Bandwidth of single carrier system.

I have a fundamental question here regarding the bandwidth of QPSK system.


For example, I've a QPSK system generates +1 and -1 in 10Mbit/s. Then, I
modulate the QPSK signals with carrier frequency 62.5 MHz. What is the
bandwidth of my system? Is it 10 MHz? Let's assume it is a single-carrier
system. 

Thank you. 

On Wed, 19 Dec 2012 15:49:23 -0600, "kbkien" <37528@dsprelated> wrote:

>I have a fundamental question here regarding the bandwidth of QPSK system.
>
>
>For example, I've a QPSK system generates +1 and -1 in 10Mbit/s. Then, I
>modulate the QPSK signals with carrier frequency 62.5 MHz. What is the
>bandwidth of my system? Is it 10 MHz? Let's assume it is a single-carrier
>system. 
>
>Thank you. 

Generally the bandwidth of a single-carrier is the symbol rate.   If
your QPSK system is uncoded at 10Mbit/s then the symbol rate is 5MHz,
which can also be used as the bandwidth.

This depends somewhat on what you mean by "bandwidth", and what sort
of pulse shaping or filtering may be applied to the signal.   If
matched filtering is used then the 3dB bandwidth will be pretty close
to the symbol rate.   The required channel width may be much more than
that in order to accommodate the slope of the band edges.


Eric Jacobsen
Anchor Hill Communications
http://www.anchorhill.com

>On Wed, 19 Dec 2012 15:49:23 -0600, "kbkien" <37528@dsprelated> wrote:
>
>>I have a fundamental question here regarding the bandwidth of QPSK
system.
>>
>>
>>For example, I've a QPSK system generates +1 and -1 in 10Mbit/s. Then, I
>>modulate the QPSK signals with carrier frequency 62.5 MHz. What is the
>>bandwidth of my system? Is it 10 MHz? Let's assume it is a
single-carrier
>>system. 
>>
>>Thank you. 
>
>Generally the bandwidth of a single-carrier is the symbol rate.   If
>your QPSK system is uncoded at 10Mbit/s then the symbol rate is 5MHz,
>which can also be used as the bandwidth.
>
>This depends somewhat on what you mean by "bandwidth", and what sort
>of pulse shaping or filtering may be applied to the signal.   If
>matched filtering is used then the 3dB bandwidth will be pretty close
>to the symbol rate.   The required channel width may be much more than
>that in order to accommodate the slope of the band edges.
>
>
>Eric Jacobsen
>Anchor Hill Communications
>http://www.anchorhill.com
>

symbol rate is a completely different concept from bandwidth though they
are related.
In QPSK system say you start with data bitrate of 20Mbps, you then map bits
to symbols(2 bits become one symbol) which leads to symbol rate of 5Msps.
If you send these symbols directly out of digital system(through
DAC,RF))then it will contain power at very wide range of frequencies
centred on your carrier and you are going to have lot of problems with RF
engineers and other users of spectrum though it can be done in local short
distance cases.

In most cases you need to limit the frequency range to practical limits
that allows you to send enough information so that symbols will be
recovered by receiver and at same time you do not upset other users of
spectrum(legal and cost requirements) or upset RF designers.

In the case of FDM based systems such as your qpsk carried onto one carrier
then the symbols are shaped such that you pass a bandwidth numerically
equivalent to 1/2 symbol rate on either side of carrier i.e. you need to
pass a final bandwidth of 10 MHz and thus = symbol rate in Msps numerically
but remember they are two different things.

The criteria of passband width being half is based on -3dB point or so i.e.
you pass frequencies such that 5MHz is at -3dB. There will be some excess
bandwidth controllable through shaping filter.

Why 5MHz is enough is quite simple. If channel passes a 5MHz sinusoid then
you can imagine it will pass worse case symbol sequenc of
+1,-1,+1,-1...etc
draw it on paper and you will see.

The crieria for ofdm(as opposed to FDM) is different and worth comparing.

