>Eric Jacobsen <eric.jacobsen@ieee.org> wrote:
>> On Thu, 11 Apr 2013 16:30:39 -0500, "rudykeram" <51467@dsprelated>
>> wrote:
>
>>>I am little unsure about line coding in the digital baseband transmission.
>
>>>I know that in passband modulation, for most cases we do apply some
>>>pulse-shaping signal to make sure that we are limiting the bandwidth of the
>>>symbols before modulation, so that we avoid ISI, and even from spilling to
>>>another the adjacent channel in Multi-channel modulation systems.
>
>(snip)
>
>> Yes, but if you have the whole channel to yourself (i.e., the wire
>> pair or whatever wired medium you're transmitting over), why would you
>> care how much spectrum you take up? In integrate-and-dump filter in
>> the receiver is exactly matched to square pulses, so you don't lose
>> any performance from transmitting the squarest pulses that you can.
>
>Well, for one, interference. For coaxial ethernet, where signals are
>pretty good at staying inside a coaxial cable, maybe not. However,
>the collision detect for coaxial ethernet depends on measuring the
>voltage on the cable, appropriately filtered, and so might be
>sensitive to the signal filter.
>For UTP ethernet, RFI goes up with frequency. However, attenuation
>(mostly due to skin effect) goes up, too, so the higher frequencies
>don't make it through a long cable, anyway. But to meet the RFI
>requirements, some filtering might be needed.
Are you worried about interference to the square-pulse signal or
interference from it? Square pulses will likely cause less
interference in a broad sense since the energy is spread out more than
it is with pulse filtering.
>> Even if the channel rounds off the pulses a bit, if it isn't enough to
>> degrade the performance (which it often isn't in a short-range wired
>> channel), then why would it matter?
>
>Yes. That presumably goes into the calculations on the appropriate
>coding, cables, and distance for each system.
>
>> I still suggest that you find a good basic comm text (like Sklar) so
>> that you can see how the signal chain works through a complete
>> architecture. It is possible to get easily confused by piecemealing
>> individual articles.
>
>-- glen
>Eric Jacobsen <eric.jacobsen@ieee.org> wrote:
>> On Thu, 11 Apr 2013 16:30:39 -0500, "rudykeram" <51467@dsprelated>
>> wrote:
>
>>>I am little unsure about line coding in the digital baseband
transmission.
>
>>>I know that in passband modulation, for most cases we do apply some
>>>pulse-shaping signal to make sure that we are limiting the bandwidth of
the
>>>symbols before modulation, so that we avoid ISI, and even from spilling
to
>>>another the adjacent channel in Multi-channel modulation systems.
>
>(snip)
>
>> Yes, but if you have the whole channel to yourself (i.e., the wire
>> pair or whatever wired medium you're transmitting over), why would you
>> care how much spectrum you take up? In integrate-and-dump filter in
>> the receiver is exactly matched to square pulses, so you don't lose
>> any performance from transmitting the squarest pulses that you can.
>
>Well, for one, interference. For coaxial ethernet, where signals are
>pretty good at staying inside a coaxial cable, maybe not. However,
>the collision detect for coaxial ethernet depends on measuring the
>voltage on the cable, appropriately filtered, and so might be
>sensitive to the signal filter.
>
>For UTP ethernet, RFI goes up with frequency. However, attenuation
>(mostly due to skin effect) goes up, too, so the higher frequencies
>don't make it through a long cable, anyway. But to meet the RFI
>requirements, some filtering might be needed.
>
>> Even if the channel rounds off the pulses a bit, if it isn't enough to
>> degrade the performance (which it often isn't in a short-range wired
>> channel), then why would it matter?
>
>Yes. That presumably goes into the calculations on the appropriate
>coding, cables, and distance for each system.
>
>> I still suggest that you find a good basic comm text (like Sklar) so
>> that you can see how the signal chain works through a complete
>> architecture. It is possible to get easily confused by piecemealing
>> individual articles.
>
>-- glen
>
Thank a lot guys. The answers make a lot of sense. Answers are very
explanatory and very well address the question that I ask.
Thanks again,
--Rudy
Reply by glen herrmannsfeldt●April 11, 20132013-04-11
>>I am little unsure about line coding in the digital baseband transmission.
