a somewhat water permeable material in running electronics to have any business with frozen, thawing, boiling or condensing water Commented Apr 28, 2017 at 9:03
11

A Ceramic/glass/metal IC housing can offer hermeticity (gas, water and sh*t proofness) satisfying aerospace/military/industrial standards, while plastics cannot do so completely - gases and liquids will slowly diffuse, especially with plastics used for IC packaging often not being a simple blob of resin but a compound material made of more sand than plastic.

Most plastic materials also can take less of a harsh temperature regimen than the semiconductor itself - even if it doesn't melt or burn, trapped water (not truly watertight, as mentioned above) can burst or embrittle the plastic if it turns to steam on sudden temperature changes.

Usually, ceramic packaged versions of components are considered nearly synonymous with military/industrial/aerospace grade parts, which create an impression of especially high quality and reliability (in some parts deservedly so, acceptance testing for these is far stricter).

Also, small/early production runs might have been semi-manually bonded into cases after production - ceramic case kits are sold for such purposes. Also, a lidded ceramic package can be easily taken apart again without destroying everything inside, which can be valuable for defect analysis. Also, you can easily use a semiconductor die or even a multi-die arrangement that is not protected suitably for embedding into plastic.

Finally: There is aesthetics, especially with the gold plating commonly found on ceramic parts. Given that 1980s era computers were expensive anyway, making key components look highest quality was probably considered worthwile. Even electrically nearly irrelevant metal parts were gold plated on chips meant to be blown up along with missiles :) - my theory is that it was about corrosion protection.

4
  • 1
    There was certainly no intention to make the insides of a computer look good. To the average punter in the 80s any collection of chips looked modern. If you look at the military specifications of the time, systems had to work between -20 and +40 degrees. That range was only possible with ceramic packages and gold plating.
    – Chenmunka
    Commented Apr 28, 2017 at 8:12
  • 1
    +105 and more even in aerospace. And average punters weren't necessarily your target market back then :) Commented Apr 28, 2017 at 8:41
  • 2
    @Chenmunka +40C? That's not that extreme. (Nor is -20C; I've camped in -20C!) Commented Dec 12, 2020 at 23:12
  • @AlexHajnal You need to distinguish between ambient temperature outside a system (e.g. a tank) and case temperature of a single chip. I think it is likely that a tank at 40°C ambient temperature well exceeds commercial range (+75°C) on chips with a significant power dissipation (like "highly" integrated NMOS logic in the Z80 or 6502). Commented Dec 14, 2020 at 9:10
4

A major advantage of plastic over ceramic is that ceramics are slightly radioactive. An alpha particle can change the value of a bit stored in memory, and cause errors in programming. Because of work by physicists Timmothy C. May, and M.H.Woods in 1979, computers chips use plastic over ceramics, whenever possible.

4
  • I am surprised on this. Ceramics mean mostly some metal oxides in this context, are you sure that they have enough radioactive contaminants? I think most radiation causing bit flops is coming from some external source.
    – peterh
    Commented Dec 12, 2020 at 23:11
  • 2
    @peterh-ReinstateMonica I don't have a source at hand but that is my recollection as well. Early SRAMs had a real problem with this IIRC. Eventually low-radiation ceramics were developed. Commented Dec 12, 2020 at 23:20
  • 1
    Yes, radio activity is a concern about ceramics. although not as big as it may sound, as materials can be easy designed to reduce this. The problem is lessdue high levels, but rather close distance. ncbi.nlm.nih.gov/pmc/articles/PMC4872684
    – Raffzahn
    Commented Dec 12, 2020 at 23:28
  • 1
    Correction: I meant DRAMs, not SRAMs. Commented Dec 12, 2020 at 23:47
2

This applies more to microcontrollers and EPROMs than microprocessors but...

Devices containing EPROM were erased by exposing them to ultra-violet light for 20-odd minutes. All of the windowed UV-erasable devices I handled were always ceramic.

We had equivalent but non-EPROM parts for some of them and they were plastic, such as the plastic 8035 vs ceramic 8748/8749s. So I presumed the ceramic was to withstand UV damage that the plastic parts couldn't.

2
  • 2
    Nothing to do with UV damage. It's much simpler: the quartz window used to erase EPROMs could only be installed on the ceramic encapsulation. Plastic chips were typically made by potting the entire lead frame and die assembly in plastic resin, so there was no way to leave the die exposed.
    – user461
    Commented Nov 14, 2017 at 1:31
  • @duskwuff, ah I see - makes a lot more sense than my guess. Thanks :-)
    – TonyM
    Commented Nov 14, 2017 at 11:42
2

It used to be the case that ceramic packages were preferred for aerospace applications because it was believed that they were more reliable. This changed in about 1990 IIRC when researchers found that there was no significant difference.

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .