Optical computers run a million times faster than conventional computers, study reveals.

This extremely fast processing uses circularly polarized light instead of an electrical signal. The process depends a great deal on the handedness of the circularly polarized beam, through a crystal material that is sensitive to the beam of light.

Tpd's < 100fs

overview:

https://interestingengineering.com/innovation/optical-computers-run-a-million-times-faster-than-conventional-computers-study-reveals

nitty-gritty:

https://www.science.org/doi/10.1126/sciadv.abq8246

Fred Bloggs wrote:
> Optical computers run a million times faster than conventional computers, study reveals.
>
> This extremely fast processing uses circularly polarized light instead of an electrical signal. The process depends a great deal on the handedness of the circularly polarized beam, through a crystal material that is sensitive to the beam of light.
>
> Tpd's < 100fs
>
> overview:
>
> https://interestingengineering.com/innovation/optical-computers-run-a-million-times-faster-than-conventional-computers-study-reveals
>
> nitty-gritty:
>
> https://www.science.org/doi/10.1126/sciadv.abq8246
>

It's an interesting idea, for sure. I can imagine some interesting
measurement applications in time-resolved chemical sensing, which is a
pretty hot area--you can watch molecules rearranging themselves during a
chemical reaction. It's completely useless for computing, of course.

<soapbox>

Right off, in common with most such papers, the authors trash their
credibility with that grantsmanship verbiage about IoT and exponentially
increasing amounts of data, and they continue with artistic misdirection
about what they're actually doing in the measurement. We've all heard
about that sort of stuff for decades, and very nearly none of the
reported developments have borne any significant fruit. Some thoughts:

1. 100 fs gate delay is a million times faster than a processor with a
250-ps _cycle time_? _Somebody_ isn't super numerate over there. As
one swallow doth not a summer make, so one gate is not a computer. A
cycle takes many gate delays. Taken at face value, 100 fs is about 100x
faster than today's CMOS. What's four orders of magnitude among
friends? But that's just the beginning.

2. Nobody ever talks about interconnect latency in optical gates. This
one is off by a good 12 orders of magnitude--that "100 fs propagation
delay" is actually a measurement requiring at least milliseconds, and
which has little to do with the propagation delay. They're using a
lock-in(!) to measure the time-averaged nonlinear signals as a function
of the delay between the two input pulses. It's more of a
cross-correlation measurement.

3. The outputs are at widely different wavelengths, leading to all sorts
of dispersion problems plus very serious constraints on how you can
combine logic functions, even in principle.

4. Those wavelengths are all generated at once, and separating them
requires structures large compared with a wavelength. With 5 nm CMOS,
you can fit a lot, a lot of gates into that volume.

5. They shove the horrible inefficiency under a rug by quoting
everything in 'arbitrary units'. SHG and optical frequency mixing are
very inefficient in small structures. To get decent efficiency you need
many wavelengths for the nonlinear polarization to generate a travelling
wave, which has to be phase matched.

6. Optical computing is fundamentally a dead end because fast optical
elements have no gain, and elements with gain are not fast. You can
control lots of electrons with one electron, but you can't control many
photons with one photon.

This one seems to have a specially bad case of this problem. Late in
the paper, the authors do let slip that they're using a Spectra Physics
Solstice Ace laser and detecting with a PMT and a lock-in amplifier.
That is, they're hitting the poor sample with a laser that produces
>1-millijoule pulses, 60 fs long (>16 GW!) and detecting probably
picowatts.

They're sure not using the full output of the laser, because that would
likely excavate a hole in the sample, but we're talking 180 dB loss, at
very least.

</soapbox>

(Thanks, I feel much better now.) ;)

Cheers

Phil Hobbs

(Former optical interconnect person)

--
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

On Thu, 15 Dec 2022 05:54:57 -0800 (PST), Fred Bloggs
<bloggs.fredbloggs.fred@gmail.com> wrote:

>Optical computers run a million times faster than conventional computers, study reveals.

Do they run Windows yet?

