Transistors - die pictures

[T3sl4co1l]

Why nobody put a photodiode inside to protect expensive power transistors?

But they did!  C-B junction generates a small photocurrent; Widlar famously employed this trick to generate a few mV / uA negative say for biasing a single-supply op-amp's output so that it can go all the way through and below zero.

Tim

[Wolfgang]

... Widlar used the trick to protect *power* transistors ?

I heard the story but never seen a practical Op amp that uses the effect. Curious !

[T3sl4co1l]

Not to protect, of course, just that it's possible in a sense to build such a circuit around a normal device.  Obviously if you're using C-B as a photodiode, you can't very well also use it for delivering power to a load.

Interesting consequence: when C-B is reverse biased (as normal) and E-B is avalanched, the same photocurrent flows C-B, which means C leakage increases.  Need low leakage?  Clamp that base voltage, say with a diode, or a zener/TVS with rating somewhat below Vebo.  Simple enough.

Tim

[Wolfgang]

... tried this with a 2N2222A. When you set VCE=5V and then you let BE break down, collector current does not move.

[T3sl4co1l]

What magnitude?  Ic should be at least ~nA to start with; I would guess uA is reasonable to expect here?

If it doesn't actually go up, that's quite interesting.  Do you measure any negative voltage when it's open?

Tim

[Wolfgang]

Keysight B2962 SMU starting with 10nA, IIRC.
Maybe you need more collector voltage than the 5V I had ?

IIRC, I did not measure anything negative, but I can recheck next week when I'm back in the lab.

[T3sl4co1l]

Hmm. Offhand, here's a... 2N2102, nice TO-39 can.  Setup:
9.76V supply
100 ohm series resistor to E
B = GND
C = open

E measures 7.54V (rising slowly as it warms up).
C measures -380mV, also rising slowly (i.e. towards zero) as it warms up.
C shorted, I measure -4uA.  Hey, not bad!

Presumably, leakage shorts out the photocurrent (or free charge current, whichever it is, I forget exactly) as it heats up, so the efficiency of this mode drops quickly, much as solar panels do.

With a 10k pullup from +9.76V to C, I measure 9.71V... or more precisely a drop of 38.8mV.  So, 3.88uA, consistent with the shorted measurement; seemingly less, but it's only a 5% resistor.

Collector current drops to -0.01mV (over 10k) when the emitter is open-circuited.  (Meter is fluctuating between -0.02mV and 0.00 when shorted.  This is the Hi-Z range, no loading on the circuit.  Not that it would matter out of 10k.)

With a 1M pullup, it still measures +/- 0.01mV.  Dang, this must be a nice transistor.  (Brand name Central Semi, yay?)

With a 10M pullup and a 22nF bypass cap in parallel with the resistor (just in case there's rectification here?), it's reading 0.14mV, a whopping -- 14pA?  Fuck me, that's damn good for a BJT, especially this size?

And, with the transistor removed from circuit, it's 11mV drop.  So, plus or minus a lot of leakage through the breadboard, or the meter itself.  (Meter with just the 10M and no ground or supply connection reads -0.07mV.)

Gosh... I slide my be-socked foot across the wood floor and the measurement goes nuts...   (Meter is just a BM235.)

Whelp... I'm sure you'll have much more noticeable results with a big fat power transistor, especially a sloppy one like 2N3055 (depending on age of the specimen..), or at higher voltages (9V is a far cry from the 120V rating of the 2N2102).

Tim

[jaromir]

Regarding the glow of reverse biased PN junction of transistors or zeners, you state here https://www.richis-lab.de/REF03.htm

Während die Z-Diode leitet arbeitet sie zumindest zum Teil im Lawinendurchbruch. Wie bei den Versuchen mit den 2N3055-Transistoren ist dabei im Bereich der Sperrschicht ein Leuchten zu erkennen. Rekombinieren Ladungsträger in einem Siliziumhalbleiter, so emittieren sie üblicherweise kein Licht im sichbaren Bereich. Bei einem Lawinendurchbruch erfolgen allerdings relativ unkontrollierte Ionisierungen im Kristallgitter, die unter anderem auch sichtbares Licht erzeugen.

My German is rather poor, so I used google translator

While the Zener diode is conducting, it works at least in part in the avalanche breakdown. As in the experiments with the 2N3055 transistors, a glow can be seen in the area of the junction. If charge carriers recombine in a silicon semiconductor, they usually do not emit light in the visible range. In the event of an avalanche breakdown, however, relatively uncontrolled ionizations occur in the crystal lattice, which among other things also generate visible light.

I'm not sure whether this was debated here, but indeed the principle behind the glow is a bit peculiar one. I found relevant part of a book "Handbook of Silicon Photonics" here
https://books.google.sk/books?id=6zjNBQAAQBAJ&pg=PA347&lpg=PA347#v=onepage&q&f=false
In a case the link above becomes dead, attached is excerpt with relevant part.

[David Hess]

What magnitude?  Ic should be at least ~nA to start with; I would guess uA is reasonable to expect here?

10s of microamps is feasible with small signal devices.  a 4N25 used as a photovoltaic source is 10 times more efficient.

[Noopy]

I'm not sure whether this was debated here, but indeed the principle behind the glow is a bit peculiar one. I found relevant part of a book "Handbook of Silicon Photonics" here
https://books.google.sk/books?id=6zjNBQAAQBAJ&pg=PA347&lpg=PA347#v=onepage&q&f=false
In a case the link above becomes dead, attached is excerpt with relevant part.

Very interesting! 

I´m not sure if I understood the text 100%:
With a STM they were able to put a light in a silicon wafer. But they needed an energy of at least 3,2V.
In my view that is no "normal recombination", is it?
My understanding is that higher energies (=>avalanche) are necessary to get a glowing pn-junction. Normal recombination (normal conducting diode) can only generate heat, no light.