Kadhiem 

>
>symbol rate is a completely different concept from bandwidth though they
>are related.
>In QPSK system say you start with data bitrate of 20Mbps, you then map
bits

should read 10 Mbps(not 20 Mbps)

Kadhiem

>>
>>symbol rate is a completely different concept from bandwidth though they
>>are related.
>>In QPSK system say you start with data bitrate of 20Mbps, you then map
>bits
>
>should read 10 Mbps(not 20 Mbps)
>
>Kadhiem
>

Apologies, examples using numbers are so difficult to get them right in
posting.

Last go at it:

symbol rate is a completely different concept from bandwidth though they
are related. In QPSK system say you start with data bitrate of 10Mbps, you
then map bits to symbols(2 bits become one symbol) which leads to symbol
rate of 5Msps.

If you send these symbols directly out of digital system(through
DAC,RF))then it will contain power at very wide range of frequencies
centred on your carrier and you are going to have lot of problems with RF
engineers and other users of spectrum though it can be done in local short
distance cases.

In most cases you need to limit the frequency range to practical limits
that allows you to send enough information so that symbols will be
recovered by receiver and at same time you do not upset other users of
spectrum(legal and cost requirements) or upset RF designers.

In the case of FDM based systems such as your qpsk carried onto one
carrier
then the symbols are shaped such that you pass a bandwidth numerically
equivalent to 1/2 symbol rate on either side of carrier i.e. you need to
pass a final bandwidth of 5MHz and thus = symbol rate in Msps numerically
but remember they are two different things.

The criteria of passband width being half is based on -3dB point or so
i.e.
you pass frequencies such that 2.5MHz is at -3dB at baseband level. There
will be some excess bandwidth controllable through shaping filter.

Why 2.5MHz is enough is quite simple. If channel passes upto a 2.5MHz
sinusoid then you can imagine it will pass worse case symbol sequenc of
+1,-1,+1,-1...etc
draw it on paper and you will see.

The crieria for ofdm(as opposed to FDM) is different and worth comparing.

Kadhiem 

"kbkien" <37528@dsprelated> wrote in message 
news:5bOdnQe0fsL-q0_NnZ2dnUVZ_sudnZ2d@giganews.com...
>I have a fundamental question here regarding the bandwidth of QPSK system.
>
>
> For example, I've a QPSK system generates +1 and -1 in 10Mbit/s. Then, I
> modulate the QPSK signals with carrier frequency 62.5 MHz. What is the
> bandwidth of my system? Is it 10 MHz? Let's assume it is a single-carrier
> system.
>
> Thank you. 

"kaz" <37480@dsprelated> wrote:

> symbol rate is a completely different concept from bandwidth though they
> are related.

Symbol rate is completely different concept from bandwidth. They are 
independent and unrelated.
The choice of either is a matter of system preference.
(Think of UWB or PRML types of signal).

VLV

"Vladimir Vassilevsky" <nospam@nowhere.com> writes:

> "kaz" <37480@dsprelated> wrote:
>
>> symbol rate is a completely different concept from bandwidth though they
>> are related.
>
> Symbol rate is completely different concept from bandwidth. They are 
> independent and unrelated.
> The choice of either is a matter of system preference.
> (Think of UWB or PRML types of signal).

The OP was (fairly) clear about his question and said nothing of UWB or
other bandwidth-spreading techniques.
-- 
Randy Yates
Digital Signal Labs
http://www.digitalsignallabs.com

Keywords here are "pulse shaping filter" and "Nyquist ISI criterion" (see
Wikipedia on both), "root-raised cosine filter" and its "rolloff" as a
straightforward example.

"mnentwig" <24789@dsprelated> wrote in message 
news:KpqdnWspuYj8Ak_NnZ2dnUVZ_rudnZ2d@giganews.com...
> Keywords here are "pulse shaping filter" and "Nyquist ISI criterion" (see
> Wikipedia on both), "root-raised cosine filter" and its "rolloff" as a
> straightforward example.

Nyquist applies if we require symbols to be orthogonal to each other and use 
term "equivalent bandwidth".
This is particular case rather then general.


VLV