Transistors - die pictures

[David Hess]

Incidentally, 2N3904 is -- or, was originally, anyways? -- gold doped, I guess fairly lightly given the higher breakdown voltage than this part (though as it happens, typical 2369s avalanche at 70V, and 3904s at 105V; make of that what you will!), but I've never seen reference to this outside of an early (70s) National databook.

Bob Pease said to watch out for gold doped 2N3904s when using them as low leakage diodes, which implied to me that most 2N3904s were not gold doped.

I also noticed that the National Semiconductor discrete databook said that 2N3904s were made on their gold doped process, but concluded that it was a mistake because the National 2N3904s never had high leakage.

[T3sl4co1l]

Ah, so that's probably ancient history, like hometaxial-2N3055 era stuff even.

I've also found them (modern ones, anyway) to have good hFE at low currents.  Recovery time still seems lower than usual, I forget what the exact scenario was but the topic came up recently in relation to other types, and more generic like BC847 types can have much longer storage/recovery which can make surprises say for dimensioning R_BE.  Dunno how much that depends on doping vs construction.

Tim

[Noopy]

The PMBT2369 is a fast switching transistor similar to the ON Semiconductor MMBT2369. The specifications differ only minimally in the switch-on and switch-off times. However, the differences are hardly relevant.






The die is 0,39mm x 0,24mm in size. The active area is 0,11mm x 0,09mm.


https://www.richis-lab.de/BipolarA59.htm

 



By the way: It is always a lot of fun to handle such small things:
Find it in the burned epoxy.
Get it out of the burned epoxy.
Clean it as good as possible.
Place it on the probe holder.
Be sure the orientation is correct.
 

[magic]

The real active area is the two emitter strips of 70×20μm each and a few micron thick base layer under them

[Noopy]

Well perhaps "active layer" perhaps is not the best paraphrase...  But I would say around the collector base junction there is something happening too. Especially when you have to switch fast and capacity is relevant.

[floobydust]

Wow PMBT2369 f_T is 500MHz.

[dzseki]

Incidentally, 2N3904 is -- or, was originally, anyways? -- gold doped, I guess fairly lightly given the higher breakdown voltage than this part (though as it happens, typical 2369s avalanche at 70V, and 3904s at 105V; make of that what you will!), but I've never seen reference to this outside of an early (70s) National databook.

Bob Pease said to watch out for gold doped 2N3904s when using them as low leakage diodes, which implied to me that most 2N3904s were not gold doped.

I also noticed that the National Semiconductor discrete databook said that 2N3904s were made on their gold doped process, but concluded that it was a mistake because the National 2N3904s never had high leakage.

I could tell, since I have access to DLTS equipment at work.

[David Hess]

Wow PMBT2369 f_T is 500MHz.

The 5771 is 850 MHz, but all of the other fast saturated switches are 400 to 500 MHz,  In contrast a contemporary RF part like the MPSH81/MPSH10 are 600/650 MHz minimum, but even with Baker clamping cannot reach the switching speeds of the fast saturated switches.  These transistor types really are different.

[Wolfgang]

One of the best form saturated switching is 2N2501, also perfect for avalanche pulsers.

[David Hess]

One of the best form saturated switching is 2N2501, also perfect for avalanche pulsers.

20 volts, 350 MHz minimum, and 15 nanosecond maximum storage time so very nice, but out of production now for a long time.  The 2369 is a little bit faster at 13 nanoseconds maximum storage time.  The PNP devices are stuck at 20 nanosecond storage times.

Central Semiconductor lists it but then says please call.

[Wolfgang]

I think I ordered them from Microsemi, IIRC.

[Noopy]

The RCA 3N200 is a n-channel MOSFET. It is a depletion type, i.e. a self-conducting MOSFET. According to the datasheet, the maximum permissible blocking voltage is 20V. The drain current must not exceed 50mA. The 3N200 is advertised for applications with frequencies up to 500MHz. At 400MHz, an amplification of 12,5dB would typically still be possible.




The datasheet shows that the gate electrodes have an overvoltage protection. As the 3N200 is a depletion type MOSFET, additional antiserial diodes had to be integrated. This is the only way to set the gate-source voltages to negative.




The RCA Application Note AN-4018 contains a somewhat clearer circuit diagram of the 3N200 and also shows the structures on the silicon. The simplified cross-section shows how the MOSFET works. Gate electrodes 2 and 1 comprise the drain area in the middle, followed by the source area. Heavily n-doped areas form both the drain and source areas, as well as the transition zone between the gate areas.

Two p-doped elements in an n-doped well form the antiserial diodes, which serve as overvoltage protection for the gate electrodes. A p-doped shield is shown between the MOSFET structures and the protective diode wells, which ensures that the two areas always remain isolated from each other. Otherwise it would be possible for conductive channels to form due to potentials applied above them or for parasitic bipolar structures to become conductive.

