design of adptive DFE

On Jun 11, 9:04&#4294967295;am, "cpshah99" <cpsha...@rediffmail.com> wrote:
> Hi Julius
>
> So the observations are as follows:
>
> System Parameters:
>
> Delay spread of channel is 200 symbols, QPSK modultion, 511 long training
> sequence, rate 1/2 turbo code, adaptive DFE with LMS algorithm. The FF taps
> =50 and FB taps =50. The channel has 5 taps and taps are time varying.

If the channel has 5 taps, how can the delay spread be 200 symbols?

>
> So I fixed the SNR at 14 dB and after one time equalization and 3 turbo
> iterations I am getting BER of 10^-5. Which I think is quite good.
>
> Next I will be more specific when I ask any que.
>
> Thanks as always
>
> Chintan

xsong wrote:


> If the channel has 5 taps, how can the delay spread be 200 symbols?

Irrelevant.
Think of rake receiver.


VLV

On Jun 10, 10:15 pm, Vladimir Vassilevsky <antispam_bo...@hotmail.com>
wrote:
> robert bristow-johnson wrote:
>
> > i don't even know what it is?  student with a "pi" stuck in it?
>
> > curiopius,
>
> STUPIDENT is the same thing as the STUDIOT
>
> //---------------------
> Hi
> I saw a lot of posting of yours on dsprelated.com,
> I saw on your website a name of a software as MathLab. I want to know
> more about this.
> I have seen only MATLAB and MathCad, what is this mathLab.
> Please please Please reply sure
> //-----------------
> Hello
> I need Matlab code for BPSK/PM modulation. Can you send me this code?
> Please i will be very grateful.
> Thank you
> //--------------------------
> In my final year project , I need a convolutional encoder (rate = 1/2
> with constraints length = 5) and the viterbi decoder implemented in C
> for blackfin DSP(Analog devices)
> Can you help me?
> //-----------------
> hello
> i had few queries about..the concepts of non gaussian noises.
> In communication channel..what are the models for impulsive noises and
> other noises.. so that we can arrive at a more practical simulation
> while making a communication channel on matlab
> i read some work of Middleton which was relevant.
> pls help me out with the appropriate model or the links..
> thanking you
> geetika
> btech final yr student..working on simulating and comparison of
> modulation tech
> //---------------------
> I hope you don't mind me contacting you. I saw your name in a
> mathematics forum discussing DSP on sine waves. And so hope you may be
> able to help with a perhaps simple problem, but which escapes me.
> I have a clean sine wave (A) with know frequency and amplitude.
> I sampled a 2nd sine wave (B) at e.g. 4 random points. I know its
> frequency (same as A), however I wish to calculate the phase difference
> between the two sine waves(A)-(B).
> //---------------------
> I'd like to receice some advice about implementing an FM demodulator in
> a DSP proccessor (TMS320C6416T). My baseband frequency is 200Khz max on
> the input ,after downconverting from 88-108Mhz , and here are my issues.
> //---------------------
> I am doing a project on secure speech communication where i m trying
> to find out a suitable encrytion algorithm which has got low
> complexity as well high efficiency but before that i m trying to
> implement GSM 06.10 speech coder in MATLAB i ve implemented the coder
> but skipped a large no. of block this is where i got stuck up.
> //---------------------
> Hello Mr.s,
> I'm a student of eletronics and I woul like implement a voice scrambler
> using the algorithm RSA on plataform DSP tms320c6711, I have watched
> that you done some algorithms to work in a DSP plataform, maybe you can
> help me with this,If dou you have any  RSA algorithms for DSP, I'll be
> very gratefully. Thanks sincerely
> //-----------------------
> I would like to ask you something. I'm implementing a GMSK demodulator
> based on Laurent decomposition. After using the matched filter, the
> baseband signal has to be sampled. Could you recommend me a suitable
> synchronization algorithm? I read about Gardner's method but I'm not
> sure about how it works. Is the interpolation needed if the signal has
> already 5 samples per bit?
> //------------------------
> I am a student at the Ohio State Univ, USA and presently for my project
> I am trying to implement the frequency domain LMS algo.
> I have a few questions for which I seek help. They are as follows:
> 1) Do we have to start with a FD (frequency domain) weight vector that
> is all zeros (wont this give a zero output).
> 2) How do you convert the FD weight vector to time domain (My FD weight
> vector makes sense (in matlab simuation), but in time domain
> (IFFT(W(k+1)/N)it is a periodic vector (vector with periodic data elements).
> 3) How do you decide the convering factor 'mu' and or the converging
> matrix. what are its typical values.
> 4) Can the algo. be applied to complex signals as it is?
> //---------------------------
>     I am [...] an M tech student in embedded systems from Kerala, India.
> Right now I have completed my first year and now I have a full one year
> Project work. And I selected Active noise control as my topic. I have a
> good basic with digital signal processing. I would like to have ur
> suggestions on this topics.I hope u can give valuable suggestion that
> can help me to improve my work.
> With prayers
> //----------------------------
>
> VLV

A lot of people receive junk e-mail. Not everyone dumps theirs on
comp.dsp.

