[Discuss-gnuradio] Problem with Differential PSK

[Christian Sokolowski]

Hi all,

I use differential PSK with up to 4 bits per symbol. I tested transmission as 
a simulation modelling the channel with AWGN noise and phase rotation. So 
far, simulation works perfect (100% packets will be received correct) with 
all modulation orders using Costas Loop for phase recovery and 
Mueller&Mueller for clock recovery at the receiver. 

Problems occur when switching to the USRP board. I use Flex 2400 and TX and RX 
antennas are close to each other (near perfect propagation channel). In very 
few cases (~5%) the receiver recovers the signal and correct packets will be 
received. In most runs, no packet is received. Until now, I do not know if 
the problem is due to some saturation effects (too high transmission power at 
the amplifier) or if parameters of Costas Loop, Automatic Gain Control and 
Mueller&Mueller clock recovery are set properly. I played with all those 
parameters a lot, but without success. Can someone help?

In general: Does Costas Loop need to be modified when differential PSK is 
used? I read in the forum that the order needs to be 2 for BPSK and order 4 
used for QPSK in the block gr.costas_loop_cc(..., order)? What is the 
purpose? 

If someone is interested, you can try with the following source code. I 
appreciate any comments.
http://www.mobnets.rwth-aachen.de/~ade/cso/DPSK_code.zip

the block "howto_psk_packet_sink.cc" is necessary to detect packets after 
differential decoding at the receiver. One byte contains 1 to 4 bits ( which 
is the symbol) depending on the modulation order.

psk.py is extended to use 16PSK.

---------------------------------------------------------------------
Christian Sokolowski     Office: +49(0)-240-7575-7043
RWTH Aachen              FAX:    +49(0)-240-7575-7050
Wireless Networks Dept
Kackertstrasse 9         E-mail: address@hidden
D-52072 Aachen           Web: http://www.mobnets.rwth-aachen.de
---------------------------------------------------------------------