Transistors - die pictures

[RoGeorge]

Half-Life was an incredibly good game.   
Googled the year and it was released in 1998.

[Noopy]

Half-Life was definitely a highlight! 

I got some insider information about the SF137:
It was the first silicon epitaxial planar transistor built by the HFO (1967).
They introduced these many hfe bins to be able to sell as much transistors as possible. After 1970 they mostly got D and E.
Here we have an updated design because the first one consumed 1mm2 of silicon.

[Noopy]

2N2915, an old dual bipolar transistor which was built by a lot of companies. I found the first 2N2915 in a GE transistor manual dating back to 1964. This one was built by Texas Instruments.
There was a family of transistors named 2N1913 - 2N1920 with slightly different specifications. The 2N2915 can isolate 45V, can conduct 30mA and gives you a hfe of 60-240 while the hfe factor of the two transistors is higher than 0,9.






In the package there are two dies placed on a ceramic substrate. That gives you a good thermal coupling while the transistors are electrically isolated.






The die has an edge length of 0,43mm. There are four base contacts. You can see that on one contact there was a test needle some time ago.
I assume the outer metal ring is just for a proper border finalization.


https://www.richis-lab.de/Bipolar76.htm

 

[David Hess]

Dual transistors like that were very common as differential pairs until the 741 operational amplifier became popular and economical.

[Noopy]

Dual transistors like that were very common as differential pairs until the 741 operational amplifier became popular and economical.

 

I should have mentioned that. The binned and thermally coupled transistors work well in differential stages.

[bsw_m]

A dual low noise differential pair on channel field-effect transistors US006 was used on a V7-54 voltmeter as the input differential pair of the amplifier of the device. In reality, it has an ultra-low leakage current of the gate (according to the documentation, no more than 3pA, according to measurements it is less than 100fA). The actual offset of the amplifier on this micro-assembly (without offset correction) is no more than 30μV.
This micro-assembly was developed and manufactured by MNIPI.
Includes 4 crystals a channel field-effect transistor crystals and a set of resistors, with the possibility of adjustment.

[bsw_m]

A electrometric pair of P-channel MOSFETs.
Designed and manufactured by MNIPI. In production from 1978 to 19??
This pair of transistors was used in electrometers and electrometric amplifiers developed by MNIPI.
An example of devices where these transistors were used are U5-11, V7E-42.
Of interesting, datasheet say that gate leakage current value does not exceed 1fA. The measured actual value of the gate leakage current is approximately 50aA.

[David Hess]

Dual transistors like that were very common as differential pairs until the 741 operational amplifier became popular and economical.

I should have mentioned that. The binned and thermally coupled transistors work well in differential stages.

They are not the only ones, but I was thinking of various Tektronix linear regulator designs which changed from dual transistors to 741 and 1458 operational amplifiers sometime between 1981 and 1984.  They used a lot of 2N2918/2N2919s in the form of two 2N2484s in the TO-78 package as differential pairs for the voltage and current error amplifiers.  They had more than three different grades but I do not know what their matching criteria was beyond one being low noise and one being higher voltage.  I assume the offset voltage would have been somewhere between 1.5 and 5 millivolts but maybe they were making their own to get better precision, or were just being cheap.

I do not have any spares or extras, or I could mail one to Noopy to get an idea of what they were doing.

[David Hess]

A pair of P-channel MOSFETs.
Designed and manufactured by MNIPI.
This pair of transistors was used in electrometers and electrometric amplifiers developed by MNIPI.
An example of devices where these transistors were used are U5-11, V7E-42.
Of interesting, passport gate leakage current value does not exceed 1fA. The measured actual value of the gate leakage current is approximately 50aA.

How did others get those kinds of parts?  Did they grade 3N series MOSFETs or 2N4351s for gate leakage?

[bobAk]

What is 3n mosfet?

[David Hess]

What is 3n mosfet?

The original common discrete 4-lead small signal MOSFETs, which are still available, were JEDEC registered with 3N part numbers like 3N163 through 3N166, 3N170, 3N171, 3N190, 3N191, etc.

[Noopy]

BC149B, a old transistor in an old package (SOT-25) built by many manufacturers.
20V, 50mA, 350MHz but the different manufacturers had sometimes different specifications.






