Transistors - die pictures

[ocw]

magic, you seem more comfortable in reviewing Id/Vgs curves.
The attachment shows the five curves from each of the two halves of one SD2942 almost overlapping as compared to the five curves from half of another SD2942 not coming close to the first ten.

[magic]

This is kinda offtopic, but no, really not. It doesn't matter how you show it, it's just a philosophical question:

For a FET in the pentode region, is its Vds/Ids characteristic really anything other than just the drain impedance? If it takes different gate voltages to get the desired drain current on two parts, and then their drain impedance is about the same (high enough it's hardly even seen on those plots) is that really a difference in the Vds/Ids characteristic or just Vgs mismatch? That's all I have ever said, and as you see, I really don't  have that much to say

Also, I realize that drain impedance is probably not what you care about as long as it's "high enough", which probably isn't even that very high, as your schematic shows a 700Ω drain load.

If anything, it looks like the mismatched parts have different transconductance too, which might be relevant besides the different Id at given Vgs.

[T3sl4co1l]

In the pentode region, the MOS equivalent of Early effect, is channel length modulation.  Drain conductance (y_oe) is usually very small indeed.

Tim

[Noopy]

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



The package has notches to increase the creepage distance.




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




It´s a perforated Emitter MESA-Transistor just like the new BUX22 (https://www.richis-lab.de/Bipolar08.htm)






 

[exe]

I wonder what is the biggest bjt (in terms of die area) ever produced?

[Noopy]

I have a halfbridge brick sitting here (KD324510). The dies in there are much bigger. Could be something around factor three. I will take some pictures! 

[capt bullshot]

It's not a BJT anyway, I've seen IGBT dice of 11mm * 16mm and larger. You'd find them in modules like this one:



Also have a few BJT modules, but their potting is pitch black, so one can't see the dice.

[duak]

I have some old BJT six-pack modules for a servo drive.  While looking for information on them I ran across the attached paper on the development of similar devices.  Page 2 shows the pre-potted assembly but without device ID or dimensions.  Assuming the overall dimensions are at least 30 mm x 50 mm the largest dice are maybe 10 mm x 10 mm.  If memory serves, these modules are attached with M4 or M5 screws so these dice could be even larger than that.

This fellow built a 3 phase linear amplifier using similar devices: http://wunderkis.de/pwramp3/index.html

I toyed with the idea of making an electronic load with one of modules I have.  I've used one as a battery simulator to test a 50 A battery charger because it was easier than wiring a bunch of power MOSFETs or BJTs in parallel - I just needed a few zener diodes, resistors and wires with ring terminals.  The module was prone to high frequency oscillation when in the linear region at some currents so it seems that the basic devices are reasonably fast.  I think I read somewhere that these are triple diffused epitaxial to minimize switching losses.

These devices remind me of old IC engines with 5 litre cylinders and a 1000 RPM redline.  If memory serves, these run at about 2 kHz.

[capt bullshot]

This fellow built a 3 phase linear amplifier using similar devices: http://wunderkis.de/pwramp3/index.html

That's me, by the way.

I toyed with the idea of making an electronic load with one of modules I have. 

I'd recommend against that:
- they "like" to oscillate
- their DC SOA isn't that great, they can die at way lower power levels than their maximum ratings if used in DC linear mode (I had this once or twice in my amplifier trying to use is as a DC current source). With AC output, even at as low frequencies as 50Hz, the SOA gets better.

[duak]

Ah, the capt himself.  Nice job on the driver and the write up is very clear - thank you!

Yes, the DC SOAs on the modules I have are not so good - a Pdmax of 200 W for a devices with a BVCEO of 500 V and an ICMAX of 100 A, clearly optimized for switching.  Their instability was a bit of a surprise as the servo drive they are from had 10 to 30 cm long wires and bus bars all over the place carrying drive signals and switched currents.  However, because it was a PWM, the devices weren't in the linear regime long enough to encounter trouble from oscillation.  After encountering the oscillation and reading the write up, it's back to the old plan.  Too bad, because the module would be simple to mount on a heat sink and wire up.

[Wolfgang]

Ah, the capt himself.  Nice job on the driver and the write up is very clear - thank you!

Yes, the DC SOAs on the modules I have are not so good - a Pdmax of 200 W for a devices with a BVCEO of 500 V and an ICMAX of 100 A, clearly optimized for switching.  Their instability was a bit of a surprise as the servo drive they are from had 10 to 30 cm long wires and bus bars all over the place carrying drive signals and switched currents.  However, because it was a PWM, the devices weren't in the linear regime long enough to encounter trouble from oscillation.  After encountering the oscillation and reading the write up, it's back to the old plan.  Too bad, because the module be simple to mount on a heat sink and wire up.

Try linear mosfets (IXYS). They are made for electronic loads.

[David Hess]

Try linear mosfets (IXYS). They are made for electronic loads.

