Channel Matrix H

Would anyone please have any idea about how to find the h values of a
channel when you know the power delay profile of this channel

I have tried using the rayleighchan and filter commands I dont know how to
find the h  value though which one is which. I have a 6 ray model

Please help

On Dec 28, 3:07&#4294967295;pm, "leonardo232" <leosivri...@gmail.com> wrote:
> Would anyone please have any idea about how to find the h values of a
> channel when you know the power delay profile of this channel
>
> I have tried using the rayleighchan and filter commands I dont know how to
> find the h &#4294967295;value though which one is which. I have a 6 ray model
>
> Please help

From what I have heard from real engineers, all channel models are
bollocks and don't work in any case.

H.

On Dec 28, 11:07&#4294967295;am, "leonardo232" <leosivri...@gmail.com> wrote:
> Would anyone please have any idea about how to find the h values of a
> channel when you know the power delay profile of this channel
>
> I have tried using the rayleighchan and filter commands I dont know how to
> find the h &#4294967295;value though which one is which. I have a 6 ray model
>
> Please help

Since you say "channel matrix H", are you referring to the narrowband
MIMO channel matrix?
If so, you have nowhere near enough information just from the power
delay profile.

The PDP is the time-domain power expectation for a single channel (tx-
rx). Therefore firstly you use this to generate an instantaneous
(complex) channel impulse response, generated according to the
statistical model for each of your paths/rays (e.g. Rayleigh fading,
Ricean, etc).
Then from this instantaneous impulse response you can find the
frequency domain channel response by Fourier transform.
Then, assuming your MIMO is combined with a multi-carrier modulation
such as OFDM, you can take a given subcarrier value as one of the
entries in your H (you will have a separate H for every subcarrier
frequency).
Then, you need to repeat the whole process for the other channels
(e.g. Tx2 - Rx1, Tx1 - Rx2, etc), generating channel responses for
each. If you assume each channel is independently fading, this is
simple. However in some models you might want to assume some
correlation between tx and/or rx antennas, in which case you need to
model this correctly too.

-T

>On Dec 28, 11:07=A0am, "leonardo232" <leosivri...@gmail.com> wrote:
>> Would anyone please have any idea about how to find the h values of a
>> channel when you know the power delay profile of this channel
>>
>> I have tried using the rayleighchan and filter commands I dont know how
t=
>o
>> find the h =A0value though which one is which. I have a 6 ray model
>>
>> Please help
>
>Since you say "channel matrix H", are you referring to the narrowband
>MIMO channel matrix?
>If so, you have nowhere near enough information just from the power
>delay profile.
>
>The PDP is the time-domain power expectation for a single channel (tx-
>rx). Therefore firstly you use this to generate an instantaneous
>(complex) channel impulse response, generated according to the
>statistical model for each of your paths/rays (e.g. Rayleigh fading,
>Ricean, etc).
>Then from this instantaneous impulse response you can find the
>frequency domain channel response by Fourier transform.
>Then, assuming your MIMO is combined with a multi-carrier modulation
>such as OFDM, you can take a given subcarrier value as one of the
>entries in your H (you will have a separate H for every subcarrier
>frequency).
>Then, you need to repeat the whole process for the other channels
>(e.g. Tx2 - Rx1, Tx1 - Rx2, etc), generating channel responses for
>each. If you assume each channel is independently fading, this is
>simple. However in some models you might want to assume some
>correlation between tx and/or rx antennas, in which case you need to
>model this correctly too.
>
>-T
>

Hi Yes and thanks for the reply. I am actually using OFDMA therefore a
number of users accessing a number of channels ofdm. I need an array number
of users*sub-channels. Each user should have a different value of h and I
will then feed these values into shannon capacity equation
C=BWlog(1+Power*h/noise) and then a radio resource algorithm will be
created(this part works fine. So if I have six paths then I have six
complex impulse responses for each path (six pathgains) what do i do with
these to create one that is associated with each subchannel. Could someone
provide an example please? Also if I am doing this according to a specific
channel model i.e ITU PED B then how will I put the different features of
each user into the equation (i.e distances,etc) do i put some random PDP
keeping the dopller and sampling f constant?

