PDA

View Full Version : GMSK bandwidth???


bhargava
03-25-2008, 07:40 PM
hi all...
I've implemented a GMSK modem in MATLAB and also obtained the PSD of th
modulated GMSK signal...

In doing so I'm being haunted by an inconsistency...

we know that MSK and GMSK signal power spectral density theoretically hav
a main lobe width of 1.5*Rb (where Rb is the bit rate)hence theoreticall
bandwidth of GMSK and MSK signal is considered as 1.5*Rb.(here the PS
will be nearly 40 dB down compared to the value at F=Fc)

but GSM 5.04 standard specifies bitrate Rb=270.833kbps and

bandwidth/channel=200KHz now this is not as given in theory according
theoritical calculations the band width ie., 1.5*Rb will be 406.2495KH
now this means the power level at the cut off band width inGSM ha
attenuation way less than 40 dB...


now here where am I going wrong here???

and if this reasoning is true why/how does GSM specify a BW much less tha
theoretical band width??


Regards...

Bhargava

John
03-26-2008, 01:39 AM
On Mar 25, 2:40 pm, "bhargava" <[email protected]> wrote:
> hi all...
> I've implemented a GMSK modem in MATLAB and also obtained the PSD of the
> modulated GMSK signal...
>
> In doing so I'm being haunted by an inconsistency...
>
> we know that MSK and GMSK signal power spectral density theoretically have
> a main lobe width of 1.5*Rb (where Rb is the bit rate)hence theoretically
> bandwidth of GMSK and MSK signal is considered as 1.5*Rb.(here the PSD
> will be nearly 40 dB down compared to the value at F=Fc)
>
> but GSM 5.04 standard specifies bitrate Rb=270.833kbps and
>
> bandwidth/channel=200KHz now this is not as given in theory according o
> theoritical calculations the band width ie., 1.5*Rb will be 406.2495KHz
> now this means the power level at the cut off band width inGSM has
> attenuation way less than 40 dB...
>
> now here where am I going wrong here???
>
> and if this reasoning is true why/how does GSM specify a BW much less than
> theoretical band width??
>
> Regards...
>
> Bhargava

You don't need 400 kHz of bandwidth to demodulate the GSM signal.
Typical final IF bandwidth is 180 kHz or so.

John

bhargava
03-26-2008, 03:53 PM
>On Mar 25, 2:40 pm, "bhargava" <[email protected]> wrote:
>> hi all...
>> I've implemented a GMSK modem in MATLAB and also obtained the PSD o
the
>> modulated GMSK signal...
>>
>> In doing so I'm being haunted by an inconsistency...
>>
>> we know that MSK and GMSK signal power spectral density theoreticall
have
>> a main lobe width of 1.5*Rb (where Rb is the bit rate)henc
theoretically
>> bandwidth of GMSK and MSK signal is considered as 1.5*Rb.(here the PSD
>> will be nearly 40 dB down compared to the value at F=Fc)
>>
>> but GSM 5.04 standard specifies bitrate Rb=270.833kbps and
>>
>> bandwidth/channel=200KHz now this is not as given in theory accordin
o
>> theoritical calculations the band width ie., 1.5*Rb will be
406.2495KHz
>> now this means the power level at the cut off band width inGSM has
>> attenuation way less than 40 dB...
>>
>> now here where am I going wrong here???
>>
>> and if this reasoning is true why/how does GSM specify a BW much les
than
>> theoretical band width??
>>
>> Regards...
>>
>> Bhargava
>
>You don't need 400 kHz of bandwidth to demodulate the GSM signal.
>Typical final IF bandwidth is 180 kHz or so.
>
>John
>
yep..
the IF band width as specified by GSM 5.04 standard is exactly 200KHZ
but what i want to know is why are they allocating a band width which i
much smaller than the main lobe width of PSD of a GMSK signal(which i
1.5*Rb and in this case is 406KHz) will it not result in increase
adjacent channel interference??? theroritically all the books and paper
state the bandwidth to be equal to 1.5*Rb for both GMSK and MSK signals..
it is this inconsistency which is troubling me... as i'm doing a projec
based on GMSK modem i need a clear explaination regarding this so as t
defend my project...
please help me...
regards
Bhargava..

