From: Tom R. firstname.lastname@example.org
To: Bill S. email@example.com
Sent: Friday, March 6, 2009 10:55:09 AM
Subject: Re: [Discuss-gnuradio] does the number of samples per symbol
for the dbpsk case really affect the PDR performance?
On Fri, Mar 6, 2009 at 9:38 AM, Bill S.
Bill S. wrote:
Yes, you are correct. There is a 2:1 difference in the interp to decim
rates. However, the bps rate of the digital system is based on the
number of samples per symbol and the interp/decim.
r_tx = r_dac / interp / sps
r_rx = r_adc / decim / sps
Where r_dac is the DAC rate, r_adc is the ADC rate, interp is the
interpolation rate, decim is the decimation rate, and sps is the
samples per symbol.
If you look in pick_bitrate.py, we make some calculations to determine
what the actual bit rate is depending on what parameters are supplied
based on these calculations above.
If you want 500 kbps, then r_dac = 128 MSps, interp = 128, and sps = 2
r_tx = 128e6 / 128 / 2 = 500e3
At the receiver, r_adc = 64 MSps, decim = 64, and sps = 2 Sps
r_rx = 64e6 / 64 / 2 = 500e3
If you had specified a different sps for each side, then the
pick_bitrate calculation would try to attain the specified rate of 500
kbps. It would find that with the receiver using 2, the interp rate
should be 128, but at the transmitter with 4 Sps, the decim rate would
be 32 instead.
Under these conditions, both sides would be able to hit the 500 kbps
rate exactly and you should have no problem.
On the other hand, if you specified BOTH the sps and the decim/interp
rates, then the resulting rate would be different by a factor of two
and therefore your tests would be screwed up. OR if you specified the
different sps and a rate that produces a different rate for both the
transmitter and receiver, then things could be screwed up.
I suggest rerunning both scripts quickly with your exact settings you
used in your experiments and see what the decim, interp, sps, and
resulting TX and RX rates were.
Thank you! I understand what you said! Actually, in our experiment, the
bitrates and number of samples per symbol were specified at both sides
and only these two parameters were specified for our experiment. The
decim and interp will be self-adjusted! I think that is the case!