Transistors - die pictures

[Noopy]

Siliconix 2N5911 (1978)




The 2N5911 contains two sorted J-FETs with one dedicated pin contacting the package.




Hey that´s absolutely the same as the DN1682 (https://www.richis-lab.de/FET06.htm, 1986).
The dies are placed on the flattened gate-pins.






Yeah, that are exactly the same J-FETs.
As described with the DN1682 I assume the numbers are for sorting the dies.


https://www.richis-lab.de/FET10.htm


 

[David Hess]

Ha, the gate connection for a JFET is through the substrate but that is also a point of failure so they added a bond wire in parallel.

[Noopy]

Ha, the gate connection for a JFET is through the substrate but that is also a point of failure so they added a bond wire in parallel.

In addition the bond wire reduces the gate impedance.

[David Hess]

Ha, the gate connection for a JFET is through the substrate but that is also a point of failure so they added a bond wire in parallel.

In addition the bond wire reduces the gate impedance.

The connection through the substrate is much lower impedance than the bond wire, which also goes to the substrate.

The problem is that the substrate connection through the bottom of the die is difficult to make reliably.  There are examples of this in the past from National and Tektronix where the substrate connection became intermittent with temperature but the transistor still worked because of capacitive coupling, except of course when it did not.

[Noopy]

Ha, the gate connection for a JFET is through the substrate but that is also a point of failure so they added a bond wire in parallel.

In addition the bond wire reduces the gate impedance.

The connection through the substrate is much lower impedance than the bond wire, which also goes to the substrate.

The problem is that the substrate connection through the bottom of the die is difficult to make reliably.  There are examples of this in the past from National and Tektronix where the substrate connection became intermittent with temperature but the transistor still worked because of capacitive coupling, except of course when it did not.

Are you sure? I thought silicon has always more resistance than metal? OK, the die is much thicker and shorter than the bondwire but nevertheless...
It's hard to believe they bonded a wire just in case the substrate connection fails. The behaviour would change all the same.

[Noopy]

Finally we have a Dual-Gate-MOSFET: Motorla MFE122




Source is connected to the case.








The innermost metal structure is the drain contact. Around the drain we have gate 2 and gate 1 followed by the source contact.
The gate 1 electrode is smaller than the gate 2 electrode. I doesn´t look like an accident. Perhaps the gate 1 capacitance was lowered a little by using the smaller electrode.
Under the gate bondpads there are additional structures. Probably that are the gate protection diodes.


https://www.richis-lab.de/FET11.htm

 

[David Hess]

Ha, the gate connection for a JFET is through the substrate but that is also a point of failure so they added a bond wire in parallel.

In addition the bond wire reduces the gate impedance.

The connection through the substrate is much lower impedance than the bond wire, which also goes to the substrate.

The problem is that the substrate connection through the bottom of the die is difficult to make reliably.  There are examples of this in the past from National and Tektronix where the substrate connection became intermittent with temperature but the transistor still worked because of capacitive coupling, except of course when it did not.

Are you sure? I thought silicon has always more resistance than metal? OK, the die is much thicker and shorter than the bondwire but nevertheless...
It's hard to believe they bonded a wire just in case the substrate connection fails. The behaviour would change all the same.

For a JFET the substrate connection goes to the gate so any series resistance is inconsequential, unless it becomes very high.  And even when open, many circuits will still work because of capacitive coupling.

[Noopy]

For a JFET the substrate connection goes to the gate so any series resistance is inconsequential, unless it becomes very high.  And even when open, many circuits will still work because of capacitive coupling.

Basically I agree with you. I´m just not sure whether that is unlimited true for matched HF-transistors. At 100MHz there will flow some 10mA over the gate and bad or "high resistance" connections will degrade the matching.

[Noopy]

Sescosem 2N3054




Some white potting on top of the transistor die.




There are some residues but it should be ok...




Yes, it´s a MESA-Transistor.




Nothing special but...






It seems they had problems with the mask alignment. 


https://www.richis-lab.de/Bipolar36.htm

 

[Noopy]

Today a small one:




Ferranti ZTX108C




That´s an interesting contact design. The base electrodes lead around the emitter bond and contact the base area in the lower corners. I assume there is also a contact under the base bondpad. The emitter electrode is kind of U-shaped.
The edge length is 0,37mm. It´s always a lot of fun to find such a small die in the burned remains of the mold. Then you have to clean it and bring it in the front of the camera in the right orientation. If the small thing leaps down the tweezer you often can´t find it again. 


https://www.richis-lab.de/Bipolar37.htm

 

[dzseki]

Sorry I don't want to hijack the thread but it is somewhat relevant and I hope you find this interesting as well.
At work we use some temperature sensor diodes those are suitable measuring between 1.4K-500K temperatures, with reasonable precision (up to +/-0.25K at low temperatures). Needless to say these can be stupidly expensive, so interesting to take them apart

One such example is the Lakeshore DT-670 series:
https://www.lakeshore.com/products/categories/overview/temperature-products/cryogenic-temperature-sensors/dt-670-silicon-diodes

One can find these in many flavours, even in 'bare die' configuration.

