Author Topic: Why do companies obsolete so many good transistors without substitutes?  (Read 3375 times)

0 Members and 1 Guest are viewing this topic.

Offline matthuszaghTopic starter

  • Frequent Contributor
  • **
  • Posts: 402
  • Country: us
I'm designing a low residual phase noise buffer amplifier for a 10MHz clock that uses an NPN transistor as the gain element. The NPN ends up being the most significant contributor to the noise performance of this circuit. Using the noise analysis in the Art of Electronics, the input-referred noise ends up being determined by the base-spreading resistance, DC current gain, 1/f corner frequency, and transition frequency (ignoring other important parameters such as collector current, temperature and source impedance). The Art of Electronics provides a really nice table of low-noise BJTs. Several standouts (among others) are the Sanyo 2SC3601 and 2SC3955, which have very low base-spreading resistance and good high-frequency operation (low input capacitance and high transition frequency). Unfortunately, these, along with many other good transistors are obsolete.

Why did OnSemi discontinue these? Same question probably applies to other companies and transistors. I've noticed a similar trend for good RF transistors. Is the market for high-quality discrete transistors too small? Do companies not use discrete transistors for low-noise applications anymore? Why is this? Is it because they can build their own ICs with the transistors they want, or are happy with the good, but not quite as good, performance of modern opamps? Any other reasons? I know a lot of these are through-hole parts, but presumably it isn't hard to place the die in a SMT package if the problem is that people aren't buying through-hole anymore.

I'm sure I'm not the only one who feels this way, but this is incredibly frustrating as a designer without the capabilities to fabricate custom transistors. Also, I know this isn't a particularly good reason, but I like being able to use discrete transistors in designs, rather than having a lot of the interesting work done for me with opamps.

I should also mention that the ZTX618 and ZTX851 are often used as replacements in applications where low rbb is desired, but they have pretty horrible input capacitance, so the high-frequency performance is significantly worse than the Sanyo NPNs mentioned above.
« Last Edit: September 04, 2022, 04:23:28 pm by matthuszagh »
 

Online wraper

  • Supporter
  • ****
  • Posts: 17952
  • Country: lv
Re: Why do companies obsolete so many good transistors without substitutes?
« Reply #1 on: September 04, 2022, 04:30:01 pm »
Old equipment gets decommissioned for example, factory burns down, whatever. And then there is no new equipment for the same manufacturing process, or it's simply not viable to buy it just for some old parts with not that much market demand. Not to say there is no Sanyo anymore as it's gone for more than a decade. Something spun off, such as capacitor production which are now Suncon, other parts were bought by other companies such as Panasonic and Fujitsu. Obviously not all product lines survived this.
« Last Edit: September 04, 2022, 04:37:46 pm by wraper »
 

Offline Benta

  • Super Contributor
  • ***
  • Posts: 6420
  • Country: de
Re: Why do companies obsolete so many good transistors without substitutes?
« Reply #2 on: September 04, 2022, 05:53:00 pm »
30+ years of semiconductor marketing has taken its toll. It's all about MOSFETs, MOSFETs, MOSFETs and MOSFETs. BJTs are out of fashion with young engineers, although they are superior in many ways.
Just look around this forum: the knee-jerk solution to any problem is a MOSFET. People seek workarounds to the problems surrounding MOSFETs instead of looking at, eg, BJTs or other devices.

Only older, experienced engineers and audio people look a BJTs today. The same can sadly also be said for JFETs.

Sign of the times.
 

Offline coppercone2

  • Super Contributor
  • ***
  • Posts: 11336
  • Country: us
  • $
Re: Why do companies obsolete so many good transistors without substitutes?
« Reply #3 on: September 04, 2022, 06:27:22 pm »
30+ years of semiconductor marketing has taken its toll. It's all about MOSFETs, MOSFETs, MOSFETs and MOSFETs. BJTs are out of fashion with young engineers, although they are superior in many ways.
Just look around this forum: the knee-jerk solution to any problem is a MOSFET. People seek workarounds to the problems surrounding MOSFETs instead of looking at, eg, BJTs or other devices.

Only older, experienced engineers and audio people look a BJTs today. The same can sadly also be said for JFETs.

Sign of the times.

well it has to do with education, you need someone to in a class teach the differences and explain the nuances of where they actually make your life easier. The problem is it would need to be senior level class that you take right before your first job. How many universities teach with the idea that 95% of the students are going to go into making ASICs and the other 5% into code? But I agree it seems like at some point mosfets are 'hacked in' to a circuit to make it work and it looks like some kind of desperate bootleg substitution if you know the alternative.

