52

Put simply, why do some diodes such as most Zeners and Schottky diodes have a glass package as opposed to the more traditional plastic package?

Is it ease of manufacturing, thermal properties, or some other electrical phenomenon?

JYelton
  • 32,302
  • 33
  • 134
  • 249
Hugoagogo
  • 2,405
  • 3
  • 22
  • 29

2 Answers2

49

Early semiconductor diodes were mostly glass packaged which provided the advantage that they were hermetic and did not depend on passivation of the chip to survive heat and humidity. The glass package also allows a very high operating temperature. Early devices such as the 1N34A (germanium) and the 1N914 as well as the 1N7xx Zener series became very popular and inexpensive.

Plastic-packaged devices were developed to reduce costs where high performance was not so important.

For example, the glass 1N4148 has a maximum junction temperature of 200 °C compared to only 150 °C for the plastic-packaged 1N4001.

Ceramic packaged diodes have also been produced.

Community
  • 1
Spehro Pefhany
  • 376,485
  • 21
  • 320
  • 842
  • 3
    One think to look out for with glass packaged diodes is their light sensitivity. – George Herold Jul 28 '14 at 13:38
  • @GeorgeHerold Is it measurable with, say, a 1N4148? I've heard that but never been bitten (yet). I guess now that I use BAVxx all the time it's unlikely to happen. – Spehro Pefhany Jul 28 '14 at 13:43
  • 3
    I've never had issues with 1n4148's. (I just stuck one under an incandescent bulb with DMM measuring current, I got ~30nA right at the resolution limit of my meter so a bit suspect, but I'm pretty sure it's real.) (I'm too lazy to pull out the big guns.) I've had the biggest issue with (20V) Zeners that I use in a noise source. I run those right near the knee, and light will just kill all the lovely noise spikes. There the bias currents are in the 1-10 uA range. I'm not sure why the zeners seem to be more sensitive. – George Herold Jul 28 '14 at 13:55
  • 1
    I second the realness of light sensitivity, though it is a very small current indeed and significant in only the most sensitive circuits. LEDs make decent photodiodes in a pinch, and I bet (though it's just a guess) that a Zener operating close to the knee shares some mechanism with an [avalanche photodiode](http://en.wikipedia.org/wiki/Avalanche_photodiode). – Phil Frost Jul 28 '14 at 18:48
  • 1
    @PhilFrost Reportedly a flash photo of an old-style EPROM-based microcontroller (the kind with a quartz window for UV erasure) could cause it to latch up and be destroyed if it was powered at the time. – Spehro Pefhany Jul 28 '14 at 18:54
  • 1
    The issue of flash photography and silicon components hasn't gone away completely: https://www.raspberrypi.org/blog/xenon-death-flash-a-free-physics-lesson/ . I guess the lesson here is don't use bare silicon dies without enclosures. – Jack B Sep 19 '16 at 17:12
  • 1
    @JackB There are a fair number of parts in that kind of tiny package these days. Worth remembering. – Spehro Pefhany Sep 19 '16 at 17:35
  • I wonder if it was possible to have those early cat's whiskers in a plastic package. – Sredni Vashtar Mar 18 '18 at 06:52
11

Thermal properties. The glass and the semiconductor expand and contract at the same rates. This is for reliability of signal diodes. The expansions or contraction at different rates would cause damage to the semiconductor.

Relevant paper from 1961

Rev
  • 10,017
  • 7
  • 40
  • 77
Enemy Of the State Machine
  • 1,740
  • 11
  • 10
  • 2
    CTE SiO2 = 5e-7, Si = 2.6e-6. In what sense of "Matching" are these the same? – placeholder Jul 28 '14 at 04:54
  • Sorry, some clarification and some history: http://electronicdesign.com/archive/high-temperature-glass-used-seal-diodes The glass used was of a molecular structure like pyrex. – Enemy Of the State Machine Jul 28 '14 at 04:59
  • That is from 2001, glass diodes were the first packaging type. So this link is doesn't support what you think it does. – placeholder Jul 28 '14 at 05:03
  • I must be misunderstanding your comment. I was responding to the thermal expansion difference of the glass used in the process. That is all I was trying to support. – Enemy Of the State Machine Jul 28 '14 at 05:07
  • The first packages were glass, that dates from the 1950's, finding a paper from the modern era regarding matching CTE in an aspect that is off topic (it's not about packaging, it's about passivation w/o packaging), a quote from that paper ".. of silicon-diode sheets has been achieved with a technique that may provide hermetically sealed semiconductor devices without the use of cans. " Note carefully the "may". – placeholder Jul 28 '14 at 05:15
  • 1
    The paper Is from October 25th 1961. The hypothesis did contain a "may" but the last paragraph states: No devices failed in five months. Diodes tested under 20-V reverse bias and similar environmental conditions did not show any changes after two months of testing, according to IBM researchers." Summing up the results. – Enemy Of the State Machine Jul 28 '14 at 05:21
  • While not as full an answer it brings up an interesting point, the linked article looks like it is more talking about a treatment of the semi-conductor die rather than the packaging. +1 for some interesting information relevant to the question. – Hugoagogo Jul 28 '14 at 08:55
  • I saw the date at the top, 2001. But again, this is about passivation NOT the packaging, which is what the OP was about. It's a very different thing. In fact this pyrex based SiO2 passivation was never really commercialized as the industry moved to Si34 (which is an incredibly tough and impermeable material) – placeholder Jul 28 '14 at 11:34
  • The coefficient of expansion for pure SiO2 may be below that of pure silicon, but most kinds of glass have a coefficient of expansion which is orders of magnitude larger than that of pure SiO2. What's important, though, is not the relative difference but the absolute difference. The COE of borosilicate glassware is low enough that it can survive rapid heating to temperatures above the working temperatures of semiconductor junctions even though the entire piece won't heat perfectly uniformly; if the COE of silicon is less than that there should be no trouble. – supercat Sep 19 '16 at 16:28
  • @supercat Isn't the matching of the thermal coefficient of expansion between the glass envelope and the metal leads of major importance? – Andrew Morton Sep 19 '16 at 18:28
  • @AndrewMorton: Many materials are slightly elastic; what matters is whether the difference in thermal expansion is to great to be absorbed by elastic deformation. – supercat Sep 19 '16 at 18:32
  • @supercat I was under the impression that, at least for small signal diodes, the junction is held in free space within the envelope. So the most significant thermal transfer (i.e. discounting thermal radiation) from the junction to ambient is via the leads and so the difference in CoE between the glass and the leads is more significant than that between the silicon and the glass. (I think the paper EOtSM quoted is a red herring because it about passivation by coating the diode rather than a hermetically-sealed envelope.) But I have no 1N4148 to peer at through a magnifiying glass. – Andrew Morton Sep 19 '16 at 19:05
  • @AndrewMorton: For small signal diodes, the material inside is indeed floating. My main purpose in posting was to respond to the comment saying that the COE of pure SiO2 was too small to make it thermally compatible with pure Si; I don't know what kinds of glass are thermally suitable for use in semiconductors, but don't think the fact that pure SiO2 has a lower COE than pure Si should pose any problem whatsoever, since most glass has a higher COE. – supercat Sep 19 '16 at 19:14