Transistors - die pictures

[ricmm]

The following jfet was extracted from a cheap 6mm omni no-name electret mic. There was no marking at all in the package, shape like SC-72 from 2SK596S datasheet.
Die size about 275umx275um.

After acid bath:


Scraped with a bamboo stick:


After NaOH etching:

[Noopy]

Different design, same circuit I would say... 

[magic]

Must be something like this:


An ordinary N-ch JFET is created by diffusing or implanting N dopant into a P wafer and using special pattern of more P doping on the surface to cut the N region into source and drain with a narrow channel between them. The P+ implant is electrically connected with the substrate and together they form the gate. A similar three-layer process is used in bipolar transistor fabrication, so a PNP can be easily added on the die and connected as a diode. In the same manner, simple tweaks to ordinary NPN-oriented IC process can add P-ch JFETs - that's how TL072 and similar opamps are made.

The strip looks like a resistor, but it must be some uncommon process because it needs to be almost a gigaohm or so to work properly in electret mic applications.

[Noopy]

The Nexperia PMPB19 is a p-channel MOSFET with a reverse voltage of 20V and a current carrying capacity of 11A. The resistance is typically specified at 17mΩ. The package type is designated DFN2020M-6 or SOT1220-2. The dimensions are only 2,6mm x 2,3mm. The imprints of test probes can be seen on the contacts.




The PMPB19 contains a protective structure between the gate and source.






The dimensions of the die are 1,6mm x 1,0mm. It is therefore larger than the large contact of the package via which the power loss is dissipated. The imprints of seven needle contacts can be seen on the large source surface. Obviously the MOSFET was tested with a high load current. In the package, the source current is transmitted via three bondwires.




In the corner where the gate bondpad is located, there is a small square recess in the source surface. Either this measure is intended to reduce thermomechanical stress or there was a risk that the structuring of the thick metal layer could cause a short circuit in this corner.


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

 

[T3sl4co1l]

Finally, some modern trench.

Doesn't look like any evidence of stripes on top, wonder if that's because they really are narrower than is optically resolvable, or the top metal is so thick as to overshadow it.  Which seems impressive in its own way, metal isn't usually (or doesn't usually seem) that thick!  Might well be both together.

On a related note, I've realized something I think about planar types... particularly for higher voltage ratings, the channel region pinches off at a voltage that corresponds to the presumptive strip width.  That is, the Coss shows a sharp drop around 5 to 30V, then a more relaxed drop rate beyond there.

I was looking at this recently,
https://www.infineon.com/dgdl/irfp4768pbf.pdf?fileId=5546d462533600a40153562c959b2021
With a 2016 date, this is probably the latest greatest in the "HEXFET" lineage (alas, they don't note what generation it is).  Notice the inflection point and faster drop in Coss at 4-6V.  The SOA is also highly compromised at high voltages, illustrating the high power density of this design.  Compare with a much older "large" transistor like,
https://alltransistors.com/adv/pdfview.php?doc=irfps43n50k.pdf&dire=_international_rectifier
which shows the same Coss drop at 15-20V.  I wonder if pitch is proportionate, and the '4768 is fully 4x finer?

I think trench don't show the same pinch-off effect, as the drift region comes right up to the end of the (vertical) channel; but I haven't looked over too many datasheets with this aspect in mind so there may be more about this.  The above part is also rated quite low voltage; the drift region width is probably comparable to pitch, and at a relatively high doping level.  Hmm, a shame it's not rated for avalanche, even single-pulse.

Tim

[Noopy]

 

I assume the metal layer is too thick to see the structure of the "MOSFETs".
I can try to strip the metal layer. It will take some time but let´s see what is possible... ...and visible. 

[Noopy]

The Sanyo 2SC4632 is a bipolar transistor that blocks up to 1200V. The package is fully insulated and has recesses between the pins which increase the creepage distances. Despite its large design, the 2SC4632 is not really a power transistor. The collector current must not exceed 10mA continuously. For short periods 30mA are permissible. Although the thermal resistance is specified as 12,5K/W, the power dissipation must not exceed 2W. These two parameters do not seem to fit together. The amplification factor is given in the datasheet as 10-60. The cutoff frequency is typically 6MHz.