The cross-section conceals the fact that the substrate is connected to the source potential. This connection requires additional insulation measures for the protective diodes. This can be seen particularly clearly in the large metal area above the protective structures of gate 1, where the underlying contours can be seen. The square structure in the right-hand area shows a contact (smaller square) and a p-doped area (larger square). There is an n-doping in the neighbourhood, forming a diode. The same structure can be found with a slightly more elongated shape in the bottom left-hand corner of the area. This is the second diode. The area with the two diodes is not directly embedded in the substrate. The contours show two additional frame structures, the outer one of which is connected to the source potential.




In the package, the source pin is connected directly to the metal can.






The dimensions of the die are 0,63mm x 0,61mm. The structures correspond to those shown in the application note.




In the upper area, a few squares show how well the masks are aligned against each other. The structures of the protective diodes are also clearly visible here in the metal surfaces.


https://www.richis-lab.de/FET46.htm

 

[SeanB]

I used one 3N200 as a logic gate, to decode a single segment off of a VFD, as I wanted to decode if a set of channels were in use. The -20V limit was handled using some 4M7 1/8W resistors in a voltage divider, to droop the -35V drive signals down to a safer level, and source was connected to 0V, and drain had a RC network to filter the pulses, where a PNP transistor powered off the 5V rail was used to drive a 5V relay. Relay was used to switch audio input, as I wanted to record TV shows that were simulcast, and wanted the alternate audio (the original English) which was broadcast on a radio instead.

[Noopy]

One more Dual-Gate MOSFET.  The КП350 (KP350) is a Dual-Gate n-channel MOSFET. As with many Soviet semiconductors, the manufacturer cannot be identified. The device was produced in 1985. A description of the KP350 can be found in the journal Radio Fernsehen Elektronik (issues 6 and 7, 1975). With operating frequencies of up to 400 MHz, it is listed there as an alternative to the 3N140, 3N141 and TIXS35. The maximum permissible reverse voltage is 15V. The maximum permissible drain current is specified as 30mA. The indices A, B and V indicate different bins.




The KP350 has no protective diodes at the gate electrodes. These contacts must be protected against overvoltage accordingly. Before installation in a circuit, a plastic sleeve ensures that the pins remain electrically connected to each other.




The KP350 is a depletion type, i.e. a self-conducting transistor. The RFE specification shows that the MOSFET can be controlled with both positive and negative gate-source voltages.




The package of the KP350 is connected to the source potential.






The edge length is 0,74 mm. In the lower area, the fracture edge comes very close to the active structures. There are some structures on the upper edge that make it possible to assess the manufacturing quality.




The drain electrode is located in the active structures on the very inside. It is surrounded by gate electrodes 2 and 1. The source potential encompasses these areas and also contacts the substrate.


https://www.richis-lab.de/FET47.htm

 

[Noopy]

The Fairchild 2N3568 is a standard NPN transistor. This component was produced in 1972. The 2N3568 blocks up to 60V. In the same datasheet you will find the 2N3567, a slightly worse bin, whose blocking voltage is only 40V. The datasheet specifies a maximum current of 500mA. The transistor achieves the typical current gain of 80 at a current of 150mA. At a frequency of 20MHz, an amplification factor of 3 can be expected. The package is called TO-150. It is a cylindrical element on which the actual transistor is placed. A black potting serves as protection against environmental influences. At an ambient temperature of 25°C, up to 300mW of power loss can be dissipated.






The transistor was slightly damaged when it was decapped. The edge length is approximately 1mm. The structures are reminiscent of a power transistor.


https://www.richis-lab.de/BipolarA60.htm

 

[floobydust]

Was it a ceramic base with epoxy on top? I think it was one of the early plastic? transistor packages. I bought many from surplus dealers mid 1970's, it did not catch on.

The collector current curve- I don't remember the knee and rise as you approach breakdown voltage  Is this due to the process "NPN silicon PLANAR* epitaxial" *planar is a patented Fairchild process? Same as 2N2222 circa 1970.

[Noopy]

Yes, it seems to be kind of a ceramic and some epoxy on top of it.
This package type is really old. I found smaller ones with the same construction in a discrete opamp.

The rise before breakdown could be due to a rising collector base leakage current. Old process more leakage?
"Silicon Planar Epitaxial" is a normal process in my view.

[David Hess]

I still have some 2N3568s, as well as 2N3565s in the same but smaller diameter package.  I think there were PNP versions as well.  An 8 lead version of the same package was used for Fairchild RTL "Micrologic" integrated circuits.  The package has a flat to indicate orientation.  I assume the TO-92 package lacking ceramic was less expensive.