On Thursday, December 15, 2022 at 11:01:55 AM UTC-5, John Larkin wrote:
> On Thu, 15 Dec 2022 05:54:57 -0800 (PST), Fred Bloggs
> <bloggs.fred...@gmail.com> wrote:
>
> >Optical computers run a million times faster than conventional computers, study reveals.
> Do they run Windows yet?

Not if Intel has anything to say about it.

On Thu, 15 Dec 2022 10:44:47 -0500, Phil Hobbs
<pcdhSpamMeSenseless@electrooptical.net> wrote:

>Fred Bloggs wrote:
>> Optical computers run a million times faster than conventional computers, study reveals.
>>
>> This extremely fast processing uses circularly polarized light instead of an electrical signal. The process depends a great deal on the handedness of the circularly polarized beam, through a crystal material that is sensitive to the beam of light.
>>
>> Tpd's < 100fs
>>
>> overview:
>>
>> https://interestingengineering.com/innovation/optical-computers-run-a-million-times-faster-than-conventional-computers-study-reveals
>>
>> nitty-gritty:
>>
>> https://www.science.org/doi/10.1126/sciadv.abq8246
>>
>
>It's an interesting idea, for sure. I can imagine some interesting
>measurement applications in time-resolved chemical sensing, which is a
>pretty hot area--you can watch molecules rearranging themselves during a
>chemical reaction. It's completely useless for computing, of course.
>
><soapbox>
>
>Right off, in common with most such papers, the authors trash their
>credibility with that grantsmanship verbiage about IoT and exponentially
>increasing amounts of data, and they continue with artistic misdirection
>about what they're actually doing in the measurement. We've all heard
>about that sort of stuff for decades, and very nearly none of the
>reported developments have borne any significant fruit. Some thoughts:
>
>1. 100 fs gate delay is a million times faster than a processor with a
>250-ps _cycle time_? _Somebody_ isn't super numerate over there. As
>one swallow doth not a summer make, so one gate is not a computer. A
>cycle takes many gate delays. Taken at face value, 100 fs is about 100x
>faster than today's CMOS. What's four orders of magnitude among
>friends? But that's just the beginning.
>
>2. Nobody ever talks about interconnect latency in optical gates. This
>one is off by a good 12 orders of magnitude--that "100 fs propagation
>delay" is actually a measurement requiring at least milliseconds, and
>which has little to do with the propagation delay. They're using a
>lock-in(!) to measure the time-averaged nonlinear signals as a function
>of the delay between the two input pulses. It's more of a
>cross-correlation measurement.
>
>3. The outputs are at widely different wavelengths, leading to all sorts
>of dispersion problems plus very serious constraints on how you can
>combine logic functions, even in principle.
>
>4. Those wavelengths are all generated at once, and separating them
>requires structures large compared with a wavelength. With 5 nm CMOS,
>you can fit a lot, a lot of gates into that volume.
>
>5. They shove the horrible inefficiency under a rug by quoting
>everything in 'arbitrary units'. SHG and optical frequency mixing are
>very inefficient in small structures. To get decent efficiency you need
>many wavelengths for the nonlinear polarization to generate a travelling
>wave, which has to be phase matched.
>
>6. Optical computing is fundamentally a dead end because fast optical
>elements have no gain, and elements with gain are not fast. You can
>control lots of electrons with one electron, but you can't control many
>photons with one photon.
>
>This one seems to have a specially bad case of this problem. Late in
>the paper, the authors do let slip that they're using a Spectra Physics
>Solstice Ace laser and detecting with a PMT and a lock-in amplifier.
>That is, they're hitting the poor sample with a laser that produces
> >1-millijoule pulses, 60 fs long (>16 GW!) and detecting probably
>picowatts.
>
>They're sure not using the full output of the laser, because that would
>likely excavate a hole in the sample, but we're talking 180 dB loss, at
>very least.
>
></soapbox>
>
>(Thanks, I feel much better now.) ;)
>
>Cheers
>
>Phil Hobbs
>
>(Former optical interconnect person)

Come on, tell us what you really think.

