Transistors - die pictures

[floobydust]

I've heard it's the old "meets the spec" criteria. If I make a transistor/IC that meets the datasheet numbers, I can call it this or that part number. TI told me that when I bitched about their (acquisition) IC's "it meets all specs" despite different behaviour.

The 2N3568 is interesting being part of the low cost packaging move, towards plastic transistors. In the mid 1960's so many ads and marketing on new transistors. Today we pretty much have none at all which is strange because haven't we beat the 2N2222 in almost 60 years? (I'm not sure what the PN2222A is about, it seems to be heat).
I did see Fairchild succumb and also make the TO-92 package, advertising better leadframe heat dissipation as their selling feature. The epoxy on ceramic base I guess lasted maybe 10 years? Funny the ADR1001 is also on a thick slab of ceramic, although metal lid. It's the same mentality - "metal package bad".

[David Hess]

In the mid 1960's so many ads and marketing on new transistors. Today we pretty much have none at all which is strange because haven't we beat the 2N2222 in almost 60 years? (I'm not sure what the PN2222A is about, it seems to be heat).

We did beat the 2N2222, 2N3904, and 2N4401 eventually.  Zetex Semiconductors built small signal perforated or ring emitter transistors using obsolete IC processing tools and called them "Super E-Line" transistors.  To take advantage of their better performing bipolar transistors, they improved the TO-92 package by making the collector tab of the leadframe more massive, improving heat conduction through the collector pin.

Diodes Incorporated bought Zetex not too long ago and their parts are still available, and other manufacturers also make similar advanced transistors now.  They have about the same speed as the 2N2222, lower and higher voltage ratings, and much better saturation characteristics and hfe.  Their low base spreading resistance also makes them lower noise.

[T3sl4co1l]

The 2N3568 is interesting being part of the low cost packaging move, towards plastic transistors. In the mid 1960's so many ads and marketing on new transistors. Today we pretty much have none at all which is strange because haven't we beat the 2N2222 in almost 60 years? (I'm not sure what the PN2222A is about, it seems to be heat).

I mean, if you aren't shopping for transistors, you won't be seeing new ones!

BJTs are pretty well explored technology; the main innovations in recent decades have been more current in smaller packages, and with faster switching at lower Vce(sat) (finer pitch, more interleaving, I suppose?).  Typical application: PNP 40V 3A in SOT-23 for gate drive booster.  Interesting quirk: low Vce(sat) somewhat implies good inverted behavior, and indeed they usually have high inverted hFE.

That, and HV switches; not really a new application, but optimization here and there.  Main thing is MOSFETs are easier to use, and achieve lower losses overall, so there are few BJT based SMPS controllers for example (but not none; there is at least one Chinese SOT-23-6 offline flyback controller that's made for BJT switch, that I've seen).  Maaaybe some integrated ones, not sure, but as you might imagine, most such innovation is in Chinese so I'm not too aware of it.

It's testament to their usefulness that the basic 600mA 60V 150MHz general-purpose BJT has never gone away.  Technologies are rarely if ever fully obsoleted; even vacuum tubes remain useful, in niches, though arguably more as physics packages than electronic devices.  Jellybean transistors will never go away; CD4000 will never go away; 74HC will never go away; they're all too useful to glue together base level circuitry, even if it's just for building and wiring power supplies while top-tech ASICs handle everything else under the sun.

I did see Fairchild succumb and also make the TO-92 package, advertising better leadframe heat dissipation as their selling feature. The epoxy on ceramic base I guess lasted maybe 10 years? Funny the ADR1001 is also on a thick slab of ceramic, although metal lid. It's the same mentality - "metal package bad".

Well, metal and ceramic are long established hermetic and high temperature packaging.  Pay enough and you can get anything in that.  I don't see what relation that bears to economical plastic packaging.

Tim

[Wolfgang]

ZTX851 is a spectacular part with 5A, a very low VSat and a base spreading resistance thats really small, making it useful as a low noise amp.

[exmadscientist]

I wonder what you'd see if you decapped a recent 2222, 3904, 4401, and BC847 from the same (major) vendor, and same week of production (as much as possible)?

[David Hess]

BJTs are pretty well explored technology; the main innovations in recent decades have been more current in smaller packages, and with faster switching at lower Vce(sat) (finer pitch, more interleaving, I suppose?).  Typical application: PNP 40V 3A in SOT-23 for gate drive booster.  Interesting quirk: low Vce(sat) somewhat implies good inverted behavior, and indeed they usually have high inverted hFE.

Those are what Zetex pioneered with the "Super E-Line" transistors.  Their hfe and Ft are sustained better at higher currents allowing high current low power devices in small packages, and their saturation voltage is excellent.  The small signal devices are not particularly faster however, although the power devices based on the same fabrication are faster than equivalent power devices.