Does it feel like 'oh, mosfet is the one they used in a processor, it MUST be better' and 'i feel comfortable and trendy because mosfet based device lets me browse tik tok on my cell phone, lets use it for everything, that way I don't risk being confused by a new concept'
« Last Edit: September 04, 2022, 07:05:29 pm by coppercone2 »
 

Online tggzzz

  • Super Contributor
  • ***
  • Posts: 21226
  • Country: gb
  • Numbers, not adjectives
    • Having fun doing more, with less
Re: Why do companies obsolete so many good transistors without substitutes?
« Reply #4 on: September 04, 2022, 07:00:14 pm »
One explanation I've heard, but cannot verify, is that they make a large number of wafers, and store them until needed. When the wafers run out,the parts become obsolete.

That relies on there being a large number of very cheap devices on a wafer; it won't work with, say, large processors.
There are lies, damned lies, statistics - and ADC/DAC specs.
Glider pilot's aphorism: "there is no substitute for span". Retort: "There is a substitute: skill+imagination. But you can buy span".
Having fun doing more, with less
 

Offline floobydust

  • Super Contributor
  • ***
  • Posts: 7680
  • Country: ca
Re: Why do companies obsolete so many good transistors without substitutes?
« Reply #5 on: September 04, 2022, 07:40:37 pm »
Japanese semiconductor manufacturers are the worst, they constantly obsolete parts for their "latest greatest" technology. Older parts somehow were better- I miss the old 2SC2240 lol and any replacement comes and goes and then goes...
They just need to shrink their portfolio and focus on fewer optimized features, and have parts around for years.
Toshiba, Sanyo, Sanken suck the worst they've left a lot of products with no replacements.
 

Offline Benta

  • Super Contributor
  • ***
  • Posts: 6420
  • Country: de
Re: Why do companies obsolete so many good transistors without substitutes?
« Reply #6 on: September 04, 2022, 08:07:49 pm »
One explanation I've heard, but cannot verify, is that they make a large number of wafers, and store them until needed. When the wafers run out,the parts become obsolete.
Rubbish, don't post hearsay from the pub.
Every semiconductor manufacturer will have a "wafer bank" where processed wafers are stored, waiting for probing/testing, cutting, assembly. final testing and marking.
That's because it's the cheapest storage way while waiting for orders or executing planned deliveries before doing the final processing.
It has nothing to do with obsoletion.
 

Online coppice

  • Super Contributor
  • ***
  • Posts: 10034
  • Country: gb
Re: Why do companies obsolete so many good transistors without substitutes?
« Reply #7 on: September 04, 2022, 08:14:47 pm »
One explanation I've heard, but cannot verify, is that they make a large number of wafers, and store them until needed. When the wafers run out,the parts become obsolete.
Rubbish, don't post hearsay from the pub.
Every semiconductor manufacturer will have a "wafer bank" where processed wafers are stored, waiting for probing/testing, cutting, assembly. final testing and marking.
That's because it's the cheapest storage way while waiting for orders or executing planned deliveries before doing the final processing.
It has nothing to do with obsoletion.
Rubbish. Don't post hearsay from the pub.
Semiconductor manufacturers rarely keep a wafer bank, unless the process is being killed off, and they need a stock to fulfill the expected lifetime orders. As long as the process is still running its more cost effective to make wafers as necessary. Many now deal with this through people like Rochester Electronics, who specialise in keeping wafers fresh for extended periods.
 

Offline coppercone2

  • Super Contributor
  • ***
  • Posts: 11336
  • Country: us
  • $
Re: Why do companies obsolete so many good transistors without substitutes?
« Reply #8 on: September 04, 2022, 08:41:28 pm »
its like freezing pizza dough
 

Offline strawberry

  • Super Contributor
  • ***
  • !
  • Posts: 1199
  • Country: lv
Re: Why do companies obsolete so many good transistors without substitutes?
« Reply #9 on: September 04, 2022, 08:51:48 pm »
OP amps are not great for voltages above 35V
MOSFETs are made for switching applications with poor SOA, switching is not good for low noise application
interesting is that they make BJTs for switching applications , why? isnt MOSFET better without need for base resistors...
cool JFETs are still made
My guess is all good transistors go to Space and Military 8) and for those other losers they dont care
~1 wafer is not market for them
 

Online coppice

  • Super Contributor
  • ***
  • Posts: 10034
  • Country: gb
Re: Why do companies obsolete so many good transistors without substitutes?
« Reply #10 on: September 04, 2022, 08:58:24 pm »
OP amps are not great for voltages above 35V
There are great op-amps above 35V. There are great op-amps up to hundreds of volts. The great price can be an issue, though. :)
 

Offline Someone

  • Super Contributor
  • ***
  • Posts: 5155
  • Country: au
    • send complaints here
Re: Why do companies obsolete so many good transistors without substitutes?
« Reply #11 on: September 04, 2022, 10:29:16 pm »
Why did OnSemi discontinue these?
Did you order x00,000 units per year for the next n years? Well if no-one did, they aren't going to keep an effectively zero demand product in production (overheads of selling through small quantities, or would you like them to offer 10 units at $40,000 ?)
 

