How are transistors designed

[Klaus Kragelund]

On Sunday, January 12, 2020 at 3:01:14 PM UTC+1, Steve Wilson wrote:

Klaus Kragelund <klauskvik@hotmail.com> wrote:

Any good book recommendation on how the BJT was discovered?

Many excellent examples in Youtube, BSTJ, and on the web.

YOUTUBE

Transistor Full Documentary
https://www.youtube.com/watch?v=U4XknGqr3Bo

Plus others.

BSTJ

The Bell System Technical Journal (1922-1983)
https://archive.org/details/bstj-archives

Example

BSTJ 28: 3. July 1949: The Theory of p-n Junctions in Semiconductors
and p-n Junction Transistors. (Shockley, W.)
https://ia802602.us.archive.org/3/items/bstj28-3-435/bstj28-3-435.pdf

Nice, thank you

On a side note, the other day I heard a podcast with Chuck Peddle, commonly regarded as the farther of the computers

https://theamphour.com/241-an-interview-with-chuck-peddle-charismatic-chipmaking-coryphaeus/

Very interesting talk

Other important persons are also on record here:

https://www.computerhistory.org/collections/oralhistories/

Cheers

Klaus

 

[Steve Wilson]

Klaus Kragelund <klauskvik@hotmail.com> wrote:

On Sunday, January 12, 2020 at 3:01:14 PM UTC+1, Steve Wilson wrote:
Klaus Kragelund <klauskvik@hotmail.com> wrote:

Any good book recommendation on how the BJT was discovered?

Many excellent examples in Youtube, BSTJ, and on the web.

YOUTUBE

Transistor Full Documentary
https://www.youtube.com/watch?v=U4XknGqr3Bo

Plus others.

BSTJ

The Bell System Technical Journal (1922-1983)
https://archive.org/details/bstj-archives

Example

BSTJ 28: 3. July 1949: The Theory of p-n Junctions in Semiconductors
and p-n Junction Transistors. (Shockley, W.)
https://ia802602.us.archive.org/3/items/bstj28-3-435/bstj28-3-435.pdf

Nice, thank you

On a side note, the other day I heard a podcast with Chuck Peddle,
commonly regarded as the farther of the computers

https://theamphour.com/241-an-interview-with-chuck-peddle-charismatic-chi
pmaking-coryphaeus/

Very interesting talk

Other important persons are also on record here:

https://www.computerhistory.org/collections/oralhistories/

Cheers

Klaus

Ha! I used the 6502 early in my business. Very easy to program.

Brings back fond memories. Thanks.

 

[Robert Baer]

upsidedown@downunder.com wrote:

On Sat, 11 Jan 2020 22:43:52 -0800, Robert Baer
robertbaer@localnet.com> wrote:

whit3rd wrote:
On Friday, January 10, 2020 at 10:36:28 PM UTC-8, amal banerjee wrote:
A question for all you electronics gurus out there,
maybe silly, maybe not. How are transistors designed ?

The start of transistor technology was a surface-contact experiment
that had gain. William Shockley wrote a theoretical description of
a possible mechanism (not a very good fit to the experiment, but very
productive) around 1948.

Practical transistors initially were three-layer Germanium devices
created by alloying a slab, and diffusion of the alloying impurities determined
the dopant distribution.

Improvements over the years were important;
Silicon can be passivated by oxidation (foolproof compared with
Germanium passivation schemes), and has higher service
temperatures.
Ion bombardment can implant impurities at depth, with diffusion as
a finishing step.
Masking operations can print shapes for different
parasitic characteristics.
Chemical vapor deposition can add a layer of pure material overtop
a partial construction.

Finally, there are packaging considerations (you need to connect wires somehow)
with some importance in final-construction circuit parameters. For silicon, aluminum
is the universally approved way to make an ohmic contact, to either P-type or N-type silicon.

So, to design a BJT, you might want to start with a design that nearly fills your needs, and
change its scale, or its chemistry, or its manufacture process. Then, it's just a matter of knowing
the three-dimensional composition result and applying physical conditions to the model (biasing
the transistor) and predicting the results. There's a lot of physics, chemistry, and
lore involved.

There are always some surprises (fringe fields, emitter push, and cosmic ray sensitivity),
so a final assessment is done experimentally, after cooking up a few batches.

Whats his face, Shockley, was beat by over 10 years WRT invention of
the transistor.

Julius Edgar Lilienfeld' first patent "Method and Apparatus for
Controlling Electric Currents" was filed Oct 8, 1926 and published Jan
28, 1930 #1745175.

