GMSK process

三三呦

已于 2024-09-17 21:13:47 修改

阅读量521

点赞数 6

分类专栏：通信原理文章标签：算法信号处理信息与通信

于 2024-07-18 23:04:20 首次发布

本文链接：https://blog.csdn.net/qq_27837513/article/details/140380801

版权

通信原理专栏收录该内容

1 篇文章 0 订阅

订阅专栏

Gaussian filter：

1. continuous-time impulse response of the Gaussian filter:

$h(t) = \frac{1}{\sigma\sqrt{2\pi}} e^{-\frac{(t )^2}{2\sigma^2}}\textup{}$

2. the frequency respone can be shown as,

$H(f) = \frac{1}{\sigma\sqrt{2\pi}} \int_{-\infty}^{\infty} e^{-\frac{t^2}{2\sigma^2}} e^{-j2\pi ft} dt$

cause

$-\frac{t^2}{2\sigma^2} - j2\pi ft = -\frac{1}{2\sigma^2} \left(t + j{2\pi\sigma^2 f}\right)^2 - 2\pi^2 f^2 \sigma^2$

then

$H(f) = e^{-2\pi^2 f^2 \sigma^2} \frac{1}{\sigma\sqrt{2\pi}} \int_{-\infty}^{\infty} e^{-\frac{1}{2\sigma^2} (t +j2\pi f\sigma^2)^2} dt$

get

$H(f) = e^{-2 \pi^2 f^2 \sigma^2}$

the 3dB bandwidth should equal to the bandwith of singal to satisfy the requirement of filter design, so,

$H(f)_{max}= 1\rightarrow H(B) = e^{-2 \pi^2 (B)^2 \sigma^2} = \sqrt{\frac{1}{2}}(3.01dB)$

then

$\sigma^2 = \frac{ln2}{4\pi^2B^2}$

when we normalize the t and use the t/T (T is symbol period), then the $\sigma$ can be represented as,

$\sigma^2 = \frac{ln2}{4\pi^2(BT)^2}$

function h = gaussdesignImpl(bt,span,sps)

% Check if filter order is even
sps_span = sps*span;
coder.internal.errorIf(isodd(sps_span),'signal:gaussdesign:OddFilterOrder')
filtLen = sps_span+1;

% Convert to t in which to compute the filter coefficients
t = linspace(-span/2,span/2,filtLen);

% Equation 6.53 of [1], page 290 is
% a = sqrt(log(2)/2)/B, here we use alpha = a/T
alpha = sqrt(log(2)/2)/(bt);

% Equation 5.54 of [1] is
% h = (sqrt(pi)/a)*exp(-(t1*pi/a).^2); 
% We use t = t1/T, alpha = a/T.  Then
% h = (sqrt(pi)*T/alpha)*exp(-(t*pi/alpha).^2); 
% But then we normalize, so T is not needed.
h = (sqrt(pi)/alpha)*exp(-(t*pi/alpha).^2); 
 
% Normalize coefficients
h = h./sum(h);

end
% code forward from matlab internal code

BT is 3dB bandwidth bit duration product. which can totally describe the filter characteristic.

we can see the BT is smaller, the sidelobe is smaller, but the ISI is more obvious.

GMSK

the GMSK process:

the sequence of M-ary data symbols is shaped with Gaussian filter, the symbol $rect(\frac{t}{Tb})$ do convolution with h(t), shown as,

$q(t)=\int_{-\infty}^{\infty} h(\tau ) \cdot \text{rect}\left(\frac{t-\tau}{T_b}\right) \, d\tau$

$=\int_{t-\frac{T_b}{2}}^{t+\frac{T_b}{2}} \frac{1}{\sigma\sqrt{2\pi}} e^{-\frac{(\tau)^2}{2\sigma^2}}\textup{} d\tau$ $=\int_{\frac{(t-\frac{T_b}{2})}{\sigma}}^{\frac{(t+\frac{T_b}{2})}{\sigma}} \frac{1}{\sqrt{2\pi}} e^{-\frac{(u)^2}{2}}\textup{} du$