[David Hess]

I wonder what you'd see if you decapped a recent 2222, 3904, 4401, and BC847 from the same (major) vendor, and same week of production (as much as possible)?

Diotec (German) makes the 2N4401/2N4403 and 2N2222A/2N2907A in the TO-92 package which look suspiciously like different voltage grades of the same part.  They also make the slightly higher current BC337/BC327 but I wonder if it is the same part with another different grading.

[Noopy]

I should put the investigation of the "2222, 3904, 4401, and BC847" on my list. Unfortunately the list is long... 





Here we have another one of this era. BTW: I have flipped the numbers. The package of the 2N3568 and the 2N5136 is called TO-105, not TO-150. The smaller version is called TO-106.

The American semiconductor manufacturer Continental Device Corporation (CDC) emerged from the semiconductor division of Hughes Aircraft and was later acquired by Teledyne. CDC co-founded the Indian semiconductor manufacturer CDIL.

There is no separate datasheet for the CDC 2N5136. The 2N5136 variant from Fairchild is specified with a reverse voltage of 20V. At 150mA, the typical amplification factor is 100 and the datasheet specifies a factor of 2 as the minimum for high-frequency amplification at 20MHz. A power loss of up to 300mW can be dissipated via the TO-105 housing.




The epoxy encapsulation does not contain any fillers, which is why only a small amount of material remains after thermal conversion. The collector pin is bent and widened in the housing. The actual transistor is located on it. Base and emitter potentials are supplied via bondwires. The transistor is protected with additional potting.




Although the additional potting has not burnt, it has become brittle. The impression shows that the transistor is almost too large for the metal surface.






Some of the potting still adheres to the die.






The edge length of the die is 0,88 mm. Contacting the base potential is possible on two sides. The star shape of the emitter ensures a low base resistance with a slightly larger emitter surface. Smaller transistors do not require any special geometries. Power transistors, on the other hand, usually have finger-shaped interlocking structures.


https://www.richis-lab.de/BipolarA61.htm

 

[exe]

Diotec (German) makes the 2N4401/2N4403 and 2N2222A/2N2907A in the TO-92 package which look suspiciously like different voltage grades of the same part. 

Do they have their own semiconductor plant? I always had an impression they just resell stuff from China. They say they have production sites in Slovenia, China, and India. But, looking at linkedin, I see very few people working outside of Germany (https://www.linkedin.com/company/diotec-semiconductor/people/). I do see two people with titles "process engineer" and "senior engineer", but the rest is sales and marketing. So, kinda looks they just rebadge stuff. However, I can't find a equivalent of CL05M6F from another manufacturer. So, not sure.

Does it even make sense to still have own factory for cheap diodes and transistors? (Unless you are a big established factory)

[dzseki]

Diotec (German) makes the 2N4401/2N4403 and 2N2222A/2N2907A in the TO-92 package which look suspiciously like different voltage grades of the same part. 

Do they have their own semiconductor plant? I always had an impression they just resell stuff from China. They say they have production sites in Slovenia, China, and India. But, looking at linkedin, I see very few people working outside of Germany (https://www.linkedin.com/company/diotec-semiconductor/people/). I do see two people with titles "process engineer" and "senior engineer", but the rest is sales and marketing. So, kinda looks they just rebadge stuff. However, I can't find a equivalent of CL05M6F from another manufacturer. So, not sure.

Does it even make sense to still have own factory for cheap diodes and transistors? (Unless you are a big established factory)

Well, opening up their web page makes the impression that they actually make the devices.

[David Hess]

Diotec (German) makes the 2N4401/2N4403 and 2N2222A/2N2907A in the TO-92 package which look suspiciously like different voltage grades of the same part.

Do they have their own semiconductor plant? I always had an impression they just resell stuff from China. They say they have production sites in Slovenia, China, and India. But, looking at linkedin, I see very few people working outside of Germany

...

Does it even make sense to still have own factory for cheap diodes and transistors? (Unless you are a big established factory)

I wondered the same thing and saw that they had production sites outside of Germany.  It seemed unlikely that it would be economical to do production in Germany, but I do not know if their primary source is China.  It does seem likely but they must be more than a distributor with those other sites.

I am rebuilding my transistor selection and I am planning to pick up a set of transistors from them for evaluation and use:

BC546C/BC556C
2N3904/2N3906
2N4401/2N4403 or 2N2222A/2N2907A
BC337-40/BC327-40

[dzseki]

Diotec (German) makes the 2N4401/2N4403 and 2N2222A/2N2907A in the TO-92 package which look suspiciously like different voltage grades of the same part.