Please help 

Thanks a lot


>

On Sat, 27 Dec 2008 20:41:08 -0800 (PST), HardySpicer
<gyansorova@gmail.com> wrote:

>On Dec 28, 3:07&#4294967295;pm, "leonardo232" <leosivri...@gmail.com> wrote:
>> Would anyone please have any idea about how to find the h values of a
>> channel when you know the power delay profile of this channel
>>
>> I have tried using the rayleighchan and filter commands I dont know how to
>> find the h &#4294967295;value though which one is which. I have a 6 ray model
>>
>> Please help
>
>From what I have heard from real engineers, all channel models are
>bollocks and don't work in any case.
>
>H.

It just depends on what you're trying to do.   Many models were
developed from real-world data collection, so they can represent a
pretty good idea of what would be encountered for that application. 

I think people run into trouble when they try to apply a model meant
for one application in some other application.

Nevertheless, all models have limitations.   It's up to the user to
understand the limitations of the model and what applications it was
intended for.   Many complex systems have been developed and
successfully deployed strictly using models.

Eric Jacobsen
Minister of Algorithms
Abineau Communications
http://www.ericjacobsen.org

Blog: http://www.dsprelated.com/blogs-1/hf/Eric_Jacobsen.php

>On Sat, 27 Dec 2008 20:41:08 -0800 (PST), HardySpicer
><gyansorova@gmail.com> wrote:
>
>>On Dec 28, 3:07&#65533;pm, "leonardo232" <leosivri...@gmail.com> wrote:
>>> Would anyone please have any idea about how to find the h values of a
>>> channel when you know the power delay profile of this channel
>>>
>>> I have tried using the rayleighchan and filter commands I dont know
how to
>>> find the h &#65533;value though which one is which. I have a 6 ray model
>>>
>>> Please help
>>
>>From what I have heard from real engineers, all channel models are
>>bollocks and don't work in any case.
>>
>>H.
>
>It just depends on what you're trying to do.   Many models were
>developed from real-world data collection, so they can represent a
>pretty good idea of what would be encountered for that application. 
>
>I think people run into trouble when they try to apply a model meant
>for one application in some other application.
>
>Nevertheless, all models have limitations.   It's up to the user to
>understand the limitations of the model and what applications it was
>intended for.   Many complex systems have been developed and
>successfully deployed strictly using models.
>
>Eric Jacobsen
>Minister of Algorithms
>Abineau Communications
>http://www.ericjacobsen.org
>
>Blog: http://www.dsprelated.com/blogs-1/hf/Eric_Jacobsen.php
>
Hi Yes and thanks for the reply. I am actually using OFDMA therefore a
number of users accessing a number of channels ofdm. I need an array
number
of users*sub-channels. Each user should have a different value of h and I
will then feed these values into shannon capacity equation
C=BWlog(1+Power*h/noise) and then a radio resource algorithm will be
created(this part works fine. So if I have six paths then I have six
complex impulse responses for each path (six pathgains) what do i do with
these to create one that is associated with each subchannel. Could
someone
provide an example please? Also if I am doing this according to a
specific
channel model i.e ITU PED B then how will I put the different features of
each user into the equation (i.e distances,etc) do i put some random PDP
keeping the dopller and sampling f constant?