The edge length of the die is 0,35mm.






"DE9", a different datecode?






There is a little difference, there is just one alignment structure in the upper right corner.


https://richis-lab.de/Bipolar77.htm

 

[Noopy]

BC158, a complementary type to the BC149B (but not the low noise one, that would be the BC159).






The design is a little different. 


https://www.richis-lab.de/Bipolar78.htm

 

[Noopy]

SL3127, a fast transistor array built by Plessey Semiconductors. There are five transistors with a ft of 1,6Ghz at 5mA collector current. The maximum collector emitter voltage is 18V. The base emitter voltages differ no more than 5mV.
There is a substrate contact which you have to connect to the most negative voltage of the transistors. You can also use a more negative potential. That reduces the leakage current and the parasitic capacitance. Maximum collector substrate voltage is 55V typ. (but only 20V min.!).






The die is 1,1mm x 1,0mm. WH047B is the internal name. There are a lot of structures to check the process quality.
There are six transistors. The unused one is shorted with a metal strip. There is also an unused substrate contact. With a different metal layer you get a different circuit.








The brown area is the collector area (n) isolated from the green area (p). The collector potential is connected at the top and at the bottom of the transistor. It looks like there is an area around the collector contact. That is probably an area with higher doping for low resistance in the collector path.
Base and emitter are a little strange. There is a big rectangle which is connected to the emitter pin by two vias. There are three base contacts which seem to contact three smaller areas in the emitter rectangle. Such an arrangement only makes sense if it´s a perforated emitter like structure.


https://www.richis-lab.de/Bipolar79.htm

 

[Noopy]

The SL3145 and the SL3127 share the same datasheet.




The SL3145 gives you 5 transistors too. It´s packaged in a smaller package (DIL-14 vs. DIL-16). To reduce the pincount there are two emitters connected for a differential stage and the substrate is connected to these two emitters.






No surprise: The die is basically the same as the die in the SL3127 but the metal layer is different. The SL3145 uses the upper substrate contact, the transistor in the middle of the upper line is shorted and the transistors in the lower right corner are connected together (connections of collector and emitter are quite similar to get a good matching of the differential stage). The internal naming is WH047A (vs. WH047B).


https://www.richis-lab.de/Bipolar80.htm

 

[T3sl4co1l]

The substrate is connected to another emitter, rather.

A likely application for this arrangement (also CA3086, LM3086 etc.) is the substrate-emitter transistor serves as current sink, the common-emitter pair is the diff pair on top of it, and the other two can be used for whatever, maybe cascode or follower at the diff amp collectors.  (Another monolithic pair would be needed for PNP current mirrors, if using that type of design.  Would be handy if they put PNPs into the same package, but alas.)  Gain depends on bias (no external resistance between emitters), so this is useful for mixers and variable-gain (OTA) circuits as well as plain old amplifiers.

Tim

[Noopy]

The substrate is connected to another emitter, rather.

Correct, that was a mistake on my side. Sorry!

A likely application for this arrangement (also CA3086, LM3086 etc.) is the substrate-emitter transistor serves as current sink, the common-emitter pair is the diff pair on top of it, and the other two can be used for whatever, maybe cascode or follower at the diff amp collectors.  (Another monolithic pair would be needed for PNP current mirrors, if using that type of design.  Would be handy if they put PNPs into the same package, but alas.)  Gain depends on bias (no external resistance between emitters), so this is useful for mixers and variable-gain (OTA) circuits as well as plain old amplifiers.

The five transistors are ideal to built mixer stages, "Gilbert-Zelle" in Germany.
https://de.m.wikipedia.org/wiki/Gilbertzelle

[Noopy]

Wait a moment! That is strange:




In the datasheet for the SL3145 the substrate is connected to the two emitters of the differential pair transistors!
(For example here: https://datasheetspdf.com/pdf-file/270949/ZarlinkSemiconductorInc/SL3145/1)
That is odd... 

[T3sl4co1l]

Uh.. that's interesting!

I wonder how many people got burned by that!?  Or maybe they changed it along the way?

Tim

[Noopy]

I found three different datasheets but all of them showed substrate at pin 3.
Would be a strange change during.   