Do they have their own semiconductor plant? I always had an impression they just resell stuff from China. They say they have production sites in Slovenia, China, and India. But, looking at linkedin, I see very few people working outside of Germany

...

Does it even make sense to still have own factory for cheap diodes and transistors? (Unless you are a big established factory)

I wondered the same thing and saw that they had production sites outside of Germany.  It seemed unlikely that it would be economical to do production in Germany, but I do not know if their primary source is China.  It does seem likely but they must be more than a distributor with those other sites.

I am rebuilding my transistor selection and I am planning to pick up a set of transistors from them for evaluation and use:

BC546C/BC556C
2N3904/2N3906
2N4401/2N4403 or 2N2222A/2N2907A
BC337-40/BC327-40

According to their company video, they do silicon processing in Slovenia and Germany, and only "assembly" plants are in the far east.

Diotec company video

[exe]

May be I overestimate how much space a fab would need (actually, never seen one). On this video and others it seems their plant in Germany is half-empty: https://youtu.be/q1Q7Yd8u8mQ?si=KNkwI2SsKeichGOT&t=84 . Equipment, it seems, doesn't take much space at all.

They say they now sell 3.7B part annually. That is 820k part per day on average (including weekends etc). I'm surprised such a small facility can handle that. May be they don't show everything, or even re-use old footage, as I see on their video.

On the other hand, it can be just some 200-500 reels a day (depending how many parts fit one reel). So, may it's not that huge volume after all. I'll be happy to know more about manufacturing from you guys.

[dzseki]

May be I overestimate how much space a fab would need (actually, never seen one). On this video and others it seems their plant in Germany is half-empty: https://youtu.be/q1Q7Yd8u8mQ?si=KNkwI2SsKeichGOT&t=84 . Equipment, it seems, doesn't take much space at all.

They say they now sell 3.7B part annually. That is 820k part per day on average (including weekends etc). I'm surprised such a small facility can handle that. May be they don't show everything, or even re-use old footage, as I see on their video.

On the other hand, it can be just some 200-500 reels a day (depending how many parts fit one reel). So, may it's not that huge volume after all. I'll be happy to know more about manufacturing from you guys.

Consider a 6" wafer (150mm diameter), assuming 1mm^2 die area (cutting loss included), then counting with 80% yield a single wafer would give 14500 devices. For 800k devices a day they need to process slightly more than 55 wafers a day, these factories are working in 0-24h, does not sound too terrible.

[David Hess]

The American semiconductor manufacturer Continental Device Corporation (CDC) emerged from the semiconductor division of Hughes Aircraft and was later acquired by Teledyne. CDC co-founded the Indian semiconductor manufacturer CDIL.

I have a bunch of the CDC ones also.  Maybe I missed it, but how are they related to the Fairchild production using the same cases?

As I recall, the CDC ones always used white ceramic, but the Fairchild ones could be white or black.

[Noopy]

What I wanted to say was, that there is no datasheet for the CDC 2N5136 but there is a datasheet for the Fairchild 2N5136 and I assume the two 2N5136 are very similar.

The Fairchild 2N3568 and the CDC 2N5136 use the same package size (besides the color) but that of course is nothing special.

[floobydust]

A few 2N5136 datasheets show it as the slowest in town f_T 40MHz, and only 20V?  Nothing could I find for CDC.
Sprague process "JLA" does not match though. Interesting die design pics in the Sprague databook.

I think a transistor's part number is just generic JEDEC specs, everyone uses their own process it seems.
P.S>  I hate JEDEC what a mess they've turned into. TO-226AAQWWGEBSW forever lol

[Noopy]

Yeah, quite a big small signal transistor but the specs are not very impressive. Same with the Sprague variant.

I agree with you. These are JEDEC part numbers and a lot of manufacturer built their version as economical as possible. 

[quince]

watching.

[floobydust]

It could also be these lackluster transistors are just the rejects/fallouts for the root/parent part number? 1,000's of mutts are out there.

Myself and many others are using the Zetex for low noise audio discretes. It's unfortunate I will bet Diodes Inc. is clueless about that application for them and thinks they are great power transistors for CFL inverters etc. I wish they would invest in that acquisition and roll out some new low RBB' parts.

Japanese semi companies i.e. Toshiba constantly roll out new BJT's, but by time you design it in, it's discontinued. So their approach is terrible.

[David Hess]

My literal 1976 D.A.T.A BOOK says 220 mW, 40 MHz, 30 Vcbo, 20 Vceo, 500 mA, and 20 hfe @ 150 mA for the 2N5136, but says nothing about CDC making them.