Please help 

Thanks a lot

On Sun, 28 Dec 2008 11:44:01 -0600, "leonardo232"
<leosivridis@gmail.com> wrote:

>>On Sat, 27 Dec 2008 20:41:08 -0800 (PST), HardySpicer
>><gyansorova@gmail.com> wrote:
>>
>>>On Dec 28, 3:07&#65533;pm, "leonardo232" <leosivri...@gmail.com> wrote:
>>>> Would anyone please have any idea about how to find the h values of a
>>>> channel when you know the power delay profile of this channel
>>>>
>>>> I have tried using the rayleighchan and filter commands I dont know
>how to
>>>> find the h &#65533;value though which one is which. I have a 6 ray model
>>>>
>>>> Please help
>>>
>>>From what I have heard from real engineers, all channel models are
>>>bollocks and don't work in any case.
>>>
>>>H.
>>
>>It just depends on what you're trying to do.   Many models were
>>developed from real-world data collection, so they can represent a
>>pretty good idea of what would be encountered for that application. 
>>
>>I think people run into trouble when they try to apply a model meant
>>for one application in some other application.
>>
>>Nevertheless, all models have limitations.   It's up to the user to
>>understand the limitations of the model and what applications it was
>>intended for.   Many complex systems have been developed and
>>successfully deployed strictly using models.
>>
>>Eric Jacobsen
>>Minister of Algorithms
>>Abineau Communications
>>http://www.ericjacobsen.org
>>
>>Blog: http://www.dsprelated.com/blogs-1/hf/Eric_Jacobsen.php
>>
>Hi Yes and thanks for the reply. I am actually using OFDMA therefore a
>number of users accessing a number of channels ofdm. I need an array
>number
>of users*sub-channels. Each user should have a different value of h and I
>will then feed these values into shannon capacity equation
>C=BWlog(1+Power*h/noise) and then a radio resource algorithm will be
>created(this part works fine. So if I have six paths then I have six
>complex impulse responses for each path (six pathgains) what do i do with
>these to create one that is associated with each subchannel. Could
>someone
>provide an example please? Also if I am doing this according to a
>specific
>channel model i.e ITU PED B then how will I put the different features of
>each user into the equation (i.e distances,etc) do i put some random PDP
>keeping the dopller and sampling f constant?
>
>Please help 
>
>Thanks a lot
>

The relevant models usually include a stochastic description for the
channel FIR tap values and spacing, i.e., the model can be used to
randomly generate different channel instances that fit the stochastic
model.   For systems like you're describing (e.g., WiMAX), there are
usually different model configurations for different terrain types
(e.g., urban, rural).   If a single base station serves mobile
terminals in different environments (I'd think this would not be the
typical case), then perhaps even different terrain models across the
users might be appropriate.

Even if you use a single terrain model, multiple instances of the
random model outputs could be applied for each user.

Eric Jacobsen
Minister of Algorithms
Abineau Communications
http://www.ericjacobsen.org