On Thu, 15 Dec 2022 08:59:03 -0800 (PST), Fred Bloggs
<bloggs.fredbloggs.fred@gmail.com> wrote:

>On Thursday, December 15, 2022 at 11:01:55 AM UTC-5, John Larkin wrote:
>> On Thu, 15 Dec 2022 05:54:57 -0800 (PST), Fred Bloggs
>> <bloggs.fred...@gmail.com> wrote:
>>
>> >Optical computers run a million times faster than conventional computers, study reveals.
>> Do they run Windows yet?
>
>Not if Intel has anything to say about it.

The chips will need glass caps so you can see the computing flashing.
The gamers will love that.

John Larkin wrote:
> On Thu, 15 Dec 2022 10:44:47 -0500, Phil Hobbs
> <pcdhSpamMeSenseless@electrooptical.net> wrote:
>
>> Fred Bloggs wrote:
>>> Optical computers run a million times faster than conventional computers, study reveals.
>>>
>>> This extremely fast processing uses circularly polarized light instead of an electrical signal. The process depends a great deal on the handedness of the circularly polarized beam, through a crystal material that is sensitive to the beam of light.
>>>
>>> Tpd's < 100fs
>>>
>>> overview:
>>>
>>> https://interestingengineering.com/innovation/optical-computers-run-a-million-times-faster-than-conventional-computers-study-reveals
>>>
>>> nitty-gritty:
>>>
>>> https://www.science.org/doi/10.1126/sciadv.abq8246
>>>
>>
>> It's an interesting idea, for sure. I can imagine some interesting
>> measurement applications in time-resolved chemical sensing, which is a
>> pretty hot area--you can watch molecules rearranging themselves during a
>> chemical reaction. It's completely useless for computing, of course.
>>
>> <soapbox>
>>
>> Right off, in common with most such papers, the authors trash their
>> credibility with that grantsmanship verbiage about IoT and exponentially
>> increasing amounts of data, and they continue with artistic misdirection
>> about what they're actually doing in the measurement. We've all heard
>> about that sort of stuff for decades, and very nearly none of the
>> reported developments have borne any significant fruit. Some thoughts:
>>
>> 1. 100 fs gate delay is a million times faster than a processor with a
>> 250-ps _cycle time_? _Somebody_ isn't super numerate over there. As
>> one swallow doth not a summer make, so one gate is not a computer. A
>> cycle takes many gate delays. Taken at face value, 100 fs is about 100x
>> faster than today's CMOS. What's four orders of magnitude among
>> friends? But that's just the beginning.
>>
>> 2. Nobody ever talks about interconnect latency in optical gates. This
>> one is off by a good 12 orders of magnitude--that "100 fs propagation
>> delay" is actually a measurement requiring at least milliseconds, and
>> which has little to do with the propagation delay. They're using a
>> lock-in(!) to measure the time-averaged nonlinear signals as a function
>> of the delay between the two input pulses. It's more of a
>> cross-correlation measurement.
>>
>> 3. The outputs are at widely different wavelengths, leading to all sorts
>> of dispersion problems plus very serious constraints on how you can
>> combine logic functions, even in principle.
>>
>> 4. Those wavelengths are all generated at once, and separating them
>> requires structures large compared with a wavelength. With 5 nm CMOS,
>> you can fit a lot, a lot of gates into that volume.
>>
>> 5. They shove the horrible inefficiency under a rug by quoting
>> everything in 'arbitrary units'. SHG and optical frequency mixing are
>> very inefficient in small structures. To get decent efficiency you need
>> many wavelengths for the nonlinear polarization to generate a travelling
>> wave, which has to be phase matched.
>>
>> 6. Optical computing is fundamentally a dead end because fast optical
>> elements have no gain, and elements with gain are not fast. You can
>> control lots of electrons with one electron, but you can't control many
>> photons with one photon.
>>
>> This one seems to have a specially bad case of this problem. Late in
>> the paper, the authors do let slip that they're using a Spectra Physics
>> Solstice Ace laser and detecting with a PMT and a lock-in amplifier.
>> That is, they're hitting the poor sample with a laser that produces
>>> 1-millijoule pulses, 60 fs long (>16 GW!) and detecting probably
>> picowatts.
>>
>> They're sure not using the full output of the laser, because that would
>> likely excavate a hole in the sample, but we're talking 180 dB loss, at
>> very least.
>>
>> </soapbox>
>>
>> (Thanks, I feel much better now.) ;)
>>
>> (Former optical interconnect person)
>
> Come on, tell us what you really think.
>

They just need more funding for their vitally important work, on which
the future of technological civilization obviously depends. ;)

Cheers

Phil Hobbs

(Who could also use a bit more funding for side projects.)