Offline coppercone2

  • Super Contributor
  • ***
  • Posts: 11336
  • Country: us
  • $
Re: Why do companies obsolete so many good transistors without substitutes?
« Reply #12 on: September 04, 2022, 11:34:21 pm »
i think you can tweak those numbers to be a little more realistic, its not a billion gate FPGA

sometimes I feel like people start throwing scary numbers around based on some real edge case stuff to try to make a point....
« Last Edit: September 04, 2022, 11:38:51 pm by coppercone2 »
 

Offline Someone

  • Super Contributor
  • ***
  • Posts: 5155
  • Country: au
    • send complaints here
Re: Why do companies obsolete so many good transistors without substitutes?
« Reply #13 on: September 05, 2022, 12:32:17 am »
i think you can tweak those numbers to be a little more realistic, its not a billion gate FPGA

sometimes I feel like people start throwing scary numbers around based on some real edge case stuff to try to make a point....
Try getting a quote for a fresh batch of obsolete transistors run on a new fab. The figures above are nowhere close to a one off fab of a large ASIC, assuming you'd even take the first batch before they have stabilised the process...

Having been in the meetings with suppliers working through costings for semi-custom diode and transistor parts right up to large ASICs.

Setup and packaging usually dominates the cost of low volumes. Researchers take bare dies from pool services without testing, qualification, or tuning of process control, as the cheapest option:
https://mycmp.fr/price-list/
Which is about the cheapest you can get a low volume set of bare die built for, $1000 per mm2 on a 0.35um BiCMOS fab.

Compared to the retail of the largest production FPGAs:
https://au.mouser.com/c/semiconductors/programmable-logic-ics/fpga-field-programmable-gate-array/?m=Intel&series=Stratix%2010%20GX
$20,000 and up per unit
 

Online mawyatt

  • Super Contributor
  • ***
  • Posts: 4119
  • Country: us
Re: Why do companies obsolete so many good transistors without substitutes?
« Reply #14 on: September 05, 2022, 01:04:18 am »
I'm designing a low residual phase noise buffer amplifier for a 10MHz clock that uses an NPN transistor as the gain element. The NPN ends up being the most significant contributor to the noise performance of this circuit. Using the noise analysis in the Art of Electronics, the input-referred noise ends up being determined by the base-spreading resistance, DC current gain, 1/f corner frequency, and transition frequency (ignoring other important parameters such as collector current, temperature and source impedance). The Art of Electronics provides a really nice table of low-noise BJTs. Several standouts (among others) are the Sanyo 2SC3601 and 2SC3955, which have very low base-spreading resistance and good high-frequency operation (low input capacitance and high transition frequency). Unfortunately, these, along with many other good transistors are obsolete.

Why did OnSemi discontinue these? Same question probably applies to other companies and transistors. I've noticed a similar trend for good RF transistors. Is the market for high-quality discrete transistors too small? Do companies not use discrete transistors for low-noise applications anymore? Why is this? Is it because they can build their own ICs with the transistors they want, or are happy with the good, but not quite as good, performance of modern opamps? Any other reasons? I know a lot of these are through-hole parts, but presumably it isn't hard to place the die in a SMT package if the problem is that people aren't buying through-hole anymore.

I'm sure I'm not the only one who feels this way, but this is incredibly frustrating as a designer without the capabilities to fabricate custom transistors. Also, I know this isn't a particularly good reason, but I like being able to use discrete transistors in designs, rather than having a lot of the interesting work done for me with opamps.

I should also mention that the ZTX618 and ZTX851 are often used as replacements in applications where low rbb is desired, but they have pretty horrible input capacitance, so the high-frequency performance is significantly worse than the Sanyo NPNs mentioned above.

Have you considered a SiGe BJT? They might be a suitable replacement, very high ft, low noise, low 1/f corner, low rbb, high Beta which is ~flat with Ic, however are low voltage devices.

Mouser has a number of BFR, BFP and MT types in stock.