[Noopy]

The datasheet of the SL3145 explains that it can replace the SL3045. The datasheet of the SL3045 shows the substrate connected to T5:
https://www.datasheetarchive.com/pdf/download.php?id=a3525a182d942a45d57edcaa2f3f915488d2b5&type=M&term=SL3045
Nevertheless that doesn´t help developers that start with the SL3145.

[Noopy]

The Gleichrichterwerk Stahnsdorf had built a MJ3001 too.






There is the red potting we often have seen in "eastern" transistors.
In the SU111 they used two different bondwire diameters (https://www.richis-lab.de/Bipolar59.htm). In the MJ3001 they used only one diameter and two of them for the emitter contact.




The die is orientated in a way you are able to bond the wires as good as possible.




The die looks like the die in the SU111 (https://www.richis-lab.de/Bipolar59.htm). It´s quite possible they used the same die here, perhaps a different bin of the SU111. The SU111 can conduct the same current as the Motorola MJ3001 (https://www.richis-lab.de/Bipolar60.htm) but the maximum Vce is a lot higher (400V).




The edge of the die is a little different. In the SU111 the MESA structure left a small strip that is removed nearly completely in the MJ3001. Just in the corners there is a little "tower" left.


https://www.richis-lab.de/Bipolar81.htm

 

[Noopy]

Honeywell 2N1262, a very old Ge power transistor (45V, 3,5A, 32W, 0,2MHz) built by Honeywell.




A 2N1262 ad back in the year 1959 (Electronic Design, December 9, 1959).




Heat spreading with copper. The package is MT-36.




Nice! 




The 2N1262 is hard to disassemble because the contact pins are shells in which the contact wires are squeezed...




...and because of that the transistor broke. 
There is a small rectangle germanium plate contacted with a metal ring which is the base contact. In the middle there is the emitter contact.




The germanium plate sits on a copper socket that is the collector contact and removes heat.






Collector area is bigger than emitter area and the active area is a little thinner than the rest of the plate. Both facts improve the transistor specs.


https://www.richis-lab.de/Bipolar82.htm

 

[salihkanber]

Those looks like true art 👍

[Noopy]

Let´s take a look into a SMY60, a Dual-p-MOSFET built by the Funkwerk Erfurt.
The SMY60 has no protection diodes. To protect the MOSFETs against ESD there is a copper clamp above the package shorting all pins. After soldering the SMY60 into a circuit you can remove the clamp.
K stands for a production in the year 1977.
(-25V/20mA)




The datasheet shows that the bulk contacts of the two MOSFETs are connected but they are not connected to the source contacts.
Although it can conduct twice the current of the SMY51 (https://www.richis-lab.de/FET09.htm) the gate capacity is a little lower: 10pF vs. 12pF. One explanation for the lower gate charge is that the SMY51 contains protection diodes.






The die was protected with some silicone like potting. It is 1,1mm x 0,9mm, a little bigger than the SMY51 but the sizes of the transistors are quite similar. The gate electrodes are a little less overlapping in the SMY60. That could be another small share to the smaller gate capacity. In the upper right area there are four squares to check the manufacturing quality. The 2222 shows the revisions of the masks.

The two MOSFETs are integrated side by side to get them most equal in terms of production and temperature. On the left side we have the bulk connection. This potential is guided around the die and between the two MOSFETs to get some shielding.

You can see the meshed drain and source areas. They are a little unsymmetrical and it seems there are different doping levels too. The datasheet states a Source-Bulk breakdown voltage of -15V and a Drain-Bulk breakdown voltage of -25V. The source area seems to be doped more heavily.






After removing some of the metal layer we can take a closer look at the MOSFET.
The field oxide (FOX) had been removed in the area where the MOSFET was integrated. The substrate is p-doped. The area of source and drain got n-doped. Between source and drain there is the p-doped channel left. The channel width is something around 8-9µm.
Left and right of the MOSFET there are vias to contact source and drain with the metal layer.
The yellowish part is the thin gate oxide (GOX). In this area the potential of the metal layer influences the resistance of the channel. In the areas around the GOX the silicone oxide is a little thicker but still thinner than the FOX.


https://www.richis-lab.de/FET20.htm