The problem is that unless you need maximum power in fewer packages, bipolar transistors just cost less for a given die area and power dissipation is proportional to die area.  So linear MOSFETs come with a double price premium which might even make lateral MOSFETs price competitive.

[T3sl4co1l]

Try linear mosfets (IXYS). They are made for electronic loads.

The problem is that unless you need maximum power in fewer packages, bipolar transistors just cost less for a given die area and power dissipation is proportional to die area.  So linear MOSFETs come with a double price premium which might even make lateral MOSFETs price competitive.

If you can find them...

But all of the above seem to be superseded by SuperJunction types, which regularly give DC SOA curves.  I've tested a few and found them to be accurate.  I'm not sure how they do it -- SJ have higher power density than ever, but they still manage not to runaway.

Tim

[Noopy]

Today I have a 2N2857 HF-Transistor for you.
It´s a quite interesting transistor manufactured by Central Semiconductor.




Central Semiconductor builds obsolete transistors. Originally the 2N2857 was built by Motorola.
This 2N2857 has a datecode 1738! 




It has four pins. The housing is isolated.




To isolate the transistor it was package on the collector pin. 
Thermal resistance is probably a bit higher...






The die is ~350µm*350µm.
There is a second transistor on the die probably to test the die. Interesting... Why didn´t they do the testing with the 2N2857 transistor itself? 


And it glows:



1mA



5mA

 

Breakdown voltage is quite low (around -6V) because of high doping for high switching frequency.


More pictures here::

https://www.richis-lab.de/Bipolar16.htm

 

[exe]

Please excuse my ignorance, where is the second bjt located? Is it two big pads on the left on the close up shot?

[Noopy]

Please excuse my ignorance, where is the second bjt located? Is it two big pads on the left on the close up shot?

Yes it is. The two bigger pads are base and emitter and the smaller at the bottom corner is the collector.
Looking very carefully you can spot the corners of the emitter area around the emitter pad.

[RoGeorge]

I wouldn't expect it to be that smol!   

[David Hess]

There is a second transistor on the die probably to test the die. Interesting... Why didn´t they do the testing with the 2N2857 transistor itself? 

Could the tests be destructive like base-emitter breakdown voltage?

[Noopy]

There is a second transistor on the die probably to test the die. Interesting... Why didn´t they do the testing with the 2N2857 transistor itself? 

Could the tests be destructive like base-emitter breakdown voltage?

Never heard of such testing but that would explain the second transistor...

[exe]

How do they pick and place such small dies? Or even cut them...

Those metal can packages... I have an urge to buy just for the sake of owning it.

PS I have Russian МП-42 (MP-42) somewhere, an old germanium transistor. I wanted to put it into use, but may be I should crack it open and see what's inside . Can't find it atm, must be hiding from me....

[Noopy]

How do they pick and place such small dies? Or even cut them...

Those metal can packages... I have an urge to buy just for the sake of owning it.

PS I have Russian МП-42 (MP-42) somewhere, an old germanium transistor. I wanted to put it into use, but may be I should crack it open and see what's inside . Can't find it atm, must be hiding from me....

And look at these small bond wires you have to place them accurate on the bondpad.
The handling must be pretty tricky.
Fascinating engineering!

The MP-42 is afraid... 

[David Hess]

There is a second transistor on the die probably to test the die. Interesting... Why didn´t they do the testing with the 2N2857 transistor itself? 

Could the tests be destructive like base-emitter breakdown voltage?

Never heard of such testing but that would explain the second transistor...

It occurred to me because RF transistors often have a rated base-emitter breakdown voltage of only 3 volts instead of the more common 5 volts.

[Zoli]

How do they pick and place such small dies? Or even cut them...

Those metal can packages... I have an urge to buy just for the sake of owning it.

PS I have Russian МП-42 (MP-42) somewhere, an old germanium transistor. I wanted to put it into use, but may be I should crack it open and see what's inside . Can't find it atm, must be hiding from me....

Back in my times(30+years ago) I've opened quite a few from the MP38-42/P401-3 series; first, they are all Ge transistors; second, all of the structures are visible with the naked eye(die connection:1.00X0.01 mm Al or similar flat); third , don't blame my memory if there's something totally diferent inside; I just try to remember best of my impressions.

[Noopy]

Back in my times(30+years ago) I've opened quite a few from the MP38-42/P401-3 series; first, they are all Ge transistors; second, all of the structures are visible with the naked eye(die connection:1.00X0.01 mm Al or similar flat); third , don't blame my memory if there's something totally diferent inside; I just try to remember best of my impressions.

Like this one?
https://www.richis-lab.de/Bipolar06.htm
That's more mechanical engineering than electrical engineering. 


Regarding the test-transistor:
Perhaps it was easier to test a transistor with all three terminals on the surface instead of the other transistor with it's collector on the back of the die...

[magic]

These two transistors have the same collector