1 Dec
2011
1 Dec
'11
9:40 p.m.
On 2011 Dec 1, at 2:58 PM, Eric Smith wrote:
In the case of V.32 and such, the raw signal is by intention a
continuously varying (analog) signal (the carrier), done for the sake
of making it through the transmission medium. In this it is different
Without meaning to put words in your mouth, I get the feeling you
don't consider the signal quantized until it is cleaned up and in a
form acceptable to the receiving system (appropriate voltage levels
and synced to internal clocks).
In contrast, I consider it quantized if discrete information is
represented in the signal in discernible 'chunks' (quanta), be they
amplitude chunks, time chunks, or both - it is a property of the
signal, without any need for consideration of the receiving system.
The size of a given chunk may vary with noise/perturbations/
fluctuations/deformation to the signal, but none of those - within
limits - are sufficient to de-quantize the signal (to ref Dave's
reply). Any physical implementation of a discrete system will have
noise and signal deformation, even TTL signals. Sending data to an
external medium may mean a loss of reference (sometimes necessitating
the addition of self-referencing info), that still doesn't disqualify
the signal as being quantized.
On 12/01/2011 02:26 PM, Brent Hilpert wrote:
The data may not be quite the way you want it (it
includes clock
transitions)
I don't have an issue with that; that's the way it was written.
and you need a transformation to get it to the
way you want, but
that
transformation is (or primarily is) a discrete
process or
transformation.
That you can perform a process that takes non-time-quantized read
data signal from the drive and by a "discrete process" produce a
time-quantized output that is (usually) the same as the signal that
was originally written does not in any way prove that the read data
signal from the drive was time-quantized. That it is
possible to produce a data-separator that recovers the
data so reliably is
exactly the point.
That argument could be used to claim that the signal a V.32 modem
receives over a telephone line is voltage-quantized, because you
can sample it then apply a "discrete process" to reliably get the
original data transmitted from the other end.
In both cases, if the channel signal was actually quantized (in the
time domain for the magnetic disk, and in the amplitude domain for
the modem), there wouldn't be any need for a "discrete process" the
you describe. from the other signals being discussed. However, once
you analyse the
AC characteristics of that signal (demodulation), you find it is
quantised - in that AC context - in that there are a finite number of
discrete AC amplitude levels and a finite number of discrete phase
relations (which together produce the dibits or whatever they are
called for multi-bit bauds).
(Perhaps we
have different concepts of what 'time-quantised' or
simply 'quantised' means. While there are differences, one might
ask the same question of an async serial signal for a simpler
case to examine.)
It appears that you've identified the reason we disagree. I would argue that asynchronous (start/stop) serial as
received is
non-time-quantized, although the clock recovery is much simpler
than with a magnetic disk because the channel does not distort the
timing by more than a trivial amount in proportion to the transmit
rate. A UART does have to oversample the async receive data and
use a decision process (generally based on detection of the leading
edge of the start bit) to identify the appropriate bit cell
boundaries to time-quantize the data.
And of course I argue the opposite, the signal is still time-
quantized because the information is still in discernible chunks.
With knowledge of the bit rate, a given event within a limited range
allowing for noise has the same discrete meaning. (e.g. a positive
level at [x +/-e uS] after the start bit means bit 5 is 1.)
There may be an analog relationship between the signal start bit (and
hence following bits) and the receiver bit clock, but that does not
mean the signal is not quantized.
A synchronous serial channel provided with an explicit
clock signal
is time quantized.