--
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

On Thursday, December 15, 2022 at 8:01:55 AM UTC-8, John Larkin wrote:
> On Thu, 15 Dec 2022 05:54:57 -0800 (PST), Fred Bloggs
> <bloggs.fred...@gmail.com> wrote:
>
> >Optical computers run a million times faster than conventional computers, study reveals.
> Do they run Windows yet?

Never. Phil touched on that when he noted 'many wavelengths' required for
a typical optical element. To achieve the billion-transistor complexity that
a Windows machine has, with multiple 400 nm wavelengths per element, implies you need
a BIG box size. Meters or more, just for the CPU chip.

Phil Hobbs wrote:
>
> They just need more funding for their vitally important work, on which
> the future of technological civilization obviously depends. ;)
>
> Cheers
>
> Phil Hobbs
>
> (Who could also use a bit more funding for side projects.)

It's surprising that Musk hasn't created his own Bell Labs or PARC.

It could have "visiting" researchers.

--
Defund the Thought Police

Tom Del Rosso wrote:
> Phil Hobbs wrote:
>>
>> They just need more funding for their vitally important work, on which
>> the future of technological civilization obviously depends. ;)
>>
>> Cheers
>>
>> Phil Hobbs
>>
>> (Who could also use a bit more funding for side projects.)
>
> It's surprising that Musk hasn't created his own Bell Labs or PARC.
>
> It could have "visiting" researchers.
>
Another PARC or Sarnoff we can do without. They did a lot of good stuff
that other people made money from. Historically, Bell and IBM Research
were better at producing value for their parent organizations.

Bell Labs never really recovered from the 1984 breakup of AT&T, but the
kiss of death was the combination of the dotcom bust and the Jan Hendrik
Scho:n scandal. IBM Research took a big hit in 1992 when the company
nearly went belly-up, but survived and more or less thrived since then.

Cheers

Phil Hobbs

(Who spent 21 years at IBM Research in the latter part of its glory days)

--
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

On 2022-12-17, Tom Del Rosso <fizzbintuesday@that-google-mail-domain.com> wrote:
> Phil Hobbs wrote:
>>
>> They just need more funding for their vitally important work, on which
>> the future of technological civilization obviously depends. ;)
>>
>> Cheers
>>
>> Phil Hobbs
>>
>> (Who could also use a bit more funding for side projects.)
>
> It's surprising that Musk hasn't created his own Bell Labs or PARC.

Not really, he doesn't seem to have any long term plans. His short
term plans seems to be promise the world in 2 years reap the stock
bubble and then fail to deliver, and hope everone is distracted.

--
Jasen.

On Sat, 17 Dec 2022 14:22:31 -0500, Phil Hobbs
<pcdhSpamMeSenseless@electrooptical.net> wrote:

>Tom Del Rosso wrote:
>> Phil Hobbs wrote:
>>>
>>> They just need more funding for their vitally important work, on which
>>> the future of technological civilization obviously depends. ;)
>>>
>>> Cheers
>>>
>>> Phil Hobbs
>>>
>>> (Who could also use a bit more funding for side projects.)
>>
>> It's surprising that Musk hasn't created his own Bell Labs or PARC.
>>
>> It could have "visiting" researchers.
>>
>Another PARC or Sarnoff we can do without. They did a lot of good stuff
>that other people made money from. Historically, Bell and IBM Research
>were better at producing value for their parent organizations.
>
>Bell Labs never really recovered from the 1984 breakup of AT&T, but the
>kiss of death was the combination of the dotcom bust ...

The Dotcom bubble took a lot of companies out, almost including
Corning, whence came fiber optics.

> and the Jan Hendrik Schön scandal.

Whatever happened to Jan Hendrik Schön afterwards? I recall something
about him taking an unspecified engineering job somewhere.