Here you can see this is in fact a transistor's C-B diode that is used for temperature sensing. It would be nice to find out which transistor could this be -roughly. It is possible that this is somewhat a custom part made by 3rd party vendor, but then why would they hassle with transistor structure when only a diode is needed?
Obviously the very precise pieces are heavily binned so I don't think there is a free meal here, but still it would be nice to get a rough idea on the source.
Here is a self made shot too of such die (not as nice as you are use to it, sorry):

[Noopy]

That´s an interesting story.

Here you can see this is in fact a transistor's C-B diode that is used for temperature sensing. It would be nice to find out which transistor could this be -roughly. It is possible that this is somewhat a custom part made by 3rd party vendor, but then why would they hassle with transistor structure when only a diode is needed?
Obviously the very precise pieces are heavily binned so I don't think there is a free meal here, but still it would be nice to get a rough idea on the source.

Well I´m definitely no expert for low temperature meassurements but I agree with you. Probably that´s a "normal" transistor. Of course the transistor is chosen so it is well suited for the temperature meassurement and I also assume they did a lot of binning but probably it´s a "normal" transistor.

Up to now I haven´t seen such a structure... 

[SilverSolder]

It seems a surprisingly big transistor for temperature measurement, wonder what drove that decision?

And why the C-B diode instead of the B-E one?

[dzseki]

Well, I have seen similar structure on your pictures! On your OPA676 die pictures.

I only assume that the C-B diode is used because the Collector itself represents the die, so if there is some gradient (and at low temperatures there is) the the bigger sensing area might give more sensible reading. To reach cryogenic temperatures from room temperature takes usualy at least a hour or so, so fast response time is less of a concern.

[Noopy]

Hm... The transistors look related but on this level all transistors look related... I wouldn´t state these are brothers... 

[Noopy]

Hitachi 2SK1317, breakdown voltage 1500V, 2,5A continuously, 7,5A peak.




Well I have taken better pictures... The package is quite tough and the die has a polyimid like coating.
The edge length is 5,4mm.




We have the usual potential steering but less rings than in the 1000V-BUK466 (https://www.richis-lab.de/FET07.htm).




There is quite some free space between the transistors.


https://www.richis-lab.de/FET12.htm

 

[exe]

Could you please remind me what are those rings around the transistor?

[Noopy]

Could you please remind me what are those rings around the transistor?

That is field steering to get a more uniform electrical field and in consequence less isolation stress. 

[Noopy]

Philips MPSA56, a pnp transistor.
They marked emitter, base and collector, nice! 




0,6mm x 0,6mm, quite a big die for a small signal transistor. Well, it can conduct 500mA, 1Apeak. 


https://www.richis-lab.de/Bipolar38.htm

 

[David Hess]

Philips MPSA56, a pnp transistor.

On Semiconductor makes it also and isn't MPS or MPSA a Motorola prefix?

It is so similar to the 2N4403 that I wonder why it even exists.  It is slightly slower but slightly higher voltage so I wonder if it is just a graded 2N4403.

[Noopy]

Philips MPSA56, a pnp transistor.

On Semiconductor makes it also and isn't MPS or MPSA a Motorola prefix?

It is so similar to the 2N4403 that I wonder why it even exists.  It is slightly slower but slightly higher voltage so I wonder if it is just a graded 2N4403.

There are also MPSA transistors built by Philips. The one who gave me these transistors stated it is a Philips... 

[exe]

It is so similar to the 2N4403 that I wonder why it even exists.

Can it be a "European" equivalent of 2n4403? Like 2n3904 and bc548. Anyway, I'm pretty sure there are more part numbers than design variations. I bought BD135G, BD137G and BD139G (don't remember the vendor, there are several companies producing supplying them) and measured them with transistor tester and with siggen for gain vs freq. They all had very close gain and base drop voltage, which I interpret that they are the same design.

[Noopy]

Here we have a complementary pair BC550C / B560C built by Philips:








The BC550C-die is 0,26mm x 0,27mm. Luckily the die sticked on the leadframe so it was easier to put it in front of the camera.








The BC560C die is bigger than the NPN BC550C: 0,32mm x 0,32mm. That´s plausible because PNP transistors in principle have worse specifications than NPN transistors. Well, we don´t know if these two transistors are different generations but the different size is plausible. You also can see different collector capacities in the datasheets: BC550C: 1,5pF / BC560C: 4pF

https://www.richis-lab.de/Bipolar39.htm

 

[exe]

For some reason I like small bjts. They are small, fast and cheap. And if you need more power just use them in parallel (I'm kidding).

[David Hess]

There are also MPSA transistors built by Philips. The one how gave me these transistors stated it is a Philips... 

I know others make MPS and MPSA prefix transistors.  I just thought that prefix originated with Motorola and others are second sources.

Can it be a "European" equivalent of 2n4403? Like 2n3904 and bc548. Anyway, I'm pretty sure there are more part numbers than design variations. I bought BD135G, BD137G and BD139G (don't remember the vendor, there are several companies producing supplying them) and measured them with transistor tester and with siggen for gain vs freq. They all had very close gain and base drop voltage, which I interpret that they are the same design.

I do not think so.  As far as I know, the BC series may be similar in function and application but are of a completely different origin of design, at least originally, then the 2N series.  The specifications for the 2N3904 and BC548 do not match at all.