>>I know that in passband modulation, for most cases we do apply some
>>pulse-shaping signal to make sure that we are limiting the bandwidth of the
>>symbols before modulation, so that we avoid ISI, and even from spilling to
>>another the adjacent channel in Multi-channel modulation systems.
(snip)
> Yes, but if you have the whole channel to yourself (i.e., the wire
> pair or whatever wired medium you're transmitting over), why would you
> care how much spectrum you take up? In integrate-and-dump filter in
> the receiver is exactly matched to square pulses, so you don't lose
> any performance from transmitting the squarest pulses that you can.
Well, for one, interference. For coaxial ethernet, where signals are
pretty good at staying inside a coaxial cable, maybe not. However,
the collision detect for coaxial ethernet depends on measuring the
voltage on the cable, appropriately filtered, and so might be
sensitive to the signal filter.
For UTP ethernet, RFI goes up with frequency. However, attenuation
(mostly due to skin effect) goes up, too, so the higher frequencies
don't make it through a long cable, anyway. But to meet the RFI
requirements, some filtering might be needed.
> Even if the channel rounds off the pulses a bit, if it isn't enough to
> degrade the performance (which it often isn't in a short-range wired
> channel), then why would it matter?
Yes. That presumably goes into the calculations on the appropriate
coding, cables, and distance for each system.
> I still suggest that you find a good basic comm text (like Sklar) so
> that you can see how the signal chain works through a complete
> architecture. It is possible to get easily confused by piecemealing
> individual articles.
-- glen
Reply by Eric Jacobsen●April 11, 20132013-04-11
On Thu, 11 Apr 2013 16:30:39 -0500, "rudykeram" <51467@dsprelated>
wrote:
>I am little unsure about line coding in the digital baseband transmission.
>
>I know that in passband modulation, for most cases we do apply some
>pulse-shaping signal to make sure that we are limiting the bandwidth of the
>symbols before modulation, so that we avoid ISI, and even from spilling to
>another the adjacent channel in Multi-channel modulation systems.
>
>But, I was reading some information about line coding
>(http://en.wikipedia.org/wiki/Line_coding), and it appears that you mostly
>apply pulse-shaping for RF transmissions.
>But wouldn't we have the same case for digital baseband modulation as
>well?
>
>let's say even if we apply Manchester encoding, we would still have sharp
>transitions when changing symbols from 0 to 1 or 1 to 0 (because even
>Manchester encoding is still a square wave), and wouldn't we have very high
>frequency components in those transitions? And wouldn't this require a very
>huge bandwidth?
>Then how come it is okay? Is there something in the RS422 transceivers that
>somehow smooths this out (e.g. apply some pulse-shaping, or reducing the
>fall-rise time between symbols to overcome the high frequency
>transitions)?
>
>I understand that Manchester encoding is a constant-weight code ( such that
>the average level over each code word is zero), but it is still a square
>waveform, and if we don't apply some sort of low-pass filter or
>pulse-shaping, we are going to have the high frequency components in the
>square waves, correct?
Yes, but if you have the whole channel to yourself (i.e., the wire
pair or whatever wired medium you're transmitting over), why would you
care how much spectrum you take up? In integrate-and-dump filter in
the receiver is exactly matched to square pulses, so you don't lose
any performance from transmitting the squarest pulses that you can.
Even if the channel rounds off the pulses a bit, if it isn't enough to
degrade the performance (which it often isn't in a short-range wired
channel), then why would it matter?
I still suggest that you find a good basic comm text (like Sklar) so
that you can see how the signal chain works through a complete
architecture. It is possible to get easily confused by piecemealing
individual articles.
Eric Jacobsen
Anchor Hill Communications
http://www.anchorhill.com
Reply by glen herrmannsfeldt●April 11, 20132013-04-11
rudykeram <51467@dsprelated> wrote:
> I am little unsure about line coding in the digital baseband transmission.
> I know that in passband modulation, for most cases we do apply some
> pulse-shaping signal to make sure that we are limiting the bandwidth of the
> symbols before modulation, so that we avoid ISI, and even from spilling to
> another the adjacent channel in Multi-channel modulation systems.
