Potentially a new idea for storing information in time itself and thus a new memory chip.

[Skybuck Flying]

"Uwe Hercksen" wrote in message

The idea is as follows:

A signal of zero's and one's is sent across the wire.

"
Hello,

it is not a new idea, this was used many decades ago, also with
ultrasonic waves in glas or metall wires.
It is no random access memory.
"

Multiple wires could be used to represent "memory cells".

This way it does become random memory access for the wires

Bye,
Skybuck.

 

[HVAC]

What are you? Like 12?

What we could do is the following:

1. We can design a spaceship/spacecraft, like voyager

2. and send it into space with it's sole purpose to function as a
reflector for us.

3. And then we can use it to keep sending laser beams at it.

4. Then it becomes our way of storing information in the galaxy.

5. The further away it gets the more information we can store.

Hopefully it will then be possible to eventually store years of data.

First 1 year then 10 years then 100 years and then 1000's of years.

That would be cool.

And all we would need for it to recollect data is laser beams and such

Perhaps if the laser beam starts to act weird we might also discover
some weird fluctuations in space ! For new theories or something or new
discoveries !

Bye,
Skybuck.

 

[[email protected]]

What we could do is the following:

1. We can design a spaceship/spacecraft, like voyager

2. and send it into space with it's sole purpose to function as a reflector
for us.

Voyagers (and more modern interplanetary probes) use two way ranging
(or even three way ranging).

In this system, a well known data sequence is sent towards the probe
at an exactly known frequency. The interplanetary probe receives the
stream, performs an _exact_ fractional frequency change (say 1/2 or
240/221) and sends the data immediately back to Earth.

When the signal is received on Earth, the distance to the probe can be
calculated with a fraction of the wavelength (at least in principle).
Since the frequency change ratio is exactly known, the doppler shift
also gives a very accurate radial speed (regardless of probe local
oscillator stability). To get the tangential position and speed,
celestial mechanics can be used to reduce the possible solutions.

The frequency change at the probe is required, since if it would
transmit the response back on the same frequency, it would block the
receiver.

Anyway, this principle would make "active" mirrors on the lunar
surface feasible at giga or terabit speeds with somewhat sensible
power levels.

3. And then we can use it to keep sending laser beams at it.

This would still need some frequency shifting system on visible
wavelengths.

If the receiving and transmitting stations could be installed at
wildly separate locations on the Moon, receiving the signal from with
one telescope, amplify with erbium amplifiers, transfer it to some
location in an other crater with optical fiber and somehow amplify it
to kW level and send it back to Earth with an other telescope. Most
likely, the sending and receiving stations on Earth would also have to
be quite far from each other, possibly on different continents, in
order to avoid direct leakage from Tx to Rx.

 

[Uwe Hercksen]

Multiple wires could be used to represent "memory cells".

This way it does become random memory access for the wires

Hello,

if you store a lot of bits in each wire, it is still sequential access.
One bit per wire is useless.

Bye

 

[Skybuck Flying]

"Uwe Hercksen" wrote in message

Multiple wires could be used to represent "memory cells".

This way it does become random memory access for the wires

"
Hello,

if you store a lot of bits in each wire, it is still sequential access.
One bit per wire is useless.
"

No just the bits in the single wire would be sequential but that's ok.

That's what computers do anyway to ram. They like storing bytes, and groups
of bytes.

Thus such a system which stores bits is usuable. If it must be then
everything could be load into the computer, manipulated and send back.

So there are possibilities plenty.

So for example the first wire stores a million bits, the second wire stores
a million bits, the third wire stores a million bits.

This means the million bits within the wire are indeed sequential, but the
wires themselfes could be seeked and would be like random access memory.

Each wire has their own sender and receiver to keep it looping. So each wire
is a loop which can be tapped into.

Each wire can be though of as a "seekable memory cell"

Ofcourse the sequential bits can also be manipulated but that would be a
more expensive operation...

Then again if the timing is just right... then perhaps even single bits
could be manipulated and replaced

Bye,
Skybuck.

 

[Jamie]

What I want to know is...

