EEVblog Electronics Community Forum
Electronics => Beginners => Topic started by: codingwithethanol on July 25, 2019, 09:41:10 pm
-
Hi, I've prepared a image detailing my query, but long story short, I would like to know if it would be a feasible idea to manufacture different sized pieces of pure p or n semiconductors for educational purposes. Said pieces could be enclosed in IC-like packages and breadboarded to create transistors, diodes, thyristors, etc. Would this work, and if not, why?
-
No. The PN junction is created only when the P and N areas are in "direct contact" (with no impurities in between).
-
But wouldnt a short length of wire create a path of neglible resistance? Shouldnt there still be a force experienced by electrons in a chunk of n type, since there is positive charge just across the wire?
edit: Ok I thought about it and I suppose the electrons in the wire itself would pose an issue, couldnt this still be solved via clever engineering?
-
This was something I didn't get, way back in the early days. You see diagrams with squares pushed together, and no one explains what that abstraction is really derived from; because most of them don't even know. Worse still, sometimes you see (internal diagrams of) ICs with back-to-back diodes, being used as diodes, and sometimes as BJTs, and vice versa; so which is it?!
Well, as it turns out, all the magic happens by pushing blocks together, yes, BUT -- the magic only happens over a very short distance, less than 10 micrometers or thereabouts.
So when you see an NPN stack structure, they're really showing a very highly magnified core sample through the device -- what's really made is a stack of pancakes, very thin and relatively wide. They're also curled up at the edges, because it's a planar process, everything applied to the top of a crystal. Or, uh... to take the pancake analogy a little too far, it's more like a large block of cake, with a sequence of different syrups absorbed into the surface, giving different characteristics at different depths.
More specifically, junctions have to do with quantum energy bands and electron/hole flows; it only works within a material that is 99.999%+ the same stuff (i.e., high purity silicon in the typical case), with only extremely nuanced changes distinguishing those layers. If you stick in a piece of metal, well metals have completely different electronic structures (they're literally flooded with free electrons, so holes are recombined instantly on contact, and there's an unlimited supply of free electrons). So the semiconductor physics stops at the door, so to speak; you can't put two blocks of semiconductor on either side of any piece of metal, and maintain those weird semiconducting current flows between the two, the metal just destroys that behavior inbetween.
So, in short, that's why you can't wire two diodes together and make a transistor -- or, to such extent as you can even call it that, it's one with hFE < 10^-10, as far as you'll be able to measure. It's a bit disingenuous to describe a transistor as "two diodes stuck together", and this is why it's not quite that simple.
Hope this helps. :)
Tim
-
You might search youtube for "jerri ellsworth homemade transistor" to see how she did it and what's involved.
-
@T3sl4co1l
Thank you for the detailed reply, is there a free resource which goes into detail on semiconductor fabrication?
@ledtester
Jeri was actually one of the first people I ever subscribed to on youtube, I should probably go back and rewatch her diy fab vids. Thank you for the reminder.
-
It's a bit disingenuous to describe a transistor as "two diodes stuck together", and this is why it's not quite that simple.
I had a boss once who was convinced that was how transistors worked. I had a design with an NPN transistor operating as a saturated switch in the ground leg of an amplifier, the whole circuit running off a 1.2V NiCd cell. My boss insisted that it wouldn't work, since the transistor "was two diodes back to back" and it was impossible to get the collector-emitter voltage under 0.6 volts. I tried to explain, and he didn't get it. I wired up a transistor with base and collector resistors and a power supply, showing him how we could drive the collector voltage to well under 0.1V. He claimed that I was "pulling the ground up". Somehow I got him to accept the design, but he often changed other designs behind my back for truly stupid reasons. I eventually quit. In his defense, he was pretty good with vacuum tubes (this was in the 1970s).
-
You can't wire pieces of doped semiconductor together and get a transistor for roughly the reason you can't wire two vacuum tube diodes together back to back and get a triode. A triode only works because the middle grid electrode is a mesh and electrons can fly right past it and on to the anode. Likewise a NPN transistor only works because the base region is thin enough and free enough of scattering sites that electrons injected from the emitter can make it through the P type base (normally dominated by hole conduction) safely to the collector without recombining. Not only would a wire be too long, metals have no bandgap between "valence" and "conduction" states so there is no separation of holes and electrons and effectively have a recombination time of zero.
