Anyone with experience/knowledge of if output transistors gets old and slow and needs to be replaced to make the amplifier sound its best?
(I have a pair of JR M7 with MJ15022/23 (=MJ15024/25) built 1990.)
Thanks,
Ulf
(I have a pair of JR M7 with MJ15022/23 (=MJ15024/25) built 1990.)
Thanks,
Ulf
Getting old?
I don't believe this happens if the devices work under normal operating conditions and temperature.
However excessive heat will kill it eventually.
But performance figures sliding down like tubes is probably not possible. Devices could go bad however due to device ( manufacturing ) imperfections . You cannot predict these either. Thankfully this does not happen often.
Cheers.
I don't believe this happens if the devices work under normal operating conditions and temperature.
However excessive heat will kill it eventually.
But performance figures sliding down like tubes is probably not possible. Devices could go bad however due to device ( manufacturing ) imperfections . You cannot predict these either. Thankfully this does not happen often.
Cheers.
The bathtub curve
The reliability textbooks talk about a bathtub curve, essentially a wide "U" with sloped sides, to describe the lifetime of a semiconductor device. The failure rate is high for the first 1-2 years, called infant mortality rate, due to manufacturing defects like thin metalization or contaniments. Then a properly used semiconductor device has a very low failure rate for about 20 years. After about 20 years of use, the failure rate increases from wear out mechanisms. Power-up and power-down cycles accelerate wear-out. Check out the MIL 217 handbook on the web.
A semiconductor that is power-cycled daily should aveage 20 years of life. A semiconductor with little active use, and few power-up power-down cycles, should average 40 years of life.
The reliability textbooks talk about a bathtub curve, essentially a wide "U" with sloped sides, to describe the lifetime of a semiconductor device. The failure rate is high for the first 1-2 years, called infant mortality rate, due to manufacturing defects like thin metalization or contaniments. Then a properly used semiconductor device has a very low failure rate for about 20 years. After about 20 years of use, the failure rate increases from wear out mechanisms. Power-up and power-down cycles accelerate wear-out. Check out the MIL 217 handbook on the web.
A semiconductor that is power-cycled daily should aveage 20 years of life. A semiconductor with little active use, and few power-up power-down cycles, should average 40 years of life.
Stress.
Yes , stress is also a major contributor. This is therefore one good reason to leave systems on all the time ! This was brought up in another thread on this forum. Check it out.
Cheers.
Yes , stress is also a major contributor. This is therefore one good reason to leave systems on all the time ! This was brought up in another thread on this forum. Check it out.
Cheers.
As I see it, is that everything that heats and expands, then retracts when cooling down, moves and therefore ages....
But, then will it outlive most us before changing the sonics of an amplifier, I dont know.
But, then will it outlive most us before changing the sonics of an amplifier, I dont know.
Re: The bathtub curve
I hate TO-3's.
I think I read somewhere that a metal can transistor is better than a plastic one in this regard.LineSource said:
I hate TO-3's.
Re: Re: The bathtub curve
Very likely. In metal cans the die is just glued or soldered on the base and the bonding wires just have a nice bend from pin to die. For the rest the inside is empty.
The bonding wires in plastic transistors allways have some stress caused by pressure of the epoxy that is injected for the housing. Normally there is some gel over the die to prevent the epoxy to touch the crystal, otherwise the bonding wires could break because of thermal stress, but you cannot be sure that the gel covers the whole die. Therefore in military applications only metal cans are used, or ceramic housings (as far as I know). Ceramic housings are built from two layers (top/bottom) with some cement keeping them apart; also no stress from epoxy on the bonding wires.
Actually this is also an issue for opamps, not because of heat, but pressure on the die may cause diffences between transistors and matching of differential pairs will be worse. Also here a gel is put on the die and the connection point of the bonding wires before plastic injection.
Another issue might be moisture/water getting into the housing. Epoxy is not completely impenetrable for moisture, unlike metal cases and ceramic cases.
Steven
Circlotron said:
Very likely. In metal cans the die is just glued or soldered on the base and the bonding wires just have a nice bend from pin to die. For the rest the inside is empty.