Joe Gwinn

> IBM Research took a big hit in 1992 when the company
>nearly went belly-up, but survived and more or less thrived since then.
>
>Cheers
>
>Phil Hobbs
>
>(Who spent 21 years at IBM Research in the latter part of its glory days)

On Sunday, 18 December 2022 at 19:56:39 UTC, Joe Gwinn wrote:
> >Bell Labs never really recovered from the 1984 breakup of AT&T, but the
> >kiss of death was the combination of the dotcom bust ...
>
> The Dotcom bubble took a lot of companies out, almost including
> Corning, whence came fiber optics.

Are you sure about that?
opticalfibrehistory.co.uk/who/charles-kuen-kao/
opticalfibrehistory.co.uk/where/

John

On Sunday, 18 December 2022 at 21:36:59 UTC, John Walliker wrote:
> On Sunday, 18 December 2022 at 19:56:39 UTC, Joe Gwinn wrote:
>
> > >Bell Labs never really recovered from the 1984 breakup of AT&T, but the
> > >kiss of death was the combination of the dotcom bust ...
> >
> > The Dotcom bubble took a lot of companies out, almost including
> > Corning, whence came fiber optics.
> Are you sure about that?
> opticalfibrehistory.co.uk/who/charles-kuen-kao/
> opticalfibrehistory.co.uk/where/
>
> John

Here is the original paper:
https://digital-library.theiet.org/docserver/fulltext/piee/113/7/19660189.pdf?expires=1671400591&id=id&accname=guest&checksum=4F929B734AB903E69AE9DB5CB963633F

John

On Sunday, 18 December 2022 at 21:44:13 UTC, John Walliker wrote:
> On Sunday, 18 December 2022 at 21:36:59 UTC, John Walliker wrote:
> > On Sunday, 18 December 2022 at 19:56:39 UTC, Joe Gwinn wrote:
> >
> > > >Bell Labs never really recovered from the 1984 breakup of AT&T, but the
> > > >kiss of death was the combination of the dotcom bust ...
> > >
> > > The Dotcom bubble took a lot of companies out, almost including
> > > Corning, whence came fiber optics.
> > Are you sure about that?
> > opticalfibrehistory.co.uk/who/charles-kuen-kao/
> > opticalfibrehistory.co.uk/where/
> >
> > John

> Here is the original paper:
This link works better:-
http://digital-library.theiet.org/deliver/fulltext/piee/113/7/19660189.pdf

John

John Walliker wrote:
> On Sunday, 18 December 2022 at 21:44:13 UTC, John Walliker wrote:
>> On Sunday, 18 December 2022 at 21:36:59 UTC, John Walliker wrote:
>>> On Sunday, 18 December 2022 at 19:56:39 UTC, Joe Gwinn wrote:
>>>
>>>>> Bell Labs never really recovered from the 1984 breakup of AT&T, but the
>>>>> kiss of death was the combination of the dotcom bust ...
>>>>
>>>> The Dotcom bubble took a lot of companies out, almost including
>>>> Corning, whence came fiber optics.
>>> Are you sure about that?
>>> opticalfibrehistory.co.uk/who/charles-kuen-kao/
>>> opticalfibrehistory.co.uk/where/
>>>
>>> John
>
>> Here is the original paper:
> This link works better:-
> http://digital-library.theiet.org/deliver/fulltext/piee/113/7/19660189.pdf
>
> John
>

1966 is an implausibly late date for the invention of optical fibre.

There's an old Latin tag to the effect that many claim credit for a
victory, but defeat is blamed on one alone. The number of folks from
around then that claim to have invented fibre optics is quite large.