Best,
« Last Edit: September 05, 2022, 01:09:12 am by mawyatt »
Curiosity killed the cat, also depleted my wallet!
~Wyatt Labs by Mike~
 
The following users thanked this post: Someone, matthuszagh

Offline coppercone2

  • Super Contributor
  • ***
  • Posts: 11336
  • Country: us
  • $
Re: Why do companies obsolete so many good transistors without substitutes?
« Reply #15 on: September 05, 2022, 03:32:06 am »
i think you can tweak those numbers to be a little more realistic, its not a billion gate FPGA

sometimes I feel like people start throwing scary numbers around based on some real edge case stuff to try to make a point....
Try getting a quote for a fresh batch of obsolete transistors run on a new fab. The figures above are nowhere close to a one off fab of a large ASIC, assuming you'd even take the first batch before they have stabilised the process...

Having been in the meetings with suppliers working through costings for semi-custom diode and transistor parts right up to large ASICs.

Setup and packaging usually dominates the cost of low volumes. Researchers take bare dies from pool services without testing, qualification, or tuning of process control, as the cheapest option:
https://mycmp.fr/price-list/
Which is about the cheapest you can get a low volume set of bare die built for, $1000 per mm2 on a 0.35um BiCMOS fab.

Compared to the retail of the largest production FPGAs:
https://au.mouser.com/c/semiconductors/programmable-logic-ics/fpga-field-programmable-gate-array/?m=Intel&series=Stratix%2010%20GX
$20,000 and up per unit

do you need a 0.35um fab for these parts? I guess we might not be thinking of the same chip.
 

Offline Someone

  • Super Contributor
  • ***
  • Posts: 5155
  • Country: au
    • send complaints here
Re: Why do companies obsolete so many good transistors without substitutes?
« Reply #16 on: September 05, 2022, 04:06:02 am »
i think you can tweak those numbers to be a little more realistic, its not a billion gate FPGA

sometimes I feel like people start throwing scary numbers around based on some real edge case stuff to try to make a point....
Try getting a quote for a fresh batch of obsolete transistors run on a new fab. The figures above are nowhere close to a one off fab of a large ASIC, assuming you'd even take the first batch before they have stabilised the process...

Having been in the meetings with suppliers working through costings for semi-custom diode and transistor parts right up to large ASICs.

Setup and packaging usually dominates the cost of low volumes. Researchers take bare dies from pool services without testing, qualification, or tuning of process control, as the cheapest option:
https://mycmp.fr/price-list/
Which is about the cheapest you can get a low volume set of bare die built for, $1000 per mm2 on a 0.35um BiCMOS fab.

Compared to the retail of the largest production FPGAs:
https://au.mouser.com/c/semiconductors/programmable-logic-ics/fpga-field-programmable-gate-array/?m=Intel&series=Stratix%2010%20GX
$20,000 and up per unit
do you need a 0.35um fab for these parts? I guess we might not be thinking of the same chip.
Okay, so you could link some pricing for other cheaper processes? The 0.35um BiCMOS is not an additional constraint/limitation I added, but just the datapoint for those who aren't going to go through and read the detail in the link (and a reasonable choice for making a discrete RF transistor). Find any cheaper low volume fab option, that example above is minimum order $5000-$6000
« Last Edit: September 05, 2022, 04:09:49 am by Someone »
 

Offline GlennSprigg

  • Super Contributor
  • ***
  • Posts: 1259
  • Country: au
  • Medically retired Tech. Old School / re-learning !
Re: Why do companies obsolete so many good transistors without substitutes?
« Reply #17 on: September 05, 2022, 12:07:50 pm »
I guess, just like 'Medicines' on the market. We would like to think they just want to do 'nice things', but the reality is that
they are a Business that has to make money, for them & shareholders.  Sometimes, certain 'Transistors' etc, are not always
used as intended/designed too. So can be hard to find a replacement. At the same time, I'm sure that many are 'specified',
where the system will work perfectly well with some 'replacement'.  Sometimes, the field is 'specialized', and using old technology.
One that comes to mind, is old TTL circuits as opposed to CMOS.   Ok, TTL didn't suffer from 'spikes/static', but drew a lot of power
and was locked in to 5v.  CMOS suffers from spikes etc, but uses very little power, and works over 3 - 15 v.  The old stuff is out!
Try an tell my Chemist(s) where my old Meds are gone, and why??    8)
Diagonal of 1x1 square = Root-2. Ok.
Diagonal of 1x1x1 cube = Root-3 !!!  Beautiful !!
 


Share me

Digg  Facebook  SlashDot  Delicious  Technorati  Twitter  Google  Yahoo
Smf