The bonding wires in plastic transistors allways have some stress caused by pressure of the epoxy that is injected for the housing. Normally there is some gel over the die to prevent the epoxy to touch the crystal, otherwise the bonding wires could break because of thermal stress, but you cannot be sure that the gel covers the whole die. Therefore in military applications only metal cans are used, or ceramic housings (as far as I know). Ceramic housings are built from two layers (top/bottom) with some cement keeping them apart; also no stress from epoxy on the bonding wires.
Actually this is also an issue for opamps, not because of heat, but pressure on the die may cause diffences between transistors and matching of differential pairs will be worse. Also here a gel is put on the die and the connection point of the bonding wires before plastic injection.
Another issue might be moisture/water getting into the housing. Epoxy is not completely impenetrable for moisture, unlike metal cases and ceramic cases.
Steven
If the base emitter junction of a transistor has been zenering, for instance because the (dis)charging of electrolytics move the base and emitter in opposite directions during switch on/off, the noise of the transistor may get higher because of some slight damage of the crystal.Richard C said:
This can be 'repaired' by running the transistor hot for a moment.
I'm not sure whether it will be as good as before. Anyway, but that's besides the topic, it can be a good idea to use a diode across the base emitter junction to prevent the zenering of the transistor in low-noise applications.
Steven
Follow up on "Does output transistors ages?"
Thanks, for your assistance.
The reason for my question is that after repairing my two Jeff Rowland M7, by replacing two pairs of output transistors in each of my mono blocks (the faulty ones were automatically disconnected when the bas and emitter resistors burned up preventing them from having an active negative impact), my M7 sounded better and livelier than I can remember. As some of you may know the M7 have a pair of 12 parallel output transistors. My belief was that replacing two of 12 could have some importance when playing very loud but shouldn’t have the significant impact I experienced if the remaining transistors perform to 100% on all their parameters. The sound is now more lively, heavier and transparent on all levels.
What troubles my mind now is: Having replaced eight faulty output transistors of totally 48 (=2*2*12), will replacing the 40 left (that still works) have any positive impact on the sound at all. An MJ15024/25 is about $6 each. To waist $240 plus the repair work to get the knowledge that “output transistors always perform their maximum until they suddenly one day are totally dead” is nothing that appeal to me if I can find the knowledge about it easier.
If I summarize your view on the subject: You are in favor of the sudden death theory, but Richard C have some experiences that would be interesting to hear more about.
This topic is something the silicon industry or research institutes should have an answer to. Anyone knowing who to ask at e.g. Motorola that makes the MJ15024/25?
Thanks for your time,
Ulf
Thanks, for your assistance.
The reason for my question is that after repairing my two Jeff Rowland M7, by replacing two pairs of output transistors in each of my mono blocks (the faulty ones were automatically disconnected when the bas and emitter resistors burned up preventing them from having an active negative impact), my M7 sounded better and livelier than I can remember. As some of you may know the M7 have a pair of 12 parallel output transistors. My belief was that replacing two of 12 could have some importance when playing very loud but shouldn’t have the significant impact I experienced if the remaining transistors perform to 100% on all their parameters. The sound is now more lively, heavier and transparent on all levels.
What troubles my mind now is: Having replaced eight faulty output transistors of totally 48 (=2*2*12), will replacing the 40 left (that still works) have any positive impact on the sound at all. An MJ15024/25 is about $6 each. To waist $240 plus the repair work to get the knowledge that “output transistors always perform their maximum until they suddenly one day are totally dead” is nothing that appeal to me if I can find the knowledge about it easier.
If I summarize your view on the subject: You are in favor of the sudden death theory, but Richard C have some experiences that would be interesting to hear more about.
This topic is something the silicon industry or research institutes should have an answer to. Anyone knowing who to ask at e.g. Motorola that makes the MJ15024/25?
Thanks for your time,
Ulf
>I think I read somewhere that a metal can transistor is better than a plastic one in this regard.
Plastic IC's degrade (as opposed to metal) because of a less perfect seal. Air gets in and contaminates the innards.
Plastic IC's degrade (as opposed to metal) because of a less perfect seal. Air gets in and contaminates the innards.
This _was_ true but nowadays the chip is covered with glass => not so sensitive against bad environment.hitsware said:
Re: Follow up on "Does output transistors ages?"
Interesting question. However, maybe the new parts perform better (have better actual specs) than the original parts.
