06/22/2021 at 23:26 •
One reason we love old equipments/parts/technologies is they are more "graspable", easier to understand. They are not a magic black box and we can "see" them working. That's why we love relays, flip-dots, and other electromechanical devices so much, despite their ridiculously poor performance or efficiency. This is a bit less so for vacuum tubes, though "microphonics" (?) effect can be observed in some conditions, but for transistors, which are usually metal-canned or epoxy-potted, forget it. You can't "see" a circuit working, unless you add LEDs for examples (extra kudos for Tim with his #LCPU - A CPU in LED-Transistor-Logic (LTL)). But still, this is far from a completely direct observation of the circuit's working.
Then I got a batch of OCP71 on eBay.
They are a variation of the vintage OC71, Germanium PNP in glass housing, with 2 modifications to make them useful as photo-sensitive transistors, or even phototransistors : no black paint, and the filler silicon is transparent (unlike the diffusing and opaque sludge of the standard OC71). This makes any circuit using them sensitive to light, so expect a lot of drift and weird bugs in "normal conditions". Already the OC71s' characteristics are altered when exposed under intense light (the black paint can't block everything, particularly infrared).
But the photo-electric effect is reversible ! This means that with the right equipment, one could observe the junctions' emissions when the circuit is working. The new problem is that it is not visible by the naked eye.
For example, https://www.rp-photonics.com/band_gap.html says that Ge is an indirect bandgap semiconductor with an energy of about 0.67eV at ambient temperature, translated to a 1.84 μm wavelength. To visualise the circuit, one would need a camera that is sensitive to the 2µm-1.8µm band. This is in the medium infrared, larger than 1µm (the near infrared that your TV remove uses) but smaller than the far IR, or thermal infrared, that is used by thermal cameras for example. Germanium is in theory active around 95THz or an equivalent black-body of 1600K. So it will not be enough to remove the IR filter of a CMOS camera, or use a temperature imager...
The cameras for 1.8µ are much less easy to source cheaply. Worse, mid-IR is a band with military interests : PbS at 3.34µm was used as the first "heat seeking" sensor in Sidewinder missiles, for example. And the exotic chemistries make such cameras expensive AND probably suspiciously considered.
I would love to have a small, cheap camera to record images of a circuit made of the transparent OCP71 transistors and make videos, or even live demos (to kids and old kids) but sourcing the old glass-germanium PNPs was only the easiest part.
I need your help to find, source and get such a camera, if it exists. I have found some but they are either "out of band" or very expensive ("industrial grade"). And I don't see the point in buying expensive gear that I'm not going to use a lot or regularly.
09/02/2018 at 17:10 •
Part 0 - The Rationale for Early 1950s Transistors
Before you ask, “Why make your own transistors at home??” – read my Manfiesto for Why.
Not Just MOSFETs!
Most of the homebrew community has been focused on fabricating Field Effect Transistors (FETs) at home. Sam Zeloof and Jeri Ellsworth are probably wiser to try and ‘etch FETs’ because they are much better suited to fabricating Integrated Circuits (ICs). ICs were a real revolution in electronics because they miniaturised sometimes enormous circuits into small, convenient packages. Discrete circuits also do not last as long as integrated circuits, because it is expensive to render them mechanically inert. ICs can be completely encased in plastic, shielding their components from dust, and heat, and other kinds of physical mechanical interference.
But, from studying the work of Ellsworth and Zeloof, and following up on their references, it seems that manufacturing the kind of transistors they have, in the ways they have, may still be too expensive and difficult for hackers. So in this article I am going to argue that one option for hackers is to fabricate not silicon, planar process FETs, but germanium, alloy-junction, Bipolar Junction Transistors.
Alloy-junction BJTs are a much older and more primitive type of transistor to fabricate than the planar process transistors that Ellsworth and Zeloof talk about, but I will argue that alloy-junction transistors present themselves as an attractive option for hackers who cannot afford expensive equipment and materials, and who have to push most of the cost of hacking onto using their own labour in order to get things done.
