bandwidth question

[Tony Hammitt]

All interesting points, but a little inaccurate.  To be very very
technical here's the following:

What is being sent over fiber optic lines is light (you all knew that)
and light has a certain maximum bandwidth.  This bandwidth maximum is
due to the number of waves that can pass a certain point in a given
time, i.e. the frequency.  For this discussion, light is to be assumed
to be only a wave.  The particlate nature is not relevant since we are
not talking about how much energy is represented in the individual
photons. (A very small amount, something that can't be measured
accurately  enough to count individual photons.  Gross affects are more
measurable.)

So, let's say that we're interested in putting a random string of bits
over a single frequency of light that is bright enough to see without
instruments.  How much info can we transmit?

It all comes down to how we set up the transmitter in the end, but lets
assume that we have some infinitely fast warm up time on a solid state
laser, i.e. we can turn it on or off literally anytime we want to, and
we can choose exactly how intensely the light comes out.  This
translates to having an analog signal, which, if it were really just a
wave, would mean that we could encode infinite amounts of information. 
Given a totally lossless transmission media, we could recieve and
decode all of the data, also assuming a totally accurate reciever.

Now let's look at reality.  There is a maximum switching speed with any
system, esp. one based on solid state devices.  The warm up time is not
zero.  The media loses photons because of bends, twists, impurities and
changes in the diameter of the fiber.  The reciever is only so
accurate. We can't exactly control how bright the light is (or how many
photons there are, which ever makes you feel better).  The faster the
laser switches intensities, the more the frequency spreads out due to
the uncertainty principle (delta E times delta t must be greater than
Plank's constant divided by 2*pi).  All of these limit the bandwidth.

So, what do we get?  Any reason listed above cuts our theoretically
infinite bandwidth down to finite.  But, if the light is bright enough,
we can ignore the more subtle effects of the lossy medium.  Solid state
lasers can turn on and off tens billions of times per second.  They
send out a narrowly defined range of frequencies, so as long as the
frequency ranges don't overlap, we can use several frequencies on the
same fiber.  (They will theoretically always overlap somewhat.) 
Splitting the frequencies back out on the recieving end is fairly
straightforward and we can interpret the signal as accurately as we can
send it (for now at least).  Destructive interference could be said to
be occuring within the fiber at just about any point but that doesn't
matter as long as the signal is _changing_ and we're not measuring it
where the signals are still multiplexed.

In the end, a single fiber of the best possible quality can carry about
a hundred frequencies easily with each frequency carrying up to about a
terabit with tomorrow's switching technologies.

On to copper...

Copper wires used to carry things like ethernet signals should not be
thought of as sending their information via having an electron travel
down the wire.  This can occur but it is excedingly slow, megavolt DC
transmission lines still could measure their electron flow in
centimeters per second.  What really occurs is that the _signal_
travels down the wire like sound through water.  The water molecules
disturb their neighboring molecules' electrical fields (their electron
clouds) with their motion, which was previously induced by some other
neighbors' motion that was ultimately caused by whatever is making the
sound.  Air molecules transmit sound a bit differently considering
their much greater space between particles.

Even the wiring in your house has to be considered to be sending a
signal instead of just having a voltage like with DC.  Sure, it's a low
frequency signal, but it is sent with quite a lot of strength.

Signals travel down copper at almost the speed of light (At the speed
of light in fact, just the speed of light in copper.  It comes out to
about 1/9 of the speed of light in vacuum/air = c.  Copper is opaque to
visible EM radiation (light) but is transparent to some of the
spectrum, depending on the thickness of the copper).  The speed of
light in fiber is something like 3/5 c.

The reason we use twisted pairs of wires for networking (or coaxial
cables) is to keep the signal integrity up.  Waves can be guided down a
line much easier if they have something to reflect off of, to put it in
a sort of wierd way.  Twisted pairs have one wire have a signal going
from high to low positively while the other goes negatively.  Coaxial
lines just change the voltage in the wire which induces a negative
signal in the shielding.  Things like gigabit copper ethernet send and
recieve signals on all four pairs of wires, each pair having 250MHz
signals.  Crosstalk only matters near the ends of the line.  If you
don't believe me, get a friend and a slinky and try sending two
different signals at once, it always works independently of what the
net result is in the middle of the slinky.

Light also has to have a signal to bounce off of, luckily it carries
its own with it, putting energy back and forth between electrical and
magnetic fields.  The frequency depends on the amount of energy just
because of the way the universe is set up.  More energy means that it
has to change back and forth faster.  It's much easier to explain with
math.

So, what is comes down to is that signals can be sent reliably over
copper or fiber.  Choice of media depends on how much you want to spend
per end.  Staying in the electrical signal space will keep costs down
but it sends signals in much the same way as with sending light down a
fiber.  It's the signal that is the important part.