Blog: http://www.dsprelated.com/blogs-1/hf/Eric_Jacobsen.php

>On Sun, 28 Dec 2008 11:44:01 -0600, "leonardo232"
><leosivridis@gmail.com> wrote:
>
>>>On Sat, 27 Dec 2008 20:41:08 -0800 (PST), HardySpicer
>>><gyansorova@gmail.com> wrote:
>>>
>>>>On Dec 28, 3:07&#65533;pm, "leonardo232" <leosivri...@gmail.com> wrote:
>>>>> Would anyone please have any idea about how to find the h values of
a
>>>>> channel when you know the power delay profile of this channel
>>>>>
>>>>> I have tried using the rayleighchan and filter commands I dont know
>>how to
>>>>> find the h &#65533;value though which one is which. I have a 6 ray model
>>>>>
>>>>> Please help
>>>>
>>>>From what I have heard from real engineers, all channel models are
>>>>bollocks and don't work in any case.
>>>>
>>>>H.
>>>
>>>It just depends on what you're trying to do.   Many models were
>>>developed from real-world data collection, so they can represent a
>>>pretty good idea of what would be encountered for that application. 
>>>
>>>I think people run into trouble when they try to apply a model meant
>>>for one application in some other application.
>>>
>>>Nevertheless, all models have limitations.   It's up to the user to
>>>understand the limitations of the model and what applications it was
>>>intended for.   Many complex systems have been developed and
>>>successfully deployed strictly using models.
>>>
>>>Eric Jacobsen
>>>Minister of Algorithms
>>>Abineau Communications
>>>http://www.ericjacobsen.org
>>>
>>>Blog: http://www.dsprelated.com/blogs-1/hf/Eric_Jacobsen.php
>>>
>>Hi Yes and thanks for the reply. I am actually using OFDMA therefore a
>>number of users accessing a number of channels ofdm. I need an array
>>number
>>of users*sub-channels. Each user should have a different value of h and
I
>>will then feed these values into shannon capacity equation
>>C=BWlog(1+Power*h/noise) and then a radio resource algorithm will be
>>created(this part works fine. So if I have six paths then I have six
>>complex impulse responses for each path (six pathgains) what do i do
with
>>these to create one that is associated with each subchannel. Could
>>someone
>>provide an example please? Also if I am doing this according to a
>>specific
>>channel model i.e ITU PED B then how will I put the different features
of
>>each user into the equation (i.e distances,etc) do i put some random
PDP
>>keeping the dopller and sampling f constant?
>>
>>Please help 
>>
>>Thanks a lot
>>
>
>The relevant models usually include a stochastic description for the
>channel FIR tap values and spacing, i.e., the model can be used to
>randomly generate different channel instances that fit the stochastic
>model.   For systems like you're describing (e.g., WiMAX), there are
>usually different model configurations for different terrain types
>(e.g., urban, rural).   If a single base station serves mobile
>terminals in different environments (I'd think this would not be the
>typical case), then perhaps even different terrain models across the
>users might be appropriate.
>
>Even if you use a single terrain model, multiple instances of the
>random model outputs could be applied for each user.
>
>Eric Jacobsen
>Minister of Algorithms
>Abineau Communications
>http://www.ericjacobsen.org
>
>Blog: http://www.dsprelated.com/blogs-1/hf/Eric_Jacobsen.php

Ok thanks 

sorry to keep this going but how would one calculate the channel gains h
for each sub-channel?
>

On Dec 29, 2:15&#4294967295;am, "leonardo232" <leosivri...@gmail.com> wrote:
> >On Dec 28, 11:07=A0am, "leonardo232" <leosivri...@gmail.com> wrote:
> >> Would anyone please have any idea about how to find the h values of a
> >> channel when you know the power delay profile of this channel
>
> >> I have tried using the rayleighchan and filter commands I dont know how
> t=
> >o
> >> find the h =A0value though which one is which. I have a 6 ray model
>
> >> Please help
>
> >Since you say "channel matrix H", are you referring to the narrowband
> >MIMO channel matrix?
> >If so, you have nowhere near enough information just from the power
> >delay profile.
>
> >The PDP is the time-domain power expectation for a single channel (tx-
> >rx). Therefore firstly you use this to generate an instantaneous
> >(complex) channel impulse response, generated according to the
> >statistical model for each of your paths/rays (e.g. Rayleigh fading,
> >Ricean, etc).
> >Then from this instantaneous impulse response you can find the
> >frequency domain channel response by Fourier transform.
> >Then, assuming your MIMO is combined with a multi-carrier modulation
> >such as OFDM, you can take a given subcarrier value as one of the
> >entries in your H (you will have a separate H for every subcarrier
> >frequency).
> >Then, you need to repeat the whole process for the other channels
> >(e.g. Tx2 - Rx1, Tx1 - Rx2, etc), generating channel responses for
> >each. If you assume each channel is independently fading, this is
> >simple. However in some models you might want to assume some
> >correlation between tx and/or rx antennas, in which case you need to
> >model this correctly too.
>
> >-T
>
> Hi Yes and thanks for the reply. I am actually using OFDMA therefore a
> number of users accessing a number of channels ofdm. I need an array number
> of users*sub-channels. Each user should have a different value of h and I
> will then feed these values into shannon capacity equation
> C=BWlog(1+Power*h/noise) and then a radio resource algorithm will be
> created(this part works fine. So if I have six paths then I have six
> complex impulse responses for each path (six pathgains) what do i do with
> these to create one that is associated with each subchannel. Could someone
> provide an example please? Also if I am doing this according to a specific
> channel model i.e ITU PED B then how will I put the different features of
> each user into the equation (i.e distances,etc) do i put some random PDP
> keeping the dopller and sampling f constant?
>
> Please help
>
> Thanks a lot
>
>