Why Choose Alloy-Junction Transistors
Alloy-junction transistors are not the earliest and most primitive kinds of transistors, but they are one of the earliest and most primitive. These kinds of transistors are necessarily discrete transistors. I argue they present themselves as an attractive kind of transistor to fabricate at home in a DIY, homebrew setting because:
- Their die-size is much larger than your average planar transistor, which means they are far better suited to making in small batches, by hand, one-at-a-time. This Wikipedia Reference explains that the 600mW Mullard OC81 transistor wafer size is 2.4mm x 2.44mm. OC44 and OC45 transistors have a circular wafer size of 1.45mm diameter. The actual transistor die is created by melting pellets of indium or antimony into the very thin germanium wafer. The pellets are relatively simple to make, conceptually. The pellets are also quite large – they are visible individually to the naked eye. See the following image below:
- The temperatures of furnaces required for fabrication are much, much lower, in the order of hundreds of degrees Celsius, and not thousands.
- The techniques of transistor fabrication are much more primitive, and are therefore much more suited to beginners, and require the knowledge of far less complex chemistry.
- Most of their materials seem inexpensive to gather (germanium, indium, antimony).
- Most of the materials for the fabrication of these transistors seem relatively safe to be exposed to in reasonable amounts, with some notable exceptions, like indium. But even indium is a lot safer than the etching fluid Sam Zeloof recommends– Hydrofluoric acid, HF.
- The process of their fabrication does not require photo-lithography. So, making these transistors does not require expensive projection equipment to be fabricated at “small” sizes, and does not require complex proprietary materials in order to etch.
Part 1 - The Actual Manufacture of Alloy-Junction Transistors
Now, I will describe the rough process of how to fabricate this primitive type of transistor.
The Creation of Monocrystalline Wafers
First, a wafer of germanium made of a single crystal is formed. This can be done by yourself with great amounts of heat, or more conveniently, and be obtained online. This is not the most important process to consider when making these transistors.
Dice your germanium wafers into wafers required to manufacture the base of the transistors.
The wafer of germanium forms the alloy-junction in this way. It is sandwiched between two melted pellets of semiconductor alloy:
The Wikipedia reference mentioned above suggests that an “ultrasonic drill” can satisfactorily dice germanium wafers into the right size.
The wafers of germanium then need to be “etched” into the correct thickness in order to satisfy the operating conditions of the transistor. This is not photo-lithographic “etching”, it is the chemical erosion of the diced germanium wafer into a much smaller thickness than previous.
(ndYG: what about recovering the etched germanium to recycle for next time ?...)
Pellets of indium or antinomy need to be fabricated in order to dope the “etched” germanium wafer which is intended to be the base of the transistor. These pellets of semiconductive precious metals form the collector and emitters of the transistor.
The Wikipedia reference above describes the process of forming the small pellets of metal:
Indium wire or strip is cut into portions containing the amount of material required for the pellets. The pellet which forms the Collector is three to five times the size of the one used for the emitter, according to the type of transistor.
The process for shaping or ‘balling up’ the pellets bears some resemblance to that used for making lead shot. The pieces of indium are dropped down a glass tube about three feet high and filled with liquid. At the top the liquid is sufficiently hot to melt the pieces of indium into droplets. Further down the liquid is cooler and the drops of indium solidify into spherical pellets.
Using a jig, you then alloy to the wafer of germanium first the emitter of the transistor, and then the collector. This is the part of the fabrication that requires a furnace that can reach a high temperature.
The temperature of the furnace is much lower than that required for the fabrication of planar process FETs, and the actual stages of heating are also much simpler. The chemistry of the alloying process is also less mission-critical than the chemical process of FET construction.
The furnace temperature is below the melting point of germanium, but above the melting point of the semiconductor alloy – so below 650 degrees Celsius.
Part 2 - Conclusion
I will not discuss the soldering of lead connections to the finished germanium transistor, because they’re not different from any other transistor fabrication process – either Zeloof’s or Ellsworth’s. But, these transistors are much larger and easier to solder than the ICs of Zeloof etc.
Anyway that’s it!
04/20/2018 at 00:00 •
I got a germanium disc from ebay, and I did a quick test by and found that pointing a scalpel to it it works as a point contact diode!
It is a bit rough but it works!
02/23/2018 at 12:26 •
From someone who discussed with the author of "Instruments of Amplification":
"Friedrichs also had tinkered with cuprous oxide transistors.
What he didn't tell in the book: only one of 10 cuprous oxide transistors had worked as intended,OK that also explains why they are not in common use :-D
and they tend to die at 5V+ or so. He wasn't aware of this when testing them with a curve tracer."
But still, 5V is not a crazy working voltage and many circuits use 3.3V.