Now what you do with the signal is up to you.  My earlier post about
ATM is more relevant to the original subject.  If this was too long of
an explanation, sorry.  Those degrees in Physics and EE have some
practical uses sometimes.

Regards,

Tony Hammitt
Jeffrey Watts wrote:
> 
> On Wed, 10 May 2000, Jeff McCright wrote:
> 
> > I am not sure what you are asking, but Electricity travels at the
> > speed of light, or so I'm told. As to the bandwidth, since the carrier
> > signal of the fiber optic(light has a much smaller wavelength,
> 
> That's incorrect.  It's always a small fraction.  Remember, electricity is
> the flow of electrons, and they are matter that have mass.  Light is a
> bizarre particle-wave, and is energy.
> 
> Einstein tells us interesting things happen when you accelerate mass to
> near c speed.  Electrons at near-light speed would do very interesting
> things.  Can you say atom smasher?
> 
> > thus carry more data and because it is light, it is unaffected by
> > crosstalk and other forms of RF interference.
> 
> But it is very susceptable to impurities and damage to the medium.  Not to
> mention destructive interference.  Remember the wave behaviour of light?
> 
> > Thus you have Fiber that can allow more data to flow simultaneously
> > with much less retransmissions of data allowing for greater or
> > Bandwidth ("Speed"). Fiber will out perform Copper, period.
> 
> Bandwidth is capacity, not latency (speed).  Fiber has infinite capacity
> (through dense wave division multiplexing).  Copper does not, since data
> is transmitted through copper as a series of voltage lows and highs and it
> needs the frequency of the shifts to be _much_ less than the speed of
> light, assuming that you want to make networking equipment that doesn't
> require a government-sized budget to purchase.
> 
> Also, your statement about retransmissions is misleading.  If you are
> running Ethernet over fiber, you'll still have collisions assuming you are
> using single-frequency fiber.  I also haven't seen any hard evidence that
> states that fiber gives more reliable data transmission than adequately
> shielded and properly installed coax or cat5.  Enlighten me if I'm
> mistaken.
> 
> Fiber is used because it has near-c transmission speed (latency) and it
> has much higher capacity (well, infinite).  Copper is used because it is
> cheap to manufacture and manipulate and it is resilient to damage.
> 
> To answer the original question:
> 
> > so..if we setup a piece of fiber, and broke down all the packets (A
> > and B) transmitted into two smaller units (A1, A2, B1, and B2) and
> > then transmitted them alternating between packets A and packet B.
> > (sending A1, then B1, A2, B2) which would essentially increase latency
> > for each user, but at the same time, allow two users to transmit at
> > the same time.
> 
> This is incorrect.  You are _not_ allowing transmission at the same time.
> You are allowing packets to be broken down into smaller packets, which are
> then sent alternatively.  This will actually _decrease_ capacity (as you
> are still sending the same amount of data down the wire, but you'll
> require additional information in the little packets to recontruct them
> into the bigger ones).  It would only increase latency in that the users'
> machines would have to spend an extra step to recombine the
> mini-packets.  This latency is not due, however, to the method of
> transmission.  Light through fiber always travels at a fixed
> speed.  Nothing you do (programmatically) can change that.
> 
> > appearing to run at the copper speed for both (2:1) essentially
> > halfing the speed of fiber and doubling the bandwidth... in
> > theory...would the overall network load be better (if the network
> > segment were supporting a large user base) or would the result be a
> > complicated network that would run at the same speed as the original?
> > note: im only speaking in theory, not practical..
> 
> Light over fiber is faster (higher latency) and has more capacity
> (bandwidth) than electricity over a metal.  Period.  There currently are
> fiber solutions that can transmit faster (NIC to NIC) and with more
> capacity than any copper solution.  I'm sure there are also slower fiber
> solutions than some copper ones.  But that is the speed of the _solution_,
> NIC to NIC, not the capacity or potential speed of light over fiber.
> 
> So the answer is yes, the network load would be better on _any_ size user
> base if you were using a fiber solution that was faster and had more
> capacity than any copper solution available.
> 
> I kind of see what you were trying to get at, but you have to see that
> your question answers itself -- it's basically saying "if a light-based
> network is faster and has more capacity than copper, does a network based
> on it have more capacity and more speed than a copper-based one?"
> 
> J.
> 
> o-----------------------------------o
> | Jeffrey Watts                     |
> | watts at jayhawks.net         o-----------------------------------------o
> | Systems Programmer         | "Outside of the killings, Washington    |
> | Network Systems Management |  has one of the lowest crime rates in   |
> | Sprint Communications      |  the country."                          |
> o----------------------------|  -- Mayor Marion Barry                  |
>                              o-----------------------------------------o