hmmm...I personally would not touch the remaining 40+ transistors that are still working.it makes no sense...unless their failures would cause you to need to replace more transistors...
The only slow deterioration effect I know about in bipolar transistors is an increase in non-ideal base current, causing hFE reduction at low current densities and increased noise, when the emitter-base voltage is anywhere near the point where avalanche breakdown starts. Like Steven said, this can be cured by heating up the transistors to a high temperature (few hundreds of degrees C) for the right amount of time (provided the package doesn't melt).
MOSFET's also have a slow deterioration effect, known as hot carrier degradation. It causes a slow change in threshold voltage and transconductance. For symmetrical MOSFET's, you usually get the fastest hot carrier degradation when the drain-source voltage is near the maximum rating and the gate source voltage is about half way. I don't know what gate-source voltage is worst for typical power MOSFET's.
The failure rates for most catastrophic failure mechanisms increase exponentially with increasing temperature (Arrhenius' law), so I doubt if keeping equipment switched on continuously is good for reliability. Keeping it switched off continuously is much better and has the added advantage that no-one will notice any failures.
MOSFET's also have a slow deterioration effect, known as hot carrier degradation. It causes a slow change in threshold voltage and transconductance. For symmetrical MOSFET's, you usually get the fastest hot carrier degradation when the drain-source voltage is near the maximum rating and the gate source voltage is about half way. I don't know what gate-source voltage is worst for typical power MOSFET's.
The failure rates for most catastrophic failure mechanisms increase exponentially with increasing temperature (Arrhenius' law), so I doubt if keeping equipment switched on continuously is good for reliability. Keeping it switched off continuously is much better and has the added advantage that no-one will notice any failures.
Holy mackeral..
Military packages....
All are hermetic, meaning they do not allow molecules to pass from outside to inside. This is measured by introducing helium during the sealing operation, then measuring the rate of helium expulsion within a vacuum..
The hermeticity of a package does not directly affect the life of a semi device..the caveat with that has to do with ionic contamination..hermetic packages prevent that failure mechanism by preventing the contamination of the semi junctions by potassium, sodium, chlorine, etc...
A typical semi has a lifetime that is based on the temperature at which the device operates..from what I recall, the lifetime doubles for every 10 degrees celsius change in operating temperature..don't hold me to that, I've had a glass of wine, and cannot be held responsible for my actions...
All power transistors are typically held down with a solder or braze...some lower power devices are epoxied down with silver filled epoxy, but they do not have the heat transfer capabilities of the metal die attach technologies..
Pb/In...150 C to 295 C...
Pb/Sn...183 C to 395 C..
Au/Sn...280 C...
Au/Ge...356 C...
Gold tin and gold germanium are hard solder attaches..not susceptible to the power cycling issue...
Wire bonds are aluminum wires ultrasonically welded to the top aluminum metal on the chip..they are not typically an issue..only if they are bonded to gold will purple plague be an issue..but that requires the setup alloy, something that is well known and avoided..
For plastic devices, the main issue is the rate at which ionic contamination can reach the semi surface..most of the current manu's have very good track records in that regard..Yah, occasionally a hiccup occurs, but otherwise, all is good (in oz)...
Disregarding quality issues, the main concern is the temperature the device is at...the higher the temp, the shorter the lifetime..
To date, the only life reliability issue I've had is the torque left in the 6-32 screws holding to-3 devices to the heat sink..
IMHO..the output die temps is the only critical parameter..everything else is just fluff..
Cheers, John
Military packages....
All are hermetic, meaning they do not allow molecules to pass from outside to inside. This is measured by introducing helium during the sealing operation, then measuring the rate of helium expulsion within a vacuum..
The hermeticity of a package does not directly affect the life of a semi device..the caveat with that has to do with ionic contamination..hermetic packages prevent that failure mechanism by preventing the contamination of the semi junctions by potassium, sodium, chlorine, etc...
A typical semi has a lifetime that is based on the temperature at which the device operates..from what I recall, the lifetime doubles for every 10 degrees celsius change in operating temperature..don't hold me to that, I've had a glass of wine, and cannot be held responsible for my actions...
All power transistors are typically held down with a solder or braze...some lower power devices are epoxied down with silver filled epoxy, but they do not have the heat transfer capabilities of the metal die attach technologies..
Pb/In...150 C to 295 C...
Pb/Sn...183 C to 395 C..
Au/Sn...280 C...
Au/Ge...356 C...
Gold tin and gold germanium are hard solder attaches..not susceptible to the power cycling issue...
Wire bonds are aluminum wires ultrasonically welded to the top aluminum metal on the chip..they are not typically an issue..only if they are bonded to gold will purple plague be an issue..but that requires the setup alloy, something that is well known and avoided..
For plastic devices, the main issue is the rate at which ionic contamination can reach the semi surface..most of the current manu's have very good track records in that regard..Yah, occasionally a hiccup occurs, but otherwise, all is good (in oz)...
Disregarding quality issues, the main concern is the temperature the device is at...the higher the temp, the shorter the lifetime..
To date, the only life reliability issue I've had is the torque left in the 6-32 screws holding to-3 devices to the heat sink..
IMHO..the output die temps is the only critical parameter..everything else is just fluff..
Cheers, John
Sanken has some info on the first pages in their power transistor catalogue (can be downloaded from thir site) about ageing and expected life of semiconductors.