Wasn't that a FET ?

* The first one appears to be a FET; look at the drawings and
description in the second one; clearly junction EBC.
The third patent expands on the technology.

And it was no fluke; he clearly knew what was happening and his
second patent 1677140 published Sept 13, 1932 teaches about a junction
transistor.
His THIRD patent 1,900018 was published March 7, 1933.

Clearly way before whats his name.

 

[Phil Hobbs]

On 2020-01-12 19:45, Robert Baer wrote:

upsidedown@downunder.com wrote:
On Sat, 11 Jan 2020 22:43:52 -0800, Robert Baer
robertbaer@localnet.com> wrote:

whit3rd wrote:
On Friday, January 10, 2020 at 10:36:28 PM UTC-8, amal banerjee wrote:
A question for all you electronics gurus out there,
maybe silly, maybe not. How are transistors designed ?

The start of transistor technology was a surface-contact experiment
that had gain.   William Shockley wrote a theoretical description of
a possible mechanism (not a very good fit to the experiment, but very
productive) around 1948.

Practical transistors initially were three-layer Germanium devices
created by alloying a slab, and diffusion of the alloying impurities
determined
the dopant distribution.

Improvements over the years were important;
Silicon can be passivated by oxidation (foolproof compared with
Germanium passivation schemes), and has higher service
temperatures.
Ion bombardment can implant impurities at depth, with diffusion as
   a finishing step.
Masking operations can print shapes for different
parasitic characteristics.
Chemical vapor deposition can add a layer of pure material overtop
a partial construction.

Finally, there are packaging considerations (you need to connect
wires somehow)
with some importance in final-construction circuit parameters.   For
silicon, aluminum
is the universally approved way to make an ohmic contact, to either
P-type or N-type silicon.

So, to design a BJT, you might want to start with a design that
nearly fills your needs, and
change its scale, or its chemistry, or its manufacture process.
Then, it's just a matter of knowing
the three-dimensional composition result and applying physical
conditions to the model (biasing
the transistor) and predicting the results.   There's a lot of
physics, chemistry, and
lore involved.

There are always some surprises (fringe fields, emitter push, and
cosmic ray sensitivity),
so a final assessment is done experimentally, after cooking up a few
batches.

   Whats his face, Shockley, was beat by over 10 years WRT invention of
the transistor.

   Julius Edgar Lilienfeld' first patent "Method and Apparatus for
Controlling Electric Currents" was filed Oct 8, 1926 and published Jan
28, 1930 #1745175.

Wasn't that a FET ?
* The first one appears to be a FET; look at the drawings and
description in the second one; clearly junction EBC.

Nonsense. There are no PN junctions there--it's all
metal/semiconductor/metal.

Cheers

Phil Hobbs


--
Dr Philip C D Hobbs
Principal Consultant
ElectroOptical Innovations LLC / Hobbs ElectroOptics
Optics, Electro-optics, Photonics, Analog Electronics
Briarcliff Manor NY 10510

http://electrooptical.net
http://hobbs-eo.com

 

[Robert Baer]

Phil Hobbs wrote:

On 2020-01-12 01:43, Robert Baer wrote:
whit3rd wrote:
On Friday, January 10, 2020 at 10:36:28 PM UTC-8, amal banerjee
wrote:
A question for all you electronics gurus out there, maybe silly,
maybe not. How are transistors designed ?

The start of transistor technology was a surface-contact
experiment that had gain.   William Shockley wrote a theoretical
description of a possible mechanism (not a very good fit to the
experiment, but very productive) around 1948.

Practical transistors initially were three-layer Germanium devices
created by alloying a slab, and diffusion of the alloying
impurities determined the dopant distribution.

Point-contact transistors were still made until about 1970, iirc.  They
weren't very reliable or repeatable, but were much much faster than the
old alloy-junction devices.


Improvements over the years were important; Silicon can be
passivated by oxidation (foolproof compared with Germanium
passivation schemes), and has higher service temperatures.

Yup.  SiO2 is a very useful material, whereas GeO2 is water soluble.
(Diamond also doesn't have a good oxide.)

Ion bombardment can implant impurities at depth, with diffusion as a
finishing step.

You don't need to do diffusion if you have ion implantation--that's more
or less the point of the exercise.  You do need to do a rapid thermal
anneal (RTA) to get rid of the lattice damage caused by the
implantation.  (It needs to be rapid precisely to _avoid_ significant
dopant diffusion.)