If you think it sounds lively enough, leave as is. If needs to be more lively, add a few more new ones. Find a good balance in sound (and cost). The old parts may contribute to other aspects of the sound you like.
JF
Ulf Eliasson said:
Interesting question. However, maybe the new parts perform better (have better actual specs) than the original parts.
If you think it sounds lively enough, leave as is. If needs to be more lively, add a few more new ones. Find a good balance in sound (and cost). The old parts may contribute to other aspects of the sound you like.
JF
" Therefore in military applications only metal cans are used, or ceramic housings (as far as I know). "
A few years ago I had some dealings with a person with the AirForce who earlier in her career was responsible foe EMP hardening. Her specialty was ceramics. The military requirements for semicoductor packaging result largely from her findings and protection from EMP was one of the highest priority considerations.
Therefore, if you select your gaineclone IC package to be ceramic it will still function after a nuclear detonation - provided it wasn't vaporized and you are still around to hear it.
A few years ago I had some dealings with a person with the AirForce who earlier in her career was responsible foe EMP hardening. Her specialty was ceramics. The military requirements for semicoductor packaging result largely from her findings and protection from EMP was one of the highest priority considerations.
Therefore, if you select your gaineclone IC package to be ceramic it will still function after a nuclear detonation - provided it wasn't vaporized and you are still around to hear it.
In some papers of Pass Labs I was read something like " ... when universe will be cold, this transistors will be still working..." - Mr. Pass is poet and have special supplyer😎
Life time of the power transistors is significantly reduced when operated mostly near maximum allowed temperature.
parameter matching more important than wear-out
" What troubles my mind now is: Having replaced eight faulty output transistors of totally 48 (=2*2*12), will replacing the 40 left (that still works) have any positive impact on the sound at all. An MJ15024/25 is about $6 each. To waist $240 plus the repair work to get the knowledge that “output transistors always perform their maximum until they suddenly one day are totally dead” is nothing that appeal to me if I can find the knowledge about it easier. "
Most hi-end amplifiers use matched transistors, or at least transistors from the same manufacturing lot. This gives a tight range of gain(hfe), input capacitance, and Vce, with the goal of getting the lowest distortion and best sharing of the output load. If one transistor has significantly higher gain than the others, it can increase the distortion and also be more prone to failure due to overheatting from shouldering a high percentage of the load, which in the worst case gives possible thermal run-away.
The small resistor in series with each transistor emitter helps to equalize the current delivered by each tansistor and helps to reduce the need for extremely tight matching.
Comparing the electrical characteristics of one of the new transistors with one of the old transistors would be one way to guage whether the march of technology over the years has created an almost entirely different set of parameters. The difference would have to be significant ... well over 25% hfe, before it would concern most designers.
Personally, in an amp with 48 output transistors, I doubt I could hear the difference between matched and un-matched output transistors, and I would not worry about current hogging by a few high gain devices affecting reliability.
" What troubles my mind now is: Having replaced eight faulty output transistors of totally 48 (=2*2*12), will replacing the 40 left (that still works) have any positive impact on the sound at all. An MJ15024/25 is about $6 each. To waist $240 plus the repair work to get the knowledge that “output transistors always perform their maximum until they suddenly one day are totally dead” is nothing that appeal to me if I can find the knowledge about it easier. "
Most hi-end amplifiers use matched transistors, or at least transistors from the same manufacturing lot. This gives a tight range of gain(hfe), input capacitance, and Vce, with the goal of getting the lowest distortion and best sharing of the output load. If one transistor has significantly higher gain than the others, it can increase the distortion and also be more prone to failure due to overheatting from shouldering a high percentage of the load, which in the worst case gives possible thermal run-away.
The small resistor in series with each transistor emitter helps to equalize the current delivered by each tansistor and helps to reduce the need for extremely tight matching.
Comparing the electrical characteristics of one of the new transistors with one of the old transistors would be one way to guage whether the march of technology over the years has created an almost entirely different set of parameters. The difference would have to be significant ... well over 25% hfe, before it would concern most designers.
Personally, in an amp with 48 output transistors, I doubt I could hear the difference between matched and un-matched output transistors, and I would not worry about current hogging by a few high gain devices affecting reliability.
- Status
- Not open for further replies.