-
Hi,
You can make your own "point contact" diode.
Get a rusty piece of steel or iron, position a needle so that it just barely touches the rust on the metal.
You've got a device that acts to some extent as a diode.
You have to play around with the pressure applied to the needle point as it touches the surface and of course the needle metal has to be electrically isolated from the rusty metal. The rusty metal forms one terminal, the needle the other.
Rust is like iron oxide.
You might experiment by using two needles with points very closely spaced to see if you can get any amplification.
-
Rhere seems to be a common misconception that a transistor is the same as two back to back diodes, however that is not exactly the case.
To be precise the two junctions (or rather their status) determines what region the BJT is in, whether is of, saturation or active. The current gain and thus the whole BJt’s characteristic are given by the transistor effect.
That is to say the electrons (or holes) that diffuse through the forward biased junction are than swept away from the base towards by the electric field of the reverse biased junction that sweeps them away to the other terminal. For this effect to work as T3sl4co1l said the base region needs to be thin (unless we are talking about high voltage power BJTs that are almost nonexistent today) and atomically pure, apart from the soaping) otherwise the band structure gets mangled up and you get no transistor at all
-
Thanks to everyone in the thread for sharing their knowledge and expertise, I appreciate it.
@MrAl
I might actually try this once i get through my current backlog of unfinished projects, why would rust exhibit this effect?
-
I also wondered for years if I can make a bjt from two diodes. For me the key to understand how transitors work is "depletion region". I'm still not sure if I got correctly all the details, but at least now I don't try to build bjts from diodes and also not surprized that Vce can be less than 0.6V. I think the Art of Electronics explicitly mentions that conduction of a bjt is not an action of a (intrinsic?) diode.
Oh, btw, diode drop is not fixed at 0.6, it's rather a function of voltage, current, and other parameters. I think this is another "myth" widely spread.
I'd say explanations I was given at university were not helpful. I had to watch youtube and some forum posts to understand it better.
-
If you want to know more, get "Integrated Electronics: Analog and Digital Circuits and Systems" by Millman & Halkias.
PDFs are available online, but secondhand hard-copies aren't very expensive. A classic text that covers all the theory in great depth, highly recommended by many on this forum.
The origin of the "0.6V" diode drop is covered more fully in one of the later chapters, it's affected by the level of doping of the junction (and temperature), so is just a rule of thumb since many different devices have differing doping level and operating conditions.
From a practical perspective, it's where the diode starts to go linear on its VI curve. For the most part a PN junction can be viewed as a logarithmic resistance, i.e. it's a straight line if you plot V against ln(I). Most diodes and BJTs will operate down to 0.1V or lower, just with exponentially less current. Once you get enough current to drop ~0.6V the PN depletion region has all-but vanished, so no more logarithmic response and it operates more like a normal linear resistor.
-
You might search youtube for "jerri ellsworth homemade transistor" to see how she did it and what's involved.
Thanks for the jerri info. Never heard of her. Watched the video of making a mosfet. Then watched an interview...very interesting story.
Making a mosfet at home...too funny.
-
think the Art of Electronics explicitly mentions that conduction of a bjt is not an action of a (intrinsic?) diode.
Well, the emitter current is legitimately due to diode behavior on the BE junction. That is why the shockley diode equation and the ebers-moll model are the same formula. It is the collector current, and by extension the lack of base current that can't be explained that way.
-
And to make things even more confusing, you have to think in terms of actual electron flow rather than conventional current flow. So everything is backwards, with electrons coming out of the emitter (hence the name). Some get captured by the thin base, the rest go straight through to the collector - which is usually at a higher potential than the base and attracts the electrons more strongly. (It's very similar to how a triode vacuum tube works, but it still takes a bit of getting used to.)
