On Sun, Apr 22, 2012 at 3:48 AM, George Sklivanitis <[email protected]> wrote: > Hello everyone, > > Lately, I have been trying to tweak the way both digital_bert_tx and > digital_bert_rx are working by introducing in the transmitter side my own > generated random .dat file instead of transmitting an infinite vector of > ones (default) while at the receiver side I consider the transmitted data > known and I am making use of the error_rate block for evaluating the Bit > Error Rate. > > However, the file source that I am using as a reference input in my error > rate block at the receiver seems to introduce overflows after some time at > the receiver as I am repeatedly giving as an input a predefined randomized > sequence of bytes. Moreover, using a scope sink at the output of the > error_rate block I get probability of error close to 1/2 either transmitting > or not. > > Does anybody have an idea of what could be the problem in the above scenario > or even better has tried something similar in the past? > > Thanks, > George > > The experiment takes place in the latest revision of both gnuradio and UHD, > Ubuntu 11.04-64bit in Intel core-i7 laptops.
Generally speaking, your hard drive is probably the slowest component of your computer; this is especially true in laptops (unless you have an SSD or hybrid). It's likely that your file read/write issues are to blame, which is causing samples to be dropped and your BER to go up. Also, make sure you are using differential mods/demods. For the file issue, memory is abundant and cheap these days. Read the file into memory and try it from there. This won't work well if your file is huge, but it will at least tell you if that's the problem. Tom _______________________________________________ Discuss-gnuradio mailing list [email protected] https://lists.gnu.org/mailman/listinfo/discuss-gnuradio
