Continue to Site

Welcome to

Welcome to our site! is an international Electronics Discussion Forum focused on EDA software, circuits, schematics, books, theory, papers, asic, pld, 8051, DSP, Network, RF, Analog Design, PCB, Service Manuals... and a whole lot more! To participate you need to register. Registration is free. Click here to register now.

MMSE standard form ???

Not open for further replies.


Full Member level 2
Nov 29, 2003
Reaction score
Trophy points
Activity points
do anyone has got the proof or have any idea of getting this proof that MMSE equation i.e. min E [|| WY-X ||^2] can be standerized to ( (H'H+σ²)^-1 )H'

where ' ----> hermettian transpose
^2 ---> rasie to the power 2
^-1 ----> raise to the power -1
σ ------> noise variance
E [] -----> mean

tthe system considered above is Y = HX+N , where X is transmiites signal, Y is recieved, H is the channel matrix and N is white gaussian noise. W is the MMSE filter coefficent.

I got this principle expression for MMSE receiver is

G=H*(H*H + No/Po I )^-1 *x

Where H -- channel matrix

I --- identity matrix

X --- received vector

Iam not understand about the above expression That is what is No/Po and why multiplying with I , how can we achevied MMSE in the above expression etc.... Can you please help me ...

From MMSE sense, you can induce the following relationship.
E[|WY - X|^2] ==> E[|WY - X|Y'] = 0 ==> WE[YY'] - E[XY'] = 0 ==> W = E[XY']/E[YY']

Here, E[XY'] is the cross-correlation, while E[YY'] is the auto-correlation.

So, E[XY'] = XX'H + XN' and E[YY'] = XX'HH' + XHN' + X'H'N + NN'
Since X and N are independent and uncorrelated, XN' = X'N = 0, and NN' = sigma^2

Therefore, W = XX'H'/(XX'HH' + sigma^2).

If you send 2-PAM, then XX' = 1. Now you get W = H'/(HH' + sigma^2)

G=H*(H*H + No/Po I )^-1 *x

Can you help me to understand the above expression...

From the orthogonal principle, E{(WY-X)Y^H} = 0, Y=HX+n
we could get WE(YY^H)=E(XY^H), assuming E(XN^H)=0, i.e., signal and noise are independent, and mean of the noise is zero

E(YY^H)=HE(XX^H)H^H+E(nn^H), Assumped that E(XX^H)=PI, and E(NN^H)=NI, i.e., signal is an independent random vector and noise is also an independent random vector, W=H^H(HH^H+N/PI)^(-1). where P/I reprsents the signal-to-noise ratio. Here, ^H=^(*')

It is clear that the MMSE is one of the Equalization technique, and the equation is as follow:
Where Y(n) is the received signal
X(n) is the transmitted signal
H(n) is the Faded signal (channel Impulse resp[onse)
Es is a symbol energy for QPSK
N0 is the noise.
I hope this will help

I have the same problem as you, but stumble across this book:
Multiantenna wireless communication systems by Sergio Barbarossa (Artech House)
refer pg 94-96 will help you to understand more

Not open for further replies.

Part and Inventory Search

Welcome to