Blind demodulation of things like V.34

Hi all,

I have never implemented blind demodulation of QAM, and seen all the 
problems first hand. However, the material I have seen on the subject is 
all based on despinning the constellation by looking for the corners of 
a square (or squarish) array sticking out. In schemes like V.34 where 
the constellation has been completely rounded, to avoid high energy 
states, are there any practical schemes what will let you latch on to 
the constellation, and track the carrier? I can see some non-practical 
ones, based on masses of trial and error, but what about something you 
could put in a product?

Regards,
Steve

Steve Underwood wrote:

> I have never implemented blind demodulation of QAM, and seen all the 
> problems first hand. However, the material I have seen on the subject is 
> all based on despinning the constellation by looking for the corners of 
> a square (or squarish) array sticking out. 

(snip)

My understanding of QAM, mostly from the early days, was the you square
the signal which allows you to find the transitions, assuming a 
sufficient density of transitions.  Scramblers are used to generate 
transitions in low entropy input data. (Like many zeros in a row.)

--  glen

glen herrmannsfeldt wrote:

> Steve Underwood wrote:
>
>> I have never implemented blind demodulation of QAM, and seen all the 
>> problems first hand. However, the material I have seen on the subject 
>> is all based on despinning the constellation by looking for the 
>> corners of a square (or squarish) array sticking out. 
>
>
> (snip)
>
> My understanding of QAM, mostly from the early days, was the you square
> the signal which allows you to find the transitions, assuming a 
> sufficient density of transitions.  Scramblers are used to generate 
> transitions in low entropy input data. (Like many zeros in a row.)

In the early days of QAM it was thought that blind demodulation was 
impractical.

Regards,
Steve