Multipath ISI removal

IEEE Transaction papers, at least in Signal Proc and Comms (and their many many many many many different journals) are almost always technically correct. Sometimes they are interesting and relevant. Rarely they are groundbreaking. 

The ability to quickly and efficiently go through the chaff to find the few good ones qualify as "technical expertise". I call it "job security". 

On Monday, May 12, 2014 9:28:29 AM UTC+12, robert bristow-johnson wrote:
> On 5/11/14 3:49 PM, gyansorova@gmail.com wrote:
> 
> > On Monday, May 12, 2014 1:29:59 AM UTC+12, robert bristow-johnson wrote:
> 
> >> On 5/11/14 12:12 AM, gyansorova@gmail.com wrote:
> 
> >>
> 
> >>>
> 
> >>
> 
> >>> That's an IEEE Coms transactions paper - won't be rubbish.
> 
> >>
> 
> >>
> 
> >>
> 
> >> no IEEE Transactions are rubbish?
> 
> >>
> 
> >>
> 
> >
> 
> > They can be irrelevant but full Transactions papers are rarely if ever rubbish.
> 
> 
> 
> so, how many irrelevant objects are there in your life that are not rubbish?
> 
> 
> 
> > very hard to get published.
> 
> 
> 
> that may be, but there are lotsa reasons IEEE Transactions are so 
> 
> exclusive (but as each new Transaction is created, it becomes less so), 
> 
> and not all have to do with the quality of scholarship.  there's a lotta 
> 
> mutual backscratching among academics going on there.
> 
> 
> 
> 
> 
> -- 
> 
>
Well there are a lot of very sophisticated channel models out there which very few if any real engineers would ever consider using for example. A mountain of Ph.Ds. so I might call them irrelevant. I had a friend who said that if its in a textbook then it's either Elementary,Irrelevant or Wrong!

This is one of the most classic papers in the field of blind equalization,
and introduces what is now known as CMA (constant modulus algorithm). The
idea was to blindly equalize a constant modulus signal (like 4-QAM) by
minimizing the distance around a circle resulting in dispersion
compensation, and it works completely independently of the phase. It worked
remarkably well with non-constant modulus signals like 16-and higher-QAMs.
Almost used in every practical blind modulator.

Having said that, does the idea of adjusting the modulus irrespective of
the phase to remove ISI make sense in FSK systems when frequency is
actually the rate of change? I think it should. But I'm not sure.	 

_____________________________		
Posted through www.DSPRelated.com

On 5/13/14 3:53 PM, gyansorova@gmail.com wrote:
> On Monday, May 12, 2014 9:28:29 AM UTC+12, robert bristow-johnson wrote:
>> On 5/11/14 3:49 PM, gyansorova@gmail.com wrote:
>>
>>> On Monday, May 12, 2014 1:29:59 AM UTC+12, robert bristow-johnson wrote:
>>
>>>> On 5/11/14 12:12 AM, gyansorova@gmail.com wrote:
>>>>> That's an IEEE Coms transactions paper - won't be rubbish.
>>
>>>> no IEEE Transactions are rubbish?
>>> They can be irrelevant but full Transactions papers are rarely if ever rubbish.
>>
>> so, how many irrelevant objects are there in your life that are not rubbish?
>>
>>> very hard to get published.
>>
>> that may be, but there are lotsa reasons IEEE Transactions are so
>> exclusive (but as each new Transaction is created, it becomes less so),
>> and not all have to do with the quality of scholarship.  there's a lotta
>> mutual backscratching among academics going on there.
>>
> Well there are a lot of very sophisticated channel models out there which very few if any real engineers would ever consider using for example. A mountain of Ph.Ds. so I might call them irrelevant. I had a friend who said that if its in a textbook then it's either Elementary,Irrelevant or Wrong!

well, i wouldn't agree with your friend.  textbooks ain't the same as 
journals.  hopefully, things get digested a bit more before they're 
canonized into a decent textbook.  textbooks may be a little behind the 
curve, but i don't think it hurts any DSPer to know the fundamentals in O&S.

-- 

r b-j                  rbj@audioimagination.com

"Imagination is more important than knowledge."