Hello

Firstly a general comment - it seems you are still getting to grips
with the basic concepts of MIMO. Therefore I suggest you try to study
these a bit more before dealing with difficult issues such as multi-
user OFDMA and resource allocation.
One very good online resource, which is a PhD thesis (in English,
despite the first few pages) on MIMO-OFDM by Albert van Zelst of Airgo
Networks, provides a very clear description of the basics:
http://www.avzelst.nl/dissertation_avanzelst.pdf

Now, as to your specific question...

You have a 6 path model for one channel (between 1 tx and 1 rx) -
probably a line-of-sight path plus 5 multipaths. #Together#, these
form the impulse response for that channel.

Basically the impulse response (time domain) for the channel is formed
of a delta function for each path - the position of each delta (in
time) equals the delay of the corresponding path, the complex
amplitude of the delta equals the complex gain of the corresponding
path. (In practice we often use a baseband representation which is
bandwidth limited).

Note there is only one channel response per channel, #not# one per
path like you suggest above.

One you have this time domain impulse response, you can take the
Fourier transform to get the frequency domain channel response. The
value of this channel response at the frequencies of each OFDM
subcarrier determine the entry in H for this channel at this
frequency.

e.g. if you have 2x2 MIMO, each H is size 2x2. You have a total of N
such matrices, one for each subcarrier. If in the above you calculated
the channel reponse between TX1 and RX1, you place the frequency
domain channel response values at each frequency into the
corresponding matrices H at element (1,1).

You then repeat this for all the other channels (e.g. TX2 - RX1, TX1 -
RX2, TX2 - RX2).
Note you must ensure in your model that these 4 channels are #not#
identical, otherwise the matrices H will be singular and MIMO will not
be possible.

In a multiuser case, you then need to generate a set of matrices for
each of the other users too. Therefore you need (non-identical)
channel models (for each tx-rx) for each of these users.
How to model these channels is not trivial and there are various
methods - stochastic, physical/geometric approaches etc.
I am not familiar enough with the particular channel model you cite in
order to comment how to implement its simulation (in any case it is
probably too complicated to easily describe here), but suggest if you
study the basics first you will eventually be able to work it out by
yourself.

-T

Ok i see i am going to do look at the channel models in more detail to
create the instances, but could someone please tell me this one h value for
each channel. THE one h value is it the filtercommand. i mean in the time
domain. The one h value associated with each channel in the time domain 

If there is a specific channel profile (PDP) and in an OFDM system how
would you get the h for each channel? 

c=ifft(d); 
d=rayleighhan(fd,fs,PDP etc) 

where is the h? 
is h=filter(d,c)? 
which command will give me the values of h please?