You might find it usefull...
You might find it usefull...
Do you mean permanenly, or while still hot and that cooling restores Hfe ?.Richard C said:
Eric.
Follow up #2 on "Does output transistors ages?"
Some comments to yours:
RichardC: Please, share your experience on reduced hfe by running transistors hot.
Joachim_B: Sanken is just talking about the bath tub curve not the ageing question. It’s like when you are 50 you can’t run as fast as when you where 20, but you live until you are 75. Does the same apply to transistors?
JohnFerrier: Your thoughts are interesting: Maybe Motorola makes MJ15024/25 with better specifications now than for 12 years ago. I will try to ask Motorola about that. And maybe we have better output transistors on the market now that are compatible and would be a better option if I should replace all 48.
Thanks,
Ulf
Some comments to yours:
RichardC: Please, share your experience on reduced hfe by running transistors hot.
Joachim_B: Sanken is just talking about the bath tub curve not the ageing question. It’s like when you are 50 you can’t run as fast as when you where 20, but you live until you are 75. Does the same apply to transistors?
JohnFerrier: Your thoughts are interesting: Maybe Motorola makes MJ15024/25 with better specifications now than for 12 years ago. I will try to ask Motorola about that. And maybe we have better output transistors on the market now that are compatible and would be a better option if I should replace all 48.
Thanks,
Ulf
I hate to throw everyone off, BUT if you run a transistor very hot, it will wear out in a relatively short time. It is most probably migration of the metalization used to make the connection to the outside. I have experienced this, in a real design.
john curl said:
From what I recall, the newer generations of CPU's have that problem because of current density, but I'm not sure about power chips..The metalization is incredibly thick in comparison, and the current density is much lower..But I do recall some diffusion things happening at the silicon to aluminum interface of the emitter tubs or fingers..The wirebond welds to the silicon aluminum also takes a beating.
Off topic: Haven't seen a post from you recently, was concerned about you w/r to the so. cal wildfires..I assume all is well?
Cheers, John
Hi
I remember some old data of lifetime of a 2N3055 once.
When run with max junction temp the lifetime has been specified
as 4000h but every 10deg of lowering the temp would about double
that.
This data was for metalcan 2N3055 from the 70th but I cant remember the manufaturer
I remember some old data of lifetime of a 2N3055 once.
When run with max junction temp the lifetime has been specified
as 4000h but every 10deg of lowering the temp would about double
that.
This data was for metalcan 2N3055 from the 70th but I cant remember the manufaturer
Regarding electromigration, most integrated circuit manufacturers aim for 10 year life time (assuming continuous use) at the maximum recommended junction temperature (recommended operating condition, not absolute maximum rating). It gets better at lower temperatures. I don't know about the usual life time goal for discrete power devices.
MarcelvdG said:
Electromigration is pretty much no longer an issue since the early 90's, when complex metallisation systems were developed and put into production. The power device failures today are mostly due to "hot spots" around crystalline defects (in bipolar devices/ICs) and gate oxide failures (in MOS devices/ICs). Power devices most encountered failure mechanism today is due to thermal/mechanical stresses.
Assuming they are used within the datasheet specs, today's power devices should have a life expectancy of 25-50 years. Of course, the bathtub failure frequency law still applies.
I'm no expert on power devices, either discrete or integrated, but for normal analogue IC's ten years has been the usual electromigration life time goal since I started working as an analogue integrated circuit designer in the mid nineties. Of course that is at maximum recommended junction temperature and continuous use, so it can easily become fifty years when the temperature is lower and/or the current doesn't flow continuously.
MarcelvdG said:
Electromigration is typical for Al or AlSi based metallisation systems. Mode advanced schemas are completely immune; a typical example (and a very old one) is TiNiAg. Titanium for adherence to silicon/silicon dioxide, nickel as an electromigration barrier and silver for conductivity.
Ulf Eliasson said:
Output devices can be weakened by handling (static discharge).
This can cause them to fail prematurely.
A colleague of mine worked for a semiconductor company and they did research into this.
Posted by syn08:
"Electromigration is typical for Al or AlSi based metallisation systems. Mode advanced schemas are completely immune; a typical example (and a very old one) is TiNiAg. Titanium for adherence to silicon/silicon dioxide, nickel as an electromigration barrier and silver for conductivity."
That probably explains it. All processes I've ever worked in used AlSi wiring or (the last couple of years) copper wires. Copper is much better in current handling at low temperatures, but degrades faster when temperatures get high.
Is TiNiAg what is normally used for power semiconductor processes? Or is this only used for fancy high-reliability stuff?
"Electromigration is typical for Al or AlSi based metallisation systems. Mode advanced schemas are completely immune; a typical example (and a very old one) is TiNiAg. Titanium for adherence to silicon/silicon dioxide, nickel as an electromigration barrier and silver for conductivity."
That probably explains it. All processes I've ever worked in used AlSi wiring or (the last couple of years) copper wires. Copper is much better in current handling at low temperatures, but degrades faster when temperatures get high.
Is TiNiAg what is normally used for power semiconductor processes? Or is this only used for fancy high-reliability stuff?
MarcelvdG said:
Used to, 20 years ago when I was in the business 😀. Today I don't know, but I have no reason to believe it's anywhere worse than the old TiNiAg.
- Status
- Not open for further replies.