On 5/13/14 10:25 PM, commsignal wrote:
> This is one of the most classic papers in the field of blind equalization,
> and introduces what is now known as CMA (constant modulus algorithm). The
> idea was to blindly equalize a constant modulus signal (like 4-QAM) by
> minimizing the distance around a circle

would "4-QAM" be something similar to QPSK and, when you toss in the 
"constant modulus", you get OQPSK as in 
http://en.wikipedia.org/wiki/Phase-shift_keying#Offset_QPSK_.28OQPSK.29 
?  you can stay on the circle (constant modulus) and move between the 
different paired bits with a minimum of movement around the circle with 
OQPSK.

-- 

r b-j                  rbj@audioimagination.com

"Imagination is more important than knowledge."

>would "4-QAM" be something similar to QPSK and, when you toss in the 
>"constant modulus", you get OQPSK as in 
>http://en.wikipedia.org/wiki/Phase-shift_keying#Offset_QPSK_.28OQPSK.29 
>?  you can stay on the circle (constant modulus) and move between the 
>different paired bits with a minimum of movement around the circle with 
>OQPSK.
>
>-- 
>
>r b-j                  rbj@audioimagination.com
>
>"Imagination is more important than knowledge."

Yes. 4-QAM = QPSK. However the application of this algorithm was not to
keep the modulus constant like in OQPSK. It's just the nature of the work
it does that this name was given.	 

_____________________________		
Posted through www.DSPRelated.com

robert bristow-johnson <rbj@audioimagination.com> wrote:

(snip)

> would "4-QAM" be something similar to QPSK and, when you toss in the 
> "constant modulus", you get OQPSK as in 
> http://en.wikipedia.org/wiki/Phase-shift_keying#Offset_QPSK_.28OQPSK.29 
> ?  you can stay on the circle (constant modulus) and move between the 
> different paired bits with a minimum of movement around the circle with 
> OQPSK.

Seems to me that the idea of QAM is that you change both amplitude
and phase. (So maybe it should be PAM.) 

(I suppose I believe that QAM better describes the chrominance
subcarrier for NTSC, where it is the sum of two quadrature
carries, amplitude modulated with carrier supressed.)

For QAM, they like to draw the constellation of phase and amplitude,
so it should be possible to figure out the optimal (except for
absolute phase) combination for a given number of points.

Think of it as a 2D circle packing problem. How many circles 
(phase amplitude combinations) before you want different
amplitudes. Seems to me that it might be 5 or 6.

-- glen

On Tuesday, May 13, 2014 10:25:45 PM UTC-4, commsignal wrote:
> This is one of the most classic papers in the field of blind equalization,
> and introduces what is now known as CMA (constant modulus algorithm). The
> idea was to blindly equalize a constant modulus signal (like 4-QAM) by
> minimizing the distance around a circle resulting in dispersion
> compensation, and it works completely independently of the phase. It worked
> remarkably well with non-constant modulus signals like 16-and higher-QAMs.
> Almost used in every practical blind modulator.
> 
> Having said that, does the idea of adjusting the modulus irrespective of
> the phase to remove ISI make sense in FSK systems when frequency is
> actually the rate of change? I think it should. But I'm not sure.	 

Yes, we all know this, as should the OP. Considering he/she asked pretty much the same question a few months ago.

https://groups.google.com/forum/#!topic/comp.dsp/kqkhsB3Uy-U

As far as I know CMA for FSK is used as a pre-EQ, it itself doesn't completely remove ISI in the regime of interest. CR Johnson had a nice paper where he used CMA for CPFSK/GSM as a pre-conditioner so that a DD-LMS can use the output and have a hope to demodulate the signal. 

One can go to the trouble of considering either coherent or noncoherent FSK, depending on the tone/symbol spacing, write out the signal space, and see whether CMA makes sense in and by itself. But beyond that point the full answer really depends on the intention of the work: to get a system working, or to write a report, or to publish a paper, etc.?

>As far as I know CMA for FSK is used as a pre-EQ, it itself doesn't
complet=
>ely remove ISI in the regime of interest. CR Johnson had a nice paper
where=
> he used CMA for CPFSK/GSM as a pre-conditioner so that a DD-LMS can use
th=
>e output and have a hope to demodulate the signal.=20

That's always the case. CMA just about brings to the constellation to its
original shape (and phase is off too), and then depends on decision
direction to do the rest.	 

_____________________________		
Posted through www.DSPRelated.com