m**h
发帖数: 207 | 1 can anyone explain the purpose of the pre-whittening filter
in MIMO receiver? THX |
s*****2
发帖数: 117 | 2 maybe it is better for u to give more information
about pre-whitening filter in MIMO receiver.
I think the pre-whitening filter u mentioned is performed
in the time domain, and is the spatio-temporal filter.
is this right?
【在 m**h 的大作中提到】
: can anyone explain the purpose of the pre-whittening filter
: in MIMO receiver? THX
|
m**h
发帖数: 207 | 3 For example MMSE w/o PWF
estimate = (H*H'+Rnn)^(-1) * H * y
with PWF
estimate = (H1*H1'+I)^(-1) * H1 *y
someone told me it's for simplication, wondering if any other benefit?
【在 s*****2 的大作中提到】
: maybe it is better for u to give more information
: about pre-whitening filter in MIMO receiver.
: I think the pre-whitening filter u mentioned is performed
: in the time domain, and is the spatio-temporal filter.
: is this right?
|
a*******e
发帖数: 346 | 4 这两个应该是等价的吧。
【在 m**h 的大作中提到】
: For example MMSE w/o PWF
: estimate = (H*H'+Rnn)^(-1) * H * y
: with PWF
: estimate = (H1*H1'+I)^(-1) * H1 *y
: someone told me it's for simplication, wondering if any other benefit?
|
f*****g
发帖数: 30 | 5 I think it is a theoretical tool to deal with colored (spatially correlated)
noise. After the filter, the noise is white and all existing results based
on white-noise assumption can be applied.
In a real implementation, I don't think such filter is necessary. Noise
whitening will be achieved automatically as part of the MMSE optimization
process. However, there may be practical benefits (such as dynamical range)
in using a separate filter. I don't know much about those issues. |
a*******e
发帖数: 346 | 6 good point on the dynamic range.
Rnn will have a wide dynamic range as SNR varies.
I guess you will notice the difference if you implement the
filters, considering the fixed-point effect, dynamic range etc..
一直都纸上谈兵，根本没有考虑过实现，唉：（
correlated)
based
range)
【在 f*****g 的大作中提到】
: I think it is a theoretical tool to deal with colored (spatially correlated)
: noise. After the filter, the noise is white and all existing results based
: on white-noise assumption can be applied.
: In a real implementation, I don't think such filter is necessary. Noise
: whitening will be achieved automatically as part of the MMSE optimization
: process. However, there may be practical benefits (such as dynamical range)
: in using a separate filter. I don't know much about those issues.
|
