What datasheet and which figures are you looking at? I see 'difference' 12V vs 17V in only a few figure, mostly less than the width of the 'pen'. If that's a true difference, it may be level-related rather than supply related and thus working a few dB lower signal may be as good as teasing the supply limit.
Over-voltage is not just heat. In my youth I ran "28V" chips at 35V. They were cheap. They worked. For 6 years. Just quit. Replaced, worked, 6 years, quit. Replaced, and I retired (not my problem now). Research suggests electromigration-- metal actually moves over the surface of the chip under electrical stress.
Over-voltage is not just heat. In my youth I ran "28V" chips at 35V. They were cheap. They worked. For 6 years. Just quit. Replaced, worked, 6 years, quit. Replaced, and I retired (not my problem now). Research suggests electromigration-- metal actually moves over the surface of the chip under electrical stress.
If you've built a high-end mixing desk with 300+ opamps, you want each and every one of them to run reliably for decades, since you'll be sending out repair engineers each and every time one fails in the field, possibly 1/2 way round the world...
There is a lot of manufacturing spread with semiconductors, so that some devices will hold out well to abuse, others however won't - these latter are the ones defining the absolute-maximum limits, and whose behaviour you need to worry about in equipment in the field. In my hypothetical scenario you don't actually care if 99.7% of the devices are as tough as nails, its the 0.3% weakest devices that will cost you if you overdrive them. And if reliability is too low your reputation will suffer and put you out of business, even if the repair costs are manageable.
There is a lot of manufacturing spread with semiconductors, so that some devices will hold out well to abuse, others however won't - these latter are the ones defining the absolute-maximum limits, and whose behaviour you need to worry about in equipment in the field. In my hypothetical scenario you don't actually care if 99.7% of the devices are as tough as nails, its the 0.3% weakest devices that will cost you if you overdrive them. And if reliability is too low your reputation will suffer and put you out of business, even if the repair costs are manageable.
Last edited:
A lot of ICs are designed assuming a 10 year lifetime as a reference point. The maximum current density is generally calculated for critical regions to limit electromigration over the estimated lifetime of the product at the specified maximum operating temperature. The temperatures over the surface of the die varies depending on local power dissipation, which may increase at higher supply voltages. To consider specifics for a particular IC, one needs to review a particular layout and schematic. Some designers add more margin than others.
At higher voltages some of the transistor junctions will see a lower parasitic capacitance, due to a wider depletion region. There is a production distribution of junction breakdown for a process & some manufacturers guarantee to 6 sigma, but some may only use 3 sigma. Also, within a manufacturer, there may be products that have more design margin, depending on when they were designed and what standards were in place at the time. This is also true for companies that got absorbed by a larger company. Second source products (same part number, different internal design) from different manufacturers may be have different reliability margins.
Another detail, that is generally hidden from the consumer is the number of Fab plant transfers that have occurred for a specific product ID. These process/product transfers are often not “square beaker” identical process technologies, so the results may be better or worse than the original design in some areas of reliability and performance.
There are many ways, that a user or application of a particular product can be applied that will reduce the reliability. When there is a question about what is safe to do, always contact the specific manufacturer for guidance.
The number one cause of reduced product lifetime is high temperature operation. Probably one of the most problematic failures modes for finding the root cause is EOS (Electrical Over Stress) which also has ESD (Electro Static Discharge) failures as a subset of the EOS category. Automotive applications have one of the toughest reliability requirements with a target of Zero PPM (Parts Per Million) failures, but some products result in 1 -2 PPM failures.
I never want to run an IC or component near the specification limits, even if one sees a slight improvement in distortion or linearity. It is generally not worth the risk, unless one does not care about reliability. The oil industry has used semiconductors in the accelerometers of drill heads (for monitoring direction), with IC specification of 225C max operation & a lifetime of about 13 hours.
At higher voltages some of the transistor junctions will see a lower parasitic capacitance, due to a wider depletion region. There is a production distribution of junction breakdown for a process & some manufacturers guarantee to 6 sigma, but some may only use 3 sigma. Also, within a manufacturer, there may be products that have more design margin, depending on when they were designed and what standards were in place at the time. This is also true for companies that got absorbed by a larger company. Second source products (same part number, different internal design) from different manufacturers may be have different reliability margins.
Another detail, that is generally hidden from the consumer is the number of Fab plant transfers that have occurred for a specific product ID. These process/product transfers are often not “square beaker” identical process technologies, so the results may be better or worse than the original design in some areas of reliability and performance.
There are many ways, that a user or application of a particular product can be applied that will reduce the reliability. When there is a question about what is safe to do, always contact the specific manufacturer for guidance.
The number one cause of reduced product lifetime is high temperature operation. Probably one of the most problematic failures modes for finding the root cause is EOS (Electrical Over Stress) which also has ESD (Electro Static Discharge) failures as a subset of the EOS category. Automotive applications have one of the toughest reliability requirements with a target of Zero PPM (Parts Per Million) failures, but some products result in 1 -2 PPM failures.
I never want to run an IC or component near the specification limits, even if one sees a slight improvement in distortion or linearity. It is generally not worth the risk, unless one does not care about reliability. The oil industry has used semiconductors in the accelerometers of drill heads (for monitoring direction), with IC specification of 225C max operation & a lifetime of about 13 hours.
I'd add that thermal cycling, rather than just higher temperature, is a major reliability issue - thermal expansions/contraction of epoxy packages is different from the chip, the lead-frame and the bond wires, so that mechanical failure can happen when the temperature cycling is frequent and of large magnitude.
This is one reason why ceramic or metal packaging is typically used for MILSPEC devices where the temperature range is higher and reliability is more of an issue.
This is one reason why ceramic or metal packaging is typically used for MILSPEC devices where the temperature range is higher and reliability is more of an issue.
I would guess because they don't do a whole lot with the theta JC/JA of normal packages, and the thermally enhanced packages are using the underside anyway. I've seen the occasional D2PAK board mounted heatsinks in use for LED drivers and similar.
Yes, you can have delamination of the plastic package during temperature cycling. Also, if there is copper bond wires in the package, any chlorine present (10ppm), a little humidity & temperature cycling, there can be corrosion with resultant open bond wires.
Typically package qualification requires temperature cycle and humidity with bias voltage.
TI Reliability: Reliability testing | Reliability | TI.com
On-Semi: Quality and Reliability Handbook
ADI Reliability Handbook: https://www.analog.com/media/en/technical-documentation/user-guides/UG-311.pdf
Typically package qualification requires temperature cycle and humidity with bias voltage.
TI Reliability: Reliability testing | Reliability | TI.com
On-Semi: Quality and Reliability Handbook
ADI Reliability Handbook: https://www.analog.com/media/en/technical-documentation/user-guides/UG-311.pdf
Years ago, even basic ICs always had theta JC/JA listed on the datasheets. If the power dissipation of the IC is low, it is often not an issue.
The OPA1656 has theta JC/JA listed on the datasheet.
Attachments
Last edited:
Right, I didn't mean to imply they were missing from datasheets.
I meant to reply to Mark as to why I thought you may not see many standard op-amps with heatsinks on them - since the thermal resistances of standard SOIC packages aren't all that great.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.