This could be a reconditioned component. The surface of the package is somewhat scratched and the tinning of the pins looks unclean. It is not a fake, as we will see in a moment. The structures of the 2SC4632 are known and can be found in this component.








Compared to the package and the heatspreader it contains, the die is relatively small with an edge length of just 1,7 mm. A large part of this area is used to control the potential distribution so that no problematically high field strengths occur.




The active area itself is just as small as in the low voltage small signal transistors we have already seen a lot.


https://www.richis-lab.de/BipolarA57.htm

 

[exe]

The Sanyo 2SC4632

Ah, so the actual transistor is that tiny region in the middle of the die? What is the rest? Does the rest of the die carry any current?

[Noopy]

Ah, so the actual transistor is that tiny region in the middle of the die? What is the rest? Does the rest of the die carry any current?

Exactly, the active transistor is the tiny thing in the middle.

The circles don´t carry current they are just for a better electric field distribution.
Here you can see some simulations of the electric field in a transistor: https://www.researchgate.net/figure/Color-online-Electric-field-distribution-when-the-electric-field-reaches-E-C-257-MV_fig4_352335818
At low voltages the distribution of the electric field is no bigger problem but above 100V you get a very high electric field strength at the edges of the base area. You can lower the doping to get higher breakdown voltages but that is bad for other parameters. These circles are a better solution. They distribute the electric field over a larger area, lowering the peak at the base collector junction.
They are doing the same in the real high voltage area: https://journals.nipes.org/index.php/njstr/article/download/400/418

[T3sl4co1l]

The Sanyo 2SC4632 is a bipolar transistor that blocks up to 1200V.

If you were wondering what such an oddball device is for, consider dynamic focus in Trinitron etc. sets:
https://web.archive.org/web/20151229031656/http://home.arcor.de/thomasfetzer/Sanjo/SY103A_e.pdf
p.11 (12).  Spittin' image of it

Crude cross-sectional diagrams are included too; fascinating.

I clicked on this document searching about FBET actually, but the detail (however modest) they give on all is quite nice.  (Evidently part of the FBET "innovation" is simply using two metal layers. On a dumb BJT, who'd have thought? )

The doc basically reads as a who's-who of all the special transistors in my final Trinitron monitor.

Tim

[dzseki]

The Sanyo 2SC4632 is a bipolar transistor that blocks up to 1200V.

If you were wondering what such an oddball device is for, consider dynamic focus in Trinitron etc. sets:
https://web.archive.org/web/20151229031656/http://home.arcor.de/thomasfetzer/Sanjo/SY103A_e.pdf
p.11 (12).  Spittin' image of it

Crude cross-sectional diagrams are included too; fascinating.

I clicked on this document searching about FBET actually, but the detail (however modest) they give on all is quite nice.  (Evidently part of the FBET "innovation" is simply using two metal layers. On a dumb BJT, who'd have thought? )

The doc basically reads as a who's-who of all the special transistors in my final Trinitron monitor.

Tim

At one point I was really into modifying high-end CRT projectors (those were intended to display 2500x2000 pixels back in the 90's), for the CRT cathode drive, where high voltage and high speed was also needed, except one, but all other manufacturers used these Sanyo FBET transistors (or video packs based on these transistors).
Thinking back the FBET was really just a marketing slogan, because both Philips and Motorola had better spec'd HV-RF transistors (BFQ232, MRF548 both with 100VCE and >1GHz). In fact these Sanyo transistors died quite easily, whereas the Motorola MRF line seemed to be more bulletproof.

I would have been interested in those TS4159/4160 transistors though...

[Noopy]

The dual-gate MOSFET SM200 blocks up to 20V and conducts up to 30mA. A power loss of up to 300mW can be dissipated via the SOT-103 package. The typical operating frequency is in the VHF range. The transistor can be found in VHF tuners, among other things.