Masking operations can print shapes for different parasitic
characteristics. Chemical vapor deposition can add a layer of pure
material overtop a partial construction.

Finally, there are packaging considerations (you need to connect
wires somehow) with some importance in final-construction circuit
parameters.   For silicon, aluminum is the universally approved way
to make an ohmic contact, to either P-type or N-type silicon.

In 1955, maybe.  There are two basic ways of making an ohmic contact to
silicon: you either dope the contact region so heavily that the bandgap
goes away (i.e. it effectively becomes metallic) or else you pick a
metal that makes a Schottky barrier with negative barrier height.
(These exist, but I forget which ones they are.)


So, to design a BJT, you might want to start with a design that
nearly fills your needs, and change its scale, or its chemistry, or
its manufacture process. Then, it's just a matter of knowing the
three-dimensional composition result and applying physical conditions
to the model (biasing the transistor) and predicting the
results.   There's a lot of physics, chemistry, and lore involved.

There are always some surprises (fringe fields, emitter push, and
cosmic ray sensitivity), so a final assessment is done
experimentally, after cooking up a few batches.

Nah, physical simulations are both easy and accurate today, and there
are lots of well-understood knobs to twist.  The cleanliness of modern
semiconductor processes is amazing, so there's not much mystery about
what's happening inside the device.


Whats his face, Shockley, was beat by over 10 years WRT invention of
 the transistor.

Julius Edgar Lilienfeld' first patent "Method and Apparatus for
Controlling Electric Currents" was filed Oct 8, 1926 and published
Jan 28, 1930 #1745175. And it was no fluke; he clearly knew what was
happening

Did you read the patent?  He has no very deep concept of the device
physics.  From P. 2, starting at line 76:

"The basis of the invention resides appar ently in the fact that the
conducting layer at the particular point selected introduces a
resistance varying with the electric field at this point; and in this
connection it may be assumed that the atoms (or molecules) of a
conductor are of the nature of bipoles. In order for an electron,
therefore, to travel in the electric field, the bipoles are obliged to
become organized in this field substantially with their axes parallel or
lying in the field of flow.  Any disturbance in this organization, as by
heat movement, magnetic field, electro static cross-field, etc., will
serve to increase the resistance of the conductor; and in the instant
case, the conductivity of the layer is influenced by the electric field.
 Owing to the fact that this layer is extremely thin the field is
permitted to penetrate the entire volume thereof and thus will change
the conductivity throughout the entire cross-section of this conducting
portion."

"...the bipoles are obliged to become organized...?"

Pretty tough to find anything about doping, band structure, conduction
vs. valence electrons, or even any real electrostatics.  Also his
copper-oxide semiconductor is polycrystalline or amorphous.

* You don't need to do diffusion to have inherent doping.
Ever heard of Ovshinsky? He opened the scientific field of amorphous
and disordered materials.

and his second patent 1677140 published Sept 13, 1932 teaches about a
junction transistor.

Should be 1877140.  What he describes isn't anything like a BJT--it's a
metal-semiconductor-metal structure, and he apparently had no idea how
it worked, if it ever did. It may have been a reach-through device, with
back-to-back Schottky barriers, but with the vacuum technology available
in 1930, who knows?  Certainly it isn't a BJT.

His THIRD patent 1,900018 was published March 7, 1933.

That seems to be a MESFET-sort of structure, controlled by a back gate.
 It's far from clear how he achieves channel modulation, because there's
no front gate at all.


Clearly way before whats his name.


Nah, no band structure, no P- and N- doping, no
semiconductor-semiconductor junction, nada.   Shockley & Co. are safely
dead, but their fame remains secure.

Cheers

Phil Hobbs

 

[Phil Hobbs]

On 2020-01-12 19:54, Robert Baer wrote:

Phil Hobbs wrote:
On 2020-01-12 01:43, Robert Baer wrote:
whit3rd wrote:
On Friday, January 10, 2020 at 10:36:28 PM UTC-8, amal banerjee
wrote:
A question for all you electronics gurus out there, maybe silly,
maybe not. How are transistors designed ?

The start of transistor technology was a surface-contact
experiment that had gain.   William Shockley wrote a theoretical
description of a possible mechanism (not a very good fit to the
experiment, but very productive) around 1948.

Practical transistors initially were three-layer Germanium devices
created by alloying a slab, and diffusion of the alloying
impurities determined the dopant distribution.

