Throughput with Turbo Encoder

Dear all,
I am new to GNU Radio. I am building a transceiver with turbo
encoder and FQPSK modulator. At the receiver, FQPSK demodulator softly
demodulates the received signals with BCJR algorithm (FQPSK can be
viewed as a two-bit input TCM with 16 states). I simulated the system
using C++ code. The length of a frame into the turbo encoder is 1024.
The generator polynomial is (013, 015). 8 iterations are taken. The
three metrics of alpha, beta and gamma are calculated by using LUT to
speed up the simulation. But I found that it still is a little bit slow.
The turbo decoding operation of a frame takes about 0.064s (the
simulation runs on a computer with 2 CPU of 3.0GHz and 2G memory). So I
think if the turbo decoding is done by software, the throughput will be
rather slow.
I checked the mailing list in 2006-05, in which it is said that
throughput is very slow with extension to turbo decoding. But I am still
wondering with some easy improvements, the throughput can be increased
significantly. Further, if iterative phase synchronization is done by
software, the receiving time of a frame will be intolerable. So I wonder
whether the software define radio is suited for the transceiver that I
mentioned above and GNU Radio is really real-time.


The throughput depends on how optimal and fast the program you write is.
I’m not familar with Turbo decoder theory. In 2007, we used turbo
from IT++ libary ( It is
than that we wrote. It supports a 64 kbps (One CPU) and 384 kbps (maybe
CPUs, I don’t remember.) TD-SCDMA PHY layer in realtime.
Another method is to use SIMD optimization.


2010/1/16 万千 [email protected]

This forum is not affiliated to the Ruby language, Ruby on Rails framework, nor any Ruby applications discussed here.

| Privacy Policy | Terms of Service | Remote Ruby Jobs