[T3sl4co1l]

Yes, something like that.  High voltages also generally throw around charge carriers, and this can be used to inject charge through insulating barriers.  Which are just semiconductors with higher band gap than the base material, hence non-conductive at room temperature, or even elevated temperature.  But given charges with sufficient energy (or high enough temperature, same thing), well, they'll merrily grab a conduction band state and pass through.

Hence EEPROM and Flash, which is programmed by dumping a relatively high voltage through the channel, spraying some charge into the floating gate.  (Though I happen to forget how and why it's also electrically erasable.)  I suppose presumably you could see an extremely small amount of light emitted from such a chip as it's being erased or written, though as we're talking very small transistors and microamperes at best, probably not much.

And then yeah, they talk about, what, nanocrystals I suppose?  The band structure, rather than being an effectively-continuous band in a bulk material, it takes on recognizable discrete levels (effectively as many allowed levels as there are atoms along a given axis, I think?), and evidently some of these levels happen to not only correspond to lower visible wavelengths (red) but emissive states as well.

There's also something about implanting dyes or phosphors or other semiconductors (as nanodots) in silicon, that act as direct bandgap recombination centers, so there's just some free charges in the silicon that happens to diffuses over to these sites and emit light.

IIRC, something like that is what gives us modern high efficiency green LEDs, which are InGaN, normally a blue substrate (and still sporting the 3.0V drop you'd expect from it) but made to emit green instead.  These LEDs are recognizable not only from their much brighter output, but a... very slightly cyan-ier rather than yellower/lime-greenier hue?  Made a board some years ago that had a mixture of both on it for status LEDs, wasn't the most consistent appearance...

Tim

[Noopy]

Today I can show you a very cheap TO3-transistor, a 3DD15D:

https://www.richis-lab.de/Bipolar05.htm




As I said: cheap…
Anyway the datasheet seems to contain some truth: Ptot=50W, Rth=2°C/W





But I´m not sure how they built this transistor. There are more areas than I would have suspected (red and blue)… 

[Jay_Diddy_B]

Hi,

If you have access to a metal lathe, you need to build this:





It puts the T03 can on the axis of the lathe and you can turn the top off.



Regards,
Jay_Diddy_B

[Noopy]

If you have access to a metal lathe, you need to build this:
...

A good idea! 
Unfortunatelly I have no access to a metal workshop...

[Wolfgang]

I have one ! I will try this out.

[Noopy]

Finally a diode is half a transistor? 

Today I have an old BAV45 for you:

https://richis-lab.de/Diode01.htm


Not extremely interesting:

[Noopy]

Today I have the breakdown of a SS109 for you:

https://www.richis-lab.de/Bipolar01.htm







 

[RoGeorge]

Wow!   

Is that ring of light visible to the eye, or is it infra red visible only by the camera?

[Noopy]

The light is visible to the naked eye. It's dim and small but you can see it.
Usually silicon pn-junctions don't emit visible light but in avalanche breakdown the electrons are lifted to some higher energy levels.

[Noopy]

Today I have a Siemens ASY25 for you, it´s an old small signal alloy transistor:





 
I assume they use this orange slurry to transfer heat from the transistor to the housing.






Germanium, yeah! 


More Pictures on my website:

https://www.richis-lab.de/Bipolar06.htm

 

[RoGeorge]

It's interesting that thermal paste is orange.  Is that orange because with time it developed rust inside?

When I was a kid I opened a few Romanian Ge transistors (AC180, AC181, EFT323, ASZ15, etc.) in order to turn them into photo-diodes (or photo-transistors).  Also we use to never throw away the broken high power transistors.  Those were usually filled with thermal paste, and we use to harvest the thermal paste and use it later on radiators.

Big or small, all the transistors I opened have had milky white thermal paste.

Since we are talking about die pics, recently bumped into a small power Ge that I opened during the 70's or 80's, in order to build an "Electronic Eye" from a book.  Back then the image sensor chips were not yet invented, so the so called "electronic eye" circuit was in fact just a light detector with a relay.   



From left to right, the EBC of a PNP Ge transistor (EFT323 or AC180?).  E wire to junction EB was cut after opening the transistor



View from the emitter side, E wire to the middle blob was cut, on the most left of the picture is the B wire



View from the collector side, still shiny but heavily contaminated and with big residual reverse I_CB, the DMM for the BC junction shows 1.3V when reverse polarized, and about 0.13V when direct polarized.

It has been staying in open air for decades, now rusty and with an almost dead BC junction, but it still is sensitive to light.  Stubborn! 

[Noopy]

I didn't find any rust and the orange colour was very uniform. In my view the paste was orange from the start. Perhaps Siemens had a different thermal paste? But I agree with you, orange isn't normal.

Thank you for the pictures and the story behind!

[T3sl4co1l]

I would assume it's condensation, polymerization or decomposition.  Perhaps it used to be an oil or resin.

Tim

[Noopy]

Today I proudly present a BUX22:





But    is it a fake?



No, it´s an old BUX22! 
Why should anyone put a new cap on an old BUX22?
Perhaps ST did some requalification and sold old parts as new?



300V breakdown voltage
50A peak current
8A base current
 

But really interesting is this one:



The BUX22 has some defects at which the glowing of avalanche breakdown occurs first.
The defects are big enough to identify that the glowing occurs next to the defect not exactly at the defect itself. That´s important for interpreting the glowing in the LTZ1000, discussed here:

https://www.eevblog.com/forum/metrology/ultra-precision-reference-ltz1000/msg3086013/#msg3086013


Whole story and much more pictures here:

https://richis-lab.de/Bipolar07.htm

 

[Noopy]

@Wolfgang: I guess these pictures are also useful for your work.