what, *in practice*, does the temperature range mean in case of an MCU (PIC16F1709 here)? There are two variants, -40C..85C and -40C..125C. I understand that formally at 86C the first part is beyond its specs and the manufacturer does not guarantee its correct operation, but what actually makes the difference in case of a silicon chip? It would be obvious in case of e.g. an electrolytic capacitor, but an MCU? The package material is different?
Price >:-} (Although, sometimes it _is_ related to packaging. Ceramic packages can go higher in temperature than plastic.) ...Jim Thompson
--
| James E.Thompson | mens |
| Analog Innovations | et |
| Analog/Mixed-Signal ASIC's and Discrete Systems | manus |
| San Tan Valley, AZ 85142 Skype: Contacts Only | |
| Voice:(480)460-2350 Fax: Available upon request | Brass Rat |
| E-mail Icon at http://www.analog-innovations.com | 1962 |
I love to cook with wine. Sometimes I even put it in the food.
*** Not necessarily. It is beyond it's temp spec but may be within the other parametric specs out to 124C. It just isn't tested at the intermediate temps.
*** One mechanism is that the leakage current of the FETs is exponential with temperature. IRRC approx doubles with each 10C rise. Art
As temperature goes up chip speed goes down, leakage increases, and perhaps thresholds change.
So it's mostly tested in. I don't think the packages will differ much, unless you're getting _really_ high temperature devices (like 200C or some such).
Sometimes you can get higher or lower temperature operation out of a part by running it slower than its rated speed -- that both reduces dissipation and increases (you hope) timing margins.
--
Tim Wescott
Control system and signal processing consulting
www.wescottdesign.com
The datasheet temperature is ambient. The chip needs to maintain whatever limit the manufacturer specs to be OK as a junction temperature, say 150 deg C. So the theta JA, which is a function of the package, will determine the operating temperature.
The die to the best of my knowledge is always the same. I've never seen otherwise, but you can't know every chip ever made.
The only time I have seen the die to be physically different is when you do rad hard chips.
The oil exploration equipment manufacturers violate the upper temperature all the time. Electronics often "work" well at elevated temperatures, though the lifetime spec will go out the window. Cold is another story. You would think cold isn't an issue, but with MOS, sometimes the threshold shift will effect the biasing. Usually your worse case for bias will include the "weak" file at the lowest spec temp.
At >=180C plastic seems more reliable, certainly as regards solder joints. One credible theory is that the plastic can deform slightly to take up some of the differential expansion between it and the PCB, whereas ceramic won't.
As for PICs, in my experience the temperature code printed on the package doesn't matter. Most parts of most PICS will operate at over
Normal plastic packages break down at 175C. They survive a short time at that temperature (otherwise soldering would be a challenge!), but not long - you will not find any parts qualified at over 175C in plastic packages. Ceramic packages will take you to about 225C, but you can't go much higher with normal silicon (the thermal noise gets too close to the bandgap) and you need to use higher temperature semiconductors like SiC or GaN.
Many other effects cause trouble at higher temperatures. Electromigration of the different metals in the bonding (such as Al die pads, Au bond wires, and Cu legs with Ni plating) increases greatly with temperature and lowers lifetimes. Leakage in the transistors increases
- eventually the leakage will be too high for the chip to work properly. Thermal expansion is obviously going to be an issue, especially with mismatches between the device and the board. Tolerances of many sorts change, as do resistances and capacitances. Different types of solders and pastes, as well as glues, plastics, filler materials, etc., begin to weaken or fail at higher temperatures (especially with long-term exposure to temperatures, or rapid temperature cycling).
Absolutely - a lot of silicon can run fine at higher temperatures but lower speeds, and with poorer tolerances. For many chips, the wafers get spot-tested at different temperatures and classified as "slow, but higher temperature" or "fast but lower temperature".
Don't tell the oilfield instrumentation guys, they'd be very upset to read that their equipment doesn't work!
If a chip doesn't work at 200C - and most won't - it's almost certainly not because of its plastic package.
There are plenty of plastic parts qualified at over 175C. It's true you won't generally find them because the equipment manufacturers do their own qualification at great expense and don't tend to share the results.
Most of it doesn't have to work very long. Downhole instruments lead short and difficult lives.
Cheers
Phil Hobbs
--
Dr Philip C D Hobbs
Principal Consultant
ElectroOptical Innovations LLC
Optics, Electro-optics, Photonics, Analog Electronics
160 North State Road #203
Briarcliff Manor NY 10510
hobbs at electrooptical dot net
http://electrooptical.net
I know a bit about this branch, even though we have not (yet) made 200C qualified boards. When I say you will not find plastic parts qualified for over 175C, I mean you won't find any manufacturer who will sell you such parts. People /do/ use them, but with their own qualifications and testing.
There are also different types of plastic packages - most will have short lifetimes at 200C, but some work longer. And of course you are right that most chips won't work at 200C regardless of the packaging.
But if you look at high temperature devices (TI has quite a few), you'll find several devices that are rated to 175C in plastic packages and the same die rated at 225C in ceramic or bare die. This is not just for fun
- TI would love to sell the chip rated at 200C or 225C in plastic, because it would be cheaper for them.
Of course, 175C is not a hard limit - the plastic does not melt at that point. But that's the limit that manufacturers are comfortable with as a general rule. If a board manufacturer wants to test them to 200C at a short lifetime, that's fine - many high temperature boards are only rated for a lifetime of a few hundred hours.
Yes, that's usually true, praise His Noodly Appendage. When I've tested parts, typically 2000 hours at 180C continuous is enough to make them viable - in use they don't tend to see that temperature for more than a few hours at a time so will last much, much longer. However, some parts just keep giving, bless them.
But still, I've never found the plastic to be the limiting factor.
As I pointed out in my post, rad hard is a special process. Well at least where I have done it.
High temperature can be done with degraded specs. Biasing problems are more likely to occur extending the low temperature spec, at least in MOS. High temp is just a speed and noise issue.
I was told the more expensive part is the same, but the extra cost is to pay the insurance policy against liability claims. Sounds a bit cynical, maybe the sales guy was pulling my leg...
Your leg was pulled. If a part fails, you are given a free replacement. That is the limit of semiconductor reliability.
Now I don't know what is done liability wise on rad hard satellite grade parts.
Now more extensive testing does command a higher price. But this is a real cost. Cold testing is an especially big pain in the ass. The parts jam in the handler. Throughput is now no longer test time related, but you need to insure the parts have adequate soak time.
Some tests can't be done on ATE. These are often in the typical column on a datasheet. But if you want the value tested, it involves bench testing. This is done on a test rack, but some engineer needs to write the test program and the parts are inserted by hand, especially if low level current is to be tested.
While the 74xx series was originally offered at a quite limited temperature range, some manufacturers also offered 54xx chips for -55 to +125 C military temperature range. Apparently there was also some manufacturers making 64xx -40 to +85 C products.
Does the 54xx product family really exist today or are all those types some variants of 74xx products ?
ElectronDepot website is not affiliated with any of the manufacturers or service providers discussed here.
All logos and trade names are the property of their respective owners.