I believe so, but it seems to be often ignored in the discussion.
> But, I was reading some information about line coding
> (http://en.wikipedia.org/wiki/Line_coding), and it appears that you mostly
> apply pulse-shaping for RF transmissions.
> But wouldn't we have the same case for digital baseband modulation as
> well?
> let's say even if we apply Manchester encoding, we would still have
> sharp transitions when changing symbols from 0 to 1 or 1 to 0
> (because even Manchester encoding is still a square wave),
> and wouldn't we have very high frequency components in those
> transitions? And wouldn't this require a very huge bandwidth?
Yes. But the important point is that, if done right, the receiver
will still get the right value.
Maybe it is an appropriate time to note what Nyquist actually did.
We most often use Nyquist in terms of the rate needed to sample an
analog signal of a given bandwidth, but what he actually did was
figure of the bandwidth needed for a digital (telegraph) signal.
That is, with a series of dits and dahs going down a telegraph cable
with a finite bandwidth (especially high capacitance underwater cables)
how fast can they go? It is the dual problem to the sampling problem.
http://en.wikipedia.org/wiki/Duality_(mathematics)#Analytic_dualities
> Then how come it is okay? Is there something in the RS422 transceivers that
> somehow smooths this out (e.g. apply some pulse-shaping, or reducing the
> fall-rise time between symbols to overcome the high frequency
> transitions)?
For RS232, and I believe also RS422, there are requirements on both
inputs and outputs to withstand a certain voltage. For RS232, I believe
it is +/- 25V. That is usually done with a resistor on the output,
which, when feeding the capacitance of the (not impedance matched)
cable will naturally filter the signal.
> I understand that Manchester encoding is a constant-weight code ( such that
> the average level over each code word is zero), but it is still a square
> waveform, and if we don't apply some sort of low-pass filter or
> pulse-shaping, we are going to have the high frequency components in the
> square waves, correct?
If you look at the design of the receiver in a UART, it specifically
looks at the center of the expected bit. That is, after the edge of
the start bit is detected, wait one half bit time, verify the start bit,
wait one additional bit time, read the first data bit, etc. (UARTs use
a 16x (usually) clock to keep track of the timing.)
The assumption is that, after any such filtering, the center will have
the right value, or at least be closer than to any other value.
But yes, the actual filter details seem to be left out of most
discussions.
If you read the ethernet descriptions, the do note the required
bandwidth. 10 Mbit Manchester code requires 10MHz bandwidth.
(That is, twice the bandwidth that would otherwise be necessary.)
100baseTX uses a different code to limit the required (fundamental)
bandwidth to less than 35MHz. That is, if the MLT3 signal is filtered
to a bandwidth of 35MHz, the receiver will still see the right value.
With UTP cabling, the RFI would be too high if such filtering
wasn't done, but most books don't mention that.
-- glen
-- glen
Reply by rudykeram●April 11, 20132013-04-11
I am little unsure about line coding in the digital baseband transmission.
I know that in passband modulation, for most cases we do apply some
pulse-shaping signal to make sure that we are limiting the bandwidth of the
symbols before modulation, so that we avoid ISI, and even from spilling to
another the adjacent channel in Multi-channel modulation systems.
But, I was reading some information about line coding
(http://en.wikipedia.org/wiki/Line_coding), and it appears that you mostly
apply pulse-shaping for RF transmissions.
But wouldn't we have the same case for digital baseband modulation as
well?
let's say even if we apply Manchester encoding, we would still have sharp
transitions when changing symbols from 0 to 1 or 1 to 0 (because even
Manchester encoding is still a square wave), and wouldn't we have very high
frequency components in those transitions? And wouldn't this require a very
huge bandwidth?
Then how come it is okay? Is there something in the RS422 transceivers that
somehow smooths this out (e.g. apply some pulse-shaping, or reducing the
fall-rise time between symbols to overcome the high frequency
transitions)?
I understand that Manchester encoding is a constant-weight code ( such that
the average level over each code word is zero), but it is still a square
waveform, and if we don't apply some sort of low-pass filter or
pulse-shaping, we are going to have the high frequency components in the
square waves, correct?
Thanks,
--Rudy