For instance, Narinder Kapany used crude fibre bundles to form images
back around 1951. (I heard him speak when I was in grad school, 35ish
years back.)

https://en.wikipedia.org/wiki/Narinder_Singh_Kapany

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

On Sun, 18 Dec 2022 15:16:09 -0800 (PST), John Walliker
<jrwalliker@gmail.com> wrote:

>On Sunday, 18 December 2022 at 21:44:13 UTC, John Walliker wrote:
>> On Sunday, 18 December 2022 at 21:36:59 UTC, John Walliker wrote:
>> > On Sunday, 18 December 2022 at 19:56:39 UTC, Joe Gwinn wrote:
>> >
>> > > >Bell Labs never really recovered from the 1984 breakup of AT&T, but the
>> > > >kiss of death was the combination of the dotcom bust ...
>> > >
>> > > The Dotcom bubble took a lot of companies out, almost including
>> > > Corning, whence came fiber optics.
>> > Are you sure about that?
>> > opticalfibrehistory.co.uk/who/charles-kuen-kao/
>> > opticalfibrehistory.co.uk/where/
>> >
>> > John
>
>> Here is the original paper:
>This link works better:-
><http://digital-library.theiet.org/deliver/fulltext/piee/113/7/19660189.pdf>

Yes, you are right, and I should have been more precise.

What Kao et al invented in 1966 was the idea of optical fiber for
sure, but the best that could be done then was one decibel loss per
foot of fiber, which was not at all practical for communications uses,
and he appealed for better fiber.

What Corning did was to figure out how to make the necessary very low
loss fiber cheaply enough to make it practical both technically and
economically.

Current datacomm fibers (at 1310 nm) are about 0.5 dB per kilometer
and telcom fibers (at 1550 nm) are 0.4 dB/Km.

One decibel per foot is equivalent to 3.3 dB per meter, so the
comparison is about 3300 dB/Km to 0.5 dB/Km.

Here is a marketing video from Corning giving a good overview of the
manufacturing process. The keys were the bit about ultra purity, with
contaminant levels in the parts per billion, and drying the silica
preform before melting it into a solid glass "cane".

..<https://www.youtube.com/watch?v=j6Il97_gRYs>

And a more general news video on Corning's businesses and history
(including the dotcom bust):

..<https://www.youtube.com/watch?v=lopwv2ViNMw>

Joe Gwinn

Joe Gwinn wrote:
> On Sun, 18 Dec 2022 15:16:09 -0800 (PST), John Walliker
> <jrwalliker@gmail.com> wrote:
>
>> On Sunday, 18 December 2022 at 21:44:13 UTC, John Walliker wrote:
>>> On Sunday, 18 December 2022 at 21:36:59 UTC, John Walliker wrote:
>>>> On Sunday, 18 December 2022 at 19:56:39 UTC, Joe Gwinn wrote:
>>>>
>>>>>> Bell Labs never really recovered from the 1984 breakup of AT&T, but the
>>>>>> kiss of death was the combination of the dotcom bust ...
>>>>>
>>>>> The Dotcom bubble took a lot of companies out, almost including
>>>>> Corning, whence came fiber optics.
>>>> Are you sure about that?
>>>> opticalfibrehistory.co.uk/who/charles-kuen-kao/
>>>> opticalfibrehistory.co.uk/where/
>>>>
>>>> John
>>
>>> Here is the original paper:
>> This link works better:-
>> <http://digital-library.theiet.org/deliver/fulltext/piee/113/7/19660189.pdf>
>
> Yes, you are right, and I should have been more precise.
>
> What Kao et al invented in 1966 was the idea of optical fiber for
> sure, but the best that could be done then was one decibel loss per
> foot of fiber, which was not at all practical for communications uses,
> and he appealed for better fiber.
>
> What Corning did was to figure out how to make the necessary very low
> loss fiber cheaply enough to make it practical both technically and
> economically.
>
> Current datacomm fibers (at 1310 nm) are about 0.5 dB per kilometer
> and telcom fibers (at 1550 nm) are 0.4 dB/Km.
>
> One decibel per foot is equivalent to 3.3 dB per meter, so the
> comparison is about 3300 dB/Km to 0.5 dB/Km.
>
>
> Here is a marketing video from Corning giving a good overview of the
> manufacturing process. The keys were the bit about ultra purity, with
> contaminant levels in the parts per billion, and drying the silica
> preform before melting it into a solid glass "cane".
>
> .<https://www.youtube.com/watch?v=j6Il97_gRYs>
>
> And a more general news video on Corning's businesses and history
> (including the dotcom bust):
>
> .<https://www.youtube.com/watch?v=lopwv2ViNMw>
>
>
> Joe Gwinn
>

Up to the late '60s, Ma Bell was doing experiments with burying
corrugated metal waveguide for high-bandwidth transmission.

