Started by March 3, 2009
```Hi all,

I would like to know if the folling would be possible.

Lets say that my tx use a raised cosine filter to limit the bandwidth of
signal by a factor of x and by using a Viterbi algorithm I would be able to
approximate every symbol affected by ISI caused by the memory of my raised
cosine filter. By doing this would it be possible to eliminate the need of
a clock recovery system if my transmitted data frame is preceded by a
header that I would correlate on in order to indicate when valid data can
be process by the viterbi

```
```On Mar 3, 2:05 pm, "Awal" <sacha.at...@lacime.etsmtl.ca> wrote:
> Hi all,
>
> I would like to know if the folling would be possible.
>
> Lets say that my tx use a raised cosine filter to limit the bandwidth of
> my transmitted signal. At the receiver I would oversample the received
> signal by a factor of x and by using a Viterbi algorithm I would be able to
> approximate every symbol affected by ISI caused by the memory of my raised
> cosine filter. By doing this would it be possible to eliminate the need of
> a clock recovery system if my transmitted data frame is preceded by a
> header that I would correlate on in order to indicate when valid data can
> be process by the viterbi

If the timing mismatch is constant, then this should work.  Does
your channel have memory?  You are considering a complicated
(and expensive to implement) solution to what is possibly a simple
problem.

```
```>On Mar 3, 2:05 pm, "Awal" <sacha.at...@lacime.etsmtl.ca> wrote:
>> Hi all,
>>
>> I would like to know if the folling would be possible.
>>
>> Lets say that my tx use a raised cosine filter to limit the bandwidth
of
>> my transmitted signal. At the receiver I would oversample the received
>> signal by a factor of x and by using a Viterbi algorithm I would be
able to
>> approximate every symbol affected by ISI caused by the memory of my
raised
>> cosine filter. By doing this would it be possible to eliminate the need
of
>> a clock recovery system if my transmitted data frame is preceded by a
>> header that I would correlate on in order to indicate when valid data
can
>> be process by the viterbi
>
>If the timing mismatch is constant, then this should work.  Does
>your channel have memory?  You are considering a complicated
>(and expensive to implement) solution to what is possibly a simple
>problem.
>

I would start with a simple channel model: only white noise and
transmission delay. No multipath, channel memory etc..

I know it&rsquo;s a complicated solution, I'm just playing with idia at the
moment. The complexity of my viterbi would be proportional to the raised
cosine filter.

thanks
```
```Awal <sacha.atwal@lacime.etsmtl.ca> wrote:

>Lets say that my tx use a raised cosine filter to limit the bandwidth of
>signal by a factor of x and by using a Viterbi algorithm I would be able to
>approximate every symbol affected by ISI caused by the memory of my raised
>cosine filter. By doing this would it be possible to eliminate the need of
>a clock recovery system if my transmitted data frame is preceded by a
>header that I would correlate on in order to indicate when valid data can
>be process by the viterbi

Yes, it is sometimes possible to derive all of your timing
recovery from headers.  Whether you can do this may not
have anything to do with your raised cosine function or
Viterbi equalizer.

Good luck.

Steve
```
```
Awal wrote:

> Hi all,
>
> I would like to know if the folling would be possible.
>
> Lets say that my tx use a raised cosine filter to limit the bandwidth of
> my transmitted signal. At the receiver I would oversample the received
> signal by a factor of x and by using a Viterbi algorithm I would be able to
> approximate every symbol affected by ISI caused by the memory of my raised
> cosine filter. By doing this would it be possible to eliminate the need of
> a clock recovery system if my transmitted data frame is preceded by a
> header that I would correlate on in order to indicate when valid data can
> be process by the viterbi

This is possible but cumbersome and inefficient. Unless there are some
special requirements, clock recovery is better solution.

DSP and Mixed Signal Design Consultant