Point-contact transistors were still made until about 1970, iirc.  They
weren't very reliable or repeatable, but were much much faster than the
old alloy-junction devices.


Improvements over the years were important; Silicon can be
passivated by oxidation (foolproof compared with Germanium
passivation schemes), and has higher service temperatures.

Yup.  SiO2 is a very useful material, whereas GeO2 is water soluble.
(Diamond also doesn't have a good oxide.)

Ion bombardment can implant impurities at depth, with diffusion as a
finishing step.

You don't need to do diffusion if you have ion implantation--that's more
or less the point of the exercise.  You do need to do a rapid thermal
anneal (RTA) to get rid of the lattice damage caused by the
implantation.  (It needs to be rapid precisely to _avoid_ significant
dopant diffusion.)

Masking operations can print shapes for different parasitic
characteristics. Chemical vapor deposition can add a layer of pure
material overtop a partial construction.

Finally, there are packaging considerations (you need to connect
wires somehow) with some importance in final-construction circuit
parameters.   For silicon, aluminum is the universally approved way
to make an ohmic contact, to either P-type or N-type silicon.

In 1955, maybe.  There are two basic ways of making an ohmic contact to
silicon: you either dope the contact region so heavily that the bandgap
goes away (i.e. it effectively becomes metallic) or else you pick a
metal that makes a Schottky barrier with negative barrier height.
(These exist, but I forget which ones they are.)


So, to design a BJT, you might want to start with a design that
nearly fills your needs, and change its scale, or its chemistry, or
its manufacture process. Then, it's just a matter of knowing the
three-dimensional composition result and applying physical
conditions to the model (biasing the transistor) and predicting the
results.   There's a lot of physics, chemistry, and lore involved.

There are always some surprises (fringe fields, emitter push, and
cosmic ray sensitivity), so a final assessment is done
experimentally, after cooking up a few batches.

Nah, physical simulations are both easy and accurate today, and there
are lots of well-understood knobs to twist.  The cleanliness of modern
semiconductor processes is amazing, so there's not much mystery about
what's happening inside the device.


Whats his face, Shockley, was beat by over 10 years WRT invention of
 the transistor.

Julius Edgar Lilienfeld' first patent "Method and Apparatus for
Controlling Electric Currents" was filed Oct 8, 1926 and published
Jan 28, 1930 #1745175. And it was no fluke; he clearly knew what was
happening

Did you read the patent?  He has no very deep concept of the device
physics.  From P. 2, starting at line 76:

"The basis of the invention resides appar ently in the fact that the
conducting layer at the particular point selected introduces a
resistance varying with the electric field at this point; and in this
connection it may be assumed that the atoms (or molecules) of a
conductor are of the nature of bipoles. In order for an electron,
therefore, to travel in the electric field, the bipoles are obliged to
become organized in this field substantially with their axes parallel
or lying in the field of flow.  Any disturbance in this organization,
as by heat movement, magnetic field, electro static cross-field, etc.,
will serve to increase the resistance of the conductor; and in the
instant case, the conductivity of the layer is influenced by the
electric field.   Owing to the fact that this layer is extremely thin
the field is permitted to penetrate the entire volume thereof and thus
will change the conductivity throughout the entire cross-section of
this conducting portion."

"...the bipoles are obliged to become organized...?"

Pretty tough to find anything about doping, band structure, conduction
vs. valence electrons, or even any real electrostatics.  Also his
copper-oxide semiconductor is polycrystalline or amorphous.
* You don't need to do diffusion to have inherent doping.
  Ever heard of Ovshinsky? He opened the scientific field of amorphous
and disordered materials.

The ovonics guy? A genius at self-promotion, for sure. Technically,
not so much.

(Also completely unrelated to the Lilienfeld discussion.)


Cheers

Phil Hobbs

--
Dr Philip C D Hobbs
Principal Consultant
ElectroOptical Innovations LLC / Hobbs ElectroOptics
Optics, Electro-optics, Photonics, Analog Electronics
Briarcliff Manor NY 10510

http://electrooptical.net
http://hobbs-eo.com

 

[amal banerjee]

On Sunday, January 12, 2020 at 3:48:56 PM UTC+5:30, DecadentLinux...@decadence.org wrote:

amal banerjee <dakupoto@gmail.com> wrote in
news:cfe1000f-09e8-4ed1-b15a-2bf9f873fb1e@googlegroups.com:

On Saturday, January 11, 2020 at 2:18:13 PM UTC+5:30,
DecadentLinux...@decadence.org wrote:
amal banerjee <dakupoto@gmail.com> wrote in
news:d083fd7c-93b6-4f6d- addb-ec4b2fbe30bf@googlegroups.com:

A question for all you electronics gurus out there,
maybe silly, maybe not. How are transistors designed ?

First of all, the SPICE parameters for example for
a BJT are obtained by curve fitting experimental
test data.

So, if I want to design a BJT with a predefined set
of performance parameters, how do I start the design
task, i.e., start with the Ebers-Moll models, adjosy
parameters such that the desired performance specifications
are specified and then go the process guys ?


Read up on the structure of a modern power transistor...

https://engineeringtutorial.com/bipolar-junction-transistor-
construction/

Read up on the history of the fin fet...

https://blog.lamresearch.com/tech-brief-finfet-fundamentals/

There is nothing new here. My Masters's thesis included sections
on how C-V measurements of sub-micron MOSFETs may be automated. I
am fully aware of the structures of common semiconductor
devices.



"touchy" Hahahah! At the molecular level!

So your claim is that you know how they are constructed, and yet
your original post asks how they are designed...

I would have said thet they get designed in colleges like MIT and
Purdue, but You apparently did not even need to ask the question.

Did you not word your query well enough?

No I am not touchy at all. The semiconductor device
group at UT-Austin ECE department was pretty big(26 profs)
but al they did was process engineering. But at the process
engineering stage, the device has already been designed,

 

amal banerjee <dakupoto@gmail.com> wrote in
news:99290e4e-01f3-487a-9e16-13ab1abb8211@googlegroups.com:

On Sunday, January 12, 2020 at 3:48:56 PM UTC+5:30,
DecadentLinux...@decadence.org wrote:
amal banerjee <dakupoto@gmail.com> wrote in
news:cfe1000f-09e8-4ed1-b15a-2bf9f873fb1e@googlegroups.com:

On Saturday, January 11, 2020 at 2:18:13 PM UTC+5:30,
DecadentLinux...@decadence.org wrote:
amal banerjee <dakupoto@gmail.com> wrote in
news:d083fd7c-93b6-4f6d- addb-ec4b2fbe30bf@googlegroups.com:

A question for all you electronics gurus out there,
maybe silly, maybe not. How are transistors designed ?

First of all, the SPICE parameters for example for
a BJT are obtained by curve fitting experimental
test data.

So, if I want to design a BJT with a predefined set
of performance parameters, how do I start the design
task, i.e., start with the Ebers-Moll models, adjosy
parameters such that the desired performance specifications
are specified and then go the process guys ?


Read up on the structure of a modern power transistor...

https://engineeringtutorial.com/bipolar-junction-transistor-
construction/

Read up on the history of the fin fet...

https://blog.lamresearch.com/tech-brief-finfet-fundamentals/

There is nothing new here. My Masters's thesis included
sections on how C-V measurements of sub-micron MOSFETs may be
automated. I am fully aware of the structures of common
semiconductor devices.



"touchy" Hahahah! At the molecular level!

So your claim is that you know how they are constructed, and
yet
your original post asks how they are designed...

I would have said thet they get designed in colleges like MIT
and
Purdue, but You apparently did not even need to ask the question.

Did you not word your query well enough?

No I am not touchy at all.

It was a joke. "Probing chip dies"... "touchy".
You don't get it?

The semiconductor device
group at UT-Austin ECE department was pretty big(26 profs)
but al they did was process engineering. But at the process
engineering stage, the device has already been designed,

OK. Grow some highly organized precisely placed sand!

 

[Don Kuenz]

amal banerjee <dakupoto@gmail.com> wrote:

A question for all you electronics gurus out there,
maybe silly, maybe not. How are transistors designed ?

First of all, the SPICE parameters for example for
a BJT are obtained by curve fitting experimental
test data.

So, if I want to design a BJT with a predefined set
of performance parameters, how do I start the design
task, i.e., start with the Ebers-Moll models, adjosy
parameters such that the desired performance specifications
are specified and then go the process guys ?

You might get some mileage out of _Bipolar Semiconductor Devices_ by
Roulston. (You need to ask an electronics guru to be certain.) Roulston
wrote some semiconductor simulation software [1], which may prove
helpful to your quest.

Note.

[1] http://www.bipole3.com/main.html

Thank you,

--
Don Kuenz KB7RPU
There was a young lady named Bright Whose speed was far faster than light;
She set out one day In a relative way And returned on the previous night.