I have a question for you. Can you give a simple description of the differences between commercial, industrial, and military temperature grades of the same chip? Are they the same silicon, just with looser specs for the wider temperature ranges? Are they the same silicon, but tested to prove that they meet specs over the wider temperature ranges? Or, is the chip design/manufacturing process different for the different grades? (Almost never! See below. /rap) If so, what are the differences, in nonspecialist language?
For openers, not all ICs are the same. Some get different kinds of testing. Some ICs have good yields to MIL grades, while some have poor yields. We try to not let that happen, but sometimes despite our best plans, it just happens. Sometimes the demand for MIL-grade parts or A-grade parts just goes sky-high. An LM101A, an LM201A, and an LM301A are all the same chip. In almost all IC examples, the MIL grade and the commercial grade are all the same chip. (But the LM3876 is a completely different circuit than an LM2876, which also is a completely different circuit from an LM1876, so beware of these cases!)
In most cases, a MIL-grade chip (1xx-grade) is a selected part, with low errors at room temperature. So it starts out as an A-grade part and then gets extra testing to make sure it's well-behaved at high and low temperatures. This usually involves 100% testing at hot and cold (125°C and −55°C) to show that the part works well at extremes.
An LM2xx that's rated for the industrial range may be the same kind of selected best-grade part, but it may or may not get tested at 85°C or −25°C. It won't need that if it gets high yield. We can usually correlate the performance at 85°C or −25°C with some test results at room temperature. Most people love that because it helps to drive down the manufacturing costs and end-user prices.
In the old days, we brought out 100, 200, and 300 grades for many, but not all, parts. Recently, most of the older 200s were discontinued, as there wasn't a large market demand. These were mostly LM2xxHs in a hermetic package. In the last couple of years, many ICs have been brought out with just the industrial grade because that's what a lot of people want. These almost entirely come in plastic packaging, which makes quite a difference!
If you buy a 300-grade part, is it a part that flunked an LM1xx test at room temperature? Maybe, but maybe not. Is it a part that flunked an LM1xx test, hot or cold? Possible, but not likely. If we screen for good parts and sell 1000 parts a week of MIL parts, and reject 600 parts, hot and cold, these may be rejects for some very tight specs. Plus, these 600 "rejects" might be salted in with 20,000 good LM3xxs, on a typical week. So the number of LM1xx rejects in a batch of LM3xxs might be pretty small.
What are the chances of getting an LM3xx that actually meets LM1xx specs? In many cases, they're quite high. Maybe 30% or 60% of these 3xxs might meet the 1xx specs—or in some cases, 3%; maybe 90%; maybe 0%. I recall when I designed the LM333, the 3-A version of the LM137, that we selected out the better grades at room temperature—these were basically the LM333A grade parts—and sent 200 pieces through the hot and cold tests. When we got back the test results, the yield was 99%. The two parts that didn't pass might have been due to handler or contactor failures. This 99% yield isn't true for all parts.
Now you might ask, "What's the chance of an LM3xx working really badly at hot or cold temperatures?" It turns out that in many situations, the chances are less than 1%. An LM3xx might go slightly out of spec, but if its usage is in a not too critical circuit, then you may never see improper operation. For example, an LM324 might have 5 mV of VOS at room temperature, and 7 mV at 125°C. But that's not going to cause any significant failure in most applications. On the other hand, the LM368H-2.5 was released to production, even though we couldn't make an LM168H-2.5; the part would oscillate at some cold temperatures, but we let it go anyway. We decided that it wasn't worth the effort to cure the oscillation. You certainly wouldn't want to use that over a wide range! There may be some parts with big problems like that, although I don't know about very many, and that's okay with me. I don't have to be an expert.
Of course, some people want to use an LM3xx in cost-sensitive industrial systems, like in a traffic-light controller where the temperatures are quite hot and cold. We caution them that they can't get mad at us if some LM3xxs don't work perfectly, hot or cold. They should comprehensively evaluate at least their first 20 systems at all temperatures, sweeping from hot to cold. If the systems still work, and they keep working at even hotter and colder temperatures than the needed range, then that's pretty encouraging. Then we caution them to test at least a sample of their systems—5%, 10%, or 20%—at good hot and cold temperatures, to make sure nothing bad is getting in. In many cases, this probably works well—so long as you aren't putting in LM368H-2.5s!!!
I can't speak for every other IC manufacturer. Many may make parts that are just fine, hot and cold. I suspect that others make parts that are really bad outside of their rated range, but I don't know much about that. I would guess that many other IC makers do tests that are comparable to what we do, and if you ask them, they ought to tell you. I hope that's a helpful answer to your question.—RAP
All for now. / Comments invited!
RAP / Robert A. Pease / Engineer
Mail Stop D2597A
P.O. Box 58090
Santa Clara, CA 95052-8090