John
03-26-2008, 05:34 PM
On Mar 26, 10:53 am, "bhargava" <[email protected]> wrote:
> >On Mar 25, 2:40 pm, "bhargava" <[email protected]> wrote:
> >> hi all...
> >> I've implemented a GMSK modem in MATLAB and also obtained the PSD of
> the
> >> modulated GMSK signal...
>
> >> In doing so I'm being haunted by an inconsistency...
>
> >> we know that MSK and GMSK signal power spectral density theoretically
> have
> >> a main lobe width of 1.5*Rb (where Rb is the bit rate)hence
> theoretically
> >> bandwidth of GMSK and MSK signal is considered as 1.5*Rb.(here the PSD
> >> will be nearly 40 dB down compared to the value at F=Fc)
>
> >> but GSM 5.04 standard specifies bitrate Rb=270.833kbps and
>
> >> bandwidth/channel=200KHz now this is not as given in theory according
> o
> >> theoritical calculations the band width ie., 1.5*Rb will be
> 406.2495KHz
> >> now this means the power level at the cut off band width inGSM has
> >> attenuation way less than 40 dB...
>
> >> now here where am I going wrong here???
>
> >> and if this reasoning is true why/how does GSM specify a BW much less
> than
> >> theoretical band width??
>
> >> Regards...
>
> >> Bhargava
>
> >You don't need 400 kHz of bandwidth to demodulate the GSM signal.
> >Typical final IF bandwidth is 180 kHz or so.
>
> >John
>
> yep..
> the IF band width as specified by GSM 5.04 standard is exactly 200KHZ
> but what i want to know is why are they allocating a band width which is
> much smaller than the main lobe width of PSD of a GMSK signal(which is
> 1.5*Rb and in this case is 406KHz) will it not result in increased
> adjacent channel interference??? theroritically all the books and papers
> state the bandwidth to be equal to 1.5*Rb for both GMSK and MSK signals...
> it is this inconsistency which is troubling me... as i'm doing a project
> based on GMSK modem i need a clear explaination regarding this so as to
> defend my project...
> please help me...
> regards
> Bhargava..

It comes down to making the most of a finite resource, in this case
spectrum. The regulators, designers, and operators are motivated to
squeeze as many channels as they can into a fixed band of spectrum so
that more users can be supported. As long as the users can
communicate, why not? There is no law of physics that says they must
be spaced by null-to-null bandwidth.

John

bhargava
03-26-2008, 06:51 PM
>On Mar 26, 10:53 am, "bhargava" <[email protected]> wrote:
>> >On Mar 25, 2:40 pm, "bhargava" <[email protected]> wrote:
>> >> hi all...
>> >> I've implemented a GMSK modem in MATLAB and also obtained the PS
of
>> the
>> >> modulated GMSK signal...
>>
>> >> In doing so I'm being haunted by an inconsistency...
>>
>> >> we know that MSK and GMSK signal power spectral densit
theoretically
>> have
>> >> a main lobe width of 1.5*Rb (where Rb is the bit rate)hence
>> theoretically
>> >> bandwidth of GMSK and MSK signal is considered as 1.5*Rb.(here th
PSD
>> >> will be nearly 40 dB down compared to the value at F=Fc)
>>
>> >> but GSM 5.04 standard specifies bitrate Rb=270.833kbps and
>>
>> >> bandwidth/channel=200KHz now this is not as given in theor
according
>> o
>> >> theoritical calculations the band width ie., 1.5*Rb will be
>> 406.2495KHz
>> >> now this means the power level at the cut off band width inGSM has
>> >> attenuation way less than 40 dB...
>>
>> >> now here where am I going wrong here???
>>
>> >> and if this reasoning is true why/how does GSM specify a BW muc
less
>> than
>> >> theoretical band width??
>>
>> >> Regards...
>>
>> >> Bhargava
>>
>> >You don't need 400 kHz of bandwidth to demodulate the GSM signal.
>> >Typical final IF bandwidth is 180 kHz or so.
>>
>> >John
>>
>> yep..
>> the IF band width as specified by GSM 5.04 standard is exactly 200KHZ
>> but what i want to know is why are they allocating a band width whic
is
>> much smaller than the main lobe width of PSD of a GMSK signal(which is
>> 1.5*Rb and in this case is 406KHz) will it not result in increased
>> adjacent channel interference??? theroritically all the books an
papers
>> state the bandwidth to be equal to 1.5*Rb for both GMSK and MS
signals...
>> it is this inconsistency which is troubling me... as i'm doing
project
>> based on GMSK modem i need a clear explaination regarding this so a
to
>> defend my project...
>> please help me...
>> regards
>> Bhargava..
>
>It comes down to making the most of a finite resource, in this case
>spectrum. The regulators, designers, and operators are motivated to
>squeeze as many channels as they can into a fixed band of spectrum so
>that more users can be supported. As long as the users can
>communicate, why not? There is no law of physics that says they must
>be spaced by null-to-null bandwidth.
>
>John
>
the logic is quite straight forward and reasonable.....
thank you very much John....

Regards

Bhargava

Vladimir Vassilevsky
03-26-2008, 11:19 PM
bhargava wrote:


> the IF band width as specified by GSM 5.04 standard is exactly 200KHZ
> but what i want to know is why are they allocating a band width which is
> much smaller than the main lobe width of PSD of a GMSK signal(which is
> 1.5*Rb and in this case is 406KHz) will it not result in increased
> adjacent channel interference???

1. IIRC the GSM is tolerant to the interferrence of up to -9dB. The
channel spacing is selected accordingly (plus some headroom, of course).

2. For the any sensible frequency plan, the adjacent channels are never
used within the same cluster.

> theroritically all the books and papers
> state the bandwidth to be equal to 1.5*Rb for both GMSK and MSK signals...

They are talking about the first order null-to-null bandwidth, which is
largely irrelevant.


Vladimir Vassilevsky
DSP and Mixed Signal Design Consultant
http://www.abvolt.com