>On Dec 29, 2:15=A0am, "leonardo232" <leosivri...@gmail.com> wrote:
>> >On Dec 28, 11:07=3DA0am, "leonardo232" <leosivri...@gmail.com> wrote:
>> >> Would anyone please have any idea about how to find the h values of
a
>> >> channel when you know the power delay profile of this channel
>>
>> >> I have tried using the rayleighchan and filter commands I dont know
ho=
>w
>> t=3D
>> >o
>> >> find the h =3DA0value though which one is which. I have a 6 ray
model
>>
>> >> Please help
>>
>> >Since you say "channel matrix H", are you referring to the narrowband
>> >MIMO channel matrix?
>> >If so, you have nowhere near enough information just from the power
>> >delay profile.
>>
>> >The PDP is the time-domain power expectation for a single channel
(tx-
>> >rx). Therefore firstly you use this to generate an instantaneous
>> >(complex) channel impulse response, generated according to the
>> >statistical model for each of your paths/rays (e.g. Rayleigh fading,
>> >Ricean, etc).
>> >Then from this instantaneous impulse response you can find the
>> >frequency domain channel response by Fourier transform.
>> >Then, assuming your MIMO is combined with a multi-carrier modulation
>> >such as OFDM, you can take a given subcarrier value as one of the
>> >entries in your H (you will have a separate H for every subcarrier
>> >frequency).
>> >Then, you need to repeat the whole process for the other channels
>> >(e.g. Tx2 - Rx1, Tx1 - Rx2, etc), generating channel responses for
>> >each. If you assume each channel is independently fading, this is
>> >simple. However in some models you might want to assume some
>> >correlation between tx and/or rx antennas, in which case you need to
>> >model this correctly too.
>>
>> >-T
>>
>> Hi Yes and thanks for the reply. I am actually using OFDMA therefore a
>> number of users accessing a number of channels ofdm. I need an array
numb=
>er
>> of users*sub-channels. Each user should have a different value of h and
I
>> will then feed these values into shannon capacity equation
>> C=3DBWlog(1+Power*h/noise) and then a radio resource algorithm will be
>> created(this part works fine. So if I have six paths then I have six
>> complex impulse responses for each path (six pathgains) what do i do
with
>> these to create one that is associated with each subchannel. Could
someon=
>e
>> provide an example please? Also if I am doing this according to a
specifi=
>c
>> channel model i.e ITU PED B then how will I put the different features
of
>> each user into the equation (i.e distances,etc) do i put some random
PDP
>> keeping the dopller and sampling f constant?
>>
>> Please help
>>
>> Thanks a lot
>>
>>
>
>Hello
>
>Firstly a general comment - it seems you are still getting to grips
>with the basic concepts of MIMO. Therefore I suggest you try to study
>these a bit more before dealing with difficult issues such as multi-
>user OFDMA and resource allocation.
>One very good online resource, which is a PhD thesis (in English,
>despite the first few pages) on MIMO-OFDM by Albert van Zelst of Airgo
>Networks, provides a very clear description of the basics:
>http://www.avzelst.nl/dissertation_avanzelst.pdf
>
>Now, as to your specific question...
>
>You have a 6 path model for one channel (between 1 tx and 1 rx) -
>probably a line-of-sight path plus 5 multipaths. #Together#, these
>form the impulse response for that channel.
>
>Basically the impulse response (time domain) for the channel is formed
>of a delta function for each path - the position of each delta (in
>time) equals the delay of the corresponding path, the complex
>amplitude of the delta equals the complex gain of the corresponding
>path. (In practice we often use a baseband representation which is
>bandwidth limited).
>
>Note there is only one channel response per channel, #not# one per
>path like you suggest above.
>
>One you have this time domain impulse response, you can take the
>Fourier transform to get the frequency domain channel response. The
>value of this channel response at the frequencies of each OFDM
>subcarrier determine the entry in H for this channel at this
>frequency.
>
>e.g. if you have 2x2 MIMO, each H is size 2x2. You have a total of N
>such matrices, one for each subcarrier. If in the above you calculated
>the channel reponse between TX1 and RX1, you place the frequency
>domain channel response values at each frequency into the
>corresponding matrices H at element (1,1).
>
>You then repeat this for all the other channels (e.g. TX2 - RX1, TX1 -
>RX2, TX2 - RX2).
>Note you must ensure in your model that these 4 channels are #not#
>identical, otherwise the matrices H will be singular and MIMO will not
>be possible.
>
>In a multiuser case, you then need to generate a set of matrices for
>each of the other users too. Therefore you need (non-identical)
>channel models (for each tx-rx) for each of these users.
>How to model these channels is not trivial and there are various
>methods - stochastic, physical/geometric approaches etc.
>I am not familiar enough with the particular channel model you cite in
>order to comment how to implement its simulation (in any case it is
>probably too complicated to easily describe here), but suggest if you
>study the basics first you will eventually be able to work it out by
>yourself.
>
>-T
>