Transistors - die pictures

[Wolfgang]

Hm... Somehow I managed to kill the 2N2857...
Unfortunatelly I can´t say what went wrong. I was in a hurry. That´s never a good thing.   

I guess the poor thing needed some limiting for current and power and did not get any
Seriously, these parts are not forgiving at all. When you look how they are mounted (on a pin header)
you can imagine that Rth is very large and heat capacitance is very small, i.e. they blow up in milliseconds.
Remedy: use a SMU and set the limits small enough. Fortunaterly, they are cheap.

[Noopy]

That´s not "medium" it´s more "well done". 

@Wolfgang: SMU and cheap are two words that don´t match. 
I still haven´t set up my HP-supply... 

[David Hess]

Somewhere I have seen a table with bond wire fuse ratings for use in failure analysis.

[Wolfgang]

That´s not "medium" it´s more "well done". 

@Wolfgang: SMU and cheap are two words that don´t match. 
I still haven´t set up my HP-supply... 

Sorry I meant that the transistors are cheap. An SMU is not, and I know that after Keysight got all my money
Something affordable you could make is a current limiter circuit (I think there was a Jim Williams app note about an electronic fuse).
Thats not rocket science and will prevent a few kills.

[Noopy]

 

Up to now I use a cheap bench supply but it has a current limiter of course.
I should have worked more concentrated. The fault is mine....

[Wolfgang]

Using a normal PSU with sensitive semiconductors is a problem even with a current limiter.
The problem is the energy stored in the output cap and the slow response of the limiter.
What you need is a limiter/electronic fuse that is fast and has no big electrolytic at the output side.

[Noopy]

Yeah, I agree with that. A SMU would definitely be a nice piece of equipment.

[RoGeorge]

I proudly did once a demonstration about my latest tool purchase, a brand new Rigol DP832 power source, and how great it is, and how it can serve as a voltage source, or as a current source, upon wish.

And to prove my point, I set the current to 20mA and confidently hooked up a LED to the wires.   

The LED flashed then instantly died with a violent pop sound, I jumped back, and my friend burst into laughter. 

Those 20mA were enough to charge a few thousands uF of output capacitors up to 32V, then all the energy rushed into the LED, limited only by the ESR and the wires.  In DP832, the output filtering capacitors are located at the end side, in direct contact with the output terminals. 


Regarding the melted bonding wires, they are terminated in intriguingly round spheres of metal, and also there is a segment attached to each sphere, segment that looks like it was melted too but only at its surface, and somehow kept its initial cylindrical shape and its initial diameter, then the rest of the wire suddenly looks like brand new.

How is that possible? 

[SilverSolder]

I proudly did once a demonstration about my latest tool purchase, a brand new Rigol DP832 power source, and how great it is, and how it can serve as a voltage source, or as a current source, upon wish.

And to prove my point, I set the current to 20mA and confidently hooked up a LED to the wires.   

The LED flashed then instantly died with a violent pop sound, I jumped back, and my friend burst into laughter. 

Those 20mA were enough to charge a few thousands uF of output capacitors up to 32V, then all the energy rushed into the LED, limited only by the ESR and the wires.  In DP832, the output filtering capacitors are located at the end side, in direct contact with the output terminals. 

[...]

That's what happens when you use an instrument that is too new and inexperienced! 

[Mecanix]

Those 20mA were enough to charge a few thousands uF of output capacitors up to 32V, then all the energy rushed into the LED, limited only by the ESR and the wires.  In DP832, the output filtering capacitors are located at the end side, in direct contact with the output terminals. 

Interesting. Mine (DP832) ramps up beautifully, no inrush or spikes whatsoever on all 3CH, smooth as silk and stable. Recently had it calibrated also so add in that now impressive accuracy, its incredible now.

[Wolfgang]

Those 20mA were enough to charge a few thousands uF of output capacitors up to 32V, then all the energy rushed into the LED, limited only by the ESR and the wires.  In DP832, the output filtering capacitors are located at the end side, in direct contact with the output terminals. 

Interesting. Mine (DP832) ramps up beautifully, no inrush or spikes whatsoever on all 3CH, smooth as silk and stable. Recently had it calibrated also so add in that now impressive accuracy, its incredible now.

I have a lot of DP832s, and they perform fine. Three things to take care:

- The startup spike only occurs when the load is *very* light.

- Its not an SMU, so accuracy is what is in the datasheet. Last digit has to taken with a larger grain of salt, as usual

- They do have an output cap, and the energy in this cap gets dumped into your load with no current limit being able to do anything against that.
So, if you set voltage to 30V, current to 10mA and then connect your LED, it could still be dead because of the cap discharge current.