When a transistor is driven hard to saturation, the collector voltage goes lower than the base (minimum VCE), so more electrons get captured by the base and fewer by the collector. This is why you need lots of base current to drive a BJT to saturation. Even though we tend to think in terms of applying more current, we're actually extracting more electrons because they're not making it to the collector as readily. (Talking NPN here, and obviously simplifying things a little.)
The emitter is usually more heavily doped to ensure plenty of excess electrons and help overcome this effect. This is a key feature of making a transistor, and another reason why two diodes won't work. (Along with the very thin base region and the need for the whole thing to be a single crystal.)
Technically you also have to consider the hole movement as well, which is why they're called "bipolar" transistors. The exponential Schockley equation is a result of holes & electrons recombining. This happens at random and basically exhibits a kind of exponential half-life decay. But they're also constantly being re-created by thermal energy (or an applied voltage), so a point of equilibrium is reached where there's a constant number of holes & electrons. The number being created then matches the number being lost through recombining.
The availability of holes & electrons (charge carriers) obviously affects how well the PN junction will conduct, and this is why the effective resistance changes with both temperature and voltage. And it happens in an exponential manner, because you're counteracting the inherent exponential half-life of random recombination.
Might be going overboard at this point, but the PN-junction is certainly a fascinating beast!
-
I had a boss once who was convinced that was how transistors worked. I had a design with an NPN transistor operating as a saturated switch in the ground leg of an amplifier, the whole circuit running off a 1.2V NiCd cell. My boss insisted that it wouldn't work, since the transistor "was two diodes back to back" and it was impossible to get the collector-emitter voltage under 0.6 volts. I tried to explain, and he didn't get it. I wired up a transistor with base and collector resistors and a power supply, showing him how we could drive the collector voltage to well under 0.1V. He claimed that I was "pulling the ground up". Somehow I got him to accept the design, but he often changed other designs behind my back for truly stupid reasons. I eventually quit. In his defense, he was pretty good with vacuum tubes (this was in the 1970s).
That's another thing I used to wonder about, Vce(sat). Turns out the ~0.6V junction drop is a built-in potential, like a battery in series. The collector current goes through two junctions which cancel out, so the collector voltage does indeed get referenced to emitter voltage. The cancellation isn't perfect, and the difference arises from the doping gradient, on the order of 10s of mV. The emitter is usually more heavily doped than the collector.
This is why most transistors have a minimum saturation (for all possible Ic and Ib) in the 10s of mV.
Conversely, you might rightfully ask: if we invert the transistor, can we get that back (negative)? Well, you can't get something for nothing, of course, so Vec(sat) as it were, cannot be negative. But it does happen that it's very close to zero in this configuration, and this fact was used in early precision DACs (12-16 bit), switched with inverted BJTs (and driven by carefully balanced current sources and sinks).
Tim
-
And to make things even more confusing, you have to think in terms of actual electron flow rather than conventional current flow. So everything is backwards, with electrons coming out of the emitter (hence the name). Some get captured by the thin base, the rest go straight through to the collector - which is usually at a higher potential than the base and attracts the electrons more strongly. (It's very similar to how a triode vacuum tube works, but it still takes a bit of getting used to.)
Indeed, one can draw the quantum energy diagram for a vacuum tube; it's merely an extension of the semiconductor band diagram, where valence electrons are promoted to the conduction band. Well, if we keep going up, eventually the conduction band ends, and the unbound band begins -- that is, electrons having enough energy to leave the material entirely, namely, above the work potential.
The current flow equations then have to be changed, for the lack of hole current and recombination (of course), ballistic rather than diffusion movement, and unbalanced charge (space charge). This gives the Child-Langmuir law, and combined with electrode geometry, we can generate the family of curves of real devices.
Alternately, we can start with what we know about tubes, assume a charge-neutralizing gas (namely, the sea of ions that is a crystal), and basically invent FETs long before we have chemical processing refined enough to actually create them.
A neutral gas (with free ions and electrons; a plasma) can actually be used to some benefit in real tubes. de Forest believed gas was actually necessary, and it does increase the gain under certain conditions. It turns out, hard vacuum tubes are only a little harder to make (much better pumping, and tighter electrode tolerances), but far more general (they don't have weird glow-discharge modes).
Tim