http://www.abvolt.com
```
```Looking up references to the IS-54 cellular system from the 90's might

Awal wrote:
>> On Mar 3, 2:05 pm, "Awal" <sacha.at...@lacime.etsmtl.ca> wrote:
>>> Hi all,
>>>
>>> I would like to know if the folling would be possible.
>>>
>>> Lets say that my tx use a raised cosine filter to limit the bandwidth
> of
>>> my transmitted signal. At the receiver I would oversample the received
>>> signal by a factor of x and by using a Viterbi algorithm I would be
> able to
>>> approximate every symbol affected by ISI caused by the memory of my
> raised
>>> cosine filter. By doing this would it be possible to eliminate the need
> of
>>> a clock recovery system if my transmitted data frame is preceded by a
>>> header that I would correlate on in order to indicate when valid data
> can
>>> be process by the viterbi
>> If the timing mismatch is constant, then this should work.  Does
>> your channel have memory?  You are considering a complicated
>> (and expensive to implement) solution to what is possibly a simple
>> problem.
>>
>
> I would start with a simple channel model: only white noise and
> transmission delay. No multipath, channel memory etc..
>
> I know it&rsquo;s a complicated solution, I'm just playing with idia at the
> moment. The complexity of my viterbi would be proportional to the raised
> cosine filter.
>
>
> thanks
```
```>
>
>Awal wrote:
>
>> Hi all,
>>
>> I would like to know if the folling would be possible.
>>
>> Lets say that my tx use a raised cosine filter to limit the bandwidth
of
>> my transmitted signal. At the receiver I would oversample the received
>> signal by a factor of x and by using a Viterbi algorithm I would be
able to
>> approximate every symbol affected by ISI caused by the memory of my
raised
>> cosine filter. By doing this would it be possible to eliminate the need
of
>> a clock recovery system if my transmitted data frame is preceded by a
>> header that I would correlate on in order to indicate when valid data
can
>> be process by the viterbi
>
>This is possible but cumbersome and inefficient. Unless there are some
>special requirements, clock recovery is better solution.
>
>DSP and Mixed Signal Design Consultant
>http://www.abvolt.com
>

The reiceved signal as not cyclostationry. Which mean I cannot used
Gardner algorithm or other algorithm that use non-linear tranformation to
recover the clk.
```
```
Awal wrote:

> The reiceved signal as not cyclostationry. Which mean I cannot used
> Gardner algorithm or other algorithm that use non-linear tranformation to
> recover the clk.

What does the clock mean if the signal is not cyclostationary?

DSP and Mixed Signal Design Consultant
http://www.abvolt.com
```
```>
>
>Awal wrote:
>
>
>> The reiceved signal as not cyclostationry. Which mean I cannot used
>> Gardner algorithm or other algorithm that use non-linear tranformation
to
>> recover the clk.
>
>What does the clock mean if the signal is not cyclostationary?
>
>
>DSP and Mixed Signal Design Consultant
>http://www.abvolt.com
>

At the receiver I oversample the signal and undo the tx transformation
that render the signal non cyclostationary in parallele to viterbi
decodign.
```
```On Mar 3, 2:05&#2013266080;pm, "Awal" <sacha.at...@lacime.etsmtl.ca> wrote:
> Hi all,
>
> I would like to know if the folling would be possible.
>
> Lets say that my tx use a raised cosine filter to limit the bandwidth of
> my transmitted signal. At the receiver I would oversample the received
> signal by a factor of x and by using a Viterbi algorithm I would be able to
> approximate every symbol affected by ISI caused by the memory of my raised
> cosine filter.

Here symbol refers to the oversampled one at the receiver or the one
at the transmitter side?

By doing this would it be possible to eliminate the need of
> a clock recovery system if my transmitted data frame is preceded by a
> header that I would correlate on in order to indicate when valid data can
> be process by the viterbi

I am a little bit confused here. By using viterbi, one assumption is
that you have estimated the
channel, which means you already have the clock recovered. If you do
not know the channel, how
could you use viterbi?

```