[floobydust]

I wondered if it was Control Data Corp? but they didn't use house-labelled that I could find. I got them from a surplus store around 1976, they had bushels and couldn't give them away almost. Some on mainframe looking huge PC boards.
Minimum Cost Semiconductor Silicon Survey (April 1969) of Fairchild TO-105, TO-106 parts. 2N5136 $0.11 "Not recommended", " need not supply a data sheet for publication... none of the readers of 73 would be interested." 

"In 1963, he {Robert Noyce} convinced the notoriously conservative board of one of his companies to start the semiconductor industry’s first offshore manufacturing facility—at a site that was then completely under water, soon to be reclaimed from the bay by the government of Hong Kong. He talked a carload of traveling companions into joining him for a dip in a brackish Tibetan river, murky and, just a bit upstream, filled with crocodiles." source The man behind the microchip : Robert Noyce and the invention of Silicon Valley

"Beginning in mid-1965, Fairchild shipped silicon wafers from Mountain View to a Hong Kong plant that would produce 100,000 planar transistors per week with the help of a staff of about 12 technical and 135 production employees, each of whom was paid one dollar per day—less than their American counterparts earned in an hour."

I am surprised outsourcing semiconductors (packaging) to Asia began in 1965, and it was all about the labour costs.

[David Hess]

I wondered if it was Control Data Corp? but they didn't use house-labelled that I could find. I got them from a surplus store around 1976, they had bushels and couldn't give them away almost. Some on mainframe looking huge PC boards.
Minimum Cost Semiconductor Silicon Survey (April 1969) of Fairchild TO-105, TO-106 parts. 2N5136 $0.11 "Not recommended", " need not supply a data sheet for publication... none of the readers of 73 would be interested." 

That is what I was thinking; maybe Fairchild house numbered them for Control Data Corp.

I just checked and the small white CDC ones that I can find immediately are marked CDC CS2923 and CDC CS2924, and also I have these part numbers as 2N in the older "top hat" style TO-92, and probably in real TO-92.  The all black ones seem to all be Fairchild, and some of those are also marked Singapore.  All appear to have date codes through the 1970s.

Somewhere I am sure I have some of the larger packaged white CDC ones.  I might have pulled all of the white CDC ones off of forgotten boards, but many appear new with the original lead lengths.

[Noopy]

I wondered if it was Control Data Corp? but they didn't use house-labelled that I could find. I got them from a surplus store around 1976, they had bushels and couldn't give them away almost. Some on mainframe looking huge PC boards.
Minimum Cost Semiconductor Silicon Survey (April 1969) of Fairchild TO-105, TO-106 parts. 2N5136 $0.11 "Not recommended", " need not supply a data sheet for publication... none of the readers of 73 would be interested." 

That is what I was thinking; maybe Fairchild house numbered them for Control Data Corp.

I just checked and the small white CDC ones that I can find immediately are marked CDC CS2923 and CDC CS2924, and also I have these part numbers as 2N in the older "top hat" style TO-92, and probably in real TO-92.  The all black ones seem to all be Fairchild, and some of those are also marked Singapore.  All appear to have date codes through the 1970s.

Somewhere I am sure I have some of the larger packaged white CDC ones.  I might have pulled all of the white CDC ones off of forgotten boards, but many appear new with the original lead lengths.

Here we see a transistor built by Continental Device Corporation similar looking to ours with a "CDC" on the side of the white ceramic:

https://archive.org/details/bitsavers_ElectronicignV15N1419670705_116901282/page/72/mode/2up?q=%22Continental+Device+Corporation%22





Do you remember the Nexperia PMPB19 with the metal interface that didn´t vanish?






A further one-hour treatment with HF finally removes the last layer on the active structures.




The individual MOSFET cells or lines are too small to be resolved optically (with my equipment). However, the active area can be recognised by its rougher surface. Thin strips lead from the frame, where the gate potential is distributed via the metal layer, to the active area. The geometry appears surprisingly complex for a simple supply line.

It is conceivable that the protective diodes for the gate potential are located here. They would then be distributed over the entire frame structure. If it is possible to ensure a reasonably even current distribution, this would be an efficient solution.

However, it could also be that the supply lines are gate resistors. Gate resistors reduce the switching speed, but improve the synchronisation of the individual MOSFET areas.


https://www.richis-lab.de/FET44.htm#strip2

 

[ArdWar]

I always get confused with those "protected" MOSFETs, never sure whether those protection zeners are supposed to be continuous duty rated or just an "oh crap" intermittent rated. I don't think anyone ever actually put any rating datapoint on them; max shunted current/energy, reverse leakage, etc. Or even something simple like whether the listed V_GS(MAX) already factors the zeners or not.