[RoGeorge]

I apologize about telling the offtopic tale with the DP832.  I didn't want to hijack the topic.  Should have kept my mouth shut.  It was not the power source's fault, it was my mistake.  Just let it be.  Please let's get back on topic.

-----------------

Anybody knows more about the 2 spheres formed at the end of the melted bonded wires from the last transistor's pics?  Are they usually so round and alike looking?

[capt bullshot]

If you don't have an SMU, a bunch simple resistors should do the job.
Use the lab supply in voltage mode, roughly calculate the required resistance and put the resistor in series to the DUT. Measure the actural current with your multimeter and adjust the output voltage to set the desired current.
Disadvantage: for covering a larger range of currents, one has to switch resistors.
Your typical old-style curve tracer (e.g. Tek 576) does it this way. Pretty simple and effective. The modern ones have multiple digitally controlled SMUs to achieve the same results.

[Noopy]

Anybody knows more about the 2 spheres formed at the end of the melted bonded wires from the last transistor's pics?  Are they usually so round and alike looking?

Well the surface tension of the liquid metal forms a ball...

If you don't have an SMU, a bunch simple resistors should do the job.
Use the lab supply in voltage mode, roughly calculate the required resistance and put the resistor in series to the DUT. Measure the actural current with your multimeter and adjust the output voltage to set the desired current.
Disadvantage: for covering a larger range of currents, one has to switch resistors.
Your typical old-style curve tracer (e.g. Tek 576) does it this way. Pretty simple and effective. The modern ones have multiple digitally controlled SMUs to achieve the same results.

 
Usually my bench supply is good enough for most applications.
I just have to keep calm while meassuring.
I was in a hurry that´s never good... 

[Noopy]

Today I have a MJL21193 for you (250V/16A/30A):

[...]



The die is quite big: 3,64mm x 3,54mm

I was wrong with the die size! 
In fact the die is 5,48mm x 5,36mm! That´s quite big, 5,5 times the ST TIP3055! 

 

[Noopy]

Today I have a new 2N3055 for you, a Motorola 2N3055:

http://www.richis-lab.de/2N3055_07.htm




The characteristic old package with the characteristic low, more round cap.




I don´t think that round low thing is a heatspreader. Perhaps they needed this small plate to manufacture the transistor. 




The design is "quite modern" and uses a mesa structure.


 




11V 10mA




20mA




30mA




40mA




50mA




100mA




200mA




300mA

 

[ocw]

After prior pictures of a good SD2942W with more zoom https://www.eevblog.com/forum/projects/transistors-die-pictures/msg3138636/#msg3138636 , attached is a picture of the same SD2942W with less zoom along with one damaged by lightning.
[ Specified attachment is not available ]

[Jay_Diddy_B]

Hi,

You need to find a big transistor like this one:




I have included a TO-3 transistor in the picture for scale.

The BJT is rated at 200A 700V.

I am not ready to open this one up. The silicon will probably be 33 or 38mm in diameter.

Regards,
Jay_Diddy_B

[Noopy]

Well that's really a big one!

[Noopy]

Today I have a really old 2N3055:







Here you can see the protective coating of the die and the higher emitter since it is a hometaxial transistor.





Of course  I did some breakdown pictures. The fabrication process didn´t create very smooth structures. You can see that the breakdown light is quite non uniform even at high currents:




0,1A




0,2A




0,3A




0,4A




0,5A




0,6A




0,7A




0,8A


And there is a small structural flaw:







https://www.richis-lab.de/2N3055_08.htm

[SilverSolder]

Interesting with that old 2N3055, it seems technology has moved on and we are making much "cleaner" devices nowadays?

[Noopy]

That is for sure. 

[Noopy]

Finally we can compare the hometaxial RCA 2N3055 with the epitaxial RCA 2N3055:










The metal layer has an interesting structure. It seems that RCA had to modify the layer to do some current steering. You can see that while driving the base-emitter-junction in breakdown:




There are two very bright areas. There is no small bright point as we have seen with impurities. It´s more a line getting constantly brighter and dimmer. For sure that is a local higher current density.




In the bottom left corner of the base contact RCA has integrated two holes in the emitter area to increase the resistance and reduce the current density.




In the bottom right corner of the die RCA added even a small electrode connected to the base potential over the silicon.


https://www.richis-lab.de/2N3055_09.htm


 

[David Hess]

Interesting with that old 2N3055, it seems technology has moved on and we are making much "cleaner" devices nowadays?

Early processing was incredibly crude by modern standards.  Today you could do better in your garage.

[magic]

Maybe almost better

https://hackaday.com/2010/05/13/transistor-fabrication-so-simple-a-child-can-do-it/