Can you take a stream of entangled photons and split-route them into two separate

(very low loss) fiber optic loops, then take one of the loops a mile
away, then change

the polarity of one loop and instantaneously detect the same change in
polarization

in the distant loop?

That would be cool.

Hi,

This 'Quantum Data Buffering' experiment EricP mentioned shows that
there is no entanglement connection between the two beams once they have
been separated in the "entanglement generation" block, since they
disprove the "instantaneous spooky action at a distance" in their
experiment by slowing one beam and detecting the quantum states of each
beam at different times. Both beams may be exact mirror images of each
other, but other than that they have no more connection to each other
than any other two beams of light do.

cheers,
Jamie

 

[Mark Thorson]

Chips like the Intel 1103, of course, also had one circuit cell per
bit, because they were shift-register serial memories.

No, it wasn't. The 1103 was a 1K x 1 DRAM. Maybe
you're thinking of the 1401?

 

[John Devereux]

Hello,

it is not a new idea, this was used many decades ago, also with
ultrasonic waves in glas or metall wires.
It is no random access memory.

How about racetrack memory

<http://en.wikipedia.org/wiki/Racetrack_memory>

 

[[email protected]]

If one could implement something analogous to a delay line on a chip,
so that a small number of transistors could regenerate a larger number
of bits, then, even if the memory is slower, there would be a
potential for a significant increase in memory density.

The bucket brigade devices used in analog audio delay lines in the
1980's could store audio samples typically in 512 or 1024 stages,
clocked by a serial clock. The audio SNR was not that great, so
perhaps 4-6 bits could reliably be stored in each stage (well).

Since the charge transfer was not 100 %, some electrons remained in
the cell, which were added to the new charge transferred from the
previous stage, thus digital regeneration would be required quite
frequently, perhaps after 1024 stages. Thus, n x 1024 well
organization would make sense as a disk replacement, with relatively
short "seek" times and fast data transfers.

Of course, current CCD camera layout are quite similar and capable of
storing multiple bits in each well. If the maximum charge for each
well is 100000 electron charges and if we could detect the difference
of 4 electron charges, the cell could store 14 bits or 16384 different
states. For reliable operation, some ECC system across multiple shift
registers would be needed, to reduce the quantum effects.

 

[Jamie]

That's a shame.

Don't forget I am wrong more than 50% of the time

cheers,
Jamie

 

[Jasen Betts]

Quite right. Still, although it's not a new idea, right now DRAMs
require one transistor (or at least one diode) for every bit, where a
reverse-biased diode acts as a capacitor.

Chips like the Intel 1103, of course, also had one circuit cell per
bit, because they were shift-register serial memories.

If one could implement something analogous to a delay line on a chip,

per bit electrical delay lines are bulkier than transistors.

so that a small number of transistors could regenerate a larger number
of bits, then, even if the memory is slower, there would be a
potential for a significant increase in memory density.

sounds like "bubble memory"

I don't think there's any technology that makes this practical today,
but _in theory_, the ideal memory would be something like a Williams
tube - where a large number of bits are selected from by means of
deflection voltages, without any need to manufacture one home for each
bit. Even by a process as simple as lithography.

something like that can be done using a scanning tunelling electron
microscope (or something lik that) to manipulate individual atoms
the main problem is miniaturisation

 

[Jamie]

The best delay-line memory would be a few hundred kilometers of
single-mode optical fiber. That could store gigabits, at great expense
and absurdly slow access times.

I wonder if a terahertz radioation transceiver was used that could
actually be a shorter length of fiber with high capacity and faster
access times.

Isn't an flipflop a kind of 1bit delay line memory?

cheers,
Jamie

 

[[email protected]]

The best delay-line memory would be a few hundred kilometers of
single-mode optical fiber. That could store gigabits, at great expense
and absurdly slow access times.

To get 1 gigabit storage capacity, you would have to use DWDM. The
delay in a 200 km long fiber is about 1 ms, thus, at 10 Gbit/s, the
storage capacity is 10 Mbits or about the capacity of some old floppy
disk.