The SM200 was developed at Funkwerk Erfurt as an alternative to the BF900. Production took place in Neuhaus at VEB Mikroelektronik "Anna Seghers". The logo of Funkwerk Erfurt is shown in the centre of this package. There is also V6, which indicates production in June 1987.




The datasheet shows the pin assignment of the SM200. Zener diodes protect the gate electrodes against overvoltage.






The edge length of the die in the SM200 is 0,55 mm. On the upper edge is the designation SM200 and the logo of Funkwerk Erfurt. The numbers 84 could stand for the year 1984. The number 2 could indicate a second revision. However, this is not certain. In the left-hand area, several squares show different process steps. In the lower area there is a 1 and a 2 in different layers.




The source bondpad is located in the outer area and contacts both the substrate and the outer electrode of the actual MOSFET. The two gate electrodes enclose the drain area. The structures that represent the protective diodes of the gate electrodes are located directly on the gate bondpads.




The component shown here is supposedly also an SM200. There is no marking on the front. Only the number 59 is printed on the back.






The edge length of the die is 0,46mm. It is a dual-gate MOSFET with very similar but not the same structures as in the SM200 above. The characters DM31 are shown on the upper edge. Well-informed circles report that DM31 was the designation of a so-called test field on a wafer. In addition to test structures, prototypes were also integrated into these test fields. This is probably an early engineering sample of the SM200.


https://www.richis-lab.de/FET45.htm

 

[Noopy]

Sanyo 2SC4632
[...]
Although the thermal resistance is specified as 12,5K/W, the power dissipation must not exceed 2W. These two parameters do not seem to fit together.

My mistake:
2W is the maximum power dissipation without a heatsink.
With a heatsink the thermal resistance of the package is 12,5K/W.

[David Hess]

Sanyo 2SC4632
[...]
Although the thermal resistance is specified as 12,5K/W, the power dissipation must not exceed 2W. These two parameters do not seem to fit together.

My mistake:
2W is the maximum power dissipation without a heatsink.
With a heatsink the thermal resistance of the package is 12,5K/W.

2 watts is about the limit for any TO-220 packaged part without a heat sink, including regulators.  The plastic encapsulation on an insulated TO-220F package does not make any difference when it is only exposed to air, so it is also 2 watts.

[Noopy]

Sanyo 2SC4632
[...]
Although the thermal resistance is specified as 12,5K/W, the power dissipation must not exceed 2W. These two parameters do not seem to fit together.

My mistake:
2W is the maximum power dissipation without a heatsink.
With a heatsink the thermal resistance of the package is 12,5K/W.

2 watts is about the limit for any TO-220 packaged part without a heat sink, including regulators.  The plastic encapsulation on an insulated TO-220F package does not make any difference when it is only exposed to air, so it is also 2 watts.

Exactly!
I somehow didn´t understand what the datasheet tried to tell me.

[Noopy]

I tried to strip the metal layer of the PMPB19. Unfortunately I wasn´t very successfull:






After half an hour in HCL the surface has changed noticeably. In some places you can see a now floating protective layer, as it remained on the BUX42 after the dissolution of the metallization (https://www.richis-lab.de/Bipolar50.htm). Otherwise, no further details are visible. The orange layer is probably a buffer layer that does not dissolve in HCL. Tungsten or titanium is often used for this buffer layer. It is surprising that the bondwire contacts have remained on the die and do not fall off even while cleaning the die. They appear to be welded to the deeper layers.






The layer remaining on the source surface does not dissolve even after half an hour of treatment with HF. Only the bondpad of the gate potential has dissolved. A dot in this area shows how deep the effective range of the bonding process reaches.

Further treatment with HF for 45 minutes does not change the visual appearance. It remains unclear how the MOSFET is constructed and where the protective structures of the gate are located.