Dale B. Dalrymple

On Jun 11, 10:42&#4294967295;am, xsong <songxiao...@gmail.com> wrote:
> On Jun 11, 9:04&#4294967295;am, "cpshah99" <cpsha...@rediffmail.com> wrote:
>
> > Hi Julius
>
> > So the observations are as follows:
>
> > System Parameters:
>
> > Delay spread of channel is 200 symbols, QPSK modultion, 511 long training
> > sequence, rate 1/2 turbo code, adaptive DFE with LMS algorithm. The FF taps
> > =50 and FB taps =50. The channel has 5 taps and taps are time varying.
>
> If the channel has 5 taps, how can the delay spread be 200 symbols?
>
>

He already chose for us the modulation and symbol rate. So much for a
complete system description.

Cintan, give the information in real world values please. Not
normalized to your baseband.


>
> > So I fixed the SNR at 14 dB and after one time equalization and 3 turbo
> > iterations I am getting BER of 10^-5. Which I think is quite good.
>
> > Next I will be more specific when I ask any que.
>
> > Thanks as always
>
> > Chintan

On Jun 11, 11:31&#4294967295;am, Vladimir Vassilevsky <antispam_bo...@hotmail.com>
wrote:
> xsong wrote:
> > If the channel has 5 taps, how can the delay spread be 200 symbols?
>
> Irrelevant.
> Think of rake receiver.
>
> VLV

For any receiver including rake, the number of taps of a channel is
delta_t*W, where delta_t is
the delay spread in time and W is the sampling bandwidth. 1/W is
symbol duration T. So if the channel has 5 taps,
then the maximum delay spread is 5 symbols.

Hi Guys

I thot when I say the delay spread is 200 symbols, it will make sense.

But anyways. My delay spread in time is 50ms. and data rate is 4ksys/sec
so the delay spread in terms of symbol is 200. 

I think I clearly mentioned that modulation is QPSK.

Now, as VLV said, it is irrelevant because when I say 5 taps mean 5
distinguishable paths, that deos nt mean they r spaced with a symbol
delay.

The channel that I have modelled is realistic in the sense that it is
modelled from a given geometry of ocean. And I am supremely confident tht
the modelling is correct.

So, I think the experiment that i did, where my DFE is shorter than
channel impulse response, is not bad (if not good).

Thanks

Chintan

On Jun 11, 12:36&#4294967295;pm, "cpshah99" <cpsha...@rediffmail.com> wrote:
> Hi Guys
>
> I thot when I say the delay spread is 200 symbols, it will make sense.
>
> But anyways. My delay spread in time is 50ms. and data rate is 4ksys/sec
> so the delay spread in terms of symbol is 200.
>
> I think I clearly mentioned that modulation is QPSK.
>
> Now, as VLV said, it is irrelevant because when I say 5 taps mean 5
> distinguishable paths, that deos nt mean they r spaced with a symbol
> delay

There you go.

>
> The channel that I have modelled is realistic in the sense that it is
> modelled from a given geometry of ocean. And I am supremely confident tht
> the modelling is correct.
>
> So, I think the experiment that i did, where my DFE is shorter than
> channel impulse response, is not bad (if not good).
>

It depends on how your channel delay profile is like.  If 99% of the
power is in the first tap, you
do not even need an equalizer.


> Thanks
>
> Chintan

On Jun 11, 12:36&#4294967295;pm, "cpshah99" <cpsha...@rediffmail.com> wrote:
> Hi Guys
>
> I thot when I say the delay spread is 200 symbols, it will make sense.
>
> But anyways. My delay spread in time is 50ms. and data rate is 4ksys/sec
> so the delay spread in terms of symbol is 200.
>
[snip]

And you chose the modulation and data rate because .... ?

And the distribution of energy in the channel is  ... ?

Still not enough information for me to even start recommending what
to consider.

Julius

]
>
>And you chose the modulation and data rate because .... ?

The reason for QPSK is that for underwater applications you cant go for
higher order modulations, you will face more trouble. 

And the data rate selection is because hardly you are going to have good
channel bandwidth.

Also the carrier frequency is 12kHz. But that is not an issue because I
did simulation at baseband.

>
>And the distribution of energy in the channel is  ... ?

The distribution of the energy is not like what xsong said. It is
distributed by taking into acount reflections loss at surface and bottom. 

The direct path has 40% of total energy and rest is distributed
accordingly.

It is pretty bad channel. This modelling of channel is for short range 
where you will see that the channel is very nasty.

>
>Still not enough information for me to even start recommending what
>to consider.
>
>Julius


Chintan

On Jun 12, 4:19&#4294967295;am, "cpshah99" <cpsha...@rediffmail.com> wrote:
> ]
>
>
>
> >And you chose the modulation and data rate because .... ?
>
> The reason for QPSK is that for underwater applications you cant go for
> higher order modulations, you will face more trouble.
>
> And the data rate selection is because hardly you are going to have good
> channel bandwidth.
>
> Also the carrier frequency is 12kHz. But that is not an issue because I
> did simulation at baseband.
>
>
>
> >And the distribution of energy in the channel is &#4294967295;... ?
>
> The distribution of the energy is not like what xsong said. It is
> distributed by taking into acount reflections loss at surface and bottom.
>
> The direct path has 40% of total energy and rest is distributed
> accordingly.
>
> It is pretty bad channel. This modelling of channel is for short range
> where you will see that the channel is very nasty.
>
>
>
> >Still not enough information for me to even start recommending what
> >to consider.
>
> >Julius
>
> Chintan

Based on your other posts you are trying to replicate what Stojanovic
et al did at MIT/WHOI in the late nineties. Similar channel
description and basically same approach with DFE and RLS. But your
choice of data rate is higher than theirs. Why is this so?

Julius