Standard systems using 80 or 160 wavelengths multiplexed into a single
fiber can store 0.8-1.6 Gbits.

Using only 100 km of fiber would cut down the attenuation
sufficiently, so that optical erbium amplifiers (with have some
wavelength limitations) are not required on the way. Without band
limiting amplifiers in the system, a large wavelength range could be
used with a larger number of wavelengths.

The average "rotational latency" would be 25 us and practically zero
"track-to-track seek time" (between wavelengths). This would be a
quite expensive, but fast 100-200 MB disk replacement for PCs .

 

[josephkk]

Quite right. Still, although it's not a new idea, right now DRAMs
require one transistor (or at least one diode) for every bit, where a
reverse-biased diode acts as a capacitor.

Chips like the Intel 1103, of course, also had one circuit cell per
bit, because they were shift-register serial memories.

Nope, not serial. Bog standard DRAM. YCLIU

 

[josephkk]

The best delay-line memory would be a few hundred kilometers of
single-mode optical fiber. That could store gigabits, at great expense
and absurdly slow access times.

John

1. it has been done. 2. you can't go several hundred kilometers at a
crack, there is too much attenuation.
Land based systems go 40 km, transoceanic cables have various amplifiers
and regenerators along the way.

 

[Skybuck Flying]

The best delay-line memory would be a few hundred kilometers of

single-mode optical fiber. That could store gigabits, at great expense
and absurdly slow access times.

"
To get 1 gigabit storage capacity, you would have to use DWDM. The
delay in a 200 km long fiber is about 1 ms, thus, at 10 Gbit/s, the
storage capacity is 10 Mbits or about the capacity of some old floppy
disk.

Standard systems using 80 or 160 wavelengths multiplexed into a single
fiber can store 0.8-1.6 Gbits.

Using only 100 km of fiber would cut down the attenuation
sufficiently, so that optical erbium amplifiers (with have some
wavelength limitations) are not required on the way. Without band
limiting amplifiers in the system, a large wavelength range could be
used with a larger number of wavelengths.

The average "rotational latency" would be 25 us and practically zero
"track-to-track seek time" (between wavelengths). This would be a
quite expensive, but fast 100-200 MB disk replacement for PCs .
"

This reminds me of google which supposedly have their own fiber optic
networks.

Perhaps they might be interested in using their optic networks as some kind
of fast harddisk.

But if it would be 1 millisecond access time then they would probably not be
interested in it.

That was the figure you mentioned I think...

Though you also mentioned 25 nanoseconds or so... not sure what that figure
is about.

Would that be the access time for a single bit ?!

I'm being a bit vague/lazy though... since it doesn't really concern me and
is probably a whacky idea anyway

But maybe "you hardware people" can make something lol

Bye,
Skybuck.

 

[[email protected]]

That doesn't sound like a very good delay-line memory, so how could it
be the best delay-line memory?

If you want more massive storage, and are willing to put up with even
more absurdly slow access times, I can think of a way to do it at a
relatively small expense. Point a laser at a retroreflector on the
Moon.

This has been done since 1969. There are small reflectors on the
Apollo ALSEP packages an the laser beam is constantly used to measure
the distance between the Earth and the Moon. Typically 1-2 m
telescopes are needed both for the laser as well as for the receiver
and all you get back is a few photons.

Thus, the data rate would be a few bits/second, thus the total storage
capacity is a few bits.

To get some meaningful storage capacity, at lest 1 Gbit/s data rate
would be required, With the largest (10 m) optical telescopes
available _and_ with 100x laser power _and_ sending a new (football
field size) reflector to the moon, this might be doable .

 

[Skybuck Flying]

Here is another question for ya:

Suppose a living room is stacked full with fiber optics.

How many kilometers of wire would that be ?!?

Like micro-fiber-wire or something !

An estimation is ok

Bye,
Skybuck

 

[Skybuck Flying]

According to windows live mail, your message kinda screwed up... it was an
attachment ?!?

Was that deliberate or a gnu/linux mistake ?

Bye,
Skybuck