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

 

[Noopy]

The MMBT2369 is a fast switching transistor that was developed by Motorola as a successor to the 2N2369. The MMBT2369 is now also obsolete. Although many transistors with significantly higher cut-off frequencies can be found today, these are not necessarily suitable for use as fast switching transistors. They are usually not operated in saturation mode and are therefore not optimized for this. The switch-off times are often not specified at all. Fast switching transistors are additionally doped with gold atoms. The gold atoms support the recombination of free charge carriers. It is particularly important to reduce the lifetime of the minority charge carriers, which move relatively slowly. A negative side effect of gold doping is a slightly lower current amplification, which is why it is only used when absolutely necessary.

The original datasheet from Motorola lists the MMBT2369 with a maximum blocking voltage of 15V and a current carrying capacity of 500mA. The current amplification factor is between 40 and 120 and the output capacitance is specified with a maximum of 4pF. The switch-on and switch-off times are typically 8ns and 10ns. The newer datasheets from Fairchild and ON Semiconductor contain the same characteristic values, but only allow a continuous collector current of 200mA. The design has probably been revised several times over the years. It is not possible to determine from which generation this transistor originates.






The die in the MMBT2369 has an edge length of 0,31mm. The structures in the top left-hand corner make it possible to check the alignment of the masks. The area of the active area is just 0,07mm x 0,11mm. There are two emitter areas in the base area.


https://www.richis-lab.de/BipolarA58.htm

 

[David Hess]

Although many transistors with significantly higher cut-off frequencies can be found today, these are not necessarily suitable for use as fast switching transistors. They are usually not operated in saturation mode and are therefore not optimized for this.

RF transistors sure are not replacements.  I went down this road several years ago looking for replacements for fast saturated switching transistors and none of the RF transistors that I tested could come even close.  Even with baker clamping, they were not as fast despite Fts several times higher.

The situation is similar to that of transistors with low output capacitance like the KSA1381/KSC3503 which achieve rise and fall times which cannot be matched by conventional devices.  If you need this, then there is no substitute.

Fast saturated switching transistors are available as PNP and NPN.  If you are looking for them, then it helps to include only devices with a Vceo of 15 volts or lower.

My current list of NPN parts is MMBT3646, BSV52, MMBT2369A, and MMBT2369.  For PNPs, it is MMBT5771 and MMBT3640.

[T3sl4co1l]

The MMBT2369 is a fast switching transistor that was developed by Motorola as a successor to the 2N2369. The MMBT2369 is now also obsolete.

Is it?  Or do you mean as the non-A type?  Or by Motorola -- themselves obsolete, lol.  The -A appears current from onsemi: https://www.onsemi.com/products/discrete-power-modules/general-purpose-and-low-vcesat-transistors/MMBT2369A

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.

Tim

[Noopy]

Fast saturated switching transistors are available as PNP and NPN.  If you are looking for them, then it helps to include only devices with a Vceo of 15 volts or lower.

My current list of NPN parts is MMBT3646, BSV52, MMBT2369A, and MMBT2369.  For PNPs, it is MMBT5771 and MMBT3640.

That´s a good hint!
These transistors are really fast but they are not as fast as the MMBT2369. ...but they are coming near...

Is it?  Or do you mean as the non-A type?  Or by Motorola -- themselves obsolete, lol.  The -A appears current from onsemi: https://www.onsemi.com/products/discrete-power-modules/general-purpose-and-low-vcesat-transistors/MMBT2369A

The OnSemi site says "obsolete". 

[Noopy]

MMBT2369AL is still active!

[T3sl4co1l]

Helps if I look at the references I link yes, the -L is doing it!  They don't define what it means; Pb/halogen free, perhaps?  Makes sense.

Tim

[Noopy]

 


I just got the information that this transistor was bought from a distributor not too long ago.
So now I would say that is a ON Semiconductor part.

[iMo]

@Noopy: almost every time I open YT it jumps an evilmonkeyzdesignz  tiktok-like stuff with the die shots on me.
Do you know them?

[Noopy]

I know evilmonkey a little, just his pictures...