Prev: you guys at the compdsp conference having a good time?
Next: OFDM timing synchronization
From: kadhiem on 10 Apr 2010 10:23
We are currently simulating our final 16QAM Tx/Rx RTL design in ModelSim.
The Rx front end consists of symbol clock recover + matched filter(designed
by a colleague) followed by carrier tracking module(my design).

We first deemed the symbol clock recovery+matched filter (without
connecting carrier tracking yet) as success since all symbols were
recovered if sliced.

We also deemed the carrier tracking as successful on its own if input with
a clean vector of random symbols + a carrier(with some noise)

However when full system was connected, it failed badly with some
frequencies. To my surprise I found that the input from my colleague's
clock recovery+matched filter to my carrier tracking was very noisy(SNR no
better than 10dB??). My verdict was that we need to improve the clock
recovery output and that while slicer is resistant to noise, the carrier
tracking is not...
My first question is Am I right?
secondly, how can we improve SNR given that:
clock recovery is based on Gardner error detector + 1st order
interpolator.
The Tx signal is upsampled by 2. The sample rate is 560MHz (split in two
parralel 280MHz).

The carrier tracking is based on Costas loop.

Many thanks for any contribution.

From: John on 10 Apr 2010 12:17
On Apr 10, 10:23 am, "kadhiem" <kadhiem_ayob(a)n_o_s_p_a_m.yahoo.co.uk>
wrote:
> We are currently simulating our final 16QAM Tx/Rx RTL design in ModelSim.
> The Rx front end consists of symbol clock recover + matched filter(designed
> by a colleague) followed by carrier tracking module(my design).
>
> We first deemed the symbol clock recovery+matched filter (without
> connecting carrier tracking yet) as success since all symbols were
> recovered if sliced.
>
> We also deemed the carrier tracking as successful on its own if input with
> a clean vector of random symbols + a carrier(with some noise)
>
> However when full system was connected, it failed badly with some
> frequencies. To my surprise I found that the input from my colleague's
> clock recovery+matched filter to my carrier tracking was very noisy(SNR no
> better than 10dB??). My verdict was that we need to improve the clock
> recovery output and that while slicer is resistant to noise, the carrier
> tracking is not...
> My first question is Am I right?
> secondly, how can we improve SNR given that:
> clock recovery is based on Gardner error detector + 1st order
> interpolator.
> The Tx signal is upsampled by 2. The sample rate is 560MHz (split in two
> parralel 280MHz).
>
> The carrier tracking is based on Costas loop.
>
> Many thanks for any contribution.

One suggestion: use a higher-order interpolator.

John
 | 
Pages: 1
Prev: you guys at the compdsp conference having a good time?
Next: OFDM timing synchronization