Prev: IEEE Xplore question
Next: 64 x 64 FFT algorithm for 4096 pt. FFT
From: alberto.fuggetta on 29 Jun 2010 17:37
Hi,

I'm simulating an MSK based transmission system in Matlab.
At the TX, I generate the signal with 4 samples per symbol. It is then
passed to a rician channel + AWGN. At the receiver, after removing the
channel delay, I downsample so that to have 1 sample per symbol.
The resulting signal is equalized and demodulated.
If I use 1 sample per symbol instead of 4 at the transmitter, the BER
performance improves.
I didn't apply the correction factor -10log10(nSamp) to the SNR value while
calling the awgn function.
Is there any reasonable explanation?
Thanks

Alberto

From: xsong on 30 Jun 2010 00:17
On Jun 29, 5:37 pm, "alberto.fuggetta"
<alberto.fuggetta(a)n_o_s_p_a_m.gmail.com> wrote:
> Hi,
Did you change the channel for the two simulations? Remember, if you
want to simulate the different y sampling rate at the transmitter, the
equivalent channel should change accordingly with the sampling rate
change. Typically, you normalize the delay by the sampling rate, and
when the sampling rate changes so should the delay.

>
> I'm simulating an MSK based transmission system in Matlab.
> At the TX, I generate the signal with 4 samples per symbol. It is then
> passed to a rician channel + AWGN. At the receiver, after removing the
> channel delay,  I downsample so that to have 1 sample per symbol.
> The resulting signal is equalized and demodulated.
> If I use 1 sample per symbol instead of 4 at the transmitter, the BER
> performance improves.
> I didn't apply the correction factor -10log10(nSamp) to the SNR value while
> calling the awgn function.
> Is there any reasonable explanation?
> Thanks
>
> Alberto

From: alberto.fuggetta on 30 Jun 2010 03:32
Hi,

I used the ricianchan function. My path delays are scaled to the symbol
time, that is kept constant between the two simulations.
However the sampling time, that is also a parameter for the Matlab
function, varies from Tsym to Tsym/4 (when I use 4 samples per symbol).
However, my path delays are not integer multiples of the symbol time.
My problem is that I don't know which of the two simulations is the correct
one.

>On Jun 29, 5:37=A0pm, "alberto.fuggetta"
><alberto.fuggetta(a)n_o_s_p_a_m.gmail.com> wrote:
>> Hi,
>Did you change the channel for the two simulations? Remember, if you
>want to simulate the different y sampling rate at the transmitter, the
>equivalent channel should change accordingly with the sampling rate
>change. Typically, you normalize the delay by the sampling rate, and
>when the sampling rate changes so should the delay.
>
>>
>> I'm simulating an MSK based transmission system in Matlab.
>> At the TX, I generate the signal with 4 samples per symbol. It is then
>> passed to a rician channel + AWGN. At the receiver, after removing the
>> channel delay, =A0I downsample so that to have 1 sample per symbol.
>> The resulting signal is equalized and demodulated.
>> If I use 1 sample per symbol instead of 4 at the transmitter, the BER
>> performance improves.
>> I didn't apply the correction factor -10log10(nSamp) to the SNR value
whi=
>le
>> calling the awgn function.
>> Is there any reasonable explanation?
>> Thanks
>>
>> Alberto
>
>
From: cpshah99 on 30 Jun 2010 12:49
>Hi,
>
>I used the ricianchan function. My path delays are scaled to the symbol
>time, that is kept constant between the two simulations.
>However the sampling time, that is also a parameter for the Matlab
>function, varies from Tsym to Tsym/4 (when I use 4 samples per symbol).
>However, my path delays are not integer multiples of the symbol time.
>My problem is that I don't know which of the two simulations is the
correct
>one.
>
>>On Jun 29, 5:37=A0pm, "alberto.fuggetta"
>><alberto.fuggetta(a)n_o_s_p_a_m.gmail.com> wrote:
>>> Hi,
>>Did you change the channel for the two simulations? Remember, if you
>>want to simulate the different y sampling rate at the transmitter, the
>>equivalent channel should change accordingly with the sampling rate
>>change. Typically, you normalize the delay by the sampling rate, and
>>when the sampling rate changes so should the delay.
>>
>>>
>>> I'm simulating an MSK based transmission system in Matlab.
>>> At the TX, I generate the signal with 4 samples per symbol. It is then
>>> passed to a rician channel + AWGN. At the receiver, after removing the
>>> channel delay, =A0I downsample so that to have 1 sample per symbol.
>>> The resulting signal is equalized and demodulated.
>>> If I use 1 sample per symbol instead of 4 at the transmitter, the BER
>>> performance improves.
>>> I didn't apply the correction factor -10log10(nSamp) to the SNR value
>whi=
>>le
>>> calling the awgn function.
>>> Is there any reasonable explanation?
>>> Thanks
>>>
>>> Alberto
>>
>>
>

Although I use Matlab for my simulations, I have always tried to avoid this
inbuilt function.

As xsong said, you need to change your channel model when you go for more
samples/symbol. For example, lets say for a two path channel will one
symbol delay, the received signal should be y=conv(h,x)+n, where h=h1 h2].
So if you go for 4 samples per symbol then

y=h1*[x zeros(1,4)]+h2*[zeros(1,4) x]+n;

However, you will have to scale the noise appropriately.

Chintan
 | 
Pages: 1
Prev: IEEE Xplore question
Next: 64 x 64 FFT algorithm for 4096 pt. FFT