5

enter image description here

I cracked open this 2N3904 transistor. I lack the means to test its function until tomorrow.

The die doesn't look like anything I've ever seen before. I'm not quite sure where the silicon is.

In your opinion, does the back half of the transistor contain silicon required to function,or is the die contained behind the leads?

The reason for cracking this open was to find whether the construction allows for good thermal conductivity to the leads. The answer to that question is an obvious yes, as there are no bottlenecks inside the transistor package.

I'm curious now about the actual function. I'm about to pry the leads off to see what is underneath.

JRE
  • 67,678
  • 8
  • 104
  • 179
Aaron Butkovich
  • 353
  • 1
  • 7
  • 1
    The die is probably the dark spot in the middle of the large metal slab. Can you zoom in more? – winny May 02 '23 at 10:06
  • 2
    The die is like glass, and shattered when the case was cracked. Very little if any remains where it had been bonded to the pad. What remains is possible a wood's metal-like alloy that bonded the chip to the pad Bonding wires likely are nowhere to be seen in these photos. (Nice slice!), or did the overheated chip do the "cracking"? – glen_geek May 02 '23 at 11:21
  • 1
    @winny I will get a better zoom tomorrow when I have my 12x magnifier. Also will be back to my lab equipment. – Aaron Butkovich May 02 '23 at 11:28
  • @glen_geek My father was a geologist and taught me how to find the fault line and crack stones at a young age. This was just a tiny hammer on the stress line. Anvil was the steel safe. Definitely not a clean break but I got to see the insides. But with regards to the bonding wires... are you saying there should be thinner wires in there that bond to the die? If so, my assumptions about thermals would be incorrect. – Aaron Butkovich May 02 '23 at 11:32
  • Yes, there should be two microscopic gold-color wires leading from the smaller pads to the silicon die. They are likely inside the other plastic half. The plastic can be removed leaving the bond wires intact, but that requires some very corrosive acids. – rdtsc May 02 '23 at 11:49
  • @glen_geek Okay that makes sense. So those gold wires would be the weak point as far as thermals go then? I've often heard the old timers say that the top of TO-92 have the least thermal resistance. This split certainly seems to indicate that the lowest thermal resistance would be on the face and the leads. Not that these packages are good for thermals to begin with. This is all just for educational purposes. And yes, even as a trained biochemist I do not wish to play with hot fuming nitric acid followed by a hydrofluoric acid rinse. Hydrofluoric acid is a truly terrifying substance. – Aaron Butkovich May 02 '23 at 12:02
  • Aaron, the flat side of the package is closest to the chip and also provides a nice flat surface to mate any heat-sinking surface. Bonding wires are so tiny that little heat is conducted out from the chip. The surrounding epoxy is the major thermal path. – glen_geek May 02 '23 at 12:11
  • @glen_geek Makes sense, cool down that crystal plane. The collector lead should offer some good relief if soldered to enough cold copper too. I think it was something like 70°/W from junction vs 200°/W to ambient. – Aaron Butkovich May 02 '23 at 12:36
  • Somewhat off topic I appreciate ... but whas it the background of your photo? – jonathanjo May 02 '23 at 13:06
  • Here is some engineering porn: https://eevblog.com/forum/projects/transistors-die-pictures/ – Sredni Vashtar May 02 '23 at 13:34
  • The package is plastic and amorphous so… stress planes?? There are stress concentrators due to the metal lead frame, and you exploited those. The whole question seems rather confusing… You said you’ll “test the transistor” - after destroying it? These packages take chemical removal of plastic to keep the stuff working, usually. The important bits like bond wires are embedded in plastic. My go to method is blob the pins with epoxy to prevent relative motion, then a nitric acid dip. Gold bond wires will survive. Others may not. Then either probe it or bond new wires. – Kuba hasn't forgotten Monica May 02 '23 at 13:45
  • @Kubahasn'tforgottenMonica Well obviously I wasn't aware of these things prior to this post. But thank you for your information despite the condescension. Fuming nitric acid is just not worth playing with for basically solving a problem of curiosity. I've worked in the chemical industry and the probability of an accident occuring is always nonzero. – Aaron Butkovich May 03 '23 at 07:45

2 Answers2

6

From National Semiconductor "Discrete Semiconductor Products" Databook 1989. The chip itself seems to be missing from OP's photos.

These chips were something like 0.01" thick (if memory serves):
2N3904
Two flying bonding wires were attached to emitter (E) and base (B). No bonding wire was needed for collector, since it was electrically attached to the lead frame underneath. The lead frame mostly appears in the top photo. Silicon chip remnants may be in both photos, shattered and split. The top-surface of the chip shown in the schematic above would be facing down along with bonding wires, in the bottom photo.

The package epoxy conducts much heat to the ambient environment. However, the thermal path to the collector lead is favoured over the emitter and the base leads.

glen_geek
  • 23,591
  • 1
  • 22
  • 50
  • Amazing resource you found there. Thank you for the time. I'm well onto SMD transistors at this stage as the thermal design is just so much better. But this old stuff still fascinates me. I'd love to see how much heat this package could handle given the correct attention. I did notice that the epoxy was quite metallic. There's certainly some kind of aluminum oxide or ferrous component to that case. – Aaron Butkovich May 02 '23 at 12:07
3

The reason for cracking this open was to find whether the construction allows for good thermal conductivity to the leads.

Opening it up is a good idea to see the lead frame, but it’s not terribly hard to measure the thermal resistance to the leads. The transistor itself is a decent PTAT temperature sensor. If you wanted to get more absolute temperature accuracy, it can be calibrated.

A chip-scale package transistor can be well thermally coupled each of to the pins to measure their temperature - individually.

Due to the minuscule size of the lead frame, these transistors aren’t meant to dissipate a lot of power. Some have collector current limits that kick in well before junction temperature is too high. In general, you’ll be very hard pressed to push more than 0.5A continuous at saturation through jellybean TO-92 packaged BJTs.

The lowest impedance path is from the die to one of the pins, so soldering the thing pushed all the way down with a thin shim between the plastic and the PCB, with the pins going into large copper pouts, will get lowest junction-to-ambient resistance. But that’s a lot of effort for what’s trivial to get in a better surface mount package.

You can get a tiny bit better performance when the PCB heat sinking is supplemented by a heatsink tied to the case, but that’s IMHO way in the diminishing returns zone.

A copper lead frame is a must. Some cheap component sources use ferromagnetic lead frames and those should be an immediate reject if you care about thermal performance. They are an order of magnitude worse in conduction than copper.

You can get documentation for the lead frame from smaller scale manufacturers like Central Semiconductor I bet. Haven’t tried it but if you explain what you need it for and buy their products, I don’t see why not.

Kuba hasn't forgotten Monica
  • 32,734
  • 1
  • 38
  • 103
  • 1
    Yeah I'm definitely using SMD parts and have no issues using the smallest packages available. Much more accurate for this application. This was simply curiosity and putting some information out there for others who are still stuck on long-lead components. – Aaron Butkovich May 03 '23 at 07:40
  • 1
    @AaronButkovich Makes sense! I love through-hole designs for their retro look. In higher performance designs it's more work than SMD due to parasitics, turning the ground and power planes into Swiss cheese, and so on. It's an art for sure, and it sure doesn't make things easier. Soldering SMT using a stencil to print paste and then reflowing even with a heat gun is so much quicker and neater than hand-soldering through hole components. And SMT is much nicer when etching your own boards: there isn't so much drilling! – Kuba hasn't forgotten Monica May 03 '23 at 13:39