Robert Redelmeier wrote:
My overall point is that the effects of noise are different on
digital and analog signals. Digital is NOT immune to noise.
It just tolerates low levels continuously with impunity.
Bug digital is vulnerable to spikes/high noise levels, even
if fairly rare. Analog rides these through but is vulnerable
to continuous low levels.
Your comment about "All it takes is a single spike to ruin
one bit, then the 12 kbit packet is shot." shows you do
not understand the nature of what's on the wire
Baseband signalling voltage levels. No carrier wave.
Any signal carrying intelligence involves changing some aspect, such as
voltage, frequency or phase. If you have a digital square wave on a wire,
you will have a sine wave, at the frequency of the square wave or
fundemental frequency, along with another at 3 times the the fundemental
frequency and 1/3 the amplitude and another at 5 times the fundemental at
1/5 the amplitude, etc to infinity (assuming infinite bandwidth). While
you may have a digial circuit, you have a lot of analog signals. You can
then take a square wave at, for example 1 kHz. You will see the various
sine waves, as described above, at 1 kHz, 3 kHz, 5 kHz etc. If you then
pass that square wave through a perfect low pass filter, with the cut off
frequency somewhere between 1 kHz and 3 kHz, you will remove all the
harmonics and have only a 1 kHz sine wave left. Further, since ethernet
NICs are connected to the cable via a transformer, the signal must be
analog.
Ethernet uses modulation to create a signal that can transit the cable, NIC
transformers etc. Please read up on Manchester encoding, as used in
10baseT. Incidentally, to pass a signal through a transformer, without
distortion, you need a signal that is balance, with equal energy above and
below 0V. If there is any DC bias in that signal, it will be lost. For
mathematical proof of this, read up on differential calculus.
Dial pulses won't do it. Hanging up the phone won't do it.
Both of these may be mechanical contacts that produce
very sharp edges (and bounces).
Please explain how the interfering signal would be generated
and propagated, in a manner that stands up to mathematical
or electrical engineering scrutiny.
OK. You have a set of POTS contacts (either pulse dialer or
on-hook) that are open but have voltage across them. Then they are
mechanically closed (and probably bounce open & closed a few times
in a ms or so). Current suddnly starts to flow, rapidly building
a magnetic field. This change in magnetic field induces a current
on all nearby conductors. Twisted pair is good because the two
conductors are balanced (both generating and receiving pairs),
but it isn't perfect, and magnetics follow an inverse square law.
I'll describe pulse dialing, as it's a simpler case than the hook switch,
because it's a periodic function. With pulse dialing (assuming 50 % duty
cycle for simplicity), you will typically have a 10 pulse per second, or 10
Hz, square wave. That square wave, as I have described before, is composed
of a sine wave at 10 Hz and a series of harmonics. This is a fundamental
fact, proven both mathematically and in experiment. And again, as has been
irrefutably proven over the years, the amplitude decreases in proportion to
harmonic order. There is absolutely no way, for a square wave at 10 Hz, to
generate significant energy levels at ethernet frequencies. It is
physically impossible. Now when you take the hook switch contacts
operating, you have a similar situation WRT Fourier analasys, but since
it's not a periodic function, the calculations become more complex.
So, yes, contacts can generate noise at higher frequencies, but by the time
you get to ethernet frequencies, there is simply not sufficient energy to
cause interference. In order for such pulsing to have sufficient energy in
the range of frequencies used by ethernet, it would have to be of extremely
short duration. You will then find circuit capacitance and inductance will
not permit sufficient bandwidth for such a pulse to exist. That pulse will
then be much broader and lower amplitude, due to the low pass filtering,
caused by that capacitance and inductance.
Further, how could any phone device that generates such
interference get FCC approval for connection to the phone
line? If it's producing such interference to ethernet,
it might also do so to various radio services.
Extremely easily. I'm talking about intermittant noise,
not continuous interference. Hard to detect transients,
especially in analog signals.
Situations such as dialing and hook switch operation are very easy to
repeat. Either the phone creates such interference or it doesn't. Again,
if the phone managed to produce signals capable of interfering with
internet are produced, they won't get far, due to the low pass filter
effect, that exists in every circuit.
You have yet to describe any source of such interference. If you still
believe that pulse dialing or hook switch operation can, you're proving
that you don't know what the hell you're talking about.