The ceramic could also be white with the same black epoxy.  I think these were the older ones.

Tektronix used 2N3565s in the 1970s.  I repaired a DC505 which had two bad 2N3565s, and almost all of them tested "weak" on my curve tracer with a high saturation voltage, so I replaced all of them with a modern TO-92 part.

[T3sl4co1l]

Those parts are somewhat notorious these days for the black stuff turning to goo, I think; probably it was recognized early enough along, and replaced as better resins and fillers made TO-92 widely acceptable.

The breakdown knee is simply what breakdown looks like: as you approach it, collector current goes up, regardless of base current or voltage; simple as that.  The point at which this happens, does of course depend on base voltage; in a constant-Ib setup, Vceo applies, while in a constant-Vbe setup (namely for Vbe <= 0 in RBSOA), Vcbo applies.  It's also time-dependent as the transistor comes out of saturation (hence typical switching transistors today having "dynamic RBSOA" plots, and they're actually quite interesting, not the trivial squares usually given in IGBT datasheets ).

For intermediate values (say with a B-E resistor, so that not all base leakage is shunted to emitter), breakdown can occur somewhere inbetween [Vceo and Vcbo], and often exhibits pulse avalanche behavior: at an avalanche discharge site, local base voltage is driven up, triggering additional current flow and raising base voltage overall, and a runaway process ensues; my understanding is, within a fraction of a nanosecond, the discharge filament heats up to several hundred °C, turning the material intrinsic and highly conductive in a local area, causing C-E punchthrough and hence sudden low resistance.  A typical discharge can handle several amperes peak, without apparent wear on the device (e.g., a 2N3904 switching a 50Ω transmission line from ~110V); devices rated and tested for this operation can pulse over 20A in a few ns, from a couple hundred V prior to trigger.  Recombination (cool-down, more like it) takes some tens of µs, then normal operation resumes.

Tim

[floobydust]

I often wonder why some transistors became a great success verses others that flopped. Are 20¢ Transistors Coming? 73 Magazine 1965. It looks the race to cheap transistors had that Fairchild (package) epoxy blob top on a ceramic base, packaging and test in Hong Kong - maybe it was crap (moisture), too cheap? Heat dissipation was better with the TO-18, TO-5 etc. metal can which also could met MIL. Outsourcing semi packaging in 1965 to Asia

But it seems the 2N2222 is considered a big hit and of the same era, it won I guess. Motorola must have circumvented Fairchild's planar epitaxial patents I guess, for the 2N2222.

[David Hess]

Those parts are somewhat notorious these days for the black stuff turning to goo, I think; probably it was recognized early enough along, and replaced as better resins and fillers made TO-92 widely acceptable.

I have a lot of transistors in that package, some of the ICs, and a few pieces of 1970s equipment with them, and have never seen the epoxy turn to goo.

The TO-92 package should have been cheaper since the whole thing is molded over the leadframe.

[David Hess]

I often wonder why some transistors became a great success verses others that flopped. Are 20¢ Transistors Coming? 73 Magazine 1965. It looks the race to cheap transistors had that Fairchild (package) epoxy blob top on a ceramic base, packaging and test in Hong Kong - maybe it was crap (moisture), too cheap? Heat dissipation was better with the TO-18, TO-5 etc. metal can which also could met MIL. Outsourcing semi packaging in 1965 to Asia

But it seems the 2N2222 is considered a big hit and of the same era, it won I guess. Motorola must have circumvented Fairchild's planar epitaxial patents I guess, for the 2N2222.

The 2N2222 was higher performance and much faster.

[floobydust]

I'm still baffled there are transistors, part number 60 years old that are popular today. The packaging, the die is of course changed many times, no gold, no lead etc.

It looks like Motorola invented the TO-92?  Ref. Aug. 1965 2N3903-6 press release "annular process".
"Unibloc” (pressure-molded) unit package eliminates use of separate preformed header and poured cap (which can be separated under thermal cycling due to incompatibility at the interface."
I think this implies the Fairchild (i.e. 2N3568) epoxy/ceramic package was not so reliable. The ones I had seemed to just fade, go low h_FE.

[exe]

I'm still baffled there are transistors, part number 60 years old that are popular today. The packaging, the die is of course changed many times, no gold, no lead etc.

Yeah, I think of the same. What I don't like is that there are many different "generic" parts with different characteristics but with the same/similar name.

[David Hess]

Yeah, I think of the same. What I don't like is that there are many different "generic" parts with different characteristics but with the same/similar name.

When the registered specifications are loose, like with the 2N3055, then parts which fail more specific specifications in testing can be remarked and still sold.  Unfortunately however this leads to parts with a wide variety of hidden specifications.