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

On December 15, Phil Hobbs wrote:
>> Optical computers run a million times faster than conventional computers, study reveals.
>> This extremely fast processing uses circularly polarized light instead of an electrical signal.
>> The process depends a great deal on the handedness of the circularly polarized beam, through
>> a crystal material that is sensitive to the beam of light.
>> https://www.science.org/doi/10.1126/sciadv.abq8246
> >
> It's an interesting idea, for sure. I can imagine some interesting
> measurement applications in time-resolved chemical sensing, which is a
> pretty hot area--you can watch molecules rearranging themselves during a
> chemical reaction.

There's an optics lab at Stanford which is close to that, so
they claim. They need attosecond resolution. Watching
chemical reactions, wow -

"These are ATOMS, we're looking at GOD!"
- Richard Feynman, viewing the first atomic force microscope

They just need a bit more funding - after the withdrawal from the
Afghan adventure, there ought to be plenty of loot looking for a home -

--
Rich

On Mon, 19 Dec 2022 15:13:23 -0500, Phil Hobbs
<pcdhSpamMeSenseless@electrooptical.net> wrote:

>Joe Gwinn wrote:
>> On Sun, 18 Dec 2022 15:16:09 -0800 (PST), John Walliker
>> <jrwalliker@gmail.com> wrote:
>>
>>> On Sunday, 18 December 2022 at 21:44:13 UTC, John Walliker wrote:
>>>> On Sunday, 18 December 2022 at 21:36:59 UTC, John Walliker wrote:
>>>>> On Sunday, 18 December 2022 at 19:56:39 UTC, Joe Gwinn wrote:
>>>>>
>>>>>>> Bell Labs never really recovered from the 1984 breakup of AT&T, but the
>>>>>>> kiss of death was the combination of the dotcom bust ...
>>>>>>
>>>>>> The Dotcom bubble took a lot of companies out, almost including
>>>>>> Corning, whence came fiber optics.
>>>>> Are you sure about that?
>>>>> opticalfibrehistory.co.uk/who/charles-kuen-kao/
>>>>> opticalfibrehistory.co.uk/where/
>>>>>
>>>>> John
>>>
>>>> Here is the original paper:
>>> This link works better:-
>>> <http://digital-library.theiet.org/deliver/fulltext/piee/113/7/19660189.pdf>
>>
>> Yes, you are right, and I should have been more precise.
>>
>> What Kao et al invented in 1966 was the idea of optical fiber for
>> sure, but the best that could be done then was one decibel loss per
>> foot of fiber, which was not at all practical for communications uses,
>> and he appealed for better fiber.
>>
>> What Corning did was to figure out how to make the necessary very low
>> loss fiber cheaply enough to make it practical both technically and
>> economically.
>>
>> Current datacomm fibers (at 1310 nm) are about 0.5 dB per kilometer
>> and telcom fibers (at 1550 nm) are 0.4 dB/Km.
>>
>> One decibel per foot is equivalent to 3.3 dB per meter, so the
>> comparison is about 3300 dB/Km to 0.5 dB/Km.
>>
>>
>> Here is a marketing video from Corning giving a good overview of the
>> manufacturing process. The keys were the bit about ultra purity, with
>> contaminant levels in the parts per billion, and drying the silica
>> preform before melting it into a solid glass "cane".
>>
>> .<https://www.youtube.com/watch?v=j6Il97_gRYs>
>>
>> And a more general news video on Corning's businesses and history
>> (including the dotcom bust):
>>
>> .<https://www.youtube.com/watch?v=lopwv2ViNMw>
>>
>>
>> Joe Gwinn
>>
>
>Up to the late '60s, Ma Bell was doing experiments with burying
>corrugated metal waveguide for high-bandwidth transmission.

Yes, I recall those days. Circular microwave wave-guides made by
winding a copper wire around a plastic tube, suppressing all
propagation modes except those with circumferential currents, if I
recall. It was the kind of thing that only Bell Labs could have
gotten to work, and even then it was a stretch.

All swept away by the arrival